AVR-Chat

I'm looking at developing a signaling application for 1.5" scale ride-on
railroads (a hobby of mine).  It involves a long signal bus (up to a couple of
thousand feet) based on RS-485.  I'll also need to run power to operate
ultra-bright LEDs as well as intermittently drive a couple of low voltage, low
power electric motors.  The cables will typically be buried along the track.

My design will involve and AVR-based module at locations along the track to
detect trains present on a section of track, signal other modules about track
occupancy, drive local signal (light) heads and possibly low voltage motors to
move throw track switches.

As I see it, I'll have to bury two cables along the track, one for power, and
one for the RS-485 signal bus (either 2-wire or 4-wire with signal ground to
control common-mode voltage differences).

My question at the moment is what is the best way to power the devices along the
network.  I don't expect a large current draw at each location (module power
plus several ultra-bright LEDs on at once) and at locations that have small
intermittent motors I could place a small rechargeable battery to take the surge
and trickle charge it from the power bus.  Some thoughts:

1) Something like 14-3 house wire carrying 12vdc (since 12vdc is common for
motors, lights, etc.) but I was worried about voltage drop over 1,000's of feet.
Obviously there would be voltage regulators at each module for local logic
power.

2) Same as #1 but use a higher DC voltage (48?) to reduce the current and hence
reduce the voltage drop.  How does one get from 48vdc to 3-5 vdc without huge
losses.  Cost is a factor here.  I realize that I could use a PWM regulator but
are there inexpensive chips for this?

3) Something like 48 VAC.  Easier to step down to a reasonable level but I fear
having a transformer, etc. at each module would increase costs too much.

I'm looking to keeping module costs as low as possible and still have a reliable
system.

Comments?  Other ideas?

Cheers,

Chuck Hackett
"Good judgment comes from experience, experience comes from bad judgment"
7.5" gauge Union Pacific Northern (4-8-4) 844
http://www.whitetrout.net/Chuck

Just a thought ... you can maybe use an implementation of X-10 technology 
and power your devices at the same time. But you'd have to use AC to power 
your devices.

There is a whole set of documents at the Microchip site about implementing 
X-10 with PIC's, including one on powering them directly off 110vac lines.  
I should think it can be done similarly with AVR's. There's no reason a 
lower voltage than 110 volts can't be used. (See below)

Zack


On Tue, 1 Feb 2005, Chuck Hackett wrote:

> I'm looking at developing a signaling application for 1.5" scale ride-on
> railroads (a hobby of mine).  It involves a long signal bus (up to a couple of
> thousand feet) based on RS-485.  I'll also need to run power to operate
> ultra-bright LEDs as well as intermittently drive a couple of low voltage, low
> power electric motors.  The cables will typically be buried along the track.
> 
> My design will involve and AVR-based module at locations along the track to
> detect trains present on a section of track, signal other modules about track
> occupancy, drive local signal (light) heads and possibly low voltage motors to
> move throw track switches.
> 
> As I see it, I'll have to bury two cables along the track, one for power, and
> one for the RS-485 signal bus (either 2-wire or 4-wire with signal ground to
> control common-mode voltage differences).
> 
> My question at the moment is what is the best way to power the devices along the
> network.  I don't expect a large current draw at each location (module power
> plus several ultra-bright LEDs on at once) and at locations that have small
> intermittent motors I could place a small rechargeable battery to take the surge
> and trickle charge it from the power bus.  Some thoughts:
> 
> 3) Something like 48 VAC.  Easier to step down to a reasonable level but I fear
> having a transformer, etc. at each module would increase costs too much.

You wouldn't need it.  See the Microchip Application Note AN236 and think 
about AVR's.

http://ww1.microchip.com/downloads/en/AppNotes/00236a.pdf

> 
> I'm looking to keeping module costs as low as possible and still have a reliable
> system.
> 
> Comments?  Other ideas?
> 
> Cheers,
> 
> Chuck Hackett

Hello Mr. Chuck Hackett,

A word of caution ... any voltage used where people might contact across it
must be properly installed and shielded. Any voltage higher than about
30volts is considered potentially lethal. So some careful design work is
required to study all of the possibilities and just what is required and can
be used.

Somewhere here you must define a node, it's power requirements, power cycle,
does it just need 5 volts, or does it require other voltages, and so on.
Switching regulators are inexpensive and cheap but until the requirements
are defined you don't know that even a simple 5 volt linear regulator won't
do the job without any great problem. You must start with defining the
current requirements then you can figure how to power it by an engineering
study.

I might suggest starting with a drawing of the layout of where everything is
installed, the distances, the potential copper losses and so on. It should
quickly be apparent which direction to go when the analysis and mathematics
of the whole thing is layout out in a clear and understandable fashion.

Figure out how much current each of these nodes requires, what is the
maximum, minimum, are they all on all of the time?, only when a train
approaches, or whatever. Better yet put together a little computer
simulation to see what the power requirements maximum are, what happens when
you run the train, what is the real power needed?, and so on.

Then try 12volts, 18volts, 30volts, AC or DC or if you are going to properly
shield in conduit etc, 110vAC, solar power, or whatever and so on. Study is
required here before proceeding.

Jim

----- Original Message ----- 
From: "Chuck Hackett" <egroupscdh@WhiteTrout.net>
To: <AVR-Chat@yahoogroups.com>
Sent: Tuesday, February 01, 2005 10:12 AM
Subject: [AVR-Chat] Signal System Power


>
> I'm looking at developing a signaling application for 1.5" scale ride-on
> railroads (a hobby of mine).  It involves a long signal bus (up to a
couple of
> thousand feet) based on RS-485.  I'll also need to run power to operate
> ultra-bright LEDs as well as intermittently drive a couple of low voltage,
low
> power electric motors.  The cables will typically be buried along the
track.
>
> My design will involve and AVR-based module at locations along the track
to
> detect trains present on a section of track, signal other modules about
track
> occupancy, drive local signal (light) heads and possibly low voltage
motors to
> move throw track switches.
>
> As I see it, I'll have to bury two cables along the track, one for power,
and
> one for the RS-485 signal bus (either 2-wire or 4-wire with signal ground
to
> control common-mode voltage differences).
>
> My question at the moment is what is the best way to power the devices
along the
> network.  I don't expect a large current draw at each location (module
power
> plus several ultra-bright LEDs on at once) and at locations that have
small
> intermittent motors I could place a small rechargeable battery to take the
surge
> and trickle charge it from the power bus.  Some thoughts:
>
> 1) Something like 14-3 house wire carrying 12vdc (since 12vdc is common
for
> motors, lights, etc.) but I was worried about voltage drop over 1,000's of
feet.
> Obviously there would be voltage regulators at each module for local logic
> power.
>
> 2) Same as #1 but use a higher DC voltage (48?) to reduce the current and
hence
> reduce the voltage drop.  How does one get from 48vdc to 3-5 vdc without
huge
> losses.  Cost is a factor here.  I realize that I could use a PWM
regulator but
> are there inexpensive chips for this?
>
> 3) Something like 48 VAC.  Easier to step down to a reasonable level but I
fear
> having a transformer, etc. at each module would increase costs too much.
>
> I'm looking to keeping module costs as low as possible and still have a
reliable

> system.
>
> Comments?  Other ideas?
>
> Cheers,
>
> Chuck Hackett
> "Good judgment comes from experience, experience comes from bad judgment"
> 7.5" gauge Union Pacific Northern (4-8-4) 844
> http://www.whitetrout.net/Chuck
>
>
>
>
>
> Yahoo! Groups Links
>
>
>
>
>
>
>
>

At 12:12 PM 2/1/05 -0600, Chuck Hackett wrote:
>2) Same as #1 but use a higher DC voltage (48?) to reduce the current and 
>hence
>reduce the voltage drop.  How does one get from 48vdc to 3-5 vdc without huge
>losses.  Cost is a factor here.  I realize that I could use a PWM 
>regulator but
>are there inexpensive chips for this?

Consider 24V, It's a common industrial DC bus voltage so there is a fair 
amount of support for it (also 24VAC).

For local power storage you could also consider "supercaps".  Maxwell (I 
think that's who it is) is making them in D cell packages (2V/cell or 
so).  Recharge is easy but cost and bulk might be an issue.

What about temperature?  Do you need to consider freezing?

Robert

" 'Freedom' has no meaning of itself.  There are always restrictions,
be they legal, genetic, or physical.  If you don't believe me, try to
chew a radio signal. "

                         Kelvin Throop, III

At 05:12 AM 2/02/2005, you wrote:

>As I see it, I'll have to bury two cables along the track, one for power, and
>one for the RS-485 signal bus (either 2-wire or 4-wire with signal ground to
>control common-mode voltage differences).


Can you use the tracks as a ground wire ...just like the real thing?
You will need to make sure there is continuity, but with the size
of the tracks you should get low voltage drop on the common, so your
power wires can just be used for positive/active.

I would be inclined to use AC or you may end up with oxidation.



Regards

John Samperi

******************************************************
                         Ampertronics Pty. Ltd.
   11 Brokenwood Place Baulkham Hills, NSW 2153 AUSTRALIA
          Tel. (02) 9674-6495       Fax (02) 9674-8745
                Email: samperi@ampertronics.com.au
                  Website  http://www.ampertronics.com.au
* Electronic Design   * Custom Products   * Contract Assembly
******************************************************

(replies to multiple messages interspersed below)

> From: Javier Fiasche
> 
> Why dont you try 12 VAC (or less) and a switching regulator 
> at each location  (try using lm2575)

Thanks Javier, I just checked the data sheet on the LM2575 series.  Looks like
just the thing I need.  

(BTW: I assume you mean 12 VDC?  Otherwise I'd have to have a rectifier, etc. at
each location)

One question this brings up:  Do you think the ripple from the output would
greatly effect the A/D converters on the AVR?  The LM2575 data sheet shows an
additional ripple filter if required.  I am early in this design and I don't yet
know what my analog signal (rail occupancy detection) limits will be.  I hope
that they are in the range of 0 to VCC (as opposed to milivolts, etc.) but I
won't know until I do some field testing.  Just wondering if the ripple would
"grossly" effect the A/Ds making them unusable.  I don't think my needs will
require an external A/D reference.

Also, how sensitive is the AVR itself (ATMega16 for example) to ripple on VCC?

I realize I have a lot of characteristics to measure/calculate.  I'm just trying
to establish a basic direction to go, throwing out alternative designs which are
unworkable at a gross level.


> From: Zack Widup
> 
> Just a thought ... you can maybe use an implementation of 
> X-10 technology and power your devices at the same time. But 
> you'd have to use AC to power your devices.

Hi Zack,

I looked at X-10 but rejected it for several reasons.  One is that these ride-on
railroad tracks are many times located in public parks.  I doubt if they would
take kindly to running 120 VAC around and the conduit required would add to the
costs fast - and the grounds people have been known to sever cables when digging
in shrubs, etc. :-)

Another reason is that (to my knowledge) X-10 uses a "best effort" delivery
method (I think the message is sent three times) and there is no high level
acknowledgement.  This ACK could be added on top but my message traffic will
require a larger bandwidth than X-10 can support.


> From: James Hatley
> 
> A word of caution ... any voltage used where people might 
> contact across it must be properly installed and shielded. 
> Any voltage higher than about 30volts is considered 
> potentially lethal.

Hi James, I knew I didn't want to run 120 VAC but I thought 48 VAC was
considered "low voltage" (like in HVAC controls, etc.).  I think 12 or 24 VDC
will work fine with the switching regulators mentioned above.

> ....
> Somewhere here you must define a node, it's power 
> requirements, power cycle, does it just need 5 volts, or does 
> it require other voltages, and so on.

Understood, as I mentioned above, I'm trying to throw out some alternatives at a
gross level - like 120 VAC in the power area and, in the data bus area, methods
that don't have good noise immunity, have low bandwidth, etc. - before I start
field testing.

> .... (after determining power requirements) ....
> Then try 12volts, 18volts, 30volts, AC or DC or if you are 
> going to properly shield in conduit etc, 110vAC, solar power, 
> or whatever and so on. Study is required here before proceeding.

Something requiring conduit would probably have to be avoided (live steam clubs
don't usually have large budgets).  I do see a possibility of using solar power
to charge batteries at remote locations far from "an outlet".  Someday I can
foresee signal bus segments being "bridged" using RF links.

Thanks for the input.  I certainly agree that I need to design the module, field
test it, determine how many would be on a typical run, over what distance and
drawing what current, etc.

 
> From: Robert Adsett
> 
> ....
> Consider 24V, It's a common industrial DC bus voltage so 
> there is a fair amount of support for it 

I think 12 VDC is my current preference if the current draw requirements will
support it since then 12 volt items like light bulbs, motors, etc. can be
powered directly which would simplify things.  My next preference would be 24
VDC since, being a multiple of 12 VDC power can be supplied by common batteries
with solar cell chargers, trickle chargers, etc.

> (also 24VAC).

I'd like to stay away from AC if I can to minimize parts requirements.

> For local power storage you could also consider "supercaps".  
> Maxwell .... is making them in D cell 
> packages ....  Recharge is easy but cost and bulk 
> might be an issue.

Bulk is not an issue but cost would be.

> What about temperature?  Do you need to consider freezing?

Some clubs would have these units permanently installed so they would have to
stand freezing temps.  Others would collect them after a run day due to the
possibility of theft or vandalism.
 

> From: John Samperi
> ....
> At 05:12 AM 2/02/2005, you wrote:
> 
> >As I see it, I'll have to bury two cables along the track, one for 
> >power, and one for the RS-485 signal bus (either 2-wire or 4-wire with 
> >signal ground to control common-mode voltage differences).
>  
> Can you use the tracks as a ground wire ...just like the real thing?
> You will need to make sure there is continuity, but with the 
> size of the tracks you should get low voltage drop on the 
> common, so your power wires can just be used for positive/active.

Hi John, One of the problems with our scale rail is that it is an aluminum
extrusion in 10 or 12 foot lengths - good conductor, but typically the rails are
joined with small steel plates and steel (or SS) screws.  This leads to high
resistance joints between the rails after being in the weather for awhile.  When
using the rail as a ground these high resistance joints would play havoc with
ground potentials.  

Clubs that use relay based signaling typically have to "bond" all the joints
(attach the rails together with a wire jumper across the joints as in full
size).  As relates to detecting occupancy of a track, we also have the problem
that if the (wood) ties are wet, the conductivity of the track circuit varies
over time causing false signals.  I'm hoping to be able to get around both of
these problems by using the A/D in the AVRs to monitor the rail and detect
changes in conductivity as opposed to an absolute level of conductivity, which
is what the relay systems do (i.e.: the relays are set to pull in at a given
conductivity/current flow).

Thanks to all for your suggestions.  For those that have no idea what I'm
talking about when I say "live steam" tracks, here are some photos of my
locomotive:

http://www.whitetrout.net/Chuck/844 in the lower right you can see one of the
rail joints.  The lady behind me owns half of the locomotive - i.e.: she's my
wife :-)

http://www.whitetrout.net/Chuck/844/photos

I'm sure I'll have more specific AVR questions down the road (track?) as I
progress.  Even though I'm new to AVRs I enjoy the list and am learning a lot
just "lurking".  Hopefully some day I'll be knowledgeable enough to contribute
:-)

Cheers,

Chuck Hackett
"Good judgment comes from experience, experience comes from bad judgment"
7.5" gauge Union Pacific Northern (4-8-4) 844
http://www.whitetrout.net/Chuck

All,

  I'm currently using a MEGA16 at 16Mhz for a project, and I'm doing a lot
of floating point calculations (most exponential) and its getting a bit
slow, so I need something with a bit more 'umph'.
I know its kinda off topic, but, what I'm after (I think) is something 16Bit
and a minimum of 25MIPs (32 would be preffered). it also *must* have an
affordable C compiler (ie <£400) and an evaluation board. Surface mount is
fine (in fact preffered).
   I figured you guys would be the best place to start asking as you've got
the experience (probably), Anyone have any suggestions?

Paul

Lasse,

>Philips LPC2106

<ROFLMAO>
Sorry, I forgot to say, I want something a decent speed IO.
I got one of these (MB40 eval kit), I thought great 60Mhz internal clocking,
nice fast processor.
Problem is, the IO is limited to 3.8Mhz, Sorry I should've also said I
needed around 8Mhz IO.
Also, I had emmense problems getting the code to load into it, it kept
crashing windows and dieing, all sorts of mayhem.

Paul

----- Original Message ----- 

From: Paul Maddox
To: AVR-Chat@yahoogroups.com
Sent: Wednesday, February 02, 2005 9:35 AM
Subject: [AVR-Chat] something a bit faster


All,

  I'm currently using a MEGA16 at 16Mhz for a project, and I'm doing a lot
of floating point calculations (most exponential) and its getting a bit
slow, so I need something with a bit more 'umph'.
I know its kinda off topic, but, what I'm after (I think) is something 16Bit
and a minimum of 25MIPs (32 would be preffered). it also *must* have an
affordable C compiler (ie <£400) and an evaluation board. Surface mount is
fine (in fact preffered).
   I figured you guys would be the best place to start asking as you've got
the experience (probably), Anyone have any suggestions?

Philips LPC2000 family. Join the group:

http://groups.yahoo.com/group/lpc2000/

Leon
--
Leon Heller, G1HSM
http://www.geocities.com/leon_heller




-- 
No virus found in this outgoing message.
Checked by AVG Anti-Virus.
Version: 7.0.300 / Virus Database: 265.8.4 - Release Date: 01/02/2005

Leon,

> Philips LPC2000 family. Join the group:
>
> http://groups.yahoo.com/group/lpc2000/

thanks, but see my other email, Been there, done that, anyone want an MB40
with LCD and LPC2106?

Paul

----- Original Message ----- 

From: Paul Maddox
To: AVR-Chat@yahoogroups.com
Sent: Wednesday, February 02, 2005 10:09 AM
Subject: Re: [AVR-Chat] something a bit faster


Lasse,

>Philips LPC2106

<ROFLMAO>
Sorry, I forgot to say, I want something a decent speed IO.
I got one of these (MB40 eval kit), I thought great 60Mhz internal clocking,
nice fast processor.
Problem is, the IO is limited to 3.8Mhz, Sorry I should've also said I
needed around 8Mhz IO.
Also, I had emmense problems getting the code to load into it, it kept
crashing windows and dieing, all sorts of mayhem.


Yes, the slow I/O is rather irritating. I was quite perplexed when I tried 
toggling an output as fast as possible.

Leon 



-- 
No virus found in this outgoing message.
Checked by AVG Anti-Virus.
Version: 7.0.300 / Virus Database: 265.8.4 - Release Date: 01/02/2005

----- Original Message ----- 

From: Mike Harrison
To: AVR-Chat@yahoogroups.com
Sent: Wednesday, February 02, 2005 10:19 AM
Subject: Re: [AVR-Chat] something a bit faster


On Wed, 2 Feb 2005 10:10:57 -0000, you wrote:

>
>----- Original Message ----- 
>From: Paul Maddox
>To: AVR-Chat@yahoogroups.com
>Sent: Wednesday, February 02, 2005 9:35 AM
>Subject: [AVR-Chat] something a bit faster
>
>
>All,
>
>  I'm currently using a MEGA16 at 16Mhz for a project, and I'm doing a lot
>of floating point calculations (most exponential) and its getting a bit
>slow, so I need something with a bit more 'umph'.
>I know its kinda off topic, but, what I'm after (I think) is something 
>16Bit
>and a minimum of 25MIPs (32 would be preffered). it also *must* have an
>affordable C compiler (ie <£400) and an evaluation board. Surface mount is
>fine (in fact preffered).
>   I figured you guys would be the best place to start asking as you've got
>the experience (probably), Anyone have any suggestions?

ARM would be an obvous choice, but if it's mostly maths intensive maybe you 
should be looking at DSP
type devices - DsPIC perhaps.
Atmel also do a SRAM based AVR that runs at something like 35MHz from SRAM, 
loaded from an EEPROM -
can't remember the number but it's one of their USB parts



I've used the Analog Devices 16-bit fixed-point ADSP-2187 for mixed 
floating-point, fixed-point and integer calculations (Kalman filter), 
running at 30 MIPS. It worked very well. They are quite cheap, but awkward 
to use (PQFP and lots of pins). The I/O is nice and fast, of course. It's 
reather old now and the newer Blackfin will run at over 500 MIPS, if you 
need that sort of performance.

Leon 



-- 
No virus found in this outgoing message.
Checked by AVG Anti-Virus.
Version: 7.0.300 / Virus Database: 265.8.4 - Release Date: 01/02/2005

Paul Maddox wrote:
> All,
> 
>   I'm currently using a MEGA16 at 16Mhz for a project, and I'm doing a lot
> of floating point calculations (most exponential) and its getting a bit
> slow, so I need something with a bit more 'umph'.
> I know its kinda off topic, but, what I'm after (I think) is something 16Bit
> and a minimum of 25MIPs (32 would be preffered). it also *must* have an
> affordable C compiler (ie <£400) and an evaluation board. Surface mount is
> fine (in fact preffered).
>    I figured you guys would be the best place to start asking as you've got
> the experience (probably), Anyone have any suggestions?
> 
> Paul

A Renesas SuperH. http://www.superh.com/products/architecture.htm

On Wed, 2 Feb 2005 10:10:57 -0000, you wrote:

>
>----- Original Message ----- 
>From: Paul Maddox
>To: AVR-Chat@yahoogroups.com
>Sent: Wednesday, February 02, 2005 9:35 AM
>Subject: [AVR-Chat] something a bit faster
>
>
>All,
>
>  I'm currently using a MEGA16 at 16Mhz for a project, and I'm doing a lot
>of floating point calculations (most exponential) and its getting a bit
>slow, so I need something with a bit more 'umph'.
>I know its kinda off topic, but, what I'm after (I think) is something 16Bit
>and a minimum of 25MIPs (32 would be preffered). it also *must* have an
>affordable C compiler (ie <£400) and an evaluation board. Surface mount is
>fine (in fact preffered).
>   I figured you guys would be the best place to start asking as you've got
>the experience (probably), Anyone have any suggestions?

ARM would be an obvous choice, but if it's mostly maths intensive maybe you should be looking at DSP
type devices - DsPIC perhaps.
Atmel also do a SRAM based AVR that runs at something like 35MHz from SRAM, loaded from an EEPROM -
can't remember the number but it's one of their USB parts

I meant 12 VAC, and of course rectifing it at each location. LM2575 is a
very good switching regulator, try usin g ono of the design examples on
National Semi or the soft "switchers made simple " to design yours, i dont
thing ripple would be a problem, besides you dont seem to need to much
accurancy on the detection ( use some king of hysteresis)


-----Mensaje original-----
De: Chuck Hackett [mailto:egroupscdh@WhiteTrout.net]
Enviado el: Miercoles, 02 de Febrero de 2005 02:38 a.m.
Para: AVR-Chat@yahoogroups.com
Asunto: RE: [AVR-Chat] Signal System Power


(replies to multiple messages interspersed below)

> From: Javier Fiasche
>
> Why dont you try 12 VAC (or less) and a switching regulator
> at each location  (try using lm2575)

Thanks Javier, I just checked the data sheet on the LM2575 series.  Looks
like
just the thing I need.

(BTW: I assume you mean 12 VDC?  Otherwise I'd have to have a rectifier,
etc. at
each location)

One question this brings up:  Do you think the ripple from the output would
greatly effect the A/D converters on the AVR?  The LM2575 data sheet shows
an
additional ripple filter if required.  I am early in this design and I don't
yet
know what my analog signal (rail occupancy detection) limits will be.  I
hope
that they are in the range of 0 to VCC (as opposed to milivolts, etc.) but I
won't know until I do some field testing.  Just wondering if the ripple
would
"grossly" effect the A/Ds making them unusable.  I don't think my needs will
require an external A/D reference.

Also, how sensitive is the AVR itself (ATMega16 for example) to ripple on
VCC?

I realize I have a lot of characteristics to measure/calculate.  I'm just
trying
to establish a basic direction to go, throwing out alternative designs which
are
unworkable at a gross level.


> From: Zack Widup
>
> Just a thought ... you can maybe use an implementation of
> X-10 technology and power your devices at the same time. But
> you'd have to use AC to power your devices.

Hi Zack,

I looked at X-10 but rejected it for several reasons.  One is that these
ride-on
railroad tracks are many times located in public parks.  I doubt if they
would
take kindly to running 120 VAC around and the conduit required would add to
the
costs fast - and the grounds people have been known to sever cables when
digging
in shrubs, etc. :-)

Another reason is that (to my knowledge) X-10 uses a "best effort" delivery
method (I think the message is sent three times) and there is no high level
acknowledgement.  This ACK could be added on top but my message traffic will
require a larger bandwidth than X-10 can support.


> From: James Hatley
>
> A word of caution ... any voltage used where people might
> contact across it must be properly installed and shielded.
> Any voltage higher than about 30volts is considered
> potentially lethal.

Hi James, I knew I didn't want to run 120 VAC but I thought 48 VAC was
considered "low voltage" (like in HVAC controls, etc.).  I think 12 or 24
VDC
will work fine with the switching regulators mentioned above.

> ....
> Somewhere here you must define a node, it's power
> requirements, power cycle, does it just need 5 volts, or does
> it require other voltages, and so on.

Understood, as I mentioned above, I'm trying to throw out some alternatives
at a
gross level - like 120 VAC in the power area and, in the data bus area,
methods
that don't have good noise immunity, have low bandwidth, etc. - before I
start
field testing.

> .... (after determining power requirements) ....
> Then try 12volts, 18volts, 30volts, AC or DC or if you are
> going to properly shield in conduit etc, 110vAC, solar power,
> or whatever and so on. Study is required here before proceeding.

Something requiring conduit would probably have to be avoided (live steam
clubs
don't usually have large budgets).  I do see a possibility of using solar
power
to charge batteries at remote locations far from "an outlet".  Someday I can
foresee signal bus segments being "bridged" using RF links.

Thanks for the input.  I certainly agree that I need to design the module,
field
test it, determine how many would be on a typical run, over what distance
and
drawing what current, etc.


> From: Robert Adsett
>
> ....
> Consider 24V, It's a common industrial DC bus voltage so
> there is a fair amount of support for it

I think 12 VDC is my current preference if the current draw requirements
will
support it since then 12 volt items like light bulbs, motors, etc. can be
powered directly which would simplify things.  My next preference would be
24
VDC since, being a multiple of 12 VDC power can be supplied by common
batteries
with solar cell chargers, trickle chargers, etc.

> (also 24VAC).

I'd like to stay away from AC if I can to minimize parts requirements.

> For local power storage you could also consider "supercaps".
> Maxwell .... is making them in D cell
> packages ....  Recharge is easy but cost and bulk
> might be an issue.

Bulk is not an issue but cost would be.

> What about temperature?  Do you need to consider freezing?

Some clubs would have these units permanently installed so they would have
to
stand freezing temps.  Others would collect them after a run day due to the
possibility of theft or vandalism.


> From: John Samperi
> ....
> At 05:12 AM 2/02/2005, you wrote:
>
> >As I see it, I'll have to bury two cables along the track, one for
> >power, and one for the RS-485 signal bus (either 2-wire or 4-wire with
> >signal ground to control common-mode voltage differences).
>
> Can you use the tracks as a ground wire ...just like the real thing?
> You will need to make sure there is continuity, but with the
> size of the tracks you should get low voltage drop on the
> common, so your power wires can just be used for positive/active.

Hi John, One of the problems with our scale rail is that it is an aluminum
extrusion in 10 or 12 foot lengths - good conductor, but typically the rails
are
joined with small steel plates and steel (or SS) screws.  This leads to high
resistance joints between the rails after being in the weather for awhile.
When
using the rail as a ground these high resistance joints would play havoc
with
ground potentials.

Clubs that use relay based signaling typically have to "bond" all the joints
(attach the rails together with a wire jumper across the joints as in full
size).  As relates to detecting occupancy of a track, we also have the
problem
that if the (wood) ties are wet, the conductivity of the track circuit
varies
over time causing false signals.  I'm hoping to be able to get around both
of
these problems by using the A/D in the AVRs to monitor the rail and detect
changes in conductivity as opposed to an absolute level of conductivity,
which
is what the relay systems do (i.e.: the relays are set to pull in at a given
conductivity/current flow).

Thanks to all for your suggestions.  For those that have no idea what I'm
talking about when I say "live steam" tracks, here are some photos of my
locomotive:

http://www.whitetrout.net/Chuck/844 in the lower right you can see one of
the
rail joints.  The lady behind me owns half of the locomotive - i.e.: she's
my
wife :-)

http://www.whitetrout.net/Chuck/844/photos

I'm sure I'll have more specific AVR questions down the road (track?) as I
progress.  Even though I'm new to AVRs I enjoy the list and am learning a
lot
just "lurking".  Hopefully some day I'll be knowledgeable enough to
contribute
:-)

Cheers,

Chuck Hackett
"Good judgment comes from experience, experience comes from bad judgment"
7.5" gauge Union Pacific Northern (4-8-4) 844
http://www.whitetrout.net/Chuck




----------------------------------------------------------------------------
----
Yahoo! Groups Links

  a.. To visit your group on the web, go to:
  http://groups.yahoo.com/group/AVR-Chat/

  b.. To unsubscribe from this group, send an email to:
  AVR-Chat-unsubscribe@yahoogroups.com

  c.. Your use of Yahoo! Groups is subject to the Yahoo! Terms of Service.

______________________________
Visita http://www.tutopia.com y comienza a navegar m�s r�pido en Internet. Tutopia es Internet para todos.

At 16:07 02/02/2005, you wrote:
>(replies to multiple messages interspersed below)

ditto

> > From: James Hatley
> >
> > A word of caution ... any voltage used where people might
> > contact across it must be properly installed and shielded.
> > Any voltage higher than about 30volts is considered
> > potentially lethal.

Not quite, in the EU, US and Oz any voltage below 60Vdc is considered to be 
SELV (safety extra low voltage), in other words safe.   Above 60V the 
typical contact resistance of skin is such that currents considered 
unhealthy will start to flow if you come in to contact with them.

If you are going to design any power supply that people can inadvertently 
come into contact with be it directly or indirect via your circuits then 
your power supply shall comply with SELV requirements.   This means that 
you must comply with creepage and clearance and basic electrical safety in 
your power supply design.   Obtain a copy of your countries relevant safely 
standard and follow it explicitly.   You would not believe the difference 
between the relevant big players, US, EU, Asia & Aus/Nz :-)

> > From: Robert Adsett
> >
> > ....
> > Consider 24V, It's a common industrial DC bus voltage so
> > there is a fair amount of support for it
>
>I think 12 VDC is my current preference if the current draw requirements will
>support it since then 12 volt items like light bulbs, motors, etc. can be
>powered directly which would simplify things.  My next preference would be 24
>VDC since, being a multiple of 12 VDC power can be supplied by common 
>batteries
>with solar cell chargers, trickle chargers, etc.
>
> > (also 24VAC).
>
>I'd like to stay away from AC if I can to minimize parts requirements.

I'd give your connections about 2-3 months of reliable service then the 
corrosion gremlin will come to visit, afterwards you'll be pulling your 
hair out trying to figure out what is going on.   Having any DC flow and 
dissimilar metals present will generate/accelerate corrosion faster than 
you would consider possible.

If you are serious at using DC then you can use sacrificial anodes and 
negative rails (ie -50Vdc) to ensure that you force the anodes to be 
depleted and not your cabling.  This is what a lot of Telco's use here in 
Australia in their POTS exchanges.

Otherwise a AC supply would be a better bet.   I would seriously consider a 
24VAC system.   The HVAC industry has been using this for years and for a 
very good reason :-)   I would keep somewhere around 24-30VAC (don't forget 
to take into consideration max mains voltages on transformer primary!) that 
way your rectified DC voltage is kept below 50V.   The suggestion of the 
national simple switchers LM2574 et al is a good suggestion, get the -HV 
parts for good measure (55V max input voltage).

If an AC power system is used then make sure that you use full wave 
rectification not half wave.   This is to prevent unwanted DC biases 
flowing due to the current imbalance generated by half wave 
rectifiers.   Any DC flowing in your supply circuit will negate the 
effectiveness of the AC supply.

 > From: John Samperi
> > ....
> > At 05:12 AM 2/02/2005, you wrote:
> >
> > >As I see it, I'll have to bury two cables along the track, one for
> > >power, and one for the RS-485 signal bus (either 2-wire or 4-wire with
> > >signal ground to control common-mode voltage differences).
> >
> > Can you use the tracks as a ground wire ...just like the real thing?
> > You will need to make sure there is continuity, but with the
> > size of the tracks you should get low voltage drop on the
> > common, so your power wires can just be used for positive/active.

<Snip>

RS485 might not be a good option in this scenario for two reasons.   The 
first is the physical length of your wires are very long @1000' meaning 
lightning and other electrical discharges are not your friend.   Second I 
wouldn't bother trying to lightening proof your system from a direct 
strike, but near strikes will induce large currents in long wires buried in 
the ground.   The effect of this is very high differential voltages with 
respect to ground.   I've found that RS485 transceivers can be notoriously 
unreliable under these conditions.

For data communications why not consider the old faithful 4-20mA current 
loop ?   From what you have described your data speed requirements would be 
low, so a 75-300baud 4-20mA current loop would be well suited to your 
needs.   Did I mention this is what a number of "full sized" railways still 
use today for communication over long wire runs.   Better yet, no current 
flows in the idle state, so no corrosion problems :-)   If you keep the 
average bit ratio to 0.5 (ie successive '1010' combinations) then there is 
effective no DC flowing and the data is an AC signal.

Since we are on the topic of railways and safety if your club puts more 
than one train on the line and you are thinking of using signals to prevent 
collision then please always think safety, safety & safety when designing 
your systems.   If you have never done a Failure Mode & Effect Analysis 
(FMEA) on your circuit then have a read of the following link.

http://www.1stnclass.com/risk_analysis.htm

I've worked on Railways signalling systems here in Australia and every 
component from the lamp, electronics, software to the relay that 
electrically switches the lamp are included in the FMEA.   The most complex 
computer controlled signalling system in use here always ends up in a very 
crusty open frame relay to drive any lamp or switch/contact closure.  The 
design of this relay is some 90 years old and has yet to be ever 
changed.    The reason this relay is in use is it will *always* fail open 
circuit, 90 years of testing has supported this decision.   The effect of 
this "safe relay" is that no lamp or switch can ever be put into an unknown 
state, meaning the green lamp will never illuminate when the red should be 
shown.

I'm pretty sure that it would be rare for more than one train to be on the 
track at once, but I felt it worthy of noting, having spent a number of 
years designing safety critical systems that use embedded processors :-)

>Thanks to all for your suggestions.  For those that have no idea what I'm
>talking about when I say "live steam" tracks, here are some photos of my
>locomotive:
>
>http://www.whitetrout.net/Chuck/844

That is so cool !   What is the maximum pull under good track conditions ?

>I'm sure I'll have more specific AVR questions down the road (track?) as I
>progress.  Even though I'm new to AVRs I enjoy the list and am learning a lot
>just "lurking".  Hopefully some day I'll be knowledgeable enough to contribute
>:-)

Just keep the pictures of the trains with AVR's in 'em rolling in and we'll 
all be kept very happy.

Hope that this helps... YMMV

Regards

Matt

Leon,

> ARM would be an obvous choice, but if it's mostly maths intensive maybe
you
> should be looking at DSP type devices - DsPIC perhaps.

There are only a few times I'm using floats, but its enough to slow the
MEGA16 down quite a bit.
A DSP would be overkill.
As for ARM, yeah, they're nice, but I've yet to find a reasonbly priced
SDK/IDE that works (I've had no success with GCC and other free tools)

> Atmel also do a SRAM based AVR that runs at something like 35MHz from
SRAM,
> loaded from an EEPROM -
> can't remember the number but it's one of their USB parts

I'll take a look, thanks.

Paul

Hi Chuck,

Well, it was a thought.  As far as the voltage goes, you could probably 
use any AC voltage greater than about 10 volts or so.  I've seen others 
recommend 24v here.  That should work fine with X-10.  See the circuit in 
that PIC app note for getting 5 volts from the AC power.

I'm still fascinated by this X-10 technology and have been playing around 
with it.

X-10 is rather slow.  I didn't know what kind of data rate you needed.  If 
you're just trying to signal the location of the train every so often, it 
would work with a single byte being sent.

You might consider an RF link using one of the 380 MHz transceiver chips 
that are on the market, and eliminate the wires for the data transfer 
completely.  I don't know if they have a range of 1000 ft, though. I know 
they work up to a couple hundred feet.

Zack

On Tue, 1 Feb 2005, Chuck Hackett wrote:

> 
> > From: Zack Widup
> >
> > Just a thought ... you can maybe use an implementation of
> > X-10 technology and power your devices at the same time. But
> > you'd have to use AC to power your devices.
> 
> Hi Zack,
> 
> I looked at X-10 but rejected it for several reasons.  One is that these ride-on
> railroad tracks are many times located in public parks.  I doubt if they would
> take kindly to running 120 VAC around and the conduit required would add to the
> costs fast - and the grounds people have been known to sever cables when digging
> in shrubs, etc. :-)
> 
> Another reason is that (to my knowledge) X-10 uses a "best effort" delivery
> method (I think the message is sent three times) and there is no high level
> acknowledgement.  This ACK could be added on top but my message traffic will
> require a larger bandwidth than X-10 can support.
>

Paul Maddox wrote:

> Leon,
> 
> 
>>ARM would be an obvous choice, but if it's mostly maths intensive maybe
> 
> you
> 
>>should be looking at DSP type devices - DsPIC perhaps.
> 
> 
> There are only a few times I'm using floats, but its enough to slow the
> MEGA16 down quite a bit.
> A DSP would be overkill.
> As for ARM, yeah, they're nice, but I've yet to find a reasonbly priced
> SDK/IDE that works (I've had no success with GCC and other free tools)
> 
> 
>>Atmel also do a SRAM based AVR that runs at something like 35MHz from
> 
> SRAM,
> 
>>loaded from an EEPROM -
>>can't remember the number but it's one of their USB parts

AT76C713

Regards,

Kat.


-- 
---------------------------------------------------------------
K.A.Q. Electronics	Website: www.kaqelectronics.dyndns.org
IM: Yahoo: PinkyDwaggy  MSN: katinka@kaqelectronics.dyndns.org
For Everything Electronics     Phone: 0419 923 731
---------------------------------------------------------------	


-- 
No virus found in this outgoing message.
Checked by AVG Anti-Virus.
Version: 7.0.300 / Virus Database: 265.8.3 - Release Date: 31/01/2005

At 06:31 AM 2/3/05 +1300, Eric wrote:

Apparently nothing.  Was there any content?  All I got was many blank lines 
(some sort of obscured html)

Robert

" 'Freedom' has no meaning of itself.  There are always restrictions,
be they legal, genetic, or physical.  If you don't believe me, try to
chew a radio signal. "

                         Kelvin Throop, III

At 04:35 AM 2/2/2005, Paul Maddox wrote:

>All,
>
>   I'm currently using a MEGA16 at 16Mhz for a project, and I'm doing a lot
>of floating point calculations (most exponential) and its getting a bit
>slow, so I need something with a bit more 'umph'.
>I know its kinda off topic, but, what I'm after (I think) is something 16Bit
>and a minimum of 25MIPs (32 would be preffered). it also *must* have an
>affordable C compiler (ie <£400) and an evaluation board. Surface mount is
>fine (in fact preffered).
>    I figured you guys would be the best place to start asking as you've got
>the experience (probably), Anyone have any suggestions?

the ASIC versions of the AVR go a bit faster, like 20-40 MHz IIRC.
Would that work for you?

I was about to suggest a Renesas myself, but a different angle.  

Renesas is running a Renesas contest based on their M16
microcontrollers (yes, they are CISC, I know...).  One thing you can
do right now is sign up for a demo of their hardware and compiler on a
live computer at a remote site.  If you like it, you can apply to get
a board and a free C compiler (64K code limit).  There's 4 options,
which the 32 bit version should work for you.  

Boards provided free in the contest:

http://www.circuitcellar.com/renesas2005m16c/kits/StarterKitdescription.htm

Specific chip for the supplied demo board:

http://america.renesas.com/fmwk.jsp?cnt=m32c84_root.jsp&fp=/products/mpumcu/m16c_family/m32c80_series/m32c84_group/

Demo and entry website:
http://www.renesasinteractive.com/m16cdesigncontest.html

If you don't decide to apply for the contest, the same kit is
available for $100 straight out, although I swear I saw them listed
for $50 somewhere.  I know after I placed my order for the hardware (I
try every company eventually) it came in less than a week.

Jerry

--- In AVR-Chat@yahoogroups.com, Russell Shaw <rjshaw@n...> wrote:
> Paul Maddox wrote:
> > All,
> > 
> >   I'm currently using a MEGA16 at 16Mhz for a project, and I'm
doing a lot
> > of floating point calculations (most exponential) and its getting
a bit
> > slow, so I need something with a bit more 'umph'.
> > I know its kinda off topic, but, what I'm after (I think) is
something 16Bit
> > and a minimum of 25MIPs (32 would be preffered). it also *must*
have an
> > affordable C compiler (ie <£400) and an evaluation board. Surface
mount is
> > fine (in fact preffered).
> >    I figured you guys would be the best place to start asking as
you've got

> > the experience (probably), Anyone have any suggestions?
> > 
> > Paul
> 
> A Renesas SuperH. http://www.superh.com/products/architecture.htm

Paul ,

Certainly a faster chip will help you but not if your code is poorly 
written from a performance point of view. Before jumping ship, have you 
tried to isolate the problem more into what part of the code is slow? 
You indicated that you do a lot of floating point exponential 
calculations. Have you looked at streamlining this area of code? There 
are several methods including approximations and reductions. The best 
approach is probanly a table-driven approach (unless your FP library is 
already doing this). Here are some links:

http://www.acm.org/pubs/citations/journals/toms/1990-16-3/p185-tang/ - needs ACM access to download PDF

http://www.ece.concordia.ca/~tahar/pub/FPE_TR98.ps&e=747 - table driven algorithm in hardware example

http://www.intel.com/technology/itj/q41999/articles/art_5.htm - 
interesting background article on what Intel did

Regards,
Mike

> All,
>
>   I'm currently using a MEGA16 at 16Mhz for a project, and I'm doing a lot
> of floating point calculations (most exponential) and its getting a bit
> slow, so I need something with a bit more 'umph'.
> I know its kinda off topic, but, what I'm after (I think) is something 
> 16Bit
> and a minimum of 25MIPs (32 would be preffered). it also *must* have an
> affordable C compiler (ie <£400) and an evaluation board. Surface mount is
> fine (in fact preffered).
>    I figured you guys would be the best place to start asking as 
> you've got
> the experience (probably), Anyone have any suggestions?
>
> Paul
>
>

Do you Yahoo!?
Yahoo! Mail - Easier than ever with enhanced search. Learn more.

--- In AVR-Chat@yahoogroups.com, sami saqqa <samiceng@y...> wrote:
> i over-colck mega8535-8 to 24.5MHz 

Was it necessary to use an external clock source or did you use a 
crystal on the Xtal1/Xtal2 pins?

Don

Once I accidentally put a 16mhz crystal onto an 8mhz mega163.  It didn't
work correctly until I bumped the STK500 board voltage to 6.1v (maybe less)
but it ran reliably after that.  The baud rate was all wrong, however, and
that is how I discovered my mistake :)

-----Original Message-----
From: Don Kinzer [mailto:dkinzer@easystreet.com]
Sent: Wednesday, February 02, 2005 6:39 PM
To: AVR-Chat@yahoogroups.com
Subject: [AVR-Chat] Re: something a bit faster




--- In AVR-Chat@yahoogroups.com, sami saqqa <samiceng@y...> wrote:
> i over-colck mega8535-8 to 24.5MHz

Was it necessary to use an external clock source or did you use a
crystal on the Xtal1/Xtal2 pins?

Don






Yahoo! Groups Links

sami saqqa wrote:

> how about over-clocking the AVR
> for me
> i over-colck mega8535-8 to 24.5MHz
> my project was a digital filter
> i know it is not a good thing to do but in my case, it work just fine
>  
> note: at 30MHz the mega8535 dose not work


Before everyone gets all excited about overclocking, remember that the 
device it not necessarily rated at the increased clock speed. So 
although overclocking seems to be the norm nowadays, I would say don't 
do it for production level devices, experimentation only.

Mike

From: sami saqqa 
To: AVR-Chat@yahoogroups.com 

Sent: Thursday, February 03, 2005 12:52 AM
Subject: Re: [AVR-Chat] something a bit faster


how about over-clocking the AVR
for me
i over-colck mega8535-8 to 24.5MHz 
my project was a digital filter 
i know it is not a good thing to do but in my case, it work just fine

note: at 30MHz the mega8535 dose not work

Or, use two - parallel processing. 8-)

Leon
--
Leon Heller, G1HSM
http://www.geocities.com/leon_heller
----- Original Message ----- 


-- 
No virus found in this outgoing message.
Checked by AVG Anti-Virus.
Version: 7.0.300 / Virus Database: 265.8.4 - Release Date: 01/02/2005

Dave,

> the ASIC versions of the AVR go a bit faster, like 20-40 MHz IIRC.
> Would that work for you?

40 mhz would be superb, but I don't have the code finalised yet, makes me
wonder though, the SRAM based AVRs are quicker aswell...

Why can't I run a mega16 at 32Mhz?
Assuming I don't do any EEPROM/FLASH reading/writing.

Paul

Mike,

> Certainly a faster chip will help you but not if your code is poorly
> written from a performance point of view. Before jumping ship, have you
> tried to isolate the problem more into what part of the code is slow?

I'm doing a fair bit of floating point calculations and 16bit data
manipulation, needless to say, an 8bit micro is less than ideal...

> You indicated that you do a lot of floating point exponential
> calculations. Have you looked at streamlining this area of code? There
> are several methods including approximations and reductions. The best
> approach is probanly a table-driven approach (unless your FP library is
> already doing this). Here are some links:

:-)
Looked at tables, yes quicker, but I would need a bigger table than I could
store in flash.
I'll certainly give the links a look, cheers.

Paul

Jerry,

> Renesas is running a Renesas contest based on their M16
> microcontrollers (yes, they are CISC, I know...).  One thing you can
> do right now is sign up for a demo of their hardware and compiler on a
> live computer at a remote site.  If you like it, you can apply to get
> a board and a free C compiler (64K code limit).  There's 4 options,
> which the 32 bit version should work for you.  

Hmmm, sounds interesting, thanks..

Paul

Hi,

One approach I successfully used for remotely powering RS-485 nodes 
is to use a small step-up DC converter on each node. With a MAX1674, 
each node runs on 1.8 to 5.5V which works well on e.g. 4 AA cells or 
a long supply line. I recently have designed such a node controller 
with an ATmega32, RS-485, octal output drivers, dual dc motor 
drivers, octal servo outputs and the option for an EasyRadio RF modem 
or a BlueTooth module. There is a picture of it in the photos section 
of this forum (name: RoboCube controller). If you are interested I 
could send you one engineering sample for your field tests.

Best regards,
Hans

At 07:52 AM 2/3/2005, Paul Maddox wrote:

>Dave,
>
> > the ASIC versions of the AVR go a bit faster, like 20-40 MHz IIRC.
> > Would that work for you?
>
>40 mhz would be superb, but I don't have the code finalised yet, makes me
>wonder though, the SRAM based AVRs are quicker aswell...
>
>Why can't I run a mega16 at 32Mhz?
>Assuming I don't do any EEPROM/FLASH reading/writing.

Primarily because "overclocking" is unreliable.  You have no idea wether 
the next chip will work at that speed, or not.

On 3 Feb 2005 at 12:56, Paul Maddox wrote:

> 
> I'm doing a fair bit of floating point calculations and 16bit data
> manipulation, needless to say, an 8bit micro is less than ideal...
> 

	I did once a BELL-202 modem using an ATMEGA8 @ 6.144MHz.
	The demodulation routine used fixed-point math an did
easily some 25 multiply 16x16bit operations plus some 50 SUB/ADD
also 16 bit.

	In Assembler, of course.

	Mark Jordan

--- In AVR-Chat@yahoogroups.com, sami saqqa <samiceng@y...> wrote:
> how about over-clocking the AVR
> for me i over-colck mega8535-8 to 24.5MHz 

Why abuse a chip, when annother (ATmega168) is already specified for 
24MHz?

Stefan

On Thu, 3 Feb 2005, Stefan Wimmer wrote:

> 
> --- In AVR-Chat@yahoogroups.com, sami saqqa <samiceng@y...> wrote:
> > how about over-clocking the AVR
> > for me i over-colck mega8535-8 to 24.5MHz
> 
> Why abuse a chip, when annother (ATmega168) is already specified for
> 24MHz?
> 
> Stefan
> 

Oh good, I can try overclocking it at 35 MHz.  :-)

Zack

Stefan,

> Why abuse a chip, when annother (ATmega168) is already specified for 
> 24MHz?

it is?
I'll check out the Atmel webpage.

Paul

Stefan,

> Why abuse a chip, when annother (ATmega168) is already specified for
> 24MHz?

just checked, the site says 20Mhz, so does the datasheet...
Interesting though, it *might* be good enough, to do what I need, need to
check that my version of CV-AVR supports it and check supplies, it would be
DAMNED nice if it did run at 24Mhz..

Paul

Larry,

> If you are looking for more speed, take a look at avr-gcc.  The floating
> point library in that package is very good.  It takes about 600 or so
> instructions (depending upon what you use) and basic operations are done
in
> ~500 cycles.  For my needs, mult/div/cos/sin/add/sub I got around 1.6k
> ops/sec/mhz or about 25 kops/sec at 16mhz.

Thanks, I'll check it out.

BTW, the MEGA168 is no good, it doesn't have enough IO pins :-(

Paul

If you are looking for more speed, take a look at avr-gcc.  The floating
point library in that package is very good.  It takes about 600 or so
instructions (depending upon what you use) and basic operations are done in
~500 cycles.  For my needs, mult/div/cos/sin/add/sub I got around 1.6k
ops/sec/mhz or about 25 kops/sec at 16mhz.

-----Original Message-----
From: Paul Maddox

Stefan,

> Why abuse a chip, when annother (ATmega168) is already specified for
> 24MHz?

just checked, the site says 20Mhz, so does the datasheet...
Interesting though, it *might* be good enough, to do what I need, need to
check that my version of CV-AVR supports it and check supplies, it would be
DAMNED nice if it did run at 24Mhz..

I've run into this problem a couple times.  The full 8-bit ports have 
alternate function pins I need for some application, and the remaining 
ports have less than 8 pins available. If I need an 8-bit port, I have to 
go to a larger device.

At least there's a wide range to choose from!

Zack

On Fri, 4 Feb 2005, Paul Maddox wrote:

> Larry,
> 
> BTW, the MEGA168 is no good, it doesn't have enough IO pins :-(
> 
> Paul
>

Zack,


> I've run into this problem a couple times.  The full 8-bit ports have
> alternate function pins I need for some application, and the remaining
> ports have less than 8 pins available. If I need an 8-bit port, I have to
> go to a larger device.

yes, poor choice of where to put the alternate functions IMHO.

> At least there's a wide range to choose from!

if you want >16Mhz, there's only the one device..

Paul

Ralph,

> I downloaded an earlier datasheet which did say 24MHz and a couple of
ATmega48s
> I got were ATmege48V 12AI i.e. 12MHz so it looks like they had problems
with
> 24MHz operation and changed the spec.

yep, I read that in the amendments too. a shame, a real shame..
Can someone at atmel please give the designers a heads up? 32Mhz with four
8bit IO ports ASAP..

Paul

On Fri, 4 Feb 2005 11:03:16 -0000 you wrote:

>
>Stefan,
>
>> Why abuse a chip, when annother (ATmega168) is already specified for
>> 24MHz?
>
>just checked, the site says 20Mhz, so does the datasheet...
>Interesting though, it *might* be good enough, to do what I need, need to
>check that my version of CV-AVR supports it and check supplies, it would be
>DAMNED nice if it did run at 24Mhz..

I downloaded an earlier datasheet which did say 24MHz and a couple of ATmega48s
I got were ATmege48V 12AI i.e. 12MHz so it looks like they had problems with
24MHz operation and changed the spec.

From the new datasheet:

Changes from Rev.
2545B-01/04 to Rev.
2545C-04/0
1. Speed Grades changed:
- 12MHz to 10MHz
- 24MHz to 20MHz
--
Ralph Hilton
http://www.ralphhilton.org
C-Meter: http://www.cmeter.org
FZAOINT http://www.fzaoint.net

Russell,

> Do you really need thousands of calcs per sec, or just a few calcs
> in a very short interval?

The 16Mhz Mega16 I'm using is begining to struggle, speed wise. I mean, i
can notice it in the output.
Its doing quite a bit, and I've had to use some floating point calculations
(something I've tried very hard to avoid, but I can't).
The problem is, I need it to do a bit more than it is doing currently, and I
expect this will cause it to slow down too much.
So, yes, I need more speed, 24Mhz would be safe, 32Mhz would just be a dream
come true.

Paul

--- In AVR-Chat@yahoogroups.com, "Paul Maddox" <P.Maddox@s...> wrote:
> Stefan,
> 
> > Why abuse a chip, when annother (ATmega168) is already specified 
for
> > 24MHz?
> 
> just checked, the site says 20Mhz, so does the datasheet...

Huh?

Did they re-specify the chip??
My datasheez says:

Features
• High Performance, Low Power AVR® 8-Bit Microcontroller
• Advanced RISC Architecture
– 131 Powerful Instructions – Most Single Clock Cycle Execution
– 32 x 8 General Purpose Working Registers
– Fully Static Operation
– Up to 24 MIPS Throughput at 24 MHz <<<<<<<<<<<<<<<<<<<<<<<<<<<<
– On-chip 2-cycle Multiplier

etc.


Oh well! - I looked at the Atmel site and you're right: the spec. 
has changed...

Sorry,
Stefan

Paul Maddox wrote:
> Ralph,
> 
> 
>>I downloaded an earlier datasheet which did say 24MHz and a couple of
> 
> ATmega48s
> 
>>I got were ATmege48V 12AI i.e. 12MHz so it looks like they had problems
>>with 24MHz operation and changed the spec.
> 
> yep, I read that in the amendments too. a shame, a real shame..
> Can someone at atmel please give the designers a heads up? 32Mhz with four
> 8bit IO ports ASAP..
> 
> Paul

Do you really need thousands of calcs per sec, or just a few calcs
in a very short interval?

----- Original Message ----- 

From: Stefan Wimmer
To: AVR-Chat@yahoogroups.com
Sent: Friday, February 04, 2005 5:26 PM
Subject: [AVR-Chat] Re: something a bit faster



--- In AVR-Chat@yahoogroups.com, "Paul Maddox" <P.Maddox@s...> wrote:
> Stefan,
>
> > Why abuse a chip, when annother (ATmega168) is already specified
for
> > 24MHz?
>
> just checked, the site says 20Mhz, so does the datasheet...


I don't think that anyone has mentioned the Scenix/Ubicom SX family - PIC 
architecture, 75 MIPS and *lots* of current.

Leon 



-- 
No virus found in this outgoing message.
Checked by AVG Anti-Virus.
Version: 7.0.300 / Virus Database: 265.8.5 - Release Date: 03/02/2005

Stefan,

> Huh?
> 
> Did they re-specify the chip??

yep, check the website..

Paul

Leon,

> I don't think that anyone has mentioned the Scenix/Ubicom SX family - PIC
> architecture, 75 MIPS and *lots* of current.

good point,
Current is an issue, its going in a box with a PSU, the 'smallest' I can
find that meets my other requirements has a 5A 5v supply...

I'll have a browse around.

Paul

__________________________________________________
Do You Yahoo!?
Tired of spam? Yahoo! Mail has the best spam protection around
http://mail.yahoo.com

Paul Maddox wrote:
> Russell,
> 
> 
>>Do you really need thousands of calcs per sec, or just a few calcs
>>in a very short interval?
> 
> 
> The 16Mhz Mega16 I'm using is begining to struggle, speed wise. I mean, i
> can notice it in the output.
> Its doing quite a bit, and I've had to use some floating point calculations
> (something I've tried very hard to avoid, but I can't).
> The problem is, I need it to do a bit more than it is doing currently, and I
> expect this will cause it to slow down too much.
> So, yes, I need more speed, 24Mhz would be safe, 32Mhz would just be a dream
> come true.
> 
> Paul

But that's still not answering the Q;)

If you're doing unneccessary averaging or something, you could do it once
every 10th time around the superloop or whatever. I do a ton of realtime
dsp control stuff with floating point on an at90s8535.

Russell,

> But that's still not answering the Q;)

hehe, fair point.

> If you're doing unneccessary averaging or something, you could do it once
> every 10th time around the superloop or whatever. I do a ton of realtime
> dsp control stuff with floating point on an at90s8535.

its for a synthesiser, and as such I need to do a fair bit of exponential
maths to get frequencies, generate envelopes and so on, this is where I
think the slow down is.
I've done what I can to reduce this by updating somethings less often than
others, but I can't do much less without the effects being obvious to the
ear.

Paul

Russell,

> Because it's a musical thing, it would be hard to make shortcuts.

yeah, I've done what I can, its 'ok' now, but I'm still looking.
Renesas have a Starterkit, but unfortunatly it seems only to be a US
product, and the FAE is not returning any calls..

> An alesis dsp would be a good peripheral for that:
http://www.alesis-semi.com/products.htm

looked at them, I'm still considering if I'll go with those or not..

Paul

Paul Maddox wrote:
> Russell,
> 
>>But that's still not answering the Q;)
> 
> hehe, fair point.
> 
>>If you're doing unneccessary averaging or something, you could do it once
>>every 10th time around the superloop or whatever. I do a ton of realtime
>>dsp control stuff with floating point on an at90s8535.
> 
> its for a synthesiser, and as such I need to do a fair bit of exponential
> maths to get frequencies, generate envelopes and so on, this is where I
> think the slow down is.
> I've done what I can to reduce this by updating somethings less often than
> others, but I can't do much less without the effects being obvious to the
> ear.

Because it's a musical thing, it would be hard to make shortcuts.

An alesis dsp would be a good peripheral for that: http://www.alesis-semi.com/products.htm

----- Original Message ----- 

From: Paul Maddox
To: AVR-Chat@yahoogroups.com
Sent: Monday, February 07, 2005 10:37 AM
Subject: Re: [AVR-Chat] Re: something a bit faster


Russell,

> Because it's a musical thing, it would be hard to make shortcuts.

yeah, I've done what I can, its 'ok' now, but I'm still looking.
Renesas have a Starterkit, but unfortunatly it seems only to be a US
product, and the FAE is not returning any calls..

> An alesis dsp would be a good peripheral for that:
http://www.alesis-semi.com/products.htm

looked at them, I'm still considering if I'll go with those or not..


I've designed a little four-layer PCB with the Alesis DSP, ADC and DAC, 
controlled by a '2313.

Leon
--
Leon Heller, G1HSM
http://www.geocities.com/leon_heller





-- 
No virus found in this outgoing message.
Checked by AVG Anti-Virus.
Version: 7.0.300 / Virus Database: 265.8.5 - Release Date: 03/02/2005

Leon Heller wrote:
> ----- Original Message ----- 
> From: Paul Maddox
> To: AVR-Chat@yahoogroups.com
> Sent: Monday, February 07, 2005 10:37 AM
> Subject: Re: [AVR-Chat] Re: something a bit faster
> 
> Russell,
> 
>>Because it's a musical thing, it would be hard to make shortcuts.
> 
> yeah, I've done what I can, its 'ok' now, but I'm still looking.
> Renesas have a Starterkit, but unfortunatly it seems only to be a US
> product, and the FAE is not returning any calls..
> 
>>An alesis dsp would be a good peripheral for that:
> 
> http://www.alesis-semi.com/products.htm
> 
> looked at them, I'm still considering if I'll go with those or not..
> 
> I've designed a little four-layer PCB with the Alesis DSP, ADC and DAC, 
> controlled by a '2313.

I've thought of resurrecting the adsp gnu tools because
the sources are still around and i have a copy (i've used
adsp-2181 years ago). It's a bit high up on the project stack tho:(

On Mon, 07 Feb 2005 22:16:12 +1100, Russell Shaw <rjshaw@netspace.net.au>  
wrote:

>
> I've thought of resurrecting the adsp gnu tools because
> the sources are still around and i have a copy (i've used
> adsp-2181 years ago). It's a bit high up on the project stack tho:(
>


I've used the 2181 before for a radio tx thing (controlling a DDS) and it  
worked really well.
Beeing used to other micros it is a treat to just feed the multiplication  
unit with numbers and have a result back faster than you can want it.
Somewhere here a 21061 must be (sharc with cool heatsink pad).
Now what could one do with that? Not sure i'll ever use it becasue if i  
put a lot of work into a project where i have only one IC, well, pretty  
dangerous... i might not get a replacement or it might be too expensive.

ST

Russell Shaw said:
>
> I've thought of resurrecting the adsp gnu tools because
> the sources are still around and i have a copy (i've used
> adsp-2181 years ago). It's a bit high up on the project stack tho:(
>

There was a company selling a CD containing the compiled version of those
tools before that I have.  I don't know how well those have been
maintained in recent years.

Mike

Hi,

 well, I've found these 8051 based ones from Maxim/Dallas,
http://www.maxim-ic.com/quick_view2.cfm/qv_pk/4078/ln/

 not as well designed as the AVRs, but I've ordered one as a sample to try
and compare the speed, tey're faster (33MIPs) but don't have things like a
hardware multiply or lots of registers.

Paul

At 10:46 AM 2/7/2005, Paul Maddox wrote:

>Hi,
>
>  well, I've found these 8051 based ones from Maxim/Dallas,
>http://www.maxim-ic.com/quick_view2.cfm/qv_pk/4078/ln/
>
>  not as well designed as the AVRs, but I've ordered one as a sample to try
>and compare the speed, tey're faster (33MIPs) but don't have things like a
>hardware multiply or lots of registers.

The 8051 design is a lot more like a pic in terms of "speed", so you'll 
probably give back all you've gained, except the higher EMI profile.

Hi Paul,

my name indicates that I am a little biased ;-)
For floating point it helps a lot to have a 32-bit architecture at a
similar price. Hope I won'y get banned for helping somebody out here.

A little faster could be an ARM7 running at up to 60 MHz like the
LPC2000 family from Philips. If you go for the low end of these
devices, e.g. LPC2132 or LPC2131, you will get 32k or 64k flash with
8k RAM and much better performance. 
Info here:
http://www.semiconductors.philips.com/cgi-bin/pldb/pip/LPC2131FBD64.html

As for affordable tools, fully supported by GNU http://www.gnuarm.com,
the LPC2130 board from http://www.keil.com/mcb2130/ has a 16k compiler
coming with it. There is a free 32k compiler from IAR but I don't know
how to get it except with a "kickstart kit" which includes a different
chip LPC2106.
Both evaluation boards including the 16k / 32k compilers are $149. You
can get more information on the tools here: http://www.lpctools.com
Other lower cost C-compilers are available from Rowleys or
http://www.imagecraft.com
http://www.rowley.co.uk/
Other low cost boards if you go for the Rowley compiler would be from
Olimex who also does low cost boards for AVR offers ARM boards too.
http://www.olimex.com/dev/index.html

Lots of information on this Yahoo Group:
http://groups.yahoo.com/group/lpc2000

Hth, Bob


--- In AVR-Chat@yahoogroups.com, "Paul Maddox" <P.Maddox@s...> wrote:
> All,
> 
>   I'm currently using a MEGA16 at 16Mhz for a project, and I'm doing
a lot
> of floating point calculations (most exponential) and its getting a bit
> slow, so I need something with a bit more 'umph'.
> I know its kinda off topic, but, what I'm after (I think) is
something 16Bit
> and a minimum of 25MIPs (32 would be preffered). it also *must* have an
> affordable C compiler (ie <£400) and an evaluation board. Surface
mount is
> fine (in fact preffered).
>    I figured you guys would be the best place to start asking as
you've got

> the experience (probably), Anyone have any suggestions?
> 
> Paul

Hi,

> my name indicates that I am a little biased ;-)

thats ok :-)

> For floating point it helps a lot to have a 32-bit architecture at a
> similar price. Hope I won'y get banned for helping somebody out here.

yep, I agree 32 bit would be lovelly.

> A little faster could be an ARM7 running at up to 60 MHz like the
> LPC2000 family from Philips.

carefull with 'a little faster' ;-)

I looked at these, I even have one of the eval kits, you know why I don't
use it?
The IO is too slow, 3.4Mhz!!!! The AVR can toggle a pin at 8Mhz (it runs at
16Mhz).
So, I've got this nice fast CPU, that I have to keep stopping each time I
want to play with the IO, like this -
write a byte,
wait for 300ns,
toggle a bit,
wait for 300ns,
toggle the bit back..
thats complete insanity!!

Its a really BIG shame as I had several uses for it, but I'm sorry the IO is
just absurdly slow.
A waste of, potentially, a great chip!

So for me, its useless. BTW, if anyone is interested I have one of the MB40
eval boards with LPC2106 and LCD for sale.

Paul

Jon,

> Is that a general property of the ARM7 chips, or something specific for
> that Philips board?

after some conversations and digging,
its not a problem withthe ARM7 core, but it is a problem with the glue
logic, philips has this problems, others may also, so its 'suck it and see'.
What really annoyed me was I couldn't find mention of this on the LPC21xx
datasheets!

Paul

On Tue, 8 Feb 2005 09:05:18 -0000, "Paul Maddox"
<P.Maddox@signal.qinetiq.com> wrote:

> Its a really BIG shame as I had several uses for it, but I'm sorry the IO is
> just absurdly slow.
> A waste of, potentially, a great chip!

Is that a general property of the ARM7 chips, or something specific for
that Philips board?

Later,
Jon

--------------------------------------------------------------
   Jon Hylands      Jon@huv.com      http://www.huv.com/jon

  Project: Micro Seeker (Micro Autonomous Underwater Vehicle)
           http://www.huv.com

--- In AVR-Chat@yahoogroups.com, Jon Hylands <jon@h...> wrote:
> On Tue, 8 Feb 2005 09:05:18 -0000, "Paul Maddox"
> <P.Maddox@s...> wrote:
> 
> > Its a really BIG shame as I had several uses for it, but I'm 
sorry the IO is
> > just absurdly slow.
> > A waste of, potentially, a great chip!
> 
> Is that a general property of the ARM7 chips, or something specific 
for
> that Philips board?

It's inherent in the ARM architecture, the APB (Advanced Peripheral 
Bus) which is generally used for ordinary I/O is simple but very 
slow. The AHB (Advanced High-performance Bus) is much faster and a 
lot more complicated.

Leon

Jon,

> Have you looked at some of the products Randy sells at www.newmicros.com ?
>
> He's got a few ARM7 boards, and a bunch of Iso-whatever boards that run on
> 40 MIPS Motorola MCUs.

thanks, some interesting looking stuff, I'll have a look.

Paul

On Tue, 8 Feb 2005 12:18:21 -0000, "Paul Maddox"
<P.Maddox@signal.qinetiq.com> wrote:

> after some conversations and digging,
> its not a problem withthe ARM7 core, but it is a problem with the glue
> logic, philips has this problems, others may also, so its 'suck it and see'.
> What really annoyed me was I couldn't find mention of this on the LPC21xx
> datasheets!

Have you looked at some of the products Randy sells at www.newmicros.com ?

He's got a few ARM7 boards, and a bunch of Iso-whatever boards that run on
40 MIPS Motorola MCUs.

Later,
Jon

--------------------------------------------------------------
   Jon Hylands      Jon@huv.com      http://www.huv.com/jon

  Project: Micro Seeker (Micro Autonomous Underwater Vehicle)
           http://www.huv.com

Hi,

did not realize that you needed the port toggle and I agree that is a
significant weekness of ARM-peripheral bus. Most devices in the market
have the ports connected to the peripheral bus, ergo, some slow
toggle. Exception (although I am biased ;-)) the Atmel SAM7. This
device connects the ports to the AHB bus. Don't know exactly how much
fast the ports are (don't have HW) but the calculations come down to
approx, 2x the frequency of the LPC2000 running at the same clock.
Because these devices are limited to 50 MHz, you could probably get in
the neighborhood of 7 MHz at the ports. 

It is only a matter of time until one or more manufacturers address
this weekness by connecting the ports (or at least some fast pins) to
the local bus. This would enable the ports to be switched I think at
CPU-clock / 4. Still the AVR is better with CPU/2 BUT you get so much
more benefits with the ARM.

The AVR is great for I/O tasks and running it at 8 MHz with similar
I/O performance as an ARM with 60 MHz is also a lot more power
efficient. So for I/O intensive tasks lots of bit reading and
toggeling a 32-bit might be inappropriate.

For now, your best bet if looking at ARM would be the SAM7 for fast
ports, 6 months from now I guess that will be different.

Bob

--- In AVR-Chat@yahoogroups.com, "leon_heller" <leon.heller@d...> wrote:

> 
> --- In AVR-Chat@yahoogroups.com, Jon Hylands <jon@h...> wrote:
> > On Tue, 8 Feb 2005 09:05:18 -0000, "Paul Maddox"
> > <P.Maddox@s...> wrote:
> > 
> > > Its a really BIG shame as I had several uses for it, but I'm 
> sorry the IO is
> > > just absurdly slow.
> > > A waste of, potentially, a great chip!
> > 
> > Is that a general property of the ARM7 chips, or something specific 
> for
> > that Philips board?
> 
> It's inherent in the ARM architecture, the APB (Advanced Peripheral 
> Bus) which is generally used for ordinary I/O is simple but very 
> slow. The AHB (Advanced High-performance Bus) is much faster and a 
> lot more complicated.
> 
> Leon

--- In AVR-Chat@yahoogroups.com, "Paul Maddox" <P.Maddox@s...> wrote:
> All,
> 
>   I'm currently using a MEGA16 at 16Mhz for a project, and I'm 
doing a lot
> of floating point calculations (most exponential) and its getting a 
bit
> slow, so I need something with a bit more 'umph'.
> I know its kinda off topic, but, what I'm after (I think) is 
something 16Bit
> and a minimum of 25MIPs (32 would be preffered). it also *must* 
have an
> affordable C compiler (ie <£400) and an evaluation board. Surface 
mount is
> fine (in fact preffered).
>    I figured you guys would be the best place to start asking as 
you've got
> the experience (probably), Anyone have any suggestions?
> 
> Paul
Hi All
It's good to see AVR users aren't one eyed re processors. 

I used 8051's from Cygnal (Silicon Laboratories) a while ago and was 
very surprised. They are nothing like the 8051's of old. They are 
Code compatible with the old ones but they are genuine one clock for 
one operation; not divided by 12 as they used to be. 

They do have a couple of irritating things (like a crossbar register 
that doesn't deliver what you initially expect and register/bit 
differences between the family). But speed wise they are pretty good 
for a low end processor (suggested on the site at 25 >> 100 mips) and 
from playing with them they aren't that far off.

You'll have to follow the links as the final link is too long for 
this post. (General Purpose microcontrollers)
http://www2.silabs.com

upctoys

Hi,

> It's good to see AVR users aren't one eyed re processors.

right part for the right job :-)

> I used 8051's from Cygnal (Silicon Laboratories) a while ago and was
> very surprised. They are nothing like the 8051's of old. They are
> Code compatible with the old ones but they are genuine one clock for
> one operation; not divided by 12 as they used to be.

nice.

 > They do have a couple of irritating things (like a crossbar register
> that doesn't deliver what you initially expect and register/bit
> differences between the family). But speed wise they are pretty good
> for a low end processor (suggested on the site at 25 >> 100 mips) and
> from playing with them they aren't that far off.

hey, no we're cooking! 100MIPs!! most excellent <bill and ted guitar motion>
Thanks for the tip, I'll take a read and see what they're like.

Paul