You are here

PSOC5LP Seq SAR Noise glitche | Cypress Semiconductor

PSOC5LP Seq SAR Noise glitche

Summary: 10 Replies, Latest post by Stephen West on 16 Dec 2015 10:43 AM PST
Verified Answers: 1
Last post
Log in to post new comments.
Stephen West's picture
User
4 posts

I am looking at three 20k sine waves of different amplitudes (same wave) into a PSOC5LP SAR at 12 bits (1M freq).  I  keep getting a noise glitche (randomly) in both directions.  I have disabled everything on it and have added extra capacitance to the the analog pins.  I am using VDDA at 3.3V and VDDA/2.  All sine waves are riding on VDDA/2 (1.65V).  When I single step just before the Enable SAR, the sine wave is clean.  Once I enable, the sine wave glitches.  Any ideas on why this would happen?  I have included my project on this.

I have disabled all of the other peripherals to make sure no interactions. I have disabled the other input A/D ports as well.  I have followed the recommended capacitance in the data sheet as well.

Using Op amps ADA4528-1ARMZ-R7 as the input buffers (my own design).

None of the other A/Ds seem to have this noise issue.

Steve

hli
user_78878863's picture
User
2759 posts

There is way too much code in your project to easily see whats happening. Can you reduce to a bare minimum showing the problem?

  • how do you observe the glitches?
  • Do you use an external reference, or bypass the internal reference (on the correct pins)?
  • don't use capacitors on the analog inputs, they will change your signal
  • are you sure that your input signal is without the glitches?
  • is this your own board, or are you using a DevKit?
  • when it is your own, did you follow the recommend design guidelines (esp. for the PCB?)

 

user_1377889's picture
User
10803 posts

Welcome in the forum, Stephen.

There was a timing violation shown in your project regarding CLK_SAR and the internal ADC-clock. You could solve that issue by using a sync component as shown in the picture,

Another improvement would be to use a buffered internal reference for the SAR. Choose from SAR's property dialog and select an appropriate pin for the buffering 0.1 to 1 µF cap. Will reduce "normal" noise.

 

Bob

 

Attachments: 
hli
user_78878863's picture
User
2759 posts

With a unbuffered reference that SAR ADC should not allow to set 1Msps sample rate anyway.

user_1377889's picture
User
10803 posts

The conversion rate was set to 2.03 ksps, the clock was 1MHz which is allowed for unbuffered reference.

 

Bob

Stephen West's picture
User
4 posts

Bob,

Thanks for the help!  I did get the sync to make the error message to go away.  Also, the SAR didn't always start when jumping breakpoints.  Unfortunately, it did not make the noise glitching go away.  I also tried the internal reference just now with a little success even though it takes down my effective range.  (0-2.048V).  I could re-reference my schematics for VCC/2 (sine waves to ride on) to be 1.024V if necessary.  I have to move some pins around for the bypassed internal.  It does seem somewhat quieter as I look at my scope (internal) but why would it jump around right after the SAR starts.  Seems weird but maybe the SAR is that noisy.  Also, it doesn't seem to happen on the NTC/Temperature input but I lent my good scope to my SW guy so it is harder to do A/C component only.

Much appreciated on the help.

Steve

user_1377889's picture
User
10803 posts

Some thoughts:

When the glitches appear on one SAR input and not on the other, something is wrong with  your incoming signal. Might be something is injecting noise into the signal source (my favorite is the soldering iron) or the input leaves from time to time the allowed values as getting below GND or above Vanalog.

 

Bob

Stephen West's picture
User
4 posts

So I put the schematic on this attachment and removed (R53, C82, R52, C69, R42, C57).  And I get a 2.03kHz, 300mV impulse/decaying waveform.  Interesting finding....

 

hli
user_78878863's picture
User
2759 posts

One additional question: you are measuring a 20kHz signal, yet you are using just 2000 samples per second. What data are you expecting from that?

Stephen West's picture
User
4 posts

RMS does not require nyquist sampling, hence the 2.03k to capture all points with a repeating number. 20k/2.03k is a repeating number.

hli
user_78878863's picture
User
2759 posts

Got it, you just did never mention that you want to calculate RMS. Which still leaves the question - how do you observe these "noise glitches"?

Log in to post new comments.