ADC crosstalk when input exceeds ref/rail

[TurboBob]

With previous projects (HC11)when the input exceeded the positive supply, there were no issues as long as there was decent (10k) series impedance.

Curently I am designing with an HC908AP64 and am getting a lot of crosstalk issues, and the impedance limiting trick is not working.

Even the "resistor....diode to +5....resistor.....input" circuit is not effective.

Is there some trick in software to eliminate the crosstalk (I am using
Codewarrior) or some whiz-bang circuit trick? Generally the signals will be 0-5 volts, but I can't guarantee that......

Any ideas guys.....?

(I tried to search the board and freegeeks and found nothing....)

Thanks!

Bob

[TurboBob]

The TL7726 is a hex clamping IC.

http://focus.ti.com/docs/prod/folders/print/tl7726.html

It is designed for this exact job. Its just not cheap.....

I'm not sure the schottky diodes will clamp low enough. I was seeing issues at 5.1 volts. even with series resistance. The 908 (at least this one) seems very sensitive.


Bob

[rocco]

Bob, You are correct in that the Schottky diode probably won't work, since it will not bring you below 5.1 volts. Mac is correct that the 5.1 volt zener will mess with your linearity.

So how about the option that Mac hinted at. It may or may not be cheaper than the TL7726.

Use a cheap, but accurate enough, rail-to-rail opamp as a buffer. Depending on the amp, you may need a diode on the amp's input, but the output will not exceed the rails. However, it won't fully reach the rails, either, but could get within a few tens of millivolts.

[TurboBob]

all the r-r opamps are 'spensive....... since I am trapped in the stone age (thru-hole)


I'm going to have to brain on this for a day or so, but I don't have an elegant solution yet......


B

[TurboBob]

depending on what's connected to the other channels, the crosstalk/leakage starts at about 5.1 volts.

Clipping is not desired, I'm trying to have an accurate channel.

All of the examples and other resources I can find on the web show simple diode circuits, but most just say to stay under the rails.

I have found something that works really well, its just not cheap like diodes.... The TL7726 from Ti clamps to within .050 of the supplies. I have used it on 2 other projects with excellent results. In this project it cuts the crosstalk to ZERO.

I was hoping a simple circuit could take the place of the 7726, but it seems that I just need to spend the bux and put it in there......

Any other ideas/experience guys?

Bob

[bigmac]

Hello Bob,

I agree with Rocco about the use of Schottky diodes, and I would also use the series resistor to limit current.  The use of a 5.1V zener diode may cause additional non-linearity because it will commence conduction somewhat below its clamp voltage - there will also be a tolerance on the clamp voltage to be taken into account.

I assume the TL2776 is an op amp with rail-to-rail output.  The use of a rail-to-rail op amp will probably provide the best precision (provided you take into account its offset voltage) because it provides a low impedance drive to the ADC input.

If you need to achieve the full 10-bit maximum value, another possibility is to reduce VREFH voltage below VDDA, using an external reference (say 4.0 volts) , so that full scale will occur at a lower input voltage.  This is also likely to improve absolute accuracy and stability of the measurements.  You would still need input clamping arrangements, but would not be subject to the same non-linearity because the clamp level would be well above the FS level.

Regards,
Mac

[rocco]

Hi, Bob:

What type of diode are you using? A silicon diode will have a drop of .6 to .7 volts, so your input will still get as high as 5.6 volts.

You could try a Schottky diode or, better yet, a 5.1 volt zener diode to ground. At least the voltage won't get as high. A 4.7 volt zener to ground would insure that it never gets to 5 volts, as long as you don't mind clipping your signal a little.

What do you mean by "crosstalk"? Do you mean that the offending channel causes incorrect readings on the other channels when it gets above 5 volts? I have had channels affecting each other (on an HC908GP32), but the affected channel's reading seems to be increased when its signal is near 0 volts. I'm not sure if my problem is in the ADC or on my board.