Own one of the Denafrips R2R "ladder" dacs. Also own many classic and newer CDPs with multi-bit IC dacs (TDA1541, etc) or even Delta-Sigma or Bitstream. And also have some cheap kit dacs lying around based on ESS or Philips TDA chips.
I've heard this from reviews of R2R dacs, like Denafrips and Soekris, is that while they sound warm, clean and musical, they can sound "slow", lacking rhythm and drive. And that's what I've found, even with the Denafrips oversampling mode engaged.
Could it be all those R's in the discrete R2R (Denafrips, Soekris) topology?? That it takes some time for the signal to make it through those long R2R traces. Is that what slows the music down, taking away its sense of pace and drive?
I've heard this from reviews of R2R dacs, like Denafrips and Soekris, is that while they sound warm, clean and musical, they can sound "slow", lacking rhythm and drive. And that's what I've found, even with the Denafrips oversampling mode engaged.
Could it be all those R's in the discrete R2R (Denafrips, Soekris) topology?? That it takes some time for the signal to make it through those long R2R traces. Is that what slows the music down, taking away its sense of pace and drive?
IDK, still i think it could be more related to transient response and how the dac handles peaks, which sometimes are quite spectacular in relation to RMS voltage value, resistors shouldnt do anything to the phase of the signal, at least the is some capacitive or inductive effect somewhere. Maybe harmonic profile also changes... again IDK
The lower bits will experience more phase shift than the higher bits in an R-2R network, due to all the unintended RC low-pass sections made of the parasitic capacitances and resistors. That is bound to affect the settling and the distortion, but whether it has anything to do with your subjective impressions is a different matter.
Fortunately Soekris uses separate ladders for the positive and negative parts, so only the low bits will be switching at low volumes. I don't know if the same holds for Denafrips DACs, but it probably does, as I wouldn't know how they could get an acceptable performance out of those otherwise.
Fortunately Soekris uses separate ladders for the positive and negative parts, so only the low bits will be switching at low volumes. I don't know if the same holds for Denafrips DACs, but it probably does, as I wouldn't know how they could get an acceptable performance out of those otherwise.
Last edited:
Don't know about that exactly. May depend on the system they are in. Also, part of what you are referring to might have to do with transient response, phase alignment of the time-domain waveform. A lot of dacs are a little blurred sounding in time. That might happen due to clock timing jitter, for one possibility. Also, oversampling dacs tend to produce a lot of ultrasonic and up into RF frequency noise. The output stage typically uses opamps to LP filter and otherwise process the audio signal. Depending on details of how that is designed it might have some effect on the sound that you don't like, or that you may perceive as a little 'slow.' In particular, sharp cutoff analog LP filters starting at 20kHz my introduce some phase shift into the audio band which might be objectionable to some people.
Okay. I watched about as much of that video as I could stand. What I would say from the descriptions is that a 'slow' dac might kind of sound like a bass reflex speaker that requires a lot of resonance buildup to reproduce a low bass note. In that case the note attacks are slow in a way, that is it can take some time for the bass develop full volume, and then to fully decay at the end of a note. By way of comparison a very tight in the time-domain sealed box might have clear transient at the beginning of note and then at the end, but lack the deeper, more extended bass of a reflex design. Something more or less analogous to that type of thing. Of course, it may be rather shallow analogy, as they often tend to be.
And then in the video it turned out that a more expensive dac 'retrieved' more detail. Well, probably more like the cheap dacs failed to reproduce low level details very well. It can happen.
IMHO bottom line is that if you want a really good dac, the currently hot unit to get is one of the Holo dacs. They cost a few thousand dollars though. You are not going to get that SQ from a Topping D30 no matter how well it measures. Just not going to happen IME. So if you want something good, save up for it. If you want something almost as good for half the price, its probably going to be compromised in some way that if you are really picky you will notice. Sorry, don't have a better answer.
And then in the video it turned out that a more expensive dac 'retrieved' more detail. Well, probably more like the cheap dacs failed to reproduce low level details very well. It can happen.
IMHO bottom line is that if you want a really good dac, the currently hot unit to get is one of the Holo dacs. They cost a few thousand dollars though. You are not going to get that SQ from a Topping D30 no matter how well it measures. Just not going to happen IME. So if you want something good, save up for it. If you want something almost as good for half the price, its probably going to be compromised in some way that if you are really picky you will notice. Sorry, don't have a better answer.
Last edited:
Bare in mind that almost no one dac in the market actually is 'R2-R" ladder dac - from MSB to LSB.
Because they doing the thermomether encoding for first few bits. Mostlu 2-4 maximum.
That thermometer encoded bits are actualy going into conversion as sort of DSD bits.
All of the rest smaller significant bits are going to converison in R2R ladder.
Formula for "expand" from binnary to thermometer code is 2E(N)-1. Where N is number of first MS bits to encode.
For N=3 bit Thermometer code is long 7bits
For N=4 bit Thermometer code is long 15bits
And that is usually the border line... Because next step is for 35bits...
This method contributing in more precision.
.
So If You are talking about Diskrete DAC designs that is usually are segmented dacs with 2 segments, first "DSD" thermometer segment expanded from few first MSB, and second R2R segment from res of word bits... And these segments are clearly visible in the PCB-s in ladder form.
.
Next. You will be suprised but in the DAC ICs, marked as "R2R" is ususaly inside of PDF stated, that they are segmented koncepts. In some of them number of MSB are noted, but thay again use no more than 4 MSB for the thermometer segment. AD1865, AD1862, etc etc...
.
The 2 dac chip I know has true R2R koncept and they are TDA1540 and TDA1541A. Again, they are not employing R as passive resistors components inside the IC, because it is hard to implement them. But Transistor emulating "R" and "2R" by current sources and current sinks...
.
So Ladder is key word, it points to passive resistor network. Against just R2R (without Ladder word) that means emulated resistors by something else...
.
Conclusion:
1. Diskrete dacs usually has "Ladder", and "R" and "2R" true resistors. BUT in 99% are not true R2R because of the segmented aproach.
2. IC dacs are also in most cases segmented BUT not with "Ladder" R2R passive network inside the chip.
.
Sorry for larger post, it is just an information
Because they doing the thermomether encoding for first few bits. Mostlu 2-4 maximum.
That thermometer encoded bits are actualy going into conversion as sort of DSD bits.
All of the rest smaller significant bits are going to converison in R2R ladder.
Formula for "expand" from binnary to thermometer code is 2E(N)-1. Where N is number of first MS bits to encode.
For N=3 bit Thermometer code is long 7bits
For N=4 bit Thermometer code is long 15bits
And that is usually the border line... Because next step is for 35bits...
This method contributing in more precision.
.
So If You are talking about Diskrete DAC designs that is usually are segmented dacs with 2 segments, first "DSD" thermometer segment expanded from few first MSB, and second R2R segment from res of word bits... And these segments are clearly visible in the PCB-s in ladder form.
.
Next. You will be suprised but in the DAC ICs, marked as "R2R" is ususaly inside of PDF stated, that they are segmented koncepts. In some of them number of MSB are noted, but thay again use no more than 4 MSB for the thermometer segment. AD1865, AD1862, etc etc...
.
The 2 dac chip I know has true R2R koncept and they are TDA1540 and TDA1541A. Again, they are not employing R as passive resistors components inside the IC, because it is hard to implement them. But Transistor emulating "R" and "2R" by current sources and current sinks...
.
So Ladder is key word, it points to passive resistor network. Against just R2R (without Ladder word) that means emulated resistors by something else...
.
Conclusion:
1. Diskrete dacs usually has "Ladder", and "R" and "2R" true resistors. BUT in 99% are not true R2R because of the segmented aproach.
2. IC dacs are also in most cases segmented BUT not with "Ladder" R2R passive network inside the chip.
.
Sorry for larger post, it is just an information
I ignore reviews with such wording.
DACs can be measured with much greater precision than one can hear. All DAC measurements that I have seen are nearly perfect.
Ed
According to ESS, typical dac measurements such as SNR, DNR, THD+N are all PSS measurements (periodic steady state). Again according to ESS, such measurements do not show the real effects of signal-correlated noise which is produced by sigma-delta dacs.
In addition, ESS suggests that some people hear the effects of varying excess phase, something else not shown by typical measurements. Of course someone can choose to believe ESS or not. IMHO they know more about sigma-delta dac characteristics and how to measure them than some of the folks over at that other forum.
Moreover, according to John Westlake who visited the AKM factory sometime before the fire, he was shown a prototype of the AK4191 and AK4498 chipset. He said AKM told him they separated the dac into two chips to reduce the effects of 'substrate coupled noise.' Something else that does not show up very well in typical dac measurements (since according to John and IIUC there is signal correlated noise crosstalk between dac channels).
In short, it appears that some people feel overconfident that PSS measurements show anything and everything that constitutes dac performance. Don't know where that overconfidence comes from other than perhaps from 'social proof.' People may believe something simply because 'everybody I trust knows its true.' https://blog.crobox.com/article/social-proof-examples
In addition, ESS suggests that some people hear the effects of varying excess phase, something else not shown by typical measurements. Of course someone can choose to believe ESS or not. IMHO they know more about sigma-delta dac characteristics and how to measure them than some of the folks over at that other forum.
Moreover, according to John Westlake who visited the AKM factory sometime before the fire, he was shown a prototype of the AK4191 and AK4498 chipset. He said AKM told him they separated the dac into two chips to reduce the effects of 'substrate coupled noise.' Something else that does not show up very well in typical dac measurements (since according to John and IIUC there is signal correlated noise crosstalk between dac channels).
In short, it appears that some people feel overconfident that PSS measurements show anything and everything that constitutes dac performance. Don't know where that overconfidence comes from other than perhaps from 'social proof.' People may believe something simply because 'everybody I trust knows its true.' https://blog.crobox.com/article/social-proof-examples
Last edited:
From the ESS Hyperstream Modulator presentation:
There is a slide which initiates discussion about audiophiles with the words, "Understanding what audiophiles are hearing."
"The surprising reality is that sigma-delta DACs can be audibly distinguished from a conventional DAC despite measuring very much better than that DAC."
"...an important point: The human ear detects signals well below the noise level of the DAC."
"The ear is exauissitely sensitive to "unusual" noise sources. Your ancestors camped out by a waterfall (white noise) and yet their 'ears pricked up" when they heard a hint of a predator moving in the undergrowth. (The equivalent visual phenomenon is "seeing something out of the corner of your eye). Noise, to a large degree, can be accomodated by the ear and is not troubling, but the tiniest "anomalous" noise is raised to the conscious level."
"Sigma-delta modulators create non-periodic steady state noise (non-PSS) artifacts..."
"Periodic Steady State analysis is common in RF circuits. It means that the system is forced to repeat a pattern of behavior over and over again with a certain time period. Any artifacat is presumed to also repeat in this time period."
"Audio measurements such as THD and DNR are done in the Periodic Steady State. Therefore, they will not activate non-PSS noise. You will not find non-PSS noise by looking at THD, DNR, and SNR."
"As the audio signal moves, the noise does not remain the same."
"Non-PSS noise is the biggest issue, but experiments suggest there are more problems. For example: Audiophiles rate as inferior systems that have variable excess phase noise."
"We find that an unconditionally stable loop sounds better in listening tests."
It would have to be a signal-correlated noise crosstalk measurement. Not part of the standard AP suite of 'figure of merit' tests so far as I know.
Its in the public record. Suppose one could call ESS liars if the denial impulse were strong enough.
Besides there is more: KSTR captured the time-domain waveform of hump distortion and listened to what it would sound like if it were at a higher level. IIRC he said it sounded like 'glare' and was 'fatiguing,' exact same things audiophiles have been saying for a long time. Curious coincidence, that.
Besides there is more: KSTR captured the time-domain waveform of hump distortion and listened to what it would sound like if it were at a higher level. IIRC he said it sounded like 'glare' and was 'fatiguing,' exact same things audiophiles have been saying for a long time. Curious coincidence, that.
Markw4:
It looks like you're a fan of ESS.
Any thoughts about their new 9039?
https://www.esstech.com/wp-content/uploads/2022/05/ES9039MPRO_ES9039PRO_Product_brief_v0.2.1.pdf
It looks like you're a fan of ESS.
Any thoughts about their new 9039?
https://www.esstech.com/wp-content/uploads/2022/05/ES9039MPRO_ES9039PRO_Product_brief_v0.2.1.pdf
The TDA1540 and TDA1541 are far more advanced than just a plain old R-2R DAC. They use current sources that are switched to the output or to ground and use an R-2R network for generating the bit currents for the lower bits, but have a dynamic element matching technique for the higher currents.
I would not use 12 years old marketing material as reference.
If you use claims that are not your own do it with proper quotes.
Not so much. I tend to prefer the sound of AKM.
Yes. That one is interesting. The first one with a Hyperstream 4 modulator. All the old ones use Hyperstream 2. Wouldn't be surprised if it is intended to fix some problems the older modulators have. However, as long as the old chips are still selling well maybe better not to make too much of the new technology yet?