Not doubt subtractive tests can be made. Questions are about the limits of the test setup and equipment. Are subtractive tests highly accurate for -120dB linear distortions evaluations? Or what are the practical limits? How is the test equipment itself vetted, is it with PSS spectral measurements? With another subtractive test set? If it works well then why do we use AP machines? Just trying to get a feel for this type of methodology, its history, its limits, etc.
See
Andrew R. Collins, "Testing amplifiers with a bridge", Audio, March 1972, pages 28...32.
Peter J. Baxandall, "Audible amplifier distortion is not a mystery", Wireless World, November 1977, pages 63...66
What I wrote in post 1868 is not correct, as they did use active circuits other than the device under test, but not in the part of which the signal gets attenuated and the distortion does not.
Andrew R. Collins, "Testing amplifiers with a bridge", Audio, March 1972, pages 28...32.
Peter J. Baxandall, "Audible amplifier distortion is not a mystery", Wireless World, November 1977, pages 63...66
What I wrote in post 1868 is not correct, as they did use active circuits other than the device under test, but not in the part of which the signal gets attenuated and the distortion does not.
"The technique can obviously be implemented in various detailed ways, and Fig. 1 shows one arrangement which is suitable when the amplifier under test is of the phase -inverting type."
From Peter Baxandall's article in WW Nov 1977 "Audible amplifier distortion is not a mystery" (pages 63-66)
https://worldradiohistory.com/UK/Wireless-World/70s/Wireless-World-1977-11.pdf
Edit. A previous article from A. Collins (Quad) mentioned in the references is here (pages 30-34)
https://worldradiohistory.com/Archive-All-Audio/Archive-Audio/70s/Audio-1972-03.pdf
George
Attachments
Last edited:
If you would want to build a Nazar-style shunt regulator with my low-noise bandgap as a reference, you could try something like this. Mind you, it's just a sketch, it requires more work to ensure stability and reliable start up. Besides, the reference designators of the new parts are not consistent with the rest of the DAC schematic. The impedance of the RC filter R18-C8 has been reduced because of the large bias currents of the regulators.
Yes, please. I will cover return shipping costs. Just let me know and many thanks for your efforts☺️
Would just like to mention, looks like Marcel added a reference to one of my last summary posts in the first post of this thread. However, I was kind of hoping that anyone wanting to implement one of these dacs would start from the post at: https://www.diyaudio.com/community/threads/return-to-zero-shift-register-firdac.379406/post-7416497 which is where I described what seemed like about the worst case way of hooking it up to power and to an I2S source.
As the thread progressed from there a number of improvements to the sound were made in steps my making incremental changes to environment around dac. That includes clocking, power wiring, shielding, etc. All those sorts of things help to get the best SQ from the dac. A pic of the dac being fed I2S from and I2SoverUSB board using an adapter board to Amanero wiring format can be found at: https://www.diyaudio.com/community/threads/return-to-zero-shift-register-firdac.379406/post-7417213 The I2SoverUSB board was powered from two isolated 5v linear power supplies. That's probably about the best low cost way to drive the dac. It provides pretty good clocking and pretty good isolation from USB bus electrical noise.
Later on as things progressed, attention started to be focused on the output stage, then on the dac itself. Compared to other dacs, such as, for example, the Andrea Mori RTZ DSD dac here, Marcel's dac sounds nice, yet still a little bit 'raw.' To me the 'raw' sound is most likely some remaining distortion and or noise that sounds like distortion, the source of which was not fully resolved. Not unlikely it might have something to do with X5R bypass caps for the converter shift registers. Don't know for sure though, guess it may be up to somebody else to track it down, if anyone cares to that is.
A pic of how the dac looked with the filter board, reclocker board, power wiring, and shield case can be found at: https://www.diyaudio.com/community/...t-register-firdac.379406/page-70#post-7430757 Also, a pic of how the analog output was taken when the filter board was not in use: https://www.diyaudio.com/community/threads/return-to-zero-shift-register-firdac.379406/post-7427196
As the thread progressed from there a number of improvements to the sound were made in steps my making incremental changes to environment around dac. That includes clocking, power wiring, shielding, etc. All those sorts of things help to get the best SQ from the dac. A pic of the dac being fed I2S from and I2SoverUSB board using an adapter board to Amanero wiring format can be found at: https://www.diyaudio.com/community/threads/return-to-zero-shift-register-firdac.379406/post-7417213 The I2SoverUSB board was powered from two isolated 5v linear power supplies. That's probably about the best low cost way to drive the dac. It provides pretty good clocking and pretty good isolation from USB bus electrical noise.
Later on as things progressed, attention started to be focused on the output stage, then on the dac itself. Compared to other dacs, such as, for example, the Andrea Mori RTZ DSD dac here, Marcel's dac sounds nice, yet still a little bit 'raw.' To me the 'raw' sound is most likely some remaining distortion and or noise that sounds like distortion, the source of which was not fully resolved. Not unlikely it might have something to do with X5R bypass caps for the converter shift registers. Don't know for sure though, guess it may be up to somebody else to track it down, if anyone cares to that is.
A pic of how the dac looked with the filter board, reclocker board, power wiring, and shield case can be found at: https://www.diyaudio.com/community/...t-register-firdac.379406/page-70#post-7430757 Also, a pic of how the analog output was taken when the filter board was not in use: https://www.diyaudio.com/community/threads/return-to-zero-shift-register-firdac.379406/post-7427196
Last edited:
Mark,
Compliments for the extreme effort you have invested in testing and in trying to improve the sound to your taste, which if course could differ for others.
To resume the most eye catching points apart from the Fifo plus reclocker.
Galvanic isolation from computer seemed to be an important step in improving S/N as well as giving the I2S converter it’s own supplies instead of the 5 Volt USB supply.
And if possible feeding this converter with a SOA clock.
All these points can be perfectly realized with the JLSounds converter.
The thing that surprised me was the strong preference for SE instead of a balanced/differential output signal, altough a transformer was used to galvanicly isolating the Dac from the preamp, making the connection not a SE connection in the usual sense
It would be nice if the transformer details could be disclosed, because this could eventually replace the analog reconstruction filter.
Be aware that with the transformer solution very high frequencies are travelling through the interlink, so it must be very well shielded and terminated.
The only thing I regret is that 2x8 instead of 4x4 for the shiftregister couldn’t be tested.
But maybe in a redesing this could be made optionally switchable.
And in case of using SE, a 1x16 layout could also be considered.
Although Marcel had very good reasons for the 4x4 solution, the amount of generated HF will be substantially diminished with longer filters.
Hans
Compliments for the extreme effort you have invested in testing and in trying to improve the sound to your taste, which if course could differ for others.
To resume the most eye catching points apart from the Fifo plus reclocker.
Galvanic isolation from computer seemed to be an important step in improving S/N as well as giving the I2S converter it’s own supplies instead of the 5 Volt USB supply.
And if possible feeding this converter with a SOA clock.
All these points can be perfectly realized with the JLSounds converter.
The thing that surprised me was the strong preference for SE instead of a balanced/differential output signal, altough a transformer was used to galvanicly isolating the Dac from the preamp, making the connection not a SE connection in the usual sense
It would be nice if the transformer details could be disclosed, because this could eventually replace the analog reconstruction filter.
Be aware that with the transformer solution very high frequencies are travelling through the interlink, so it must be very well shielded and terminated.
The only thing I regret is that 2x8 instead of 4x4 for the shiftregister couldn’t be tested.
But maybe in a redesing this could be made optionally switchable.
And in case of using SE, a 1x16 layout could also be considered.
Although Marcel had very good reasons for the 4x4 solution, the amount of generated HF will be substantially diminished with longer filters.
Hans
Hans,
One point where some clarification should probably be made. I2SoverUSB uses NDK SDA clocks, buffered through a CPLD. Even under the best of circumstances they are probably not SOA. OTOH, the Acko Labs AKX-22 clocks are indeed SOA. Most of the time the Acko clocks were in use.
Regarding the SE preference, its hard to understand how it can sometimes sound better, but it can. Just means we have to figure out the physics of why. Just don't blame it on something wrong with the listener before bothering to seriously give it some listening time first. That said, not expecting everyone will agree.
One point where some clarification should probably be made. I2SoverUSB uses NDK SDA clocks, buffered through a CPLD. Even under the best of circumstances they are probably not SOA. OTOH, the Acko Labs AKX-22 clocks are indeed SOA. Most of the time the Acko clocks were in use.
Regarding the SE preference, its hard to understand how it can sometimes sound better, but it can. Just means we have to figure out the physics of why. Just don't blame it on something wrong with the listener before bothering to seriously give it some listening time first. That said, not expecting everyone will agree.
Last edited:
I have heard SE 300b tube mono block amplifiers that sounded extraordinary into classic Tannoy's. My suspicion is that signals need to be working in some window in relation to the magnitude of the SE non-linearity for the outcome to be preferred, a condition that may be difficult to realize.
I thought I'd give a quick update on my progress with the RTZ DAC. I've built two RTZ DACs so far, the 2nd one being the device I will be taking to the next "Owston" meet-up in the UK in a few weeks time. This 2nd DAC is using the Simple PCM2DSD for I2S to DSD conversion and an LVDS link of the I2S and MCLK signals from a Logitech Transporter. For the purposes of the Owston event the Transporter will be driven from my CDPro2 CD transport, since it is far easier than trying to set up a whole portable infrastructure to feed the Transporter from LMS, etc.
I recently switched back to the "simplified filter" using @nautibuoy's pcbs and am powering it now from its own diyAudio super-reg.
And the back:
And inside:
I have bodged in the +/- 15v wires to the filter board currently as I am using this as a test-bed to easily swap out the filter board for the passive one, the original filter, etc. (Although the space is not available in this box to fit the original filter, as can be seen, however I can do that on the other DAC since it's case is larger and was originally arranged for that pcb to fit.
I am very pleased with the outcome of both the DACs. I have more experimentation to do but for now I am just satisfied to listen to this day in day out.
I recently switched back to the "simplified filter" using @nautibuoy's pcbs and am powering it now from its own diyAudio super-reg.
And the back:
And inside:
I have bodged in the +/- 15v wires to the filter board currently as I am using this as a test-bed to easily swap out the filter board for the passive one, the original filter, etc. (Although the space is not available in this box to fit the original filter, as can be seen, however I can do that on the other DAC since it's case is larger and was originally arranged for that pcb to fit.
I am very pleased with the outcome of both the DACs. I have more experimentation to do but for now I am just satisfied to listen to this day in day out.
Looks very nice indeed. You aiming for high end where looks matter?
However, just in case there may happen to be any interest on the part of thread readers, it takes about 1/2" thick aluminum walls to equal 1/8" steel walls for shielding effectiveness. Either one is probably fine for hi end audio, but maybe keep especially sensitive circuitry >= 2" away from steel.
I like to get a nice result in looks. The case was from china and around 100 dollars including postage. It wasn't about screening as such, just what i could get that would fit the components. It's about compromise mostly. For me it's a voyage of discovery and a hobby. This DAC and PCM2DSD and diyAudio have been a great positive for me this year. I really love the discussions and investigations. I hope it carries on.