I have been designing with negative feedback for more than 50 years. I used to wholly believe in it as a fix for just about everything. It took me years of experience, (and rejected designs by actual listeners) before I found that it was a difficult compromise. That is when I reduced my dependence on negative feedback and became extremely successful in my field of audio design, about 46 years ago. Before that I could design for good measurements, but sometimes fail to make something 'listenable' to people who could hear the difference. Since then, I have had very few failures, and many successes. There is more to it than just how much feedback you use, but feedback is what we are talking about here.
No one said feedback is a ‘fix all’.
But we also know, emphatically, that it doesn’t go ‘round and round’ as Collums asserted in his notorious 1998 Stereophile piece.
Really John, I’d love to think more of you. Honestly. And then you come up with stuff like this.
Feedback: A Short History
There’s a link to the Collums piece in the article.
But we also know, emphatically, that it doesn’t go ‘round and round’ as Collums asserted in his notorious 1998 Stereophile piece.
Really John, I’d love to think more of you. Honestly. And then you come up with stuff like this.
Feedback: A Short History
There’s a link to the Collums piece in the article.
Last edited:
I suspect many "audiophiles" and those who design products for them are frustrated musicians/sound engineers who like to think they're in the entertainment industry 😎
I found that my new approach to audio design (Otala inspired) worked not only for the GD, but also for serious audio reproduction in the home, that was mostly classical music. It made me famous, in fact. The GD just appreciated me for my efforts, they didn't inform the World about it, that took my success with Mark Levinson, etc.
To be honest about it, I am most probably a frustrated musician, as I was in a Rock and Roll band 60 years ago, (1958, how time flies!) and I purchased a Fender Stratocaster to play in the band. Just this week, I had my later Stratocaster restrung by a pro, and ready to play, but it has been awhile, and while my head knows what to do, my fingers mostly forgot. '-)
I gave it up in about 1980.
No coordination and no rhythm. But I enjoy listening to music and funny thing, both my sons ended up being very capable musicians.
No coordination and no rhythm. But I enjoy listening to music and funny thing, both my sons ended up being very capable musicians.
John, as you know, devices, today are faster than in those times and it is quite easy to design a power amp that has a flat open loop gain up to 10kHz and a slew rate at least 10 time faster than what is required for the fastest audio signal we have to deal with.
So what ?
May-be that you don't like is amplifiers that have a very high open loop gain, that begin to decrease at low frequencies and a high feedback ratio ? May-be the ideal should be an amp with constant feedback all over the audible frequency range ?
Sometimes the ability to create is enough, I would call Eno a good musician but I believe he's said himself he can't play an instrument well
T, it isn't THAT easy to have everything in an amp. That is what I do today to get the best combination of open loop and global feedback without exotic lead-lag networks, output coils, etc. It is ALL important.
Audio certainly is easier than sending people to the moon, imaging our insides or detecting gravity waves. I can’t complain though it has been good to me.
If the issue is close in phase noise at low frequencies (both essentially say the same thing) A VCOXCO would be a good reference point. (pic below) If you are serious I can lend one to you. The pull range is small but the phase noise is quite low. It probably uses as much power a a full modern DAC but. . .
Adding an ASRC really seems unnecessary for a reasonably decent source. Streaming audio that has substantial jitter probably has dropped or resent frames from the network and is pretty compromised to start with.
The current generation of SPDIF receivers can get a clock with less that 20 pS jitter (pretty much all types, cycle to cycle, period and long term) so its not a big issue. The classic tests showed a very high sonic tolerance (nS) compared to current technology (pS) and at higher frequencies only.
Low frequency jitter/close in phase noise on any digital system is orders of magnitude less than any analog system making it something that would need really clear testing to show as important. The accepted standards for tape and turntables are .1% and sometimes .05% W&F. Minimum wow and flutter percentage that is audible? - Tapeheads Tape, Audio and Music Forums of course some are more sensitive and some music is more sensitive.
Also random jitter translates into noise in the output and deterministic jitter becomes distinct sidebands. Both can be measured pretty easily today.
I had to jump through hoops to get a decent J-Test SPDIF measurement on an AK4490 demo board. It was extremely low requiring lots of cross checking to be sure what i was seeing came from the DAC and not the ADC. Everything was in the -150 dB noise floor. This was several years and computers ago and i can't find it anymore (its probably still on DIYaudio somewhere). In any case I'm pretty convinced that jitter is a non-issue today. however I may try looking for jitter in a spectral contamination plot.
If you are using an ESS based DAC you may be encountering the "ESS HUMP" (image below) Lots more here: ESS THD ‘Hump’ Investigation | Audio Science Review (ASR) Forum Its a non-monotonic rise in distortion at low levels. The "fixes" mentioned mostly seem odd (common mode distortion, register tuning etc.) but are fertile ground for audiophile tweaking.
Here is something really distressing Tapeheads Tape, Audio and Music Forums - View Single Post - Minimum wow and flutter percentage that is audible? Universal adding flutter to "watermark" audio. Maybe this is the problem you are encountering on streaming audio (except that the flutter is encoded into the music and not affecting the audio stream).
Attachments
Not even close Chris.
The ASRC filters are half band.
The jitter rejection on AD1896 starts at about 5Hz, converseley the Mola Mola DAC has around 50 x lower corner freq for jitter rejection.
Absolutely couldn't agree more.
I've been trying to convince Mark for quite some time to dump the whole cascaded ASRC arrangement and use a simple JLsounds (isolated / reclocked) USB bridge. I2S straight into Sabre DAC running in symchronous mode (no DPLL).
This will sort out *all of the jitter ghosts and give an extremely low jitter baseline to work off.
The problem I see here is there is far too much confusion and a very unscientific methodology.
The solution is equivalent to an elimination diet. You first eliminate, as fa as possible, every thing that is a potential problem to establish a baseline, then start to introduce items back in the chain to establish which and how much they are affecting the outcome.
T
Again - this is all just going around in circles.
Setup the isolated / reclocked USB bridge and evaluate clocks from there.
You are looking for comparison of low freq jitter which will be revealed by a/ phase noise plot down to 1Hz and b/ Allan Variance plot.
The two are intrinsically linked but, IME of having custom SC-cut OCXO's made, the phase noise does not tell everything.
Mark if you wish to go down this route, PM me and I can provide details of someone who can measure your clocks for you.
This is all a pretty simple methodology to first a/ establish a very direct, low jitter topology b/ try some different clocks of *known jitter characteristics.
I'm sure this will answer a LOT of your questions WRT clocks, jitter, phase noise distribution and how it affects what you are hearing.
Further to this, WRT cooking up a custom digital filter, what characteristics are you after? The possibilities are endless and there is a lot of time to be consumed in this.
By running a software platform such as HQplayer (or equiv) you can play with endless filter types, modulators etc etc and work out a good starting point for your hardware based DF.
It's a massive time saver.
T
Hi Demian,
Thank you for the offer, but for the moment I am already overloaded with things to do. Perhaps a better time will come up at some point.
Regarding the 'hump,' John Siau said it is from insufficient common mode distortion removal in the output stage. In my first modded dac I used .01% tolerance 10k resistors for the differential summing stage. And, I don't recall exactly, but the best resistors I could find for the IV stages. Problem is that tight tolerance resistors don't come in a lot of different values, or at least they don't seem to be stocked in many values at places like Mouser and Digikey. There is that, and the fact that the output stage resistors also have to be selected for dac chip output current in the IV stages, and for filtering duty in the differential stage. Maybe easier to get in large quantities, or maybe changing the architecture a bit could work around the issues.
The reason I am using ASRC is because I would like the dac to be SPDIF compatible, and I would like to do upsampling and conversion to DSD in hardware. After I have an external interpolation filter, maybe I can skip the DSD. However, upsampling a bit helps with external interpolation filter performance since the upsampling can be used to create some frequency space for more relaxed filter transition band. Both Benchmark DAC-3 and Crane Song Solaris use that technique. I presume the new Gustard dac does too and it has beaten all the measurement records at ASR for now.
Distortion in the best ASRCs is down at -140dB or lower, better than the dac chips can do at best by around 20dB.
Regarding jitter, Crane Song has been using a clock with 45fs jitter, although it looks like they have been having some production problems wherever they source the clocks from. If everything is done right with ASRC and clocking, jitter should not be a problem, and even SPDIF jitter can be very low since its jitter is largely replaced by the ASRC VCO jitter. Already I can get jitter as it affects subjective sound quality with my second moded dac down where Allo Katana is, and reproduce low level reverb tails just as well as it can. You may recall that Katana once held the low jitter record at ASR, beating out DAC-3. The reason I still want to work on jitter is because I think I may be able to get it even a little lower than I have now, and benefit from a little more sound quality improvement at the same time. In other words, I think I might be able to best Katana in that regard and get closer to Solaris.
Regarding DSD conversion, although there may still be some issues to work out with distortion, I can beat Katana sound quality and distortion, just not DAC-3, and upsampled DSD helps with that. I do think I will probably need to go back in and do some more work with the output stage at some point. I want to experiment with a differential mode cap across the dac outputs (maybe ~ 2000pf as a starting value), and rearrange the output stage to use better matched resistors for better common mode distortion removal.
It turns out my second modded dac is now jam packed with mods that were not all planned out from the start, that there isn't much space to try more things. I will probably put together an ES9028PRO dac to experiment with next. I have a number of power supplies I want to test with it, and two or three output stages. Also, will try some different clocking schemes.
The most important thing at this point though is to get an external interpolation filter prototype going. Everybody that has a diy Sabre dac of any type could benefit from that, nobody has done one that isn't proprietary and only used as part of a commercial dac product. It will always be easy to go back and fine tune or change some of the 2nd modded dac hardware design to catch up with what I have learned so far, and as the dac keeps getting better, those little things that may have been good enough at one time will have become the bottlenecks that need attention.
Last edited:
I know he went nuts with the Mola Mola stuff. He wanted to make a statement product and certainly appears to have succeeded.
However, unless his opinion changed, he posted here once upon a time that he was pretty happy with SRC4192 and discussed it along with the AD1986 and Cirrus part. I don’t want to put words in his mouth, but he did say that he was using SRC4192 in a project. It certainly wasn’t a post about the horrors of the 1896.
I’d also point out that Benchmark believes the AD1896 is audibly transparent. Since their name is getting invoked often here. It isn’t a perfect filter, but I’m not sure the 5hz corner frequency is really an issue at all...
DAC-3 uses SRC4392, I know becuase I opened the case and read the part number. IIRC, DAC-1 may have used AD1896. Unfortunately, I sold that one, so can't check now.
- Status
- Not open for further replies.