AVR-Chat

Why not just stick with MOSFETs, Graham?  Use something like a 35 cent 
TP0610 for the the P-channel FET and a 10 cent 2N7000 for the N-channel 
FET.  Leakage currents at that point would be more than likely a 
function of board cleanliness than component leakage.  Overall circuit 
current requirements would also be lower.  Both are in TO-92 packages so 
they wouldn't take a lot of real estate on the board and their ON 
resistances are such that they would have extremely negligible affect on 
the divider accuracy.

He might also want to use some sort of gate/source protection scheme to 
be sure he doesn't let all of the smoke out of his P-channel FETs in the 
event of a transient that might occur on the supply lines.  The OP's 
description sounded like it might be an automotive environment, so 
transients can be a very likely.

Why would he not be able to expect pretty fair accuracy from your 
scheme?  Precision resistors aren't expensive these days and their 
tempcos aren't all that bad either.  But I guess we'd also have to 
define accuracy, percent or PPM.  A small capacitor could be hung across 
the low side resistor to compensate for A/D input sampling current, 
perhaps with a calibrated fudge factor in software/firmware if needed.


REB



Graham Davies wrote:

>--- In AVR-Chat@yahoogroups.com, "David VanHorn" <microbrix@...> wrote:
>
>  
>
>>For this business, you really need
>>some software, and some hardware,
>>specifically ANALOG!
>>    
>>
>
>Agreed.  So, shall we do it?  It seems like an interesting problem and 
>it isn't someone's homework.
>
>I would use a P-channel enhancement MOSFET with its source connected to 
>the supply being monitored and its drain feeding the top of the voltage 
>divider.  The gate would have a resistor to the source to keep the 
>MOSFET normally off.  To turn it on, we need an N-channel MOSFET or an 
>NPN transistor with source/emitter at ground and drain/collector 
>connected to the P-channel MOSFET's gate.  To turn the whole thing on, 
>we apply a high level to the drain of the N-channel MOSFET (or switch a 
>current into the base of the NPN transistor).  If we pick the right 
>parts for low leakage, nothing draws current when it is off.  
>Unfortunately, the whole thing may have to be replicated for each 
>supply to be measured, but if they are similar in voltage it may be 
>possible to share the N-channel/NPN part.
>
>I'd still like to hear from those advocating op-amp solutions how the 
>current draw would be less than a resistive voltage divider with, say, 
>a 10 Mohm resistor.  We are all assuming the OP doesn't need a heck of 
>a lot of accuracy and I hope that's right.
>
>Graham.
>  
>


[Non-text portions of this message have been removed]