DAC chip signature sound?

[Jorj]

How can you isolate the chips from the overall, specific, circuit designs and interconnection parameter differences by subjective listening to a device. Are you expressing this viewpoint from a knowledge-based and practiced experience or internet opinions.

Simple answer? I don't believe you can. I've yet to see a setup where you roll in DAC IC's, but if anyone has been able to do that, it's the tinkerers over at the DIY scene.

 

[Deleted member 65]

Simple answer? I don't believe you can. I've yet to see a setup where you roll in DAC IC's, but if anyone has been able to do that, it's the tinkerers over at the DIY scene.

The MKII version of my DAC enables you to roll DAC IC's (AKM AK4490, AKM AK4497 and ESS ES9038PRO) , I've no intention to upgrade mine to MKII ...

 

[Jorj]

Very cool! Musical Paradise MP-D2 MkII, that is a new one for me. Looks like a tinkerer's delight, or a sure pathway to nervosa madness.

This kind of discussion is super interesting to me. Humans are so susceptible to cognitive dissonance (confirmation bias being one form) it would be great to have empirical data to show either way that a certain brand of DAC has a signature sound, or we're just blowing smoke up each other's arses. From the data that Amir has posted on his reviews, and from other measurement sites, it seems clear to me that either:

1. The data does not support the hypothesis that DACs have 'signature sounds', or
2. Testers do not routinely run the tests that could produce an empirical analysis.

This is, of course, is different than the obvious flaws in frequency response, such as the one that the recently reviewed Schiit Modi 2 Uber displayed which would have a noticeable impact on sound. Maybe Amir or some of the others could chime in on additional tests that could be used...pre and post ringing, or perhaps tests using real driver loads and actual music as source material instead of test tones.

At the end of the day, I think we'll find that as long as the DAC does a reasonable job of converting the bits to waves, we can point to other components in the system and say with some confidence that they have more impact on sound. Transducers most of all, followed by amps, then long runs of poor quality unshielded cables. The rest of it is just hand waving and nervosa, IMO. Maybe I'm just lucky and my hearing is not good enough to tell the difference.

 

[Beherit]

I presume it's not doing the right measurements as these units do sound similar but wildly different. There tends to be bits of a consistent sound across some DACs with some chips. I've heard it with the AK4490 in the Schiit DACs, Grace M900, and the RME units myself. They lack bass punch and snare/cymbal low treble attack. Anything with the ESS 9018/9028 seems to have a universally etched treble timbre and is at least slightly warm in the bass even if the rest of the sound is dry. That's not to say the units sound the same but they do share these qualities. The Bifrost 4490 is much warmer and V-shaped "loudness button" sounding than the RME but they both still have veiling of upper bass/low mid dynamics to the point of sounding like some instruments with fundamentals there have been run through a compressor. It's certainly not "normal" or "neutral" compared to well, almost everything else. Compare a Schiit DAC to the Grace M900 to the RME ADI 2 Pro/DAC back to back with music on a detailed, neutralish setup, even a pair of Sennheiser open headphones or decent monitors in a decent room (no Redditor special of JBL 305s as computer gaming speakers), and you'll hear what I mean. They sound very different but share certain colorations.

 

[Jorj]

I've heard it with the AK4490 in the Schiit DACs, Grace M900, and the RME units myself.

Was this listening done with a sighted comparison? IOW, did you know what DAC chip was in the box before hearing it? It cannot be overstated how powerful confirmation bias and the placebo effect is. Your brain can shut off pain processing when given a placebo, why could it not fool you into hearing something it thinks it should? Again, until the empirical data comes in, I'm not sold. I have level-matched and blind tested Sabre vs AKM vs Cirrus in the past and found my guesses on which were which were no better than noise. I have had my ears tested and found that for my age and the abuse they've endured, they are not too shabby, but no way can I say that in blind listening I can tell DAC chips apart.

 

[Shazb0t]

It would be interesting for you to do a double blind test with your setup having someone else switch the DAC chips and seeing how accurate your are in identifying them.

 

[dc655321]

How about looking at the difference in energy content produced by two dac/amps?

eg: take dacA and dacB, voltage match them to a 1kHz tone, play a song through each separately (or several songs), loop that output back into a adc (soundcard?), then look at the difference in spectra produced by each (fftA - fftB = potential sonic differences).

 

[sofrep811]

I owned a Magnavox and Phillps CD player with the TDA1541 chips. I look for them at second hand stores and garage sales, and TO MY ears they sound very warm, comparable to DSD. First listen I was actually taken back at the smoothness and overall softness, and almost too soft using my standards reference CDs and WAV files --PF-Meddle CD's -- 87 Nimbus, 89 MFSL, 16 RM USA and then I use more jagged and treble with Nirvana's In Utero CD's (93 DAD CD, 97 MFSL CD, and always a 24/96 WAV unmastered Albini mix I poached from Limewire days, but mastered by myself in various states.

 

[jsrtheta]

Beherit: "...these units do sound similar but wildly different. "

How does that work, exactly? And what would be "the right measurements"?

 

[Kyle / MrHeeHo]

I'd have to guess that if it isn't just placebo it would either have to be the digital filters the DAC chips come with or more likely it's something besides the chips they are hearing. I wonder how a test would be set up to figure this out.

I have another theory about why particular audiophile circles say things like ESS chips are harsh/bad but it's a bit too mean to share here IMO

 

[TBone]

I owned a Magnavox and Phillps CD player with the TDA1541 chips. I look for them at second hand stores and garage sales, and TO MY ears they sound very warm, comparable to DSD. First listen I was actually taken back at the smoothness and overall softness, and almost too soft using my standards reference CDs and WAV files --PF-Meddle CD's -- 87 Nimbus, 89 MFSL, 16 RM USA and then I use more jagged and treble with Nirvana's In Utero CD's (93 DAD CD, 97 MFSL CD, and always a 24/96 WAV unmastered Albini mix I poached from Limewire days, but mastered by myself in various states.

I also have experience with those players, but too long ago ... I've since been looking to get a 1541 based player on the cheap to modify and compare. I still own the MASH based player that replaced my long gone Phillips 1541 chiper. It has a characteristic (which I don't really understand) that can be measured, consistently, one which I find negatively effects tonality. I wasn't that surprised (perhaps a bit surprised), that my (now smoked) Streamer had the exact same sonic quality, and upon measuring, it had the exact same characteristic.

I find it a bit silly that people compare the "sound" of DACs based on chips, but rarely based on its opamps.

 

[derp1n]

I have another theory about why particular audiophile circles say things like ESS chips are harsh/bad but it's a bit too mean to share here IMO

Do tell.

 

[maxxevv]

I find it a bit silly that people compare the "sound" of DACs based on chips, but rarely based on its opamps.

On my portable DAC, I have swapped out the dual single-channel Opamps 2 times. The differences between the 3 different chips is quite easy to spot.

Which is why, there is seldom if ever an apples to apples comparison between DAC chips sound signature as the end-stage output implementation makes such a significant difference to the "signature" of the overall device.

 

[captain paranoia]

I find it a bit silly that people compare the "sound" of DACs based on chips, but rarely based on its opamps.

There are plenty who play with op-amps, too. Including replacing their IC op-amps with amps contructed from discrete, matched JFETs.

 

[Wombat]

There are plenty who play with op-amps, too. Including replacing their IC op-amps with amps contructed from discrete, matched JFETs.

There will always be tinkerers.

 

[wbh]

The DAC-chip-swap (and/or "architecture" plug-n-play) has been around for a while. For consumer products, back in 1993, Counterpoint has a cool DAC called the DA-10:
https://www.gammaelectronics.xyz/stereoph_2-1994_Counterpoint_DA-10.html

And, very recently, Astell-Kern's A&futura SE200 features:

Equipped with the world's first Multi-DAC (2 x ESS ES9068AS + 1 x AKM AK4499EQ)
Independent AMP structure designed to compliment each DAC's characteristics

And, of course, many of latest discrete R2R dacs have OS and NOS modes. And FPGA-based dacs allow all sorts of swaps.
==============

All that said, while I've never tried any of the above, I can opine -- after having been exposed to myriad architectures (from classic Philips mulibit DEM (dynamic element matching) to the latest AKM and ESS variety -- that (a) the implementation, e.g. output analog filter design and I/V, is most important ; followed by (b) bitstream vs MASH vs D-S vs classic multi ; followed by (c) OS vs NOS.

============
Currently, although I have some high-enders in my position, I am very impressed by the sub $90 Hifime S2 USB and SPDIF DAC (Sabre ES9038q2m). This is my first experience with ESS and I'm very pleased. Not sure how much better the PRO version of the 9038 can be .... but if ESS is part of Hifime S2's "magic", consider me an ESS fanboy.

 

[vlad335]

Currently, although I have some high-enders in my position, I am very impressed by the sub $90 Hifime S2 USB and SPDIF DAC (Sabre ES9038q2m). This is my first experience with ESS and I'm very pleased. Not sure how much better the PRO version of the 9038 can be .... but if ESS is part of Hifime S2's "magic", consider me an ESS fanboy.

Sorry for quoting an older post but i was doing a search and this came up.

I have this little box and I am very impressed as well. I have since tried a couple somewhat more expensive DAC's and returned them both. Topping E30 and D50s. The E30 was eliminated right off the bat. The D50s I am certain measures better but it didn't draw me into the music like the little S2. Sure it was more detailed but these was a slight harshness to the highs, especially with distorted electric guitars at a certain volume threshold that I couldnt get past.

I don't trust myself so last week I did an impromptu test between theS2 and D50s with my wife listening. I level matched as best i could and had her sit there and tell me the difference between the two and tell me which one she liked best with a variety of music. She picked the S2 every time. i was very surprised by this and tried to trick her but never did.

Looking for another DAC to try but wondering what Amazon's threshold for returning stuff is.

 

[JustMe]

Eschew tone controls and EQ, then engage in tube and IC rolling, before or after hunting for similar but 'different' components that sound 'right'.
Seems like an expensive form of audio self-flagellation to me.

OK, that is funny. Tho i suspect cables are the #1 tools used as expensive, inflexible tone controls.
Tubes at least have very different overload characteristics compared ot most solid state devices

 

[Wes]

what about ladder designs?

 

[JustMe]

what about ladder designs?

R2R are ladders. That's what i'm talking about. All the (now old) R2R, ladder, resistor pair, call them what you will were chips. Some of the best ones were hybrids with laser trimmed thin film resistors. But "chip" means nothing frankly. A transistor is a chip, with its leads brought out to pins.

Mots of today's R2R/ ladder designs (MSB, denefrips, .....) use discrete resistors because......drum roll please ... you can no longer buy the pre-made chip based ones except for one AD chip (used by Schiit) that glitches. Mike's spcial sauce is de-glitching software.

thin films, by the way, can be very good. I am not a materials scientist - maybe one is here and would give us all a tutorial

But i design with all this stuff.

In the large, the DAC chjp is, like, the least important part fo the design in my experience. Its liek saying "i prefer the MPSA43 to the 2N3904". Yawn.

I/V concerter, power supply, timing (everything from osciallator to trace design), ground noise isolation, are all more important. As one person noted, many chips' "unqiue sounds" actually coems from companies that take the datasheet and reproduce the reference design. I'm one of those old fools who is designing every thing from scatch and evaluating many alternatives.

BTW i'm looking for "sounds gopod" not ",measures good". Anything > 80 dB is beyond our hearing -so what we ARE hearing is not characterized int eh SNR data, at least that' my opinion. -130 dB better than -120dB? Nah. Move on, ntohign to see here.

For those who dont geek out maybe i shoudl back up. Fundamentally what a DAC does is take a series of 1s and 0s, and run them through resistive dividers, that progress per binary: 0.01V, 0.02V, 0.04V, etc (arbitrary numbers, meaningful progression). This is done with pair of resistors and ohm's law. In a 16 bit DAC you have 16 pairs. The currents are summed and amplified. (actually Z-converted really, let's keep it simple). So some people call them Resistor to resistor (R2R) some call them ladders (they look like ladders - with the pairs being the rungs) or us simple types call them resistor pairs, pine pair per bit per channel.

Bitstream, on the other hand, uses ONE resistor/voltage and Pulse width/density modulation to turn it on and off a lot. Just like lightbulb dimmers. Or your car's electronic fuel injection. Or an arduino "analog" pin. Much cheaper. If full volume (2Vrms) is full volume, then 50% pulse density produces 1Vrms. and being on 10% of the time 0.2Vrms, etc. It seems fake, but once filtered it works very very well. All the chips discussed above are bitstream.

In the end that filter probably means much more than the chip. I hope so or I'm wasting a lot of time