Hello,
I'm trying to get started in DIY, and I'm looking for reccomendations on a DAC IC to use for it. The myriad of chips available makes it quite hard to choose any which one with any amount of confidence.
Anyhow, I had been looking at the Sabre32 ES9012/8, but I'd rather avoid anything which involves the signing of restrictive NDAs, mostly because I would like release the schematics of my DAC as open hardware. Essentially, I am asking if any DAC chips exist which would be of approximately the same quality of the ES9018, with more open documentation? ESS claims to have patented their so-called "HyperStream" DAC, supposedly offerring unmatched low distortion and dynamic range. I am naturally weary of such bold claims, especially when accompanied by an obligatory trademarked named. Are these boasts warranted? Are there any other DACs that offer the same degree of jitter reduction? I am not concerned with the built in DSP functions, as I plan on coupling the DAC with a dedicated DSP anyway. I am not sure if 32-bit is truly necessary, as I not even sure that any vaguely commonly available music is avaible in 32-bit.
Thanks,
-Kevin
I'm trying to get started in DIY, and I'm looking for reccomendations on a DAC IC to use for it. The myriad of chips available makes it quite hard to choose any which one with any amount of confidence.
Anyhow, I had been looking at the Sabre32 ES9012/8, but I'd rather avoid anything which involves the signing of restrictive NDAs, mostly because I would like release the schematics of my DAC as open hardware. Essentially, I am asking if any DAC chips exist which would be of approximately the same quality of the ES9018, with more open documentation? ESS claims to have patented their so-called "HyperStream" DAC, supposedly offerring unmatched low distortion and dynamic range. I am naturally weary of such bold claims, especially when accompanied by an obligatory trademarked named. Are these boasts warranted? Are there any other DACs that offer the same degree of jitter reduction? I am not concerned with the built in DSP functions, as I plan on coupling the DAC with a dedicated DSP anyway. I am not sure if 32-bit is truly necessary, as I not even sure that any vaguely commonly available music is avaible in 32-bit.
Thanks,
-Kevin
Although 24Bit/96KHz is regarded is the highest optimum sampling rate for human being, there is no formal research to support it. Since sampling rate includes two parameters of Bit rate (e.g. 16/24/32 Bit) and Frequency sampling (e.g. 44.1/48/88.2/96/176.4/192 KHz) The sensitivity of bit rate and frequency sampling to human being is just a "question mark".
=>Will bit rate (Bit) be more sensitive than frequency sampling (KHz) to human being?
=>Will 32Bit/96KHz be better than 24Bit/96KHz or 24Bit/192KHz?
Presently, the highest resolution of music is 24Bit/192KHz in the market, but not limited to future development. For this reason, your question “I am not sure if 32-bit is truly necessary, as I not even sure that any vaguely commonly available music is available in 32-bit.” is hard to answer without formal supporting research. Thus, it is now subject to your preference.
I would love to believe that ESS have a DAC with jitter rejection so good that jitter can no longer be considered to be an issue. I don't mean an issue for me, I mean an issue for everybody who obsesses about jitter.
But...
...I don't trust anybody who produces a DAC and calls it 32-bit, because I don't believe that any piece of electronics has 190-odd dB of dynamic range.
Thermal noise power (the noise floor at room temperature) in a 20kHz bandwidth is -174 + (10*log(20,000)) = -130dBm. 190 dB above this is 60dBm, or a million milliwatts, 1000 watts.
You pays your money and you takes your choice.
But...
...I don't trust anybody who produces a DAC and calls it 32-bit, because I don't believe that any piece of electronics has 190-odd dB of dynamic range.
Thermal noise power (the noise floor at room temperature) in a 20kHz bandwidth is -174 + (10*log(20,000)) = -130dBm. 190 dB above this is 60dBm, or a million milliwatts, 1000 watts.
You pays your money and you takes your choice.
In quantitative approach, it is unlimited to look for higher resolution in term of sampling frequency (i.e. it is technical feasible to design higher Bit/KHz) which is classified as “engineering”. In qualitative approach, it is limited to the audibility of human being which is classified as “Art”. Due to the constraint of budgeting in managerial level, the destination may be set to the end point at the same or slightly higher to the audibility of human being, but not the technical difficulty to attain higher specification.
DIY alternatives that are tried, tested and sound "perfect" to human ears are also avail:
PCM1794A
WM8741
AD1955
CS4398
The proper designation for "32 bit DAC" is "ACCEPTS 32 bit formated PCM". They don't have 32 bit performance. Heck, ANY DAC on market barely get close to 22-23 bit true audio performance.
WM accepts 32 bit. TI has also seven 32bit DAC's (like PCM1975) but the specs for those chips are below their present flagship 24 bit DAC.
They have moved to 32 bit more for bragging rights and, IMO, probably for ease of integration in some future low-end Bluray systems (because don't need an external DSP to process an eventual 32bit to 24bit).
PCM1794A
WM8741
AD1955
CS4398
The proper designation for "32 bit DAC" is "ACCEPTS 32 bit formated PCM". They don't have 32 bit performance. Heck, ANY DAC on market barely get close to 22-23 bit true audio performance.
WM accepts 32 bit. TI has also seven 32bit DAC's (like PCM1975) but the specs for those chips are below their present flagship 24 bit DAC.
They have moved to 32 bit more for bragging rights and, IMO, probably for ease of integration in some future low-end Bluray systems (because don't need an external DSP to process an eventual 32bit to 24bit).
Last edited:
I'm not certain if it is still posted on their website, but Meridian has researched this question and published the results in a whitepaper a number of years ago. If I correctly recall, they determined that 21-bits of resolution at a 60ksps rate was the maximum required to achieve absolute fidelity with respect to the human auditory system.
Thank you for information. It is welcome to anyone referring more supporting documents. In order to this research, you need “source + equipment + audience”. My questions are:
1) How did Meridian get 21-bit source?
2) Where is this research paper located for public reference or the ref. no. and date of Journal?
I´ve also been around to look for the best chips, to be able to make the best DAC possible. But there is more to it than just specs.
If you consider the Sabre DAC, then be aware, that it needs an I/V converter.
And the output is so hot, that your only choise in fact is an op-amp for I/V.
Then you are throwing yourself in the arms of the op-amp manufacturer.
The Sabre is not recommended with i.e. passive I/V, because performance deteriorates quickly with higher impedance.
BB1794 is another good alternative, and this might even perform pretty nicely into a passive I/V.
The AD 1955 will not, it needs an op-amp.
So to get the good performance of the top of the pop DACs further down the line, you most likely need to use op-amps for I/V. Otherwise performance will not be that good.
That could actually make some of the other makes more interesting, i.e. AKM, CS and maybe others.
CS is quite close to the performance of both BB and AD as far as regarding THD+N. The filter in the BB1794 is some steps deeper than in their competitors, but that is mainly due to AKM and CS widely professional use.
As a deeper filter will make more delay, and cannot be used on stage due to that.
Also the voltage output opens up for discrete designs, which I would prefer over op-amps at any time. Mostly due to the free choice of using NFB or not.
So after all, one should look at the final resolution in the analog output.
Stereophile are normally exited when they find something like 20 bit resolution in the analog output of a digital device.
If you consider the Sabre DAC, then be aware, that it needs an I/V converter.
And the output is so hot, that your only choise in fact is an op-amp for I/V.
Then you are throwing yourself in the arms of the op-amp manufacturer.
The Sabre is not recommended with i.e. passive I/V, because performance deteriorates quickly with higher impedance.
BB1794 is another good alternative, and this might even perform pretty nicely into a passive I/V.
The AD 1955 will not, it needs an op-amp.
So to get the good performance of the top of the pop DACs further down the line, you most likely need to use op-amps for I/V. Otherwise performance will not be that good.
That could actually make some of the other makes more interesting, i.e. AKM, CS and maybe others.
CS is quite close to the performance of both BB and AD as far as regarding THD+N. The filter in the BB1794 is some steps deeper than in their competitors, but that is mainly due to AKM and CS widely professional use.
As a deeper filter will make more delay, and cannot be used on stage due to that.
Also the voltage output opens up for discrete designs, which I would prefer over op-amps at any time. Mostly due to the free choice of using NFB or not.
So after all, one should look at the final resolution in the analog output.
Stereophile are normally exited when they find something like 20 bit resolution in the analog output of a digital device.
There are several axis upon which the audio DAC chip universe might be fundamentally divided.
Axis 1: Sigma-Delta DACs vs. the resistor-ladder class of DACs. The sigma-delta DACs almost universally have superior measured performance, well, at least up to a 20kHz measurement band-limit. They employ noise-shaping feedback to attain their very low measured THD. The Trouble is, as we all know, often in high-end audio there is little apparent correlation between measured perforformance and subjective performance.
Axis 2: Current-output DACs versus voltage-output DACs. Voltage-output DACs are usually more convenient to implement, but much depends on how they are achieving their voltage output - as far as perceived sound quality is concerned.
Axis 3: DACs with internal digital filters (most of them) vs. DACs either without an internal digital filter or one which can be bypassed (few of them). I view this as probably being the most important axis, as I feel that the sound quality of a DAC is much more determined by the digital filter implementation than by the DACs converter circuit.
I feel that the number of bits of input resolution supported by a DAC has next to no bearing on it's sound quality. I also feel this is true of audio bandwidth beyond 20kHz (or even lower). Of the many problems in attaining subjectively good digital playback, sample resolution beyond 16-bits and high-frequency reproduction beyond 20kHz are near the bottom. What I mean by this is that whatever it is that works against digital audio producing the gestalt of real instruments is not related to CD's 16-bit resolution or 20kHz audio bandwidth.
Axis 1: Sigma-Delta DACs vs. the resistor-ladder class of DACs. The sigma-delta DACs almost universally have superior measured performance, well, at least up to a 20kHz measurement band-limit. They employ noise-shaping feedback to attain their very low measured THD. The Trouble is, as we all know, often in high-end audio there is little apparent correlation between measured perforformance and subjective performance.
Axis 2: Current-output DACs versus voltage-output DACs. Voltage-output DACs are usually more convenient to implement, but much depends on how they are achieving their voltage output - as far as perceived sound quality is concerned.
Axis 3: DACs with internal digital filters (most of them) vs. DACs either without an internal digital filter or one which can be bypassed (few of them). I view this as probably being the most important axis, as I feel that the sound quality of a DAC is much more determined by the digital filter implementation than by the DACs converter circuit.
I feel that the number of bits of input resolution supported by a DAC has next to no bearing on it's sound quality. I also feel this is true of audio bandwidth beyond 20kHz (or even lower). Of the many problems in attaining subjectively good digital playback, sample resolution beyond 16-bits and high-frequency reproduction beyond 20kHz are near the bottom. What I mean by this is that whatever it is that works against digital audio producing the gestalt of real instruments is not related to CD's 16-bit resolution or 20kHz audio bandwidth.
Last edited:
sorry kurt but that is misinformation at best and complete ******** at worst. the 3vrms (which is a pretty standard balanced line level output, its others that are too low and need gain to meet the standard) output of the sabre is completely usable with any well designed discrete solid state IV stage.
the 32bits of sabre are not about using 32bit audio, although you can, its about giving the filter and volume control (if desired) headroom.
does it reject jitter perfectly? no it doesn't, you can still hear differences between different i2s and spdif sources, but it does a better job than any other dac i've used and consistently gives both measured and audible results superior to any other i have used too, even without opamps. if you can't make the 3vrms output work for you with a discrete stage you are doing something wrong
I saw you just before singing up sampling's praises
the 32bits of sabre are not about using 32bit audio, although you can, its about giving the filter and volume control (if desired) headroom.
does it reject jitter perfectly? no it doesn't, you can still hear differences between different i2s and spdif sources, but it does a better job than any other dac i've used and consistently gives both measured and audible results superior to any other i have used too, even without opamps. if you can't make the 3vrms output work for you with a discrete stage you are doing something wrong
I saw you just before singing up sampling's praises
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.