Reference DAC Module - Discrete R-2R Sign Magnitude 24 bit 384 KHz

[ynmichael]

Okay so here it is, my humble conclusion, or should I say verdict, of the sound quality of the dam1021-01 after it has been running for more than a week straight.
I have only been using the isolated I2S in via the Amanero board and I have tested both the balanced XLR outputs and the raw SE rca outputs at J7. Haven’t had the chance to use the other inputs yet. Also I’m still on the original firmware (not that it matters in terms of SQ).

Various sorts of music have been played, acoustic jazz and classical, rock, pop and some electronica. Some of these are regarded as ‘audiophile’ material and some not. Some material are genuine stereo miked recordings, others are typical modern studio layered recordings. A lot of well known music that I use when I test the SQ of different equipment. Uncompressed 44.1, 88, 96, 192 and even some higher res. as well as mp3 320Khz.

In his first post Søren aimed high, hinting at R2R-ladder dacs as totalDac and MSB. So naturally my hopes were high having owned a non-oversampling Discreet Monica R2R-ladder dac myself, as well as having heard the absolutely outstanding totalDac.

Now, the first thing that struck me with this dac was the harsh digital sound it produced. Okay, it had only been running for a few minutes so that was to be expected. Unfortunately this still hasn’t changed to a degree I had hoped for.

Most music sounds quite flat, the image is 2D, lacking depth and calmness. Not at all the 3D sensation or holographic image which are the trademarks of an R2R dac. There are however, a lot of details. But it doesn’t transcend into the sense of the acoustics of the recording room or the texture of the instruments or the sensation of space between the instruments.

Lots of details yes, but when presented with complex multilayered music such as Carl Nielsens' 5th it can't keep the grip and the composition sort of collapses. It's hard to really decipher the instrument groups and the notes each group is playing. At times the 5th sounds somewhat chaotic. The dac fares better with smaller ensembles, though the placement of each instrument is not always crystal clear. The soundstage is simply too wide IMO - very left-rightish.
With both my old Discreet Monica and the totalDac you instantaneously go ‘Yes, there it is’! Your heart is at rest. There is really no need to describe the experience of those dacs much further, as music is being played as it should be, effortlessly and naturally. Dacs like these make all other dacs, including dam-1021, sound like they are struggling. You’re listening to music while with the dam-1021 I still find myself listening to a piece of equipment.

To my ears the frequency spectrum is not flat. I hear an emphasis on the upper mid and treble region that makes, for instance brass and horn sections, and even sometimes solo saxophones, sound metallic. Very unpleasant to my ears.

Another thing that really bugs me with this dac is the aggressive in-your-face signature. I suspect there is a connection between its forwardness and the biased frequency spactrum? Maybe exchanging the caps would help in this respect?

It does however, seem to benefit from a tube amplication. I would like to hear how it performs swapping the ceramic(?) caps and maybe another clock than the Si514 if it’s applicable. Also different output stages, a Lundahl trafo or a tubed one.

It’s still not really clear to me whether there is a default FIR filter installed or if it’s bypassed? And will it be possible to deactivate oversampling altogether although Søren advises against it?

Is there anyone here who’ve tried to mess around with different FIR filters with this dac?

Anybody in this forum who have heard or own a totalDac (or any other R2R-ladder dac) are welcome to chime in after hearing the soekris dac.

I’m sorry to drop this bomb but my hopes are that we’ll all end up with a product that can really compete with the totalDac & MSB, for the benefit of all diy’ers who cannot afford a totalDac or the like - and enjoy tailormade electrical toys. Maybe there is a basis for doing some experimental modding on the existing dac or for Søren to work on a vers. 2?

Is it a bad dac then? No far from it. It just doesn't meet my expectations of an R2R-ladder dac. It’s at least on par with a lot of mid-priced commercial delta-sigma dacs.

Finally let me stress that I still highly value all the effort and hard work Søren has put into this project.

I power the dac with AC. It's a Gerth Trafo 4815-2 which was suggested by someone earlier in this thread. One of the cheapest ones I could find, just to sort of get the project started. I too am keen on a battery solution.

Just reposting some very ancient feedbacks. His comments seem to agree with my previous experiences. Of course I haven't heard a totalDac so consider my comparison relative to the two invariably different semantic systems that he and I developed over each of our audio journeys. Is it a coincidence that AC to DC inputs reduces harshness and perhaps restores the naturalness that belongs to R2R dacs (or so seems the consensus)? I can attest to the great changes in sound signature before and after the cap mod on rev4 0.012%, and second to only the huge increase in low-end volume is the improvement in naturalness in all frequencies, in the most general terms possible. What if, as previous measurements and simulations shown, the vref caps do improve ripple, just like good DC supplies, and continue to do so significantly even on rev.4? I urge you to pick up the soldering irons and see for yourself! Or better yet some measuring equipment or simulation software.

Even if I can't contribute much more to the development of this wonderfully accessible piece of technology, it's been a very interesting experience thinking about "next tweaks" rather than "next music", one that I'll probably reflect on for a quite a while to come.

I don't believe that there is any effect to sound quality when using less precise resistors in the LSB's, and selecting the 14 MSB is more a question about gut feeling, something like still having 4 MSB's at the full precision when the volume control is at -60 db.

I also believe that using 0.01% resistors (beside in the reference) is not going to affect sound quality, although it will probably measure a little better.... That's why I was originally only planning for 0.05% and 0.02% versions.

This, and the smooth strings from my cans, puts a smile on my face. No one in today's HiFi industry should be allowed to indulge their sanity and make sales at the same time.

Now to catch up on sleep

 

[Jogi]

So, the caps arrived, and to not do a dumb mistake I better ask how to. I will never be sure about how to connect caps in V-...Is it OK to add the caps as shown in the pic?
And then there are some important looking ceramic caps in the yellow box, do these belong to Vref as well?
I stole the picture from an innocent Chinese DIYer, please forgive me dude.
And yes Ynmichael, you should get some sleep from time to time, I was already wondering if you ever sleep

 

[analog_sa]

Jogi, what board is this? The temptation to rip out all these horrid ceramics and replace them with Pana ECPU is overwhelming

 

[Jogi]

That's the 1121 OEM board and you are absolutely right about those ceramics. But I think I will just solder the polymer caps over the existing ceramics, should be fine.

 

[TNT]

Please document your efforts gentlemen! Would be nice to understand the result.

//

 

[androa76]

I will not copy paste all SebastianL words but i totally agree with his words. I have also chance to test and hear R2R dacs including Totaldac, MSb, Metrum...
I have modified dac reall carefull and also step by step... I think it is not limitation in hardware (I tried to add from B1, Jumas preamp, tube output.. not big difference) but in software. I just wait for final test when my computer expert friend will help me with downgrading to old software and try some other filters and remove dc filter. If no improvement dac goes away or in basket. If anybody had similar experience and discovered the cure to natural sounding Soekris dac... welcome

 

[living sounds]

I use a simple Rev 5 board with ordinary but well filtered PSU, lot's of capacitance on VREF and the latest "soft" filter.

Sounds wonderfull via the balanced outputs. No idea about Totaldac, MSB etc., but it sounds much better than the converters used today to master most of the music you can buy.

 

[TNT]

I regard them as accurate. I have questioned and put efforts in the rest of the system - and been rewarded. A modern good orchestra recording is a treat, resembling what I hear at the hall.

//

 

[Jogi]

Ahem, anyone can confirm the polarity on the attached Pic is ok? I am about to heat the iron...

 

[analog_sa]

I think it is not limitation in hardware

The early ones responded very positively to hardware mods. Of course you can never expect the carved in stone 3d images and dynamics one of the better MSB throws, as those have a real clock... But for the price and effort the 1021 is very rewarding. Looks like the newer ones will need a similar amount of attention, but hopefully be even more rewarding.

As for filters i am stuck on the C128dp Linear Phase filter by Paul and don't much like any of the stock ones.

 

[ynmichael]

Ahem, anyone can confirm the polarity on the attached Pic is ok? I am about to heat the iron...

I believe TNT is the only one here with the 1121?... If they measure right I wouldn't worry at all. For a permanent solution you would need to consider distance to IC, so maybe something like what spikestabber has done to minimize distance from caps to shift registers. The shift regs are considerably smaller though so many a workaround is needed. I would be happy with those shift regs though because they're probably better than the ones on 1021... Soren probably didn't consider impedance all that much when he chose the LVC.

Jogi, what board is this? The temptation to rip out all these horrid ceramics and replace them with Pana ECPU is overwhelming

I wouldn't be so quick to give in to that temptation... what's your justification for replacing the Samsung caps? People reported great results with regular electrolytics...

 

[ynmichael]

The early ones responded very positively to hardware mods. Of course you can never expect the carved in stone 3d images and dynamics one of the better MSB throws, as those have a real clock... But for the price and effort the 1021 is very rewarding. Looks like the newer ones will need a similar amount of attention, but hopefully be even more rewarding.

As for filters i am stuck on the C128dp Linear Phase filter by Paul and don't much like any of the stock ones.

There appears to be a lengthy debate about this in the early days, of which you might be a part. It seems the FPGA jitter would really overwhelm any improvement you get in the clock. And rockna seems to be doing something similar to dam1021. That's as much as I was able to gather from what Soren said anyways.


Interesting observation on the filters... Perhaps time to downgrade to 1.06 again Maybe this weekend

 

[ynmichael]

I regard them as accurate. I have questioned and put efforts in the rest of the system - and been rewarded. A modern good orchestra recording is a treat, resembling what I hear at the hall.

//

Maybe for the 1121, but for 1021 there's a huge difference before and after vref caps. Separate question whether only pure undiluted distortion was added.

 

[ynmichael]

I use a simple Rev 5 board with ordinary but well filtered PSU, lot's of capacitance on VREF and the latest "soft" filter.

Sounds wonderfull via the balanced outputs. No idea about Totaldac, MSB etc., but it sounds much better than the converters used today to master most of the music you can buy.

That seems good evidence that this little effort is on the right track! Btw, I wonder if the THD-frequency graph would change with the mod...

 

[Jogi]

Wow.

 

[living sounds]

Btw, I wonder if the THD-frequency graph would change with the mod...

Not that I've noticed. And it really shouldn't matter. Most of the studio gear music is run through during production has higher THD anyway. My Lynx Aurora16 has much lower THD than the DAM1021 testing with sinewaves, but sounds much more coloured vs. the analog source... And not in a good way.

 

[ynmichael]

Not that I've noticed. And it really shouldn't matter. Most of the studio gear music is run through during production has higher THD anyway. My Lynx Aurora16 has much lower THD than the DAM1021 testing with sinewaves, but sounds much more coloured vs. the analog source... And not in a good way.

That's not bad news actually. Just wondering if there's truth in the hypothesis that the caps are adding distortion that makes things sound more analog. Like how the HD800s adds harmonic distortions in the low end to make things sound warmer and bassy. We probably want to avoid that.

Wow.

I take it that your dam1121 is still alive and well

 

[Jogi]

Yes, still alive. For whatever reason the SKlite 5V line puts out only 4,5V for the Waveio, so I had to switch to USB power from computer, which is not a good idea.
And I only added 3000µf to each of the Vref+, and left Vref- as it is as I didn't get a confirmation.Better be patient than sorry...
And I added 1500µF to the clock supply.
So...much better. Still not really there, but there is hope. The dac sounds somehow current starved and stressed in stock form, which is better now.
Our supermodel gained some weight. But still the pants are little loose around the hips.
High freq is still somehow unnatural, but on its way. Maybe the other strings and good USB power will do it.

 

[ynmichael]

Yes, still alive. For whatever reason the SKlite 5V line puts out only 4,5V for the Waveio, so I had to switch to USB power from computer, which is not a good idea.
And I only added 3000µf to each of the Vref+, and left Vref- as it is as I didn't get a confirmation.Better be patient than sorry...
And I added 1500µF to the clock supply.
So...much better. Still not really there, but there is hope. The dac sounds somehow current starved and stressed in stock form, which is better now.
Our supermodel gained some weight. But still the pants are little loose around the hips.
High freq is still somehow unnatural, but on its way. Maybe the other strings and good USB power will do it.

Great to hear that things are going well! SKLite is 300mA 5V max and I’m getting 5.00V with amanero which is 600mW max. 3000uF is a good amount as you already have 1000uF+ in stock form (I think). People have only added smallish caps to the clock but I’m not sure if 1500uf should be even better, probably. Would you like to share a photo of the mod? Good efforts for the night!

P.S. any change in bass? Huge change in dam1021 bass volume after the mod

 

[Jogi]

Yes, bass and low midrange is much improved, and this is so important to support the rest of the music. I will take a picture when finished with it.
From building and tweaking a lot of stuff the last decade or so I can say for sure, that power supply is key to good sound. Would love to supply Vref with batteries, as I am sure this would be a major step, as some folks here found out already.