PCM1792A

Hi, Glenn,

I have asked my colleague to look into this for you.

-d2

Hi Glenn,

maybe it is a good idea to use another type of DAC. DAC without oversampling and with R-2R construction can be useful for you. Specially PCM1704 or PCM1702 from TI production. There is small disadvantage, that these DACs have status NRND. PCM1704 works up to 192kHz for me. You can insipire on my page: http://www.pavouk.org/hw/modulardac/en_pcm1704.html

Best regards, Pavel.

Hi Glen,

Since this is an audio DAC, the band of interest is within 20 - 20kHz. The delta-sigma architecture implements oversampling which shapes the quatization noise to higher frequencies as a way to reduce the noise within this band. This effect can be seen in the plot in figure 20 of the datasheet:

This will influence the noise you see at the frequencies you are operating at. We have many converters specifically designed for the frequency range you are looking for. Could you let me know some of your other key specifications? I can help you narrow down to options that will fit your needs and work with you from there.

Thanks!

Regards,

Brian Wang

Yes, I completely understand the concept of a delta sigma and noise shaping. I also know that there would be more appropriate DACs, but for now I'm stuck with this one in the design and have to (somehow) get this to work.

Noise shaping pushes the quantization noise up into higher frequencies where they have no practical effect on the output. My problem is low frequency noise, <20Hz, which, in audio applications, also has no practical effect since listener's ears don't respond to frequencies below 20Hz. Unfortunately, mine is not an audio application and my users do care about low frequency noise. 

The noise I'm seeing looks like 1/f noise and is significant! Over a 20 second period it can be as large as -50dBfs (30mV p-p out of 10V). The question I have is, is this normal for this part or am I doing something wrong in my circuit design or PCB layout that is inducing this, and if so, what might it be? If it is normal for this delta sigma, I will need to do a redesign of the circuit. If it is a layout issue I will need to do a PCB change.  

My boss is growing impatient so some help with your part would be truly appreciated. I have much measured data that I can share with you that might point to the source of the problem. I would like to set up a GoToMeeting with you to share that data and discuss the problem.

Hi Glenn,

The noise specs on these audio DACs are done with 20 Hz HPF and 20kHz AES17 LPF. Higher noise levels are expected if either of these filters were taken out, but to what level we would also need to take additional measurements. One thing to try is to do another noise measurement on your board with these two filters and compare that to our datasheet numbers. That way you can see if your design is acheiving the expected performance out of the part and validate the measurement data you have.

Regards,

Brian Wang

Thanks Brian,

I understand the specs and high frequency noise is not a problem, I'm generating sines up to 80kHz and the high frequency noise is acceptable. The problem is noise below 20Hz. 

If I program a DC output there is about 3mV/V of very low frequency random noise, with a 1/f distribution. I also see this riding on top of sine waves, again, about 3mV/V of amplitude. The frequency is far below the 20Hz hearing drop off, so most audio applications don't care, but mine is not an audio application and this noise is a problem for my product. My question is, is this truly a PCM1792A issue or do I have a circuit design and/or PCB layout issue?

Glenn Fasnacht

Hi Glenn,

The performance you are seeing below 20Hz is expected of the PCM1792 (can see in figure 19 of datasheet below), the 1/f noise distribution is from flicker noise intrinsic to the part but being outside the audio band it is not specifically targetted. Please note that unlike high-frequency noise which can be eliminated relatively easily with simple R-C filters, low-frequency 1/f noise in the 0.1-Hz to 10-Hz range cannot be easily filtered without affecting dc accuracy. The only way is to minimize the 1/f noise is to ensure that you have as little of it introduced into the circuit as possible.

Regards,

Brian Wang

Hi Glenn,

Great to hear you are able to reduce the noise by that much. It may also help if you isolate the parts around the DAC and see how much noise they each generate to find the remaining noise contributors.

Regards,

Brian Wang