DAC8831EVM: help with strange noise

Hello John,

I will assist you with this issue. In order to do this, I will need some more information.

1. What is the clock rate that you are using?
2. Is there anything else occurring at the time of the "noise"? i.e. DAC output changing, CLK pulse, etc.
3. Are you measuring this at the output of the DAC or at the output of the amplifier?
4. Does it occur when the digital interface is disconnected?

At first glance it does not look like noise. Noise is usually random; both in magnitude and period.This signal looks well defined and consistent. I believe that it could be one of two things:

1. Digital feedthrough (Very likely)
   1. One of the digital signals has a current path that may be affecting the output trace.
2. RF interference (Very unlikely) 
   1. Comes from the environment around the system, and couples into the output trace.

I will test the EVM in the lab and let you know what I find. It would be most useful if you can share any variables in your system. i.e Power supplies, reference input, etc.

Thank you very much for your reply.  I am at my wit's end...

1.  I am not using any clock rate, because the digital lines are disconnected, and floating.  I initially had the digital lines connected, but disconnected them while debugging to attempt to reduce the parameters.

2.  No digital signals are applied because the digital lines are disconnected.  Since the digital lines are floating, I thought that might be a problem, so I did try to ground !LDAC and pull-up !CS, but that didn't seem to change anything, so now I am just leaving it floating.

3.  I am measuring at the output of the amplifier, but when the amplifier is connected, the DAC output gives similar response.  On my own PCB (not DAC8831EVM), I used a razor blade to cut all the traces to the output amplifier and measured the output from the DAC, pin 2.  I do NOT see the pattern.  Please see attached image of a DC-coupled measurement of the DAC Vout with the output amplifier *disconnected*.

4.  Connecting or disconnecting the digital interface doesn't seem to make a difference.

Powersupplies:  I initially used a HP E3630A, but changed to a IHAA5-1.5/OVP from http://www.internationalpower.com/ to see if it made a difference.  It did not solve the problem.

On the DAC8831EVM, I am using the onboard reference.  On my own circuit boards, I am using a REF3025 or a REF3030 from TI, similar to the DAC8831EVM reference design.

Please let me know if there is anything else you would like me to check or let you know of.

Attached is sample data taken showing ac-coupled measurement of the dac output after the bipolar amplifier (in blue) along with the ac-coupled measurement of a dc power line (in red), I think +5V.  Although the DC power is not "perfect," it doesn't seem to be carrying the same signature.

Thank you.

Yes, I am using the OPA735.  I think the period should be more like 0.62us?  However, the auto-zero seems to be a likely source.  It would also explain why the spikes don't appear when the buffer output is 0V, since discharging from 0V to 0V wouldn't cause any discontinuity, and the larger the absolute voltage difference is, the larger the spike amplitude would be, like at +/- Vref.

I tried swapping the OPA735 for a OPA353, the amplifier specified for buffering the Vref, because it was the only opamp I had with the same footprint and pinout.  The result was an oscillation at 4.1MHz, with an average value of -Vref.  I'm guessing that is because the OPA353 has a much higher gain bandwidth product, and my pcb layout is poor.  Unfortunately, the amplitude of the oscillation was too large to resolve whether a 20mV spike was still there.

Assuming the OPA735 is the problem, I would appreciate a different amplifier recommendation.  (By the way, were you able to recreate the problem in your lab on a DAC8831EVM?  If it is the OPA735, I imagine you would see this whenever the DAC output is farthest away from 0V?)

My application is to create a bipolar, analog control signal that is the input for a galvo mirror control system.  I would eventually like approximately +/-3V with an accuracy of about 1mV (possibly down to 0.5mV?) and noise small enough that the accuracy is meaningful. I believe the mirrors have a response up to several kHz.  (Thus, I think the 52us spikes could have some small, noticeable effect.)

Would you still recommend the OPA277 or a different amplifier?

Thanks again.  You have given me hope!

Hello John,

Apologies for the long delay, I have been terribly sick the past few days.

John Choi22 said:

I think the period should be more like 0.62us?

Yes, you are correct. I must have shifted the decimal point incorrectly.

John Choi22 said:

By the way, were you able to recreate the problem in your lab on a DAC8831EVM?

I have not been able to go to the lab to recreate it. I should be able to go in tomorrow.

John Choi22 said:

 If it is the OPA735, I imagine you would see this whenever the DAC output is farthest away from 0V?

Yes that is correct, this would be when the OPA input bias is the largest.

John Choi22 said:

My application is to create a bipolar, analog control signal that is the input for a galvo mirror control system.  I would eventually like approximately +/-3V with an accuracy of about 1mV (possibly down to 0.5mV?) and noise small enough that the accuracy is meaningful. I believe the mirrors have a response up to several kHz.  (Thus, I think the 52us spikes could have some small, noticeable effect.)

The LSB size that you are planning on using is 6/2^16 ≈ 0.092 mV. A 0.5 mV accuracy is definitely achievable.

As you mention, on top of this theoretical LSB size, there is noise to consider. Plus offset and gain errors.

John Choi22 said:

Would you still recommend the OPA277 or a different amplifier?

Yes. I would still recommend the OPA277. It has the GBW that you need and has much lower noise than the OPA735. Plus you don't have to worry about undesired spikes at the output.

Do keep in mind that the OPA277 requires a headroom of at least 1.2V, this means that the power supply must be at least 1.2V above the desired output voltage. In the case of your application, you will need at least ±4.2 power supplies.

I would recommend that you take a look at these articles if you want some insight as to how to calculate the errors in your DAC system.

DAC Essentials: How accurate is your DAC?

DAC Essentials: Static specifications & linearity

I can help you clarify any questions that you may have.

Hope you get well soon and are feeling better.

I have submitted a new pcb design using the OPA277, and am waiting to receive the board and parts.  Hope to try it in about a week.  Will let you know what I see.

Also, what do you think about trying a OPA237 as a test in my current design as a drop-in replacement for the OPA735?  I think the OPA237 doesn't have the zero offset feature, is cheap, and is pin compatible.  I would still use the OPA277 in the final design, but could use the OPA237 just to confirm absence of the spikes...  (I know the performance will not be as good as the OPA277.)

When you are feeling better, please do let me know what you see in your lab with the DAC8831EVM.  (I believe I also saw similar behavior with the DAC8832EVM.)

I will read the articles and let you know if I have questions.

Thanks for all your help!

Thank you very much for your concern.

After a brief inspection, the OPA237 is fairly similar to the OPA277. I think using the OPA237 should help you confirm the absence of the spikes.

I will make sure to test the DAC8831/2EVM when I have access to the lab. I will share with you my findings.

I received the OPA237 today, and tried replacing the OPA735.  So far, all our suspicions are confirmed.  Here is a capture of the ac-coupled measurement of the -3.0V power-on reset condition of the DAC8831.

The peak-to-peak measurement of the oscilloscope claims 1.6mV to 2.0mV, giving something like +/-0.9mV.  

I have not received the new pcb designed for the OPA277 yet, but hope/expect it will be similar.  The 30mV spikes are definitely gone now.

Thanks again.

Apologies for the delay. I am currently in a business international travel.

In your application the main difference between the OPA237 and OPA277 is probably the input offset voltage.

It will introduce a max offset of about 750µV (vs 20µV from OPA277).

This won't affect linearity. Meaning that your step sizes will still be accurate to your goal of 0.5mV, but it will offset the entire output transfer function by the input offset offset voltage of the amplifier.

If this is not an issue, then the OPA237 will serve your application well.

I can also review your schematic and layout if you want. We can do this privately if you allow me to contact you with your registered E2E email.

Yes, please contact me.  Let me know if I need to change any settings to allow it.

This thread has been taken offline.