What does the input signal actually stand for? I may understand your procedure wrong but - a distorted sine (no matter what harmonics) has some waveform A. If this waveform is negated (waveform B), and added to the waveform A, nothing is left (considering ideal circuits). The two waveforms cancel out, both the fundamental and all the distortions.
Or two stereo DAC channels, one fed with inverted digital sine signal, then subtracting the two SE outputs from the reconstruction filters. Since each channel has a separate component path, their distortion profiles are different. IIUC the result will be even distortions added up and fundamentals and odd distortions subtracted, but it's a mix of distortions from the two analog channels, not revealing details about distortion profiles of each output individually.
Or two stereo DAC channels, one fed with inverted digital sine signal, then subtracting the two SE outputs from the reconstruction filters. Since each channel has a separate component path, their distortion profiles are different. IIUC the result will be even distortions added up and fundamentals and odd distortions subtracted, but it's a mix of distortions from the two analog channels, not revealing details about distortion profiles of each output individually.
Argh. Of cource the whole signal is common mode if the phase is coherent. So even the even harmonic will be cancelled.
But luckily distortion that is not phase coherent will not be cancelled
I totally agree that details are not revealed.
But consider the case of me buying a -110 dB spec DAC with questionable filter components or implementation.
I have a ADC with 95 dB THD at -1 dB FS
I want to check if the dacs analog components should be changed or they ar good enough
Then I have a differential receiver with THD at -110 dB if fed with the signal + the inverted fundamental.
I could then see if the dac and filter maybe is good down to say -110 dB FS if all distortion goes down to the noise. I would then doubt there is much to gain changing filter components to bigger size or different material.
If it don't then there may be some underspeced components that could be removed or replaced.
Small fault in one channel is also reveiled.
But luckily distortion that is not phase coherent will not be cancelled
I totally agree that details are not revealed.
But consider the case of me buying a -110 dB spec DAC with questionable filter components or implementation.
I have a ADC with 95 dB THD at -1 dB FS
I want to check if the dacs analog components should be changed or they ar good enough
Then I have a differential receiver with THD at -110 dB if fed with the signal + the inverted fundamental.
I could then see if the dac and filter maybe is good down to say -110 dB FS if all distortion goes down to the noise. I would then doubt there is much to gain changing filter components to bigger size or different material.
If it don't then there may be some underspeced components that could be removed or replaced.
Small fault in one channel is also reveiled.
Last edited:
But all considering. Maybe just a -20 or -30 dB pad before the ADC solves it all.
See REW can measure THD pretty low by expanding the FFT size or maesurement time and averaging the noise away.
See REW can measure THD pretty low by expanding the FFT size or maesurement time and averaging the noise away.
For the record. the above graph is a 13 euro cirrus logic card fed with USB power. Passive output filter on codec card is removed. But 4 channels used. First two and two are balanced. Then those are again balanced to single ended. -6dB gain in both stages. Fed from spdif in in ADAU1452 card and TDM 96kHz out,
Last edited:
Wrote somewhere i post #20 that it was sallen key filter, but that was another active card, sorry. Only 4 tiny passive components.
Harmonic distortion arising from the fundamental is always coherent with the fundamental, that's why it's called harmonic.
Please what is source of the "signal" and the "inverted fundamental", where do you get them from? Analog inverted fundamental will always be distorted by the generator, unless some sharp HP filter extracts just the fundamental freq and supresses its multiples/harmonics (or the digital path adds precisely determined "antidistortions" = distortions at opposite phase - that's what my linked project does).
That is what I try to dig into.
If I generate a harmonic in the ADAU I see it has stable phase in REW. But other harmonics created by ‘overloaded’ passive components I see in REW that the phase of the harmonics changes a few degrees by the second. And the change is not predictable at least to me
I get the fundamental from REW
SPDIF into a ADAU1452
TDM into a cirrus logic 8 ch out codec
So all signal prosessing and routing/inverting is done in the ADAU
For ordinary DAC case, the inverting/non inverting can be done in configuring the outputs in REW. So left channel canselling right channel. Exept where they dont match up in amplitude or phase. So if a freq sweep is done at high volume it is a quick way to see if there are areas to investigate further
Coherent means the phase shift against the fundamental is stable, not random. That's why coherent averaging in REW (called vector averaging in other analyzers) can work and give you reasonable values of amplitude and phase shift of the fundamental harmonics.
Every non-linear phase characteristics in the signal path causes phase shifting. Look at phase characteristics of a simple RC LP filter - the phase shift is frequency-dependent. That does not make the harmonic incoherent with the fundamental, it's just phase shifted, but the phase shift does not change in time. Of course in sensitive real-world measurements in REW the measured phases fluctuate a bit due to noise.
Still I do not understand what are the two analog signals that you feed into the differential receiver : "signal" and "inverted fundamental".
I will see if i can recreate my observations. It may have been measurement faults and my observations are false.
But I was not able to propperly cancell second harmonics caused by passive filters on dac output. Maybe because the distortion was different in amplitude on the two channels https://www.diyaudio.com/community/attachments/1726083576827-png.1355200/
But I was not able to propperly cancell second harmonics caused by passive filters on dac output. Maybe because the distortion was different in amplitude on the two channels https://www.diyaudio.com/community/attachments/1726083576827-png.1355200/
Yes you were right. REW is within a degree fault on phase when amount of distortion is higher.
The whole reason for little cancelling was different amplitude (2,5 dB) on second harmonics on the two channels. (The fundamental was spot on amplitude)
The whole reason for little cancelling was different amplitude (2,5 dB) on second harmonics on the two channels. (The fundamental was spot on amplitude)
Lucky I got a board in the post today, so I could control measure with output filter.
Without filter the codec card distorts much less and the second harmonic is more or less the same on all channels. So then balancing functions for removing 2. harmonic
Without filter the codec card distorts much less and the second harmonic is more or less the same on all channels. So then balancing functions for removing 2. harmonic
You have a very good DAC, a mediocre ADC and want to measure a mediocre DAC that maybe has better THD than ADC.
In my example pure is -105 dB, mediocre is around -80-90dB
Now we put one of the good DAC channels on one input of receiver and the mediocre dac at the other. We adjust volume to be the same for both.
I want to test at 100Hz so but 100 Hz into mediocre DAC. Then i put close to 100 Hz into good DAC. Maybe 99.99Hz
Then i get a beat and cycling betveen cancellation and amplification. That is sum and difference.
These two pictures show some measurements. Cursor at 100 and 200 Hz.
We se the harmonics are fairly stable so it must all come from the DAC and not the ADC. The ADC would else add to the distortion at when difference is larger than signal and distortion should go down if it is smaller. Actually the DAC is very little loaded at max cansellation.
Cursor 100 Hz
Cursor 200 Hz
If mediocre DAC was better than ADC we should see the distortion level at max cansellation as the DAC then distorts minimal. About -30dB less fundemental in this example. But we could never measure better than good DAC
Of course for a sweep it is more difficult to get the two dacs in sync. It is easier just to wait for the min as the two freq beat as described above
Well, unless the two DACs are fed from the same master-clocked signal (e.g. SPDIF), they will each generate the fundamental sine of slightly different frequency because of their independent internal clock. May be OK for a short measurement, but in slightly longer measurements the sines will run out of phase.
Let's consider clocks precision 50ppm for each device. 50ppm means imprecision 50us per every second. A period of 1kHz signal takes 1ms. E.g. within 10 seconds (time shift 0.5ms) the second DAC's sine will be 180° shifted compared to the first DAC.
Even with the same master-clocked signal, if the signal receivers (e.g. SPDIF receivers) and DAC chips are different, they introduce a different delay and the output signals will be out of phase and likely impossible to subtract properly - see https://www.wolframalpha.com/input?i=a+sin(x)+-+a+sin(x+++c) which is not zero.