Excessive crosstalk between ADC inputs on an HC08AP64

[BobS]

I am using PA4-PA7 as analog inputs on an 'AP64.  I scan the inputs at an 8ms/channel rate.  On a prototype with an older date code part, I see a minimal and very short (50us) glitch on the PAx pins when switching channels.  On several pre-production boards, I have seen up to 100mv glitches that last the entire 8ms switching period.  The glitch shows a slight decay towards the end of the period.  The amplitude of the glitch seems to vary with the difference in input voltages to adjacent channels.  The bigger the voltage difference, the larger the glitch.  The input pins are being driven through a 1K resistor from LM224 op amps.  There is a 0.1uf filter from the micro pin to ground.  I tried eliminating the capacitor, but it did not solve the problem.

I have looked at all of the parts on the new boards, and they seem correct.  The board layout is virtually identical between the two boards.  The only thing I can surmise from this is that the later date code part has a leakage or crosstalk problem between ADC channels.

I have looked at the errata, and see no information that relates to this problem.

I will be replacing the newer part with an older part on the pre-production board to see if the problem is eliminated.

I can find no spec for crosstalk between ADC channels.  Have I just missed it in the spec sheet?

Has anyone experienced a similar problem with the 'AP64?

The date code for the older, working part is:  CTAW0420

This part has a Motorola logo.

The date code for the newer part is:  CTEL0541

This part has a Freescale logo.

Both are mask set 2L47P.

Thanks!

[bigmac]

Hello Bob S,

The LM224 op amp you are using is not a rail-to-rail type, so would need a supply voltage significantly more than the maximum voltage applied to the ATD input.  Is this the case?

I can understand that the depth of the "glitch" would depend on the voltage on the sample capacitor from the previous channel reading, but the duration should be short.  For the settling time to exceed 8 milliseconds, this suggests a much higher source resistance than 1k, or a larger capacitor.  Assuming the component values are correct, my thought was that the output resistance from the op amp might increase as it entered a non-linear region, and negative feedback had less effect.  Thus my above query.  Do you also observe the glitch at the op amp output pin?

Regards,
Mac

Message Edited by bigmac on 2006-06-07 12:10 PM

[BobS]

Thank you for the response!

The LM224 supply is 18 volts.  The diode clamp is a BAS-70 Schottky. 

Two of the ADC inputs are connected to a current source that drives a thermistor.  When the thermistor is not connected, the input to the amplifier can go to 13 volts.  There is a divider at the output of the amplifier, so the voltage driven into the clamp (through a 1K resistor) is about 6.5 volts.  This would set up the conditions I have read on other posts.

The crosstalk is larger at the micro side, but I do still see an attenuated version of the crosstlk at the output of the amplifier.

The thing that I don't understand is that we have used this circuit on several products in the past with no issues.  I have a prototype board with an older part that does not have this problem.  Has something changed in the fab process for the AP64?

[bigmac]

Hello Bob S,

The "crosstalk" or glitch due to the charging of the internal sample capacitor is not a problem provided it is of very short duration, and has settled prior to the conversion.  The problem seems to be to account for the extended duration that you are experiencing.  It could be of interest to make two measurement on the same channel, with the measurements 8ms apart, and compare the two readings - i.e. with the sequence
Reach ch1, Read ch2, Wait 8ms, Read ch2, Read ch 3, Wait 8ms, Read ch 3, etc.

Rocco's point might be supported if some of the channels do not have thermistors connected.  Assuming NTC type, at the lowest temperature of interest can I also assume that the voltage across the thermistor would be somewhat less than 5 volts?  If so, I think I would use a shunt zener diode (say 5.1 volts) at the input to the op amp, to better allow for disconnection of the thermistor.  A small amount of attenuation at the op amp output might still be present to ensure that the applied voltage, under all circumstances, does not exceed Vdd.

Another possibility could be to use an op amp type with rail-to-rail output, and supplied from Vdd, so the output can never exceed the supply rails, even with the thermistor disconnected.

Regards,
Mac

[BobS]

Many thanks to all for the help!

The solution that worked best was to clamp the inputs via a Schottky diode to a 4.7V rail that was created using an op amp and a voltage divider.  I think this was suggested in one of the posts, and it does solve the problem, as long as the max current through the Schottky is limited sufficiently.

[mjcoury]

This post may not be of much help to you, however, what you are seeing is "normal". When I worked in the Datalogging / Testing business we strayed away from chips with multiple ADC channels that required lost of switching, the transistors inside firing are the root cause for that. I don't know if there is a way to supress it or reduce it.


Mike

[BobS]

Mike,

Thanks for the information!

[rocco]

Hi, Bob:

I can't seem to find it now, but there was another post about a similar problem.

The issue was that whenever the voltage into any ADC channel exceeded Vcc by more than .1 volts, the readings on the other channels would be corrupted. The fact that the voltage that caused the problem could not be prevented by even a Schottky clamp (>.1 volts) made the problem difficult to work around.

This may not be your problem, but it might be worth looking at.

Later:
OK, I found the thread.
http://forums.freescale.com/freescale/board/message?board.id=8BITCOMM&message.id=284