How many layers does your board have? I design a ES9038Q2M board with ADP151 for AVCC and Opa1612 buffer for AVCC_L, AVCC_R on 2 layer pcb, one layer is solid gnd plane and only get one channel 2nd harmonic at about -110db, 3rd -120db, another channel is -120db, -108db respectively, opamps for IV and SE converter are also opa1612, do you have any suggestions?
My board has 4-layers. Like I mentioned earlier in this thread I did not get very good results with op amp buffer for AVCC_L/R. Also for some unknown reason I have not had great success with OPA1612 in my ES9038Q2M dacs which is why I switched to OPA1656. In my ADCs OPA1612 work better. The ADC used in loopback measurements needs to be good as well as otherwise you may be measuring ADC distortions instead. Lowering the level to -10 dBFS or below may work better with some ADCs. Good ADC is also required for THD compensations of ES9038Q2M.
Could it be that the OPA1656 is better at handling the load from the differential amplifier/filter?
The I/V op-amps look into a 215 ohm load from the next stage. Combined with the 820 ohm feedback resistor each op-amp (IC6 and IC8) will have a load of 170 ohm. Quite a tough load for an op-amp.
I think that to adjust the compensation of the DAC a notch filter is needed. This will relax the requirements for the ADC.
bohrok2610, you use 2 LPF in each channel in order to have balanced output?
IMHO this is not a good idea because of a LPF components tolerance (ok, resistors can be 0.05%, but capacitors?)
I usually use one LPF and inverter to get the full balanced outputs.
IMHO this is not a good idea because of a LPF components tolerance (ok, resistors can be 0.05%, but capacitors?)
I usually use one LPF and inverter to get the full balanced outputs.
Much good work by you!
Two questions. If you already answered these and I missed them, please abuse me accordingly...
1. Do you know at what frequencies these glitches occur?
2. How do you know that the glitches don't cause the AP to have a bad response? Maybe it's the AP that's distorting from out of band signals.
Thanks for your time.
Those are valid concerns, I agree.
1. I haven't managed to clearly measure the glitches in unloaded voltage mode right at the DAC output pins, on a Tone Board board where I completely removed the I/V. Will need to rig up my old 40lbs network analyzer, but like in my scopes I'm limited to <200MHz. I'll probaly also need to use a preamp and 75Ohm (AC-coupled) environment for this, loading the DAC.
For the moment I could check for magnetic fields hovering a simple loop probe over the ESS chip in a Topping D10 (which behaves much better in general) and saw the expected 100Mhz and mutliples of the local 100MHz clock, running from a 100MHZ crystal (ESS in async mode here). Near the DVDD line to the local decoupling cap the field was strongest and showed components at 50MHz and 100Mhz at about equal levels, plus multiples at lower levels. In the time trace I further saw spikes at ~5Mhz repetition rate.
The final DAC stage likely is running at 50MHz or a 2^N subdivision but probably not any lower than 1/4x (12.5MHz). I would guess the spikes are short compared to bit time (at the final output rate) which then again means 50Mhz++ frequencies. As you can see, a lot of guessing for the moment but I hope to clear this up if possible.
2. My old AP is known to bea bit prone to RF demodulation and therefore I almost always have a passive RC filter at the analyzer input right in the cable connector. Further, I run the levels 10..20dB lower than the auto-leveling would suggest as range setting. This reduces HD and should also reduce any RF susceptibility. The noise increase is counteracted by the excessive time-domain averaging in post processing.
Also, I have noted the same or extremely similar residual patterns when hooked up directly to the ADC, removing the fundamental with a digital filter. This, however, gives additional HD contribution from the ADC, that's why I settled using the AP's variable notch and makeup-gain, besides the general convenience benefits.
All in all I'm quite confident I'm not seing results corrupted by RF-demodulation in the AP.
The reason I ask about the AP is that I know people who measure Class-D and other pulse based amplifiers almost always have to use filters at the inputs to their AP analyzers to make reliable measurements. I've also heard that some people do the very same when they measure DACs, especially ESS based products. It doesn't even have to be RF demodulation - some input circuit could be getting killed by the transient signals due to slew rate listing or something similar. There is also the question of aliasing. But, it sounds like you may have gotten that covered already.
This all does point out the old fallacy that if a signal is outside of the frequency range of what humans can hear, it doesn't matter...
Reverse engineered version of AP AES17 (AUX0025) https://cranialcrumbs.blogspot.com/2015/05/audio-precisions-class-d-filter.html
Hello ivan did you use same voltage divider setup 10k, 6.34k resistor and 1uf cap for vref voltage same as in the evo board schematic. I am glad if you can share the schematic. If you cant i am glad if you teach us the calculation method. For example ESS uses opa1612 in latest es9068 evo board but dac output impadence is different than 9038q2m. And as far as i know opamp feedback resistor and capacitor is dependent on dac output impedance.
potstip, not sure if I understand but why VRef should depend on the DAC's impedance? Divider for the VRev setting would have pretty much any impedance, 5k, 10k, etc need to set the particular voltage and well filtered that's it.
BTW, ES9068 has noticeably worse IMD 19+20 performance vs ES9038Q2M, that was found by our discord member from Korea.
BTW, ES9068 has noticeably worse IMD 19+20 performance vs ES9038Q2M, that was found by our discord member from Korea.
I guess about .1db, because I see a total deviation within .3db but there included AVCC 1%, ES9822 some %, and resistors .5-1% as well.