ADAU1467
Recommended for New Designs
The ADAU1463/ADAU1467 are automotive qualified audio processors that far exceed the digital signal processing capabilities of earlier SigmaDSP® devices...
Datasheet
ADAU1467 on Analog.com
Hi,
An ADAU1467 evaluation board is connected to a DAC board with ~15 cm ribbon cable (through ground). Spectrum analyzer connected to the DAC output shows 2 spikes (approx 430 and 215 Hz at approx -90 db).
The spikes move a bit (+/- 10 Hz) over time probably due to warming up.
I guess it’s just a derivative of the ADAU unwanted emission and the Spectrum Analyzer ADC clock difference.
The spikes amplitude doesn’t depend on (main) signal level/frequency/presence at all, i.e the spikes exist always and stably.
If the same DAC is connected to different i2s sources (configured with the same i2s parameters) there are no spikes at all. I tried 2 different USB-2-i2s adapters on the same ribbon cable. So I guess the problem is not the DAC related.
Experimenting with the problem I’ve found that it’s enough to connect just a spare, connected to ADAU’s MCLK only (even not connected to the DAC) wire of exactly the same length to see the same spikes in the spectrum. So the MCLK wire acts as antenna and emits a signal of significant level that is “successfully” received by the DAC’s analog (OPA based) low pass filter. The problem is pretty reproducible and never happens with a different i2s MCLK source.
Thinking of such behavior I could imagine that the main difference between the ADAU’s MCLK (and other signals) and third party i2s/MCLK is the ADAU ability to change signal strength and slew rate.
I may be wrong be it seems to me that the latter is implemented via a kind of PWM at a frequency much higher that MCLK (eg ADAU main clock ~ 300 MHz) and the 15 cm wire is just a good transmitting antenna for that 300 MHz. I may be wrong of course, I have no details of the strength/slew related implementation
What I tried, unfortunately unsuccessfully
- changing ADAU filter schematics – no effect at all
- changing MCLK programmatically (in SigmaStudio) – tiny effect. The spikes level changes very slightly
- changing slew rate and signal - tiny effect as well.
- using a shielded cable and instead of ribbon cable has a noticeable effect. The spikes level is getting lower but stays much higher desired level (DAC noise level or so)
- shielding the entire DAC board itself – tiny effect. Even with the shielded MCLK board-2-board cable there is a ~3cm printed MCLK wire on the DAC board that still probably emits the same.
- 12 MHz LPF (using on board 30 Ohm “debouncer” and external 200pF cap) before the MCLK
“antenna” – tiny effect.
- an active buffer of 2 inverters (74HC04) connected sequentially – no effect on the spikes and the DAC stops working (probably due to unacceptable delay between MCLK and other i2S signals)
- 20 kHz RC LPF on the DAC output gives the best but though incomplete result.
Using other less RF sensitive DACs is not option for me. The planned device is a crossover, too many new DACs with unpredictable result will be required, and all the components are already mounted into a chassis. 20 kHz RC LPF on the DAC output looks rather “infamously”
Do you have an idea for what to do or try?
Thanks in advance,
Serge
PS Sorry, turned out too long
Hello SergeG,
Longer posts are not a problem at all!. I am the king of long posts! This is so much better than getting a post that says "It is not working! I get no sound! What is wrong?".
I think you may be dealing with multiple problems. Your early comments:
DAC output shows 2 spikes (approx 430 and 215 Hz at approx -90 db).
The spikes move a bit (+/- 10 Hz) over time probably due to warming up.
This is classic symptoms of idle tones from Sigma-Delta converters. Despite the attempt of designers to get rid of these they can still appear. It happens when the audio signal is very low so the DAC does not have much to do. Small errors and mismatches in silicon and current sources etc. inside the DAC will cause an error that gets fed back into the modulator and causes sort of a feedback or a tone. Hence the name "idle tones". They will differ from part to part and they will be sensitive to temperature.
There are a couple of ways to get rid of them. If the noise floor is high enough then it will not happen. Most people do not want to increase the noise floor for some reason. :) Another way is to add a small DC offset. If the DAC filter is cap coupled then this will work well. The downside of this is that the clipping will be non-symmetrical so you will clip slightly early. Since you have a DSP you can create a smarter project that adds some DC to the signal when the audio is really low and slowly get rid of it as it gets louder. I have not done that but it is possible.
First, to verify that they are idle tones. Setup a project with an oscillator going out to the DAC. Mute the oscillator and observe the tones on the DAC output. Then turn on the oscillator at about -70dB and see if the tones move up in frequency. Increase the level in about 5dB steps and see if the tones go up in frequency. If they are idle tones they will go up and eventually go up out of the audio band or just disappear.
If you verify they are tones then just add a little DC offset to the audio right at the output and that should take care of it. Do the same thing, start with very little DC and then increase until the tones go away.
Changing to a different ADC or I2S source probably made a difference because all ADC converters have a DC offset. Most of them have a HPF built in to get rid of it. So if you use an ADC with the filter turned off you might of had enough DC offset to get rid of the tones. In addition, boosting low frequencies in the DSP and increase the DC offset as well.
Your other experiments with the clocks and the drive strengths etc., seems to me to be showing a grounding issue with the entire system.
Modern systems are horribly grounded. Often ground comes from the USB port and most of those are horribly grounded through ferrite beads allowing high frequency currents to drop a voltage across them. Then the computer and most evaluation boards are using switching power supplies. Most of them are not grounded at all so they are adding a bunch of noise and no reference or chassis ground to give a path for return currents. You need a path for noise picked up over the air to return to the source and with no chassis ground the path will be through the audio circuits and power supplies.
We (Analog Devices) do make some great USB isolators that allow you to get power from a different source and isolate the grounds allowing you to properly ground your device. This gets so beyond this post. I probably could write a book about it!
In reality you are probably designing a speaker with a crossover. You will most likely be using a local power supply. You will just need to carefully look at how the PCB layout is done for power and ground and placement of the components. You have to keep in mind where the signal currents are going and where/how will they return? Do not have regulators and voltage references be such that it will share a return path with some digital clock signals. The audio ground paths need to be as short as possible and far from current paths and magnetic fields from those currents to prevent this crosstalk.
Also, super important is how the bypass caps are done for all the parts. Look at this document I wrote on the subject. It is for a specific codec but applies to any part.
(+) PCB Layout Best Practices for AD193x codecs - Q&A - Audio - EngineerZone (analog.com)
These are my first thoughts. See, I told you I am the king of long posts. 
Dave T
Hello Serge,
This is still somewhat ground related but also seems like it can be a transmission line issue as you are thinking. There are pin drive settings on the DSP.
Set these lower might help. There is a setting for the Clock Out pin. If you are not using it then make sure it is turned off. There is a register setting to shut it off.
The harmonics of a 12MHz square wave go quite high! Lowering the slew will help. You can also add some damping resistors on these clock lines. It is very difficult for me to really help too much here because I would really need to be in front of it to see all the connections and also to see details like PCB layouts.
I am not sure exactly which board you are using?
Dave T
