DIY Audio Analyser using CS4398 / CS5381 (CS5361)

I recently set to work the guts of the audio analyser:
- MiniDSP USB Streamer
- Interface card
- ADC
- DAC

The ADC and DAC were from an old project. the tests I had been able to do on these were done using simple test gear, I knew the distortion was fairly low - but had been unable to really bottom out the level.

I did know the DAC was down in the 0.001-3% level or below.

Imagine my chargin when I fired the system up, got all the bits running and measured something closer to 0.01% distortion at -1dBc!

I was initially convinced that there was something wrong. After cranking the level down to -10dB, the distortion dropped to 0.001%. Further tesing (I ran the DAC at -1dBc and attenuated the output) proved that it was the ADC dominating the measurements.



Much fiddling led to me concluding that the ADC buffers were the culprit. I would like to blame the chip manufacturers, as I pinched the buffers from one of their app notes. The above is a BAD idea.

If all the components are perfext, then hooking the inverting driver to the output of the voltage follower should be OK, but things are not ideal, and I am pretty sure it results in imbalance in the ADC drive.

I found this by rejigging the drivers to:


Which was much better.

Distortion at -1dBc is not perfect, but pretty close to 0.001%, and at -10dBc is heaps better:



The boards have protection on the input of the ADC (schottky diodes to ground and positive rail) which on earlier tests showed to be transparent. I was initially somewhat concerned that these would be a problem, but not so.

Observations I have made in this process:
- The first ADC board I loaded I used a CS5381 that I pinched from one of my other boards.
- This first board in the test setup measured well on the left channel, 0.009%, but 0.0012% on the right channel. This did not make sense.
- After a lot of fiddling, I ripped a CS5361 off another board and loaded this onto another of my new ADC boards.
- Ahh that is better - both left and right channels measured identically. 0.0006% at -10dBc.


I am kind of at the point where I am not sure if I am measuring the ADC or the DAC. It is apparent that the limit for the CS4398 is in this region somewhere.

So where does this leave me?
-1- I NEED to fix the ADC's in my crossover! Damn, I am actually quite miffed with myself over this - at least they are standalone boards and I can drop the ones I have made in.

-2- I think I need to get better DAC's for this test system. I have convinced myself that I am at the limit of what a CS4398 can do - but what is the right choice?

-3- I think I am close to what a CS5361 can do - though I think I need to get better DAC's to prove this to myself.


I did play with using a voltage follow buffer on the input of the ADC. I made this switchable, testing showed that this has no measureable effect on a signal showing 0.0006% distortion - so this buffer is in my mind a good thing to have. I will turn a measurement system with an input impedance of 4.7K into something with 100K input impedance.

There seemed to be some sensitivity to the capacitors, but right down low.

The boards are floating in the air, and do show sensitivity to hum pickup. They do need a box, and there is clear evidence of this at -125dBc.