AVR-Chat

Ok, this is good, it's getting meaty and I like some of your figures. 
This only takes into account a discrete BC548 and not the typical input 
transistor of the receiver which also happens to be buffered by two 
other transistors in typical inverter follower fashion.

Anyway, I enjoyed you reply and I will get back to the group with some 
observations. I question a lot of the standards in the same way I 
questions rules that require they be followed although the reasons for 
them have long since faded. Observe the principle first and the rules 
take care of themselves. Principles are hard to enforce which is why 
people make rules. Principles don't change but rules do, although they 
can be real slow to catch-up at times.

Tinkering things into existence is not the name of the game but 
tinkering them out is. This way you know where the fall-off area is and 
allow a comfortable margin. Ideally this could be designed and simulated 
but you know for the same reason that you can design and manufacture a 
BC548 that there are also BC547s and 549s, although they are the same 
transistor. Due to manufacturing tolerances these parts would fail the 
desired spec yet are "tinkered" as alternate types.

Sometimes it's not about margins but about knowing what is redundant and 
clearing away the clutter. There are a lot of things about 232 including 
the standard itself that are redundant and I get very tired of people 
quoting the decades old rules that applied back then without taking 
current usage into account. Sub 120K low-speed short-range coms to a PC 
needs high plus and minus volts??? ...come on, nobody but nobody designs 
like this except to keep the voices happy.


*Peter*


Russell Shaw wrote:
> Peter Jakacki wrote:
>> Russell Shaw wrote:
>>
>>> Peter Gargano wrote:
>>>
>>>> stevech@san.rr.com wrote:
>>>>
>>>>> *RS232C is such that if the voltage swings +/- 6V it is in compliance.*
>>>> But may receivers will work at a swing of 0 to 5 volts.
>>> Maybe, but what about max speed reliability? Negative drive removes transistor
>>> base charge faster.
>  >
>  > That was funny Russ, this may be more true of a mosfet's gate charge,
>  > not a bipolar which works quite a bit differently (minority carrier
>  > charge, electron holes, etc).
>
> If you know how BJTs work, you'll know about stored base charge in saturated
> switching.
>
>  > Anyway, what speed are we talking about?
>  > 100Mhz? 10Mhz? 1Mhz? The speeds that RS-232 operate at can be handled
>  > easily by the slowest of the slow transistors.
>
> Well, old technology without a baker clamp or gold-doped switching
> transistors (or a homebrew job) may use an NPN common-emitter input
> transistor like a BC548 (Beta=300, ft=300MHz etc).
>
> If this was driven with ~5V via a 10k resistor (with a base-emitter diode),
> the stored charge would be atleast 68pC, excluding depletion
> capacitances and excess storage in the base-collector epitaxial layer.
> The input thresh-hold would be ~0.7V.
>
> Now with 0V applied, the base charge can only be removed by 0.7V
> across 10k. The switch-off delay-time is then ~1us. 115200 baud is 8.7us/bit.
>
> Not all logic pulls down to ground or to supply, so the RS232 drive
> could be 0.8V->3.3V (TTL: http://www.interfacebus.com/voltage_threshold.html)
> which makes it impossible at any speed. If the pull-down voltage was more
> typically 0.5V, that would be a delay time of 3.4us. The noise margin is
> ratshit too.
>
>  > Have you seen the reverse diode across the base emitter? This would
>  > clamp the voltage to -0.3V internally anyway.
>
> Only after the stored base charge is rapidly removed.
>
>  > Engineers are people that get their hands dirty, right? Ok then, how
>  > about hooking-up a scope to the receiver and driving it at say 115.2K
>  > via a quick and dirty 100K series pot so that you can vary the input
>  > voltage. This will confirm the reliable turn-on threshold and by placing
>  > a series diode you can cutout the negative transition (although this is
>  > not as good as 0-5V drive) to confirm that part too. If a diode is used
>  > then it really needs a 1K pull-down resistor after it.
>
> It would be advisable to use a 1k shunt if you were attempting unipolar drive
> anyway.
>
>  > Since I suggested this and since I design all kinds of coms gear anyway,
>  > I will run some real tests and post the results to settle this matter in
>  > a less academic but in a more "engineer" like manner.
>
> I never "tinker" things into existance, which is *not* engineering.
> I write electronics simulation software and did a masters research
> degree that included radio-frequency transistor measurement, modelling, and
> circuit simulation on both BJTs and mosfets. After that, i designed and
> built radio-link fading simulators. I'm an analog/radio-frequency designer,
> and only got into software and computer hardware by necessity.
>