Hello, I use ADAU1701 as crossover for loudspeakers. Its on freeDSP Classic SMD B board. Sound from DAC0&1 is OK, no noise. But DAC2&3 have unacceptable high noise. I would describe it as white noise. When I power down DAC in SigmaStudio, noise is gone.
Here are things I already tried
There are many people using this board without problems, so I don't expect the board is bad, or badly designed. All SMD components were pre-assembled.
Amplifiers are class AB, have 27dB gain and I could not measure/hear their noise.
Is this ADAU1701 broken?
Your graph shows the noise about 94 dB below your 1 KHz tone for DACs 2-3, and 110 dB on DACs 1-2. The ADAU1701 data sheet spec is 104 dB, so really you're not doing all that badly. I'm wondering how your results compare with that of other FreeDSP Classic users. Have you tried sharing your tests while asking about the audible and measured noise at the FreeDSP forum? https://www.diyaudio.com/forums/digital-line-level/285260-freedsp-main-thread.html
The data sheet's noise specs are typical and thus likely not 100% tested on individual production chips. Still, the difference between outputs is puzzling. Check the bypass capacitors at pins 34, 36, 40, 41, and 48 to GND (pin 1) -- examine the soldering at these pins and at the caps. Since most of your measured noise is at higher frequency, perhaps setting up DACs 1, 2 for your tweeters and the other DACs for the bass drivers may help keep the perceived noise to a manageable level.
I'm afraid I'm an old dinosaur who was happy with the 60 dB S/N of cassette tape!
Hello Bob, when I saw that little difference in measurements I would not believe that I can hear it, but from DAC2&3 I can hear loud noise from distance more than 3m from 94dB/2.83V tweeter, while from DAC0&1 only slight hiss from distance of 0.5m and less. It was on the same amplifier and the same tweeter. I also tried different amplifier, results were the same.
I already wrote about my problem in that thread on diyaudio (DarkOne is my nickname there)
All these SMDs were pre-assembled, I don't expect and I don't see any bad joints.
Your suggestion about using the noisy DACs for bass and other for treble is not usable here. My aluminium cone woofers have massive peaks above 3kHz to more than 100dB/2.83V, so the noise will be even more amplified.
I've attached little sound clip comparing outputs. Recorded with ECM8000 microphone and ESI Maya22USB interface. Mic distance was 20cm, amplifier gain 27dB and loudspeaker sensitivity approx. 88dBClick here to play this audio clip
Sorry I haven't been very helpful. The math does explain why the noise is quite audible. 27 dB of amplification of the 0.9 Vrms DAC output works out to 20 Vrms, or 50 watts into eight ohms. This is 17 dB above one watt, so your full-scale SPL at one meter from the speaker is 88 + 17 = 105 dB. Subtract the 94 dB worst-case DAC S/N and this leaves 11 dB SPL, which could be audible (and quite annoying) in a quiet room. Such can often be the case with bi-amplified applications like this one -- where post-DSP, the gain is always at max. Since you tried the FreeDSP forum, I assume that the answers there, if any, were less than helpful.
Since your recorded noise appears "brighter" for outputs 2 and 3, perhaps there's an issue with the components in the active output filters for these channels. Failing that, it appears that you have a "bad" ADAU1701 -- or simply I'm out of ideas on what else could be causing two DACs to be this much noisier than the other two. Perhaps someone else here has another suggestion.
Probably, I would not complain about noise if drivers of my new loudspeakers are 88dB or less sensitive, but the woofer is 91dB and has breakups of 100dB at 4kHz. And the tweeter is around 94dB. Sensitivity of these drivers make the situation much worse.
I have nothing to lose now, hot air station is being shipped to me and I have 3 spare chips to play with.
I went in to the lab and made some measurements today. I used a larger evaluation board that has all four outputs available unlike the MINI board that has only two available on a jack.
This board has a choice of a passive filter on the DAC or a 2-pole active filter which will give you better specs but I just used the passive single pole filter.
I did two tests. One is an FFT of a signal coming into the ADC going through the DSP and out of the DACs. I only used one channel of ADC so the signal was exactly the same going to all four DACs. So you will see the distortion of the entire signal chain not just the DACs.
Then the second test I muted the signal in the DSP and ran the FFT to get just a noise floor. All four channels were pretty much the same. So something is up with the board.
Here is the SigmaStudio Project.