The guy who helps me with 9039S beta testing, reported that his result was even better than mine from the video I posted yesterday 😉
THD@1kHz is close to -148db. I don't know how to make Cosmos DAC better..
THD@1kHz is close to -148db. I don't know how to make Cosmos DAC better..
That is outstanding performance. Doing better isn't a requirement for a piece of test gear.
I think it's really a couple details for why a Cosmos DAC would be desirable.
One is the packaging. The dongle packages are optimized - very well, in fact - for use with a tablet or phone. The connectors are set for use with headphones. For bench testing, what I think most people would really like is separate output jacks for each channel, with some way to have both unbalanced and balanced connections. They don't need both at the same time - just accommodation for both types of connection.
In addition, how temperature sensitive is this distortion level? Would a temperature control system like the option for the ADC be useful? If so, I'd think that would be another reason to allow for a somewhat larger package, I'd assume matching the existing ones used for the ADC, Scaler, and APU.
I understand that the market for these three products hasn't really been what you'd hoped for. Like a lot of hobbies, DIY audio enthusiasts are often really cheap. They'll work to save seven cents on capacitors, as if they're going to build these devices by the millions like cellphones. So, I can understand why you'd be reluctant to devote the time and energy for a product that might not sell very well. I can only ask...
I think it's really a couple details for why a Cosmos DAC would be desirable.
One is the packaging. The dongle packages are optimized - very well, in fact - for use with a tablet or phone. The connectors are set for use with headphones. For bench testing, what I think most people would really like is separate output jacks for each channel, with some way to have both unbalanced and balanced connections. They don't need both at the same time - just accommodation for both types of connection.
In addition, how temperature sensitive is this distortion level? Would a temperature control system like the option for the ADC be useful? If so, I'd think that would be another reason to allow for a somewhat larger package, I'd assume matching the existing ones used for the ADC, Scaler, and APU.
I understand that the market for these three products hasn't really been what you'd hoped for. Like a lot of hobbies, DIY audio enthusiasts are often really cheap. They'll work to save seven cents on capacitors, as if they're going to build these devices by the millions like cellphones. So, I can understand why you'd be reluctant to devote the time and energy for a product that might not sell very well. I can only ask...
Usually, I use another sine gen but today I run REw's one and got THD -146db on the same sample! Completely vanished H7 for instance.
That suggests there may be some cancellation of harmonics. Its an issue when the harmonics are that low. One trick- add a low pass filter to the source, It will shift the phase of the harmonics altering the cancellations. If the harmonic signature does not just follow the LP rolloff then you have cancellations to sort out.
there is no LPF just bare 9039Q2M and I/U(OPA1612+AD8397 composite), and no any anti-harmonics are used. What I see is the two digital gens difference, and REW is better. One gen THD -143db, REW's gen -146db with the same sample.
I was not clear- the LP filter is a diagnostic tool to test for harmonic cancellations not necessarly part of the system.
A difference in distortion from digital sources suggests a math problem, truncation or a difference in dither level and implementation. Virtins give some dither options. A digital loopback may show some differences relating to the digital signals.
When you are ready to sell the Cosmos DAC I'm interested.
A difference in distortion from digital sources suggests a math problem, truncation or a difference in dither level and implementation. Virtins give some dither options. A digital loopback may show some differences relating to the digital signals.
When you are ready to sell the Cosmos DAC I'm interested.
I have noticed a difference between the macOS version of REW and the Windows version, too. There's often even a difference between the generator in REW and a saved .wav file of a similar signal played back through REW into a DAC. Not a lot, but along the numbers you suggest.
I never thought much of this, because the distortion numbers are so low. But, with a DAC like this it could be significant. Or, a trick. Just as with circuit simulation, it's possible to fool yourself by getting distortion cancellation that you can't realize in the real world. An example of that is with bridged amplifier circuits. In a simulation, you can get perfect even order cancellation when you have identical distortion performance for both polarities. How do you do that in reality?
I'd still be interested in a Cosmos DAC, though. 🙂
Last time I checked, REW's generator was flawless in a digital loopback test, both on Windows with ASIO drivers and on Linux. Same results than when doing an offline FFT from file created by the generator, saved as 24bit PCM *.wav.
Dithering has to be set properly of course, that is, to 24bits or less, otherwise harmonics at ~-150dBFS may appear depending on the ratio of the test frequency vs sample rate.
Dithering has to be set properly of course, that is, to 24bits or less, otherwise harmonics at ~-150dBFS may appear depending on the ratio of the test frequency vs sample rate.
@IVX, do you have any plans to make an integrated ADC+DAC and/or provide syncing means for separated units or for other sources and sinks, so that sample-synced record while playback is possible?
Basically the functionality we have in an RME ADI-2 Pro, which I can use to have ADC+DAC locally in full sync, but also I can feed or read a DAI signal (TOSLINK, SPDIF, AES3). This allows me to do sample-synced measurement with any DAC or ADC that has an (additional) DAI input or output. And it allows to apply a scale factor of 4 on the DAI, that is, I can capture a DAC's output at 4x the rate but still in full clock sync which is a really killer feature.
I prefer to use sample-synced recording while playback whenever possible because it allows direct time-domain averaging, and rectangular windows for FFTs.
Basically the functionality we have in an RME ADI-2 Pro, which I can use to have ADC+DAC locally in full sync, but also I can feed or read a DAI signal (TOSLINK, SPDIF, AES3). This allows me to do sample-synced measurement with any DAC or ADC that has an (additional) DAI input or output. And it allows to apply a scale factor of 4 on the DAI, that is, I can capture a DAC's output at 4x the rate but still in full clock sync which is a really killer feature.
I prefer to use sample-synced recording while playback whenever possible because it allows direct time-domain averaging, and rectangular windows for FFTs.
KSTR, I do all work by myself and things go slower than I wish. For example, 9038S I tried >10 PCB versions before I got a decent THD+N performance under 16ohm load. The first version I had in Jul 2023, and only in Apr 2024 I'll be ready for production. The next model is Cosmos DAC which has to be even better, and I have no idea how. My design reminds a slow fine-polishing process, I can't say if I have plans for something else until I finish the current model.
Yes. For example, 1khz (exact, not nearest bin center) with 96kHz SR looks like that in a 4M FFT, without dithering. The sample train consists of only 96 different values and those are truncated/rounded to 24 bit, hence the accumulated error.
One small caution- using an exact submultiple of the sample rate won't "exercize" all the bit levels. 997 Hz would be enough off to run through all the levels.
Much like KSTR, I think it would be a great idea to see an integrated ADC+DAC to ensure the performance is consistent.