Eliminating jitter "completely" - Benchmark DAC1 approach

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
I was intrigued by the apparently *excellent* AP One measurement results for the Benchmark Media DAC1 (available at: http://www.benchmarkmedia.com/digital/dac1/DAC 1 - Manual - Rev A.PDF ).

I got a reply from their designer and I'm including some information from that reply (I don't want to quote him in entirety as I have not yet received permission to do so). The information provided here is however not a trade secret.

Please look at the measurements before you read further (PDF link above). The jitter immunity this DAC provides is in my personal understanding phenomenal. I have never seen such Audio Precision results for any commercial DAC for which I've seen properly done measurements. The other measures specs are excellent as well.

And yes, I know that jitter immunity isn't be-all-end-all in DAC peformance, but it can be one very important factor.

Avoiding AID is another and I'll get to that one as well later on. Having minimal DC drift in the DA stage is another.

I think these three factors: 1) jitter immunity, 2) AID filtering & 3) DC drift minimisation are perhaps the three most important DA-conversion properties that can make or break the sound of a DAC theoretically. I'm now excluding for the sake of this discussion the analog stage that comes right after the DAC (just trying to concentrate on the actual DA stage).

Please note that I haven't built a DAC like so many of you, my comments are highly theoretical (and not based on a wide experience on various DACs) and I'm willing to accept that they are 'wrong' in real life as opposed to theory.

This is exactly the reason I'm soliciting your more knowledgeable comments on these issues.

Ok, the specifics of DAC1.

- It uses AD1853 DAC ICs in a mode that achieves 115 dB stopband attenuation

- It operates AD1853 at constant 97.65625 kHz sample rate, regardless of input input sample rate, thus shifting the transition band higher (with input sample rate of 48 kHz) than with a sampling rate of 48 kHz would do

- At 96 kHz input rate the fixed clock it only gives DAC1 a transition band shift of c. 860 Hz upwards, still enough to eliminate most transition band problems

- DAC1 uses a FIR low pass filter to incoming PCM stream (before AD1853) with a passband at 44.29 kHz and stopband at 53,37 kHz (achieving 125 dB attenuation)

- It claims to eliminate almost all transition band effects (including aliasing intermodulation distortion effects) for signals with input sample rate of 96 kHz or less. For 192 kHz the data is (I assume) downsampled to 96 kHz and processed further from that.

- I think the measurements (if valid/repeatable) prove this is not just mere marketing hyperbole. The measurements are excellent IMHO. I understand however that measurements are not equal to real life perceived performance.

- It also has a variable volume headphone output (just off the DAC) (op-amps?) with an output impedance of near 0 Ohm to minimize distortion to various headphone loads.

This is coming from the designer, paraphrased by me.

Now, my questions are:

1) Are the approaches used the above 'common' within the DIY DAC community here? Esp. the fixed non-standard sample rate and relate FIR filtering at shifted transition band?

2) Has anybody here actually measure his/her DAC with AP One/Two or Miller Audio Research QC units to assess technical performance? Or does anybody have any measurements from any other DAC to compare?

3) Would $799 USD + courier + VAT of 22% be a too high a price for a professional audio DAC like this, which has gotten a warm response from the few Pro Audio mixers / producers I have been able to get a personal comment from? Do you think the DIY community could equal or outperform this TODAY, that is, is there a DIY design that at least in technical performance is better than what DAC1 claims?

Anything else you wanna add or ask?

I'm asking, because I'm in a market for a new DAC, it needs to be faultless even with jittery sources and can't cost an arm and a leg (well, that's all relative as $800 USD is a quite a chunk of money to me personally).

I'd really appreciate comments on this ready-made DAC approach, even though I understand this is a DIY forum.

All comments welcome.

best regards,
Halcyon
 
halcyon said:

- It uses AD1853 DAC ICs in a mode that achieves 115 dB stopband attenuation

No big deal there

- It operates AD1853 at constant 97.65625 kHz sample rate, regardless of input input sample rate, thus shifting the transition band higher (with input sample rate of 48 kHz) than with a sampling rate of 48 kHz would do

This is simply the result of using a 25.0000MHz clock instead of the standard 24.5760MHz.

- At 96 kHz input rate the fixed clock it only gives DAC1 a transition band shift of c. 860 Hz upwards, still enough to eliminate most transition band problems

- DAC1 uses a FIR low pass filter to incoming PCM stream (before AD1853) with a passband at 44.29 kHz and stopband at 53,37 kHz (achieving 125 dB attenuation)

Almost every DAC uses a FIR filter. This is how all oversampling and sample rate converting circuits are implemented.

- It claims to eliminate almost all transition band effects (including aliasing intermodulation distortion effects) for signals with input sample rate of 96 kHz or less. For 192 kHz the data is (I assume) downsampled to 96 kHz and processed further from that.

Again everybody makes this claim. High performance digital filters have extremely good stopband attenuation and intermodulation distortion.


1) Are the approaches used the above 'common' within the DIY DAC community here? Esp. the fixed non-standard sample rate and relate FIR filtering at shifted transition band?


I think oversampling is a very commong method in DIY DACs, and any circuit using sample rate conversion will have a fixed sample rate input to the DAC stage. The non-standard sample rate is just an artifact of the clock speed.
 
jwb,

Of course almost all DACs use a FIR filter and have a transition band and many use Analog Devices DAC ICs :)

But I was asking the specfics (level of attenuation at transition band, the amount of transition band shifting and the resulting measured performance).

Can you design a DAC that measures on the bench equally well or better than DAC1?

Can an inexpensive DIY DAC achieve the level of jitter immunity that DAC1 does, which according to measurements I've read beats out Apogee, Aphex, Aardvark, Mark Levinson, Anagram, etc... gear.

The measurements are phenomenal, I'm sure nobody can deny that?

So, are you in fact saying that all this is 'old news' and that most of the oversampling digital filtering DIY DACs around here offer the same level or better performance?

Can you state one example? Can you offer proof in actual AP measurement data?

I really want to know, as I can't buy/build all the DACs and compare them for myself.

This is not meant as a flame, but as an honest question.

regards,
Halcyon

PS It's very easy to mouth off, but to prove what one claims... let's see :)
 
I just don't see anything special about this design. They are using off-the-shelf oscillators, standard three-terminal regulators, and so forth. I'd be very surprised if their UltraLock(TM)(R)(C) technology was anything more than a normal asynchronous sample rate converter.

Also to claim that induced jitter has no effect on the output is meaningless. It could mean the DAC has high jitter rejection, or it could mean that the DAC's self-induced jitter is greater than the jitter on the input.
 
halcyon said:
I was intrigued by the apparently *excellent* AP One measurement results for the Benchmark Media DAC1 (available at: http://www.benchmarkmedia.com/digital/dac1/DAC 1 - Manual - Rev A.PDF ).
...
...
Halcyon


Hi all,

In the mentioned white paper, they say (or claim?) something like that digital filters are jitter-sensitive. I'm quite a bit confused now 'cause till now I thought (or imagined??) they were not. Could anybody knowledgeable on the issue enlighten me?

Below is an excerpt from the white paper (page 19).

-----------------------------------------------------------------------
Problem #2: Jitter can severely degrade the anti-alias filters in an oversampling converter. This is a little known
but easily measurable effect. Most audio converters operate at high oversampling ratios. This allows the use of
high-performance digital anti-alias filters in place of the relatively poor performing analog anti-alias filters. In
theory, digital anti-alias filters can have extremely sharp cutoff characteristics, and very few negative effects on
the in-band audio signal. Digital anti-alias filters are usually designed to achieve at least 100 dB of stop-band
attenuation. But, digital filters are designed using the mathematical assumption that the time interval between
samples is a constant. Unfortunately, sample clock jitter in an ADC or DAC varies the effective time interval
between samples. This variation alters the performance of these carefully designed filters. Small amounts of
jitter can severely degrade stop-band performance, and can render these filters useless for preventing aliasing.
-------------------------------------------------------------------------

Regards,
SKK
 
Right SKK. This claim is absolutely false. It doesn't matter when or how the bits and words arrive inside the digital filter, as long as data integrity is maintained. The only point where jitter matters in the system is the word clock at the DAC itself. Even the bitclock is irrelevant here.
 
Digital anti-alias filters... Another joke??

SKK said:

...
In theory, digital anti-alias filters can have extremely sharp
cutoff characteristics, and very few negative effects on
the in-band audio signal. Digital anti-alias filters are usually
designed to achieve at least 100 dB of stop-band
attenuation.
...

They also talk about "digital anti-alias filters". Is it a sheer joke
or a real thing? I mean "THE DIGITAL anti-alias filter".

Shouldn't an anti-alias filter be placed before an AD converter in the signal chain? Am I missing something??

If what I think is right, how can possibly an anti-alias filter do its
job *digitally* with analog input? I'm curious..

Regards,
SKK
 
jwb said:
The only point where jitter matters in the system is the word clock at the DAC itself. Even the bitclock is irrelevant here.

Not quite true in all cases. In the PCM1704, the output changes on the 2nd rising edge of the bitclock, after the word clock has gone low. So the output jitter performance is dependant on the bitclock, not the word clock
 
When I discussed jitter with regards to the AD1853 with the board designer from Analog Devices about 5 years ago, we talked about jitter sensitivity from various sources. It is relatively clear to me that for the DAC chip itself, master clock would likely be the deepest root problem, followed by bitclock and L/R clock.

In audio signals, typically the L/R clock is the biggest problem. The S/PDIF interface is terrible in that the clock is embedded and so there will be significant jitter on both the recovered bit clock and L/R clock. Moreover, the jitter of one will be a function of the jitter of the other.

Kiwi has a good point that indicates that the problem seen by you will depend on the device at hand.

Petter
 
Hi Halcyon,

I have just come across your 'Benchmark DAC-1' enquiry in diyAudio.

Did you construct or buy, and did you obtain anything good that you think betters the Benchmark ?

Should very much appreciate your reply so that I might head straight for good digital playback using output from a Pioneer DVD into home made power amps/spkrs.

Thanking you in advance ............. Graham Maynard.
 
I've had a benchmark apart here,

akm spdif reciever, ad1896 asrc with a basic crystal oscillator hanging off it, ad1853 and opamps for output stage with some ic for headphone out


3 pin regulators, layout is long way from optimal for a mixed digital/analog pcb


no rocket science, love their trademarks they have cooked up, the marketing spew is a great rehash of the akm/ad pdf's




Mark
 
Konnichiwa,

halcyon said:
I was intrigued by the apparently *excellent* AP One measurement results for the Benchmark Media DAC1 (available at: http://www.benchmarkmedia.com/digital/dac1/DAC 1 - Manual - Rev A.PDF ).

I am not sure why you call them "excellent", I'd call them "par for the course" and of course, measurements like those taken for the DAC1 have yet to be shown to have a good (or in fact ANY!) orrelation with percieved "good sound".

halcyon said:
The jitter immunity this DAC provides is in my personal understanding phenomenal. I have never seen such Audio Precision results for any commercial DAC for which I've seen properly done measurements.

I have never seen measurements of Jitter done in the way Benchmark do either. They seem to have invented a specific new type, probably because it offers good measurements with what they got.

That said the measurements taken by stereophile also show low clock jitter. How this is achieved we cover soon.

halcyon said:
The other measures specs are excellent as well.

They are? I thought they where within the limits provided by the Chip makers standard application spec's. No skill needed, just use the standard eval circuit.

halcyon said:
- It uses AD1853 DAC ICs in a mode that achieves 115 dB stopband attenuation

So do many other DAC's.

halcyon said:
- It operates AD1853 at constant 97.65625 kHz sample rate, regardless of input input sample rate, thus shifting the transition band higher (with input sample rate of 48 kHz) than with a sampling rate of 48 kHz would do

Or in plain english, Benchmark wanted to avoid paying a slight premium for an oscillator with the "correct" 24.576MHz frequency for their ASRC (Asyncronous Sample Rate Converter) and used a 25MHz one which is cheaper. Someone in marketing then said "Hey, we can make that a "feature"!" and so they did.

halcyon said:
- At 96 kHz input rate the fixed clock it only gives DAC1 a transition band shift of c. 860 Hz upwards, still enough to eliminate most transition band problems

I may be forgiven for saying so, but when using semi-stochastic signals, such as music, the ASRC used to implement the DAC1's jitter rejection and "fixed clock" have a demonstrated negative sonic impact. I am not sure what "transition band problems" are being alluded to, but the shift by less than 1KHz or even the shift by 53KHz does not eliminate anything, it merely moves any transtition bandproblems from the DAC's Oversampling filter (which is of a very good spec, if poor sound) into that in the ASRC, which is commonly of lower quality. If the DAC's OS Filter or that of the ASRC are better is a question specific to the chips used.

halcyon said:
- DAC1 uses a FIR low pass filter to incoming PCM stream (before AD1853) with a passband at 44.29 kHz and stopband at 53,37 kHz (achieving 125 dB attenuation)

In other words, it uses a bog standard digital filter, liek everyone else.

halcyon said:
- It claims to eliminate almost all transition band effects (including aliasing intermodulation distortion effects) for signals with input sample rate of 96 kHz or less.

One would need to add that the same is true for any other DAC in the world using the same ASRC & DAC, both of which are commodity items.

halcyon said:
For 192 kHz the data is (I assume) downsampled to 96 kHz and processed further from that.

The DAC does not accept 192KHz input.

halcyon said:
- I think the measurements (if valid/repeatable) prove this is not just mere marketing hyperbole.

I think they are nothing but, because there is nothing outstanding in these measurements, for the basic chipset that is being used. There is no technology present except those made available as commodity by the various chipmakers.

halcyon said:
- It also has a variable volume headphone output (just off the DAC) (op-amps?) with an output impedance of near 0 Ohm to minimize distortion to various headphone loads.

In other words it has a headphone output. Most headphone outputs are like that, you know.

halcyon said:
1) Are the approaches used the above 'common' within the DIY DAC community here? Esp. the fixed non-standard sample rate and relate FIR filtering at shifted transition band?

No, most DIY'er pay more attention to getting things right according to their particular espoused principles and do not go out of their way to save a penny or two. As said, the entire tirade you produced is the "marketing speak" description of a system entierly traditional in design and execution with the only "features" that diverge are marketing spin on money saving measures.

halcyon said:
2) Has anybody here actually measure his/her DAC with AP One/Two or Miller Audio Research QC units to assess technical performance? Or does anybody have any measurements from any other DAC to compare?

The APOne does not measure jitter. The Miller analyser is very expensive and rare. I have done simpler tests (FM demodulator on WCLK) and found that that low jitter seems to correlate with better sonic transparency and an absence of edginess.

halcyon said:
3) Would $799 USD + courier + VAT of 22% be a too high a price for a professional audio DAC like this,

The cost is resonable for the package. However, I would suggest to only buy the DAC without audition if you have the right to return the unit if you do not like the way it sounds.

halcyon said:
Do you think the DIY community could equal or outperform this TODAY, that is, is there a DIY design that at least in technical performance is better than what DAC1 claims?

Much will depend on your definition of "better perfornamce". A DIY DAC using AD1853 & SRC1896 and a generic 25MHz clock module will provide the same performance at a notional cost. I believe PCB's for such are available from several chinese sources.

halcyon said:
Anything else you wanna add or ask?

I beleive in terms of subjective "good sound" even the most basic TDA1543 based Non Oversampling DAC will outperform the Benchmark, because of the use of non-traditional engineering principles that correlate well with percieved "good sound", despite poor measurements in the traditional sense.

halcyon said:
Can you design a DAC that measures on the bench equally well or better than DAC1?

Any semi-competent designer can.

halcyon said:
Can an inexpensive DIY DAC achieve the level of jitter immunity that DAC1 does,

As long as it uses a similar ASRC, of course....

halcyon said:
The measurements are phenomenal, I'm sure nobody can deny that?

Are they? There are by now many DVD & CD Players of quite low cost which have as low or lower Jitter.

halcyon said:
So, are you in fact saying that all this is 'old news'

It is. Look at the release date of the various ASRC's....

Sayonara
 
Konnichiwa,

Bricolo said:
Can someone explain me the advantage of running the ASRCs'output (AD1986) and the DAC (AD1853) at a 97.65625 kHz clock instead of a 96 kHz one?

The ability to use Computer grade (and thus very high volume and low cost) 25MHz oscillator and thus saving a buck or two on the whole thing.

Sayonara
 
“Or in plain english, Benchmark wanted to avoid paying a slight premium for an oscillator with the "correct" 24.576MHz frequency for their ASRC (Asyncronous Sample Rate Converter) and used a 25MHz one which is cheaper. Someone in marketing then said "Hey, we can make that a "feature"!" and so they did”

You’re very mistaken on your above statement – while some of Benchmarks “comments” are questionable - It’s correct and preferable not to operate an ASRC at a “Drifting”1:1 ratio.

SRC can produce odd effects when operating very close to a 1:1 ratio – OK if fixed, but if the frequency’s between the primary and secondary side of the SRC drift around a 1:1 ratio then you can get “strange” artifacts. Some SRC claim to be OK with 1:1 ratios – but I still prefer to avoid this situation.

John
 
I'm about to send out parts/PCB orders for a headphone DAC which contains a SRC. I originally picked a 49.152MHz oscillator with excellent phase noise characteristics - it turns out I can get a 50, 48, 27 or 25 with the same characteristics from the same manufacturer for less.

And I just might do that, considering going with 50MHz will save $0.15 CDN on a ~$300 project. :D

Interesting point about the 1:1 ratio...
 
Re: Re: Eliminating jitter "completely" - Benchmark DAC1 approach

Kuei Yang Wang said:
I may be forgiven for saying so, but when using semi-stochastic signals, such as music, the ASRC used to implement the DAC1's jitter rejection and "fixed clock" have a demonstrated negative sonic impact.
I'm curious here. Where, how and when was the AD1896 "demonstrated" to affect negatively the sound ?
 
Anything else you wanna add or ask?

I'm asking, because I'm in a market for a new DAC, it needs to be faultless even with jittery sources and can't cost an arm and a leg (well, that's all relative as $800 USD is a quite a chunk of money to me personally).

-----------------------------------------------------------------------------------
There is no dac I have heard that:

1. Is not cable dependent;

2. Has input circuitry that can deal with jittery sources and 'cure' them.

3. Sounds 'perfect' because measurements are 'perfect'.

4. Those with srcs or asrcs all sound individualistic due to resampling or unpasmapling.

5. The best I have heard is dCS stuff with upsampling to 174.4/24.

6. Sounds good with NE5532 opamps for extended periods.

The Benchmark uses NE5532s. For $1000, one would have thought they coulc put in some decent sounding opamps that may not measure as well.

As an eaxample, the Assemblage dejitterer/src with I2S out measures very well but sounds hifi. So does their dacs.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.