Made quite a few measurements and got to see some interesting features of third harmonic distortion.
This measurement is on a syntetic a*x^3 harmonic. The a is fixed but input is varied 30 dB
We can see that adjusting a with 3 dB is the same as adjusting input by 1 dB. No suprise there.
But it is only until a low level is reached. Then it is not very predictable
This measurement is on a syntetic a*x^3 harmonic. The a is fixed but input is varied 30 dB
We can see that adjusting a with 3 dB is the same as adjusting input by 1 dB. No suprise there.
But it is only until a low level is reached. Then it is not very predictable
Also put a 3250 HP filter on the input of the second analog diff amp (2.2 nF before INA)
Then a whole lot of action came to display at the different frequencies and levels.
Some of the measurements took a few hours so noise in each frequency bin is way down.
Example of setting:
Then a whole lot of action came to display at the different frequencies and levels.
Some of the measurements took a few hours so noise in each frequency bin is way down.
Example of setting:
Code:
Stepped THD measurement at 2,5 kHz
From -10,0 to 0,0 dBFS in 0.1 dBFS steps
4 averages, FFT length 4194304Even if the analog diff amp has quite high ohm feedback (about 20k) we can "see through the noise"
Both analog 3250 Hz filter on 01 and optional digital filter on 23
The syntetic 3rd harmonic distorion is for reference how the analog filter affect the 3rd harmonic.
This is the reference without HP filter and all harmonics
Both analog 3250 Hz filter on 01 and optional digital filter on 23
The syntetic 3rd harmonic distorion is for reference how the analog filter affect the 3rd harmonic.
This is the reference without HP filter and all harmonics
Last edited:
The -6dB - 6dB = -12dB on diff amps helps to attunate the signal before DAC so that really helps to se the DAC distortion and not only the ADC distortion.
The HP filter attunated further, but more difficult to translate the measurement.
The HP filter attunated further, but more difficult to translate the measurement.
Last edited:
Just bought me a clean fundamental😉
So made some measurements un a SMSL SU-9 PRO into the Xonar U7. No differential amps. SPDIF as input to SMSL
Max resolution on REW
Noise floor 4M 8averages
Distortion, THD 0-9, at -4dBFS and 6dB steps down to -22dBFS
There is really something to gain for distortion measurements by decreasing the ADC input.
Don't have access to daily setup so can't make any subtractive measurements for some days.
So made some measurements un a SMSL SU-9 PRO into the Xonar U7. No differential amps. SPDIF as input to SMSL
Max resolution on REW
Noise floor 4M 8averages
Distortion, THD 0-9, at -4dBFS and 6dB steps down to -22dBFS
There is really something to gain for distortion measurements by decreasing the ADC input.
Don't have access to daily setup so can't make any subtractive measurements for some days.
Back to the setup with the -6dB + -6dB gain differential amps.
Source is the CS codec card stripped of the output filter and ADAU1452 as the TDM source,
Output 0 ad 1 are sent to differential inputs, output 1 inverted.
This is 300 Hz stepped THD
The distortion is not linear with input as showed over. dips at -6 and 0 dB
So this is a reference where ADC is better than DAC with respect to THD
Source is the CS codec card stripped of the output filter and ADAU1452 as the TDM source,
Output 0 ad 1 are sent to differential inputs, output 1 inverted.
This is 300 Hz stepped THD
The distortion is not linear with input as showed over. dips at -6 and 0 dB
So this is a reference where ADC is better than DAC with respect to THD
This is generator at 0, -2.5dB and -6dB, so about -5 dB lower in graph above
So now I am interested in investigating if the distortion raises abow 400 Hz also on DAC codec or if it is only in the ADC
So now I am interested in investigating if the distortion raises abow 400 Hz also on DAC codec or if it is only in the ADC
1. is CS DAC as above. -1 dB from generator. So -6dBFS into ADC
2. is clean fundamental (SMSL DAC) time alligned down to closest sample.
3. is CS DAC inverted and clean fundamental (time alligned) into differential amps. So fundamental is down max about 40 dB compared to the two sources.
The fundamental is under -20dBFS up to 5k. The common mode rejection is bad in HF because of not perfect time alignment and a 6 dB/oct 13kHz analog LP filter after the SMSL DAC
2. is clean fundamental (SMSL DAC) time alligned down to closest sample.
3. is CS DAC inverted and clean fundamental (time alligned) into differential amps. So fundamental is down max about 40 dB compared to the two sources.
The fundamental is under -20dBFS up to 5k. The common mode rejection is bad in HF because of not perfect time alignment and a 6 dB/oct 13kHz analog LP filter after the SMSL DAC
This is 2.5 kHz stepped response with fundamental cancelled with SMSL dac as red trace above
No problem to recognice the distortion pattern when comparing to 300 Hz without cancelling, post #86
No problem to recognice the distortion pattern when comparing to 300 Hz without cancelling, post #86
Just to document a series of time consuming measurements at -1, -2,5 and -6 (5) dB, but I had the13k LP filter on the CS output. Then I had some problem with hf cancelling of 3rd harmonics.
At the 3rd harmonic dip at -6dB it is interesting to see how 3rd harmonic "disapear" down in the -130dB area and the CS codec measure the same as the reference DAC on the U7 ADC
At the 3rd harmonic dip at -6dB it is interesting to see how 3rd harmonic "disapear" down in the -130dB area and the CS codec measure the same as the reference DAC on the U7 ADC
And the fundamentals for the above. Have no answer why cansellation of fundamental is higher with lower output of signal genarator
Just a comment on the repeatability on the setup. This is using a ADAU DSP to make the delay for the perfect fundemental.
.It is SPDIF copied as input the SMSL DAC.
The codec 01 outputs on the CS board are delayed about 675 samples.
But since it is real DSP with TDM out the dalay is the same every time. No dalay variation ever. The prosessing are deterministic. Unlike some linux DSP setups
.It is SPDIF copied as input the SMSL DAC.
The codec 01 outputs on the CS board are delayed about 675 samples.
But since it is real DSP with TDM out the dalay is the same every time. No dalay variation ever. The prosessing are deterministic. Unlike some linux DSP setups