DAC11001A: 1/f noise

Hi Lukas,

Thank you for your query. I am checking with the internal team if the noise density data is captured below 500Hz or not. I will get back by tomorrow.

Could you please provide more information on the application?

Regards,

Uttam Sahu

Applications Engineer, Precision DAC

Hi Uttam,

Thank you for your response.
I need precision DACs for a very stable programmable DC voltage source. Currently I use the AD5790 from Analog Devices. But it's noise below 10Hz is rather high compared to the DAC11001A (1.1uVpp vs 0.4uVpp). The source must be very stable for a few hundreds seconds and the stability for this interval is mainly dominated by the 1/f noise. Therefore having low 1/f noise is a key point.

Best regards,
Lukas

Hi Uttam,

One addidional thing is I would like to know the (1/f) noise of the DAC11001A only. To me it's not clear if the figures 41 and 42 includes the noise of the reference, reference buffer and output buffer.
Can you clarify this?

Best regards,
Lukas

Hi Lukas,

The 1/f noise plot in the datasheet specifies the noise for DAC11001A only after subtracting the noise contributed by the reference and amplifiers.

We haven't collected the noise density data from < 500Hz. 

Let me know if you can proceed with the 1/f noise spec at the moment.

Regards,

Uttam

Hi Uttam,

Thanks for the clarification. This means that the figure 42 in the datasheet only specifies the DAC noise.
Unfortunately, this specification is not sufficient for me. Figure 42 says nothing about the 1/f corner frequency and it's not clear what is the contribution of the 1/f noise and white noise. I would like to compare the noise density with the one of the AD5790 (which has a noise density plot down to 0.1Hz)

Can you collect some data to specify this?

Best regards,
Lukas

Hi Lukas,

Thanks for the clarification. We have started preparing the setup for this measurement. It will take some time as it needs to be an all-battery power based test. The standard power supplies complicate the measurement due to the noise coupling.

I will get back with my update by the end of next week. Apologies for the delay - there are some logistics involved.

Hope you understand.

Regards,

Uttam

Hi Uttam,

Thank you very much for doing the measurements. I appreciate this very much.
I understand that the measurements will take some time.

Important is that you capture enough data to get a meaningful plot of the spectral noise density down to at least 0.1Hz (better 0.01Hz).

Thank you very much and best regards,
Lukas

Hi Lukas,

We are still in the process of getting the test set up ready. I wanted to update that we are on track. I will get back with the status update next week.

Regards,

Uttam

Hi Uttam,

Any progress on this?

Best regards,
Lukas

Hi Uttam,

Thank you very much for the plots. Of course I'm interested having the final results.
How reliable is the simulation data below 10Hz? Do you have any experience (maybe from other devices)?

Best regards,
Lukas

Hi Lukas,

The simulation data is reliable as you can see it is following the measured results. The problem below 10Hz for the measurement setup is that the spectrum analyzer is reaching its resolution limits. So, we will need to design multiple narrow bandpass filters and measure the data at lower frequencies.

However, the simulation data matches with the flicker noise measurement we have done with the oscilloscope (0.4uVpp). Hence, we are confident that the simulation results are accurate.

Hope that answers your question.

Regards,

Uttam

Hi Uttam,

Thank you for your answer. One last thing: Can you tell the exact 1/f corner. From the simulation data you sent me, its hard to extract, because the plot stops at 0.1Hz and its not in a log-log scale.I estimated the corner frequency around 0.5Hz.

Best regards,
Lukas

"However, the simulation data matches with the flicker noise measurement we have done with the oscilloscope (0.4uVpp)"

I must have got something wrong but by my calculations, integrating the spreadsheet values from 0.1 to 10Hz, gives approx 24nV rms or 158nV p-p which is a long way from the datasheet 400nV p-p. Unless of course the spreadsheet values are based on a reference voltage << 10V (resistor 1/f noise is proportional to the applied voltage)?

Frankly the spreadsheet numbers don't seem credible - the AD5791 has a 1/f corner of approx 7Hz so 0.3Hz for the DAC11001A is quite oustanding! I'd be surprised if those numbers could be achieved with wirewound resistors! Resistor excess noise (1/f) is, I believe, inversely proportional to the amount of resistive material which would suggest the DAC11001A resistors are enormous compared to the AD5791, which doesn't seem likely.

Please don't get me wrong, the DAC11001A looks to be a very exciting part but I think we need to wait for some definitive 1/f noise measurements.

Hi Tony,

Please dont look at sim data which we shared earlier thats pure DAC noise irrespective of reference buffer voltage, buffer etc.

Our test set up is not yet up to measure way below 0.1Hz.  We are confident in our measurement for this DAC upto 0.4Hz. If you look at the sim data , noise is way too small, in order to measure that very decisively , we need to make sure that our measurement is not corrupted by any external factors.

Having said that, looking at the actual data in 0.1Hz to 10Hz in scope mode (DAC followed by ultra low noise amp stage with 0.1Hz to 10Hz filter ) is approx 400nVpp.

I will get back to you for corner frequency value soon.

Regards,

AK