Quick update on my EMI fried dam1021s - they now no longer lock on to I2S at all, whereas before they sometimes acquire an I2S signal lock upon boot even when there is no signal input and churn out awful noises. SPDIF signal lock is fine, but the dam1021s refuse to make any sound in these modes.
Are 1.21 and 1.23 using the same factory filters? I wasn't quite sure the 1.23 was using the same 4k filters as 1.21, so i used my old favourite filters in both 1.19 and 1.23. For completeness only listened to 1.19, 1.21, 1.23 using the default filters and agree fully with your findings.
No idea, sorry. I only use a certain custom filter and I make sure to load it into the DAM1021 after any OS update.
Did you hear a difference between 1.19 and 1.21, too?
The audible flaws with the reclocking kind of defeat the purpose of the entire FIFO buffer... if I were the developer getting this right would be my top priority...
To be fair, this is a massive problem even with commercial pro audio converters. Even those with a complex 2-stage PLL/buffer combination and arrangements to "listen" to the incoming clock over longer time frames to sync up precisely sound worse clocked externally than internally.
But the DAM1021, as an R2R DAC, by itself has much lower latency than a DS DAC, so there is room for this very comfortable 1ms buffer. In my latest test with the re-clocking stopped I found less than 3 samples drift over a good 10 minutes at 44.1 khz. So if at this time the DAC were to make one gentle course correction, we would never have been in danger of an over/under situation - 22 (= half of 44) -3 - there were still 19 samples left.
My guess is that the DAM1021 lacks functionality to monitor and compare the clocks over longer time frames, making a cruder faster/more frequent adjustment neccessary.
So again, I'd like to ask Soeren to elaborate what is going on. Because now I am actually thinking about building a multiplier for the serial connector to allow all my DAM1021 to enter Tera Term at the same time in order to stop the re-clocking from ruining the glorious sound the DAC is capable of reproducing...
But the DAM1021, as an R2R DAC, by itself has much lower latency than a DS DAC, so there is room for this very comfortable 1ms buffer. In my latest test with the re-clocking stopped I found less than 3 samples drift over a good 10 minutes at 44.1 khz. So if at this time the DAC were to make one gentle course correction, we would never have been in danger of an over/under situation - 22 (= half of 44) -3 - there were still 19 samples left.
My guess is that the DAM1021 lacks functionality to monitor and compare the clocks over longer time frames, making a cruder faster/more frequent adjustment neccessary.
So again, I'd like to ask Soeren to elaborate what is going on. Because now I am actually thinking about building a multiplier for the serial connector to allow all my DAM1021 to enter Tera Term at the same time in order to stop the re-clocking from ruining the glorious sound the DAC is capable of reproducing...
I have no clue, shouldn't be any difference to hear, there is nothing to fix.... It's probably a case of the usual wanting to hear something.... Come back when you have done a double blind test.
Sometimes, dedicated engineers just want to do the best and most elegant solution even if there are only a few who could appreciate them. In this case we have a special application where it would be favourable with a low short term clock drift where the owner claim that he can also hear a difference. Maybe its an ilusion or the listner running hes own studio as a profession indeed is more sensitive than the engineer himself and his reference listening group. Perhaps we can allow us to nurture a small hope that in this case the dedicated engineer will, after all, find inspiration and a meaning to walk that extra mile to make the product fulfil that SOTA promise on page 1? - in all aspects...
Heja Danmark!!! 🙂
//
Heja Danmark!!! 🙂
//
The whole premise of the DAM1021 goes against current technological orthodoxy - that a resistor ladder with inherently more THD than a modern DS converter actually sounds better to the human ear. So I find it suprising that now the argument turns to there "shouldn't be any difference to hear".
I don't think that it is possible to really put into words the amount of listening experience gained in a studio setting. Regardless, I can compare the DAC to the direct analog signal and I can process the converted signal at will. This makes differences very obvious. But it is audible all by itself, for instance, last week I had a friend who is an acoustics engineer listen to the DAC at 1.24, and after two minutes he told me there was something wrong with the high end.
I would like nothing better to have the DAC behave perfectly and be done with it. If I were the kind of person that takes things personal I would be offended by the suggestion of "wanting to hear something"...
I hope you can find it in your heart to look into the issue without prejudice and without feeling personally offended just because somebody finds room for improvement in your designs. It is not personal, we are just trying to help improve the product. And if I may say so, in the future people very likely will ask how mixes and productions got to sound so good, and your DAC will get mentioned.
I don't think that it is possible to really put into words the amount of listening experience gained in a studio setting. Regardless, I can compare the DAC to the direct analog signal and I can process the converted signal at will. This makes differences very obvious. But it is audible all by itself, for instance, last week I had a friend who is an acoustics engineer listen to the DAC at 1.24, and after two minutes he told me there was something wrong with the high end.
I would like nothing better to have the DAC behave perfectly and be done with it. If I were the kind of person that takes things personal I would be offended by the suggestion of "wanting to hear something"...
I hope you can find it in your heart to look into the issue without prejudice and without feeling personally offended just because somebody finds room for improvement in your designs. It is not personal, we are just trying to help improve the product. And if I may say so, in the future people very likely will ask how mixes and productions got to sound so good, and your DAC will get mentioned.
Sorry, just don't have time currently.... Spending way too much time just trying to find parts for current productions, having to redesign to use parts I can get.... "shouldn't be any difference to hear", I was talking about the clocking.
Also, I do have the dam1121, where you can have them clocked syncronous, and with the better oscillator....
Also, I do have the dam1121, where you can have them clocked syncronous, and with the better oscillator....
There you go! Stop wasting time with the "reference dac" dam1021 and just buy 18 new dam1121's. (nonexistent) Problem solved!
Soeren, I paid over 4000 EUR for your products and waited years for you to finally come up with a new OS version for this "reference" design. Will you take the DAM1021 back and exchange them for DAM1121s?
If you were a customer, would you want to be treated this way?
If you were a customer, would you want to be treated this way?
I asked this very same thing 6-7 years ago after buying 4 of the "reference design" dam1021 to use in an 8 channel dac, only to find out about a whole slew of issues, and mine had yet to be removed from their original packaging. No no no no NO! "There is nothing wrong with the dam1021"...Been there. Customer service is clearly not the strong point of this company. Doesn't feel good to be an unannounced beta tester for a so-advertised "reference quality" product.
It's a small thing we are asking: Allow us to set the parameters of the re-clocking ourselves. At this point, I would even pay someone to write the code if it were possible.
To be fair, I don't think that what you guys want was a specified property for the product at hand - or?
//
//
For the re-clocking to be optimized to minimally impact the sound quality (which is by far the most important thing in a reference DAC) of the converter?
I am not talking about syncronous clocking (which from my testing does seem to actually work, not by design, but by all paralleld converters behaving closely enough). The original claim was that adjustments of the clock would be transparent. That is clearly not the case. So the goal would be to make re-adjustments as infrequent as possible.
I also would like some more information what is actually going on with the different versions, since they seem to result in comparable drift behaviour.
The boards works as designed and advertised, I can't do special firmware for one customer.... And I can't take perfectly fine boards back, especially old version after long time.... If I started doing that I would soon go bankrupt....
My products are not priced to have that level of service.
I have no desire to return them. I have myself invested countless hours into it. Among other things I paid someone to build PCBs to get proper AES/EBU inputs and soldered them myself... What remains is only the comparably small thing with the clocking.
I am talking about sound quality, which is relevant for every single user of the DAM1021 (and probably other soekris DACs, too).
You can try this yourself by entering and exiting Tera Term with either version. Auditioning with quality monitors (ideally something with good impulse response) and proper amplifiers (not Class D) it is very easibly noticeable. I wouldn't persist like this if it were otherwise.
At least one person here has made the same test and agrees with me:
https://www.diyaudio.com/community/...-magnitude-24-bit-384-khz.259488/post-6926044
I am talking about sound quality, which is relevant for every single user of the DAM1021 (and probably other soekris DACs, too).
You can try this yourself by entering and exiting Tera Term with either version. Auditioning with quality monitors (ideally something with good impulse response) and proper amplifiers (not Class D) it is very easibly noticeable. I wouldn't persist like this if it were otherwise.
At least one person here has made the same test and agrees with me:
https://www.diyaudio.com/community/...-magnitude-24-bit-384-khz.259488/post-6926044
Lol. I'm still stuck with the EMU0404 and I have to say the f-ing thing sucks compared to the dam1021. Regardless, I think it's not unreasonable for Soren to ask for double-blind test results since you already claim to hear the difference from the clock drift (or other firmware issues) with ease. You might actually be right, but why not prove it if it's so easy?... Personally, I have absolute pitch but I seriously, seriously doubt that I can hear anything close to the long-term frequency shift in a dam1021 designed to compensate clock drift. I completely understand why Soren didn't go for a larger buffer size; otherwise, people like me wouldn't be able to use it as a daily driver (to watch stupid youtube videos, among other things).
I was able to easily produce a near perfect double-blind test result for a 5ms inter-channel delay. Surely, you could at least try to perform some kind of double-blind test? Remember that an ABX test will never disprove the possibility that you did actually hear a difference, and since you can probably only do double-blind tests on recorded outputs from the dam1021, I'd say it'd be surprising if any subtle, yet audible, difference could be detected in the recorded material. You might also be able to do real double-blind tests since you have 18 boards... Either way, I think it makes sense for Soren to ask for scientific proof before he invests significant amount of time developing a new firmware version, when he has strong reason to believe that it should be otherwise.
I should also add that I'd love to see a negative double-blind test result, from you, between the dam1021 and some random studio gear. You say you hear a big difference, and I have been hearing a big difference too. But I can't do a proper ABX test because I don't have the equipment... even if my dam1021s weren't fried already...
Last edited: