oscillation at the embed analog front end circuits of SAADC in nRF52832

[Kiyoshi Iwai]

Hello,

I found the strange phenomena on nRF52832.

By using the SAADC, I have some noise on ADC Data only at a certain temperature range.

Temperature range is depend on individual IC.

I might be the circuit oscillation at analog front end of SAADC, I think.

(internal reference voltage or op amp etc,)

I tried to switch DCDC on/off, but there's no difference on phenomena.

Anyone can resolve it?

I attach the files of real data and measurement condition.

Best regards,

Kiyoshi Iwai

3426.SAADC_noise_devzone.zip

[Kiyoshi Iwai]

Hi Kenneth,

marking is following.

CIAA chips are "N52832  CIAAE1 1803AA" all.

QFAA chips are "N52832 QFAAB0 1740JK" and "N52832 QFAAB0 1701FH".

And why don't you use our project?

Our software engineer  took a long time (almost 3 days) to modify the project which can be submitted to other company. And Avnet also used this project  and confirmed to recreate this issue.

You are going to waste all things ?

I strongly recommend to compile our project, use our project and reproduce this issue at your Lab.

This problem is almost on Hardware obviously, I think.

Could you find the root cause of Hardware as soon as possible?

(And please return this to public mode.)

 

Best regards,

Kiyoshi Iwai

 

[Kenneth]

Hi Kiyoshi,

Thank you for providing the chip markings.

The reason for not using your project is that the test lab must perform the tests in a controlled manner, to ensure we can identify if this is a hardware issue. If we test using your software, then we will not know if it's caused by software or hardware. This is currently the highest priority task to find out.

The test lab are preparing test firmware, and hopefully the automated test can run during tonight, so we have results by tomorrow. That is the plan. I will let you know if there is any delay.

Best regards,
Kenneth

[Kenneth]

Hi Kiyoshi,

The noise in internal reference is temperature dependent, and likely the main cause of your issue.

The frequency of the internal RC oscillator is also temperature dependent, and will influence measurement.

You wrote you also experienced noise when using external reference(VDD/4), and we believe the cause of that is CR2032 is not linear over temperature, we can see that there is noise from CR2032 (VDD and analog input), the noise depends on current consumption (burst sampling) and temperature.

Best regards,
Kenneth

[Kenneth]

Kiyoshi Iwai said:

You mean that nobody can use the internal reference 0.6V with CR2032, right?

I think that will depend on the required accuracy of the ADC measurements, but for 12(14-bit with oversampling) you likely will not be able use the internal reference or a CR2032 directly no (but possibly with an LDO).

Best regards,
Kenneth

[Kenneth]

Hi Kiyoshi,

It seems that we still measure noise using all the suggested improvements. 

We will need to investigate further. 

Best regards,
Kenneth

[Kiyoshi Iwai]

Hi Kenneth,

What is the noise of CR2032?  You mean CR2032 can make such higher frequency noise?

Unbelievable.

If the cause is the internal resistance of CR2032 as you said, there will be disadvantage at lower temperature.

The lower temperature, the more issue will occur.

Of course the characteristic of internal resistance for temperature is not linear, but it  will not go up and down. It will go in one direction gradually.

I alread found that 15 degree is bad, but -5 degree is OK, or only 35 degree is bad, but other all lower temperatures are OK.

I don't think the cause is CR2032.  Even though, you doubt CR2032?

Best regards,

Kiyoshi Iwai

[Kiyoshi Iwai]

Hi kenneth,

Is this the final conclusion of you Nordic Semiconductor?

I have a big disappointment to read this.

Do you intend to terminate the investigation anymore?

By the way,

Could you tell me what power source do you use on the circuit of internal reference 0.6V?

Do you connect VDD to internal reference directly?

Or via the embedded LDO?

I have no block diagram of internal reference voltage, so could you provide it in detail?

Best regards,

Kiyoshi Iwai

[Kenneth]

Hi Kiyoshi,

I assume you also have a deadline.

The team are still investigating and doing measurements, but it may seem that there is some short burst noise in the SAADC peripheral that is device specific and temperature dependent. The error is about 0.2% of full scale measurement when it occurs. There are two hypothesis of root cause: 

- Accumulated burst noise in the auto-zero loop could cause an error.

- Burst noise on the reference buffer signal could potentially cause noise as we see here.

They plan to do a bit more measurements, but it is not likely there is a workaround to the problem you are experiencing.

I notice you have a few questions on the internal IP, I don't know the answers to those, typically this is not something we share.

I can only apologize for the time this has taken.

Best regards,
Kenneth

 

 

 

[Kiyoshi Iwai]

Hi Kenneth,

Do you have any progress?

I really want to know you Plan.

BTW, what is the auto-zero loop?

Is that same thing with CALIBRATEOFFSET?

Best regards,

Kiyoshi Iwai

[Kenneth]

Hi Kiyoshi,

I know the test lab are doing some measurements to understand the noise, but to answer your questions and summarize:

The auto-zero loop is internal design, which I believe zero the circuitry between each sampling. If there is noise here it could be accumulated. Regarding CALIBRATEOFFSET I have seen in my own measurements (using your firmware) that sometimes the calibrate offset can also cause an error similar to what you see, in that case 12 samples in a row have an offset until for instance next calibration occur.

There is a temperature dependency on the error, this is also something we have stated in the datasheet:

So I am sorry to say, but we likely can't meet your requirement in this case if you need 12bit accuracy over temperature. I can only apologize for the time this has taken.

Best regards,
Kenneth

 

 

[Kiyoshi Iwai]

Hi Kenneth,

I heard you Nordic Semi has abandoned to investigate the root cause of this issue  from Avnet today.

That shouldn't happen.

Why don't you want to know the root cause?  I can't believe it.

Your own IC has obviously the fault and you Nordic Semi shall make a improvement.

If you were an engineer, you ought to demand the root cause naturally, I think.

Are there no engineer who want to find out the truth in Nordic Semi?

If so, you Nordic Semi will never to be able to become the IC supplier.

The IC supplier should take a responsibility for supplying the higher quality products to all customers.

When the customer point out the problem, the supplier has to investigate the root cause and answer the counter measure to the customer soon.

I hope you will re-start the investigation and find the root cause out in a short time.

First of all, I strongly need the root cause, you know.

Best regards,

Kiyoshi Iwai

[Hung Bui]

Hi Kiyoshi, 

 

Kenneth is on vacation. He will be back in about 2 weeks. I'm sorry for the delay in our response.

 

Best Regards, 

 

Hung