Inside the latest Benchmark DAC1

[Iron_Dreamer]

Quote:

Originally Posted by jsiau Don't underestimate the NE5532. We have just completed a very comprehensive set of tests on the NE5532 and on various substitutes. We have not found anything that will equal the tranparency provided by the NE5532. Some have suggested susbstituting OPA2134 op amps in place of the 5532s. This is a bad idea! Our tests show an increase in both 2nd and 3rd harmonic distortion, and the addition of higher order (4th, 5th, 6th, and 7th) harmonics that are virtually absent from a stock DAC1. In addition, IMD will increase, and SNR will degrade. The NE5532 is power hungry, it has high input bias currents, and high offset voltage, but it can drive high-level low impedance circuits with ease. The 5532 should not be used with low signal levels, and it should only be used in low gain circuits. Also, the offset voltage must be managed with appropriate design techniques. I believe the NE5532 has aquired a bad reputation because it has often been missapplied. The DAC1 is carefully designed specifically for the NE5532 op amp. It may surprise you that the NE5532 was selected for transparency and not on the basis of cost. John Siau Director of Engineering Benchmark Media Systems, Inc. www.benchmarkmedia.com

Well that is what I was hoping. I like the sound so much, I think the problem I am having now, with fatigue, is from the CD3000, so I will change it before I change the DAC. I guess it's HD650 here I come.

 

[aos]

SRC4192 indeed doesn't list the jitter attenuation curve unlike AD1896 (which incidentally looks very similar to an asynchronous reclocker). Since ports can be set as slaves, you can obviously eliminate any effect of clock jitter by using a low jitter oscillator since the output clock will be that same clock (you're supplying it, not the unit). However there are still effects of data jitter and the question is what the SRC4192 does with it (does it make its way into re-calculated, output data samples). Unlike AD1896, which theory of operation I've read in detail, I didn't see the same section in SRC4192 so we can't look from theoretical perspective to see what it does to jitter. But yeah, if it doesn't attenuate, I am surprised that Analog devices is not advertising that fact everywhere, since I can't imagine they would allow to lose business based on misleading data from a competitor.

It's true that 5532 are even today specified in many datasheets, including Burr-Brown's top of the line DAC. They are being recommended in order to preserve SNR and dynamic range (according to datasheets). These chips do measure great so from that perspective they are still the top dog. However I really can't recall anybody who didn't replace one of them in just about anything and wasn't rewarded by "better" sound. On the other hand it's being guessed that RA-1 uses it and lots of people love its sound. I would certainly experiment with opamp were it my own device, but given the ease of damaging the board I'd certainly not recommend it for anyone else, especially if you like the sound.

 

[jsiau]

Quote:

Originally Posted by ampgalore Wow, representative from Benchmark is here While you are here, any scoops on future upgrades to the DAC1?

I would suggest subscribing to our "e-update" newsletter at www.benchmarkmedia.com
Our subscibers receive immediate notification when new products are released.

 

[Glassman]

jsiau, I hope you'll stay here long long time, we really need people like you! this is truly wonderful place as you'll find out for youself soon

why are you running the DAC at non-audio-standard clock? is that to prevent resampling to very close rates? like receiving 96.01kHz and resampling to 191.99kHz (just an illustration).. don't you think a good pll with long lock without async upsampler would be better for the transparency? what's your opinion on the new super fast opamps?


aos, RA-1 uses NJM4556 found on front channels of Audigy2 cards too btw

 

[ampgalore]

One suggestion for the DAC1 is the unbalanced analog out RCA jacks. You can space them out a bit more. With RCA cables using large RCA connectors, it gets very tight.

 

[pbirkett]

Quote:

Originally Posted by Iron_Dreamer Well that is what I was hoping. I like the sound so much, I think the problem I am having now, with fatigue, is from the CD3000, so I will change it before I change the DAC. I guess it's HD650 here I come.

Or you could stick with the DT531, its magical

(sorry ampgalore, I really couldnt resist)

 

[aos]

Analog's chip operates by (conceptually) interpollating at very high frequency and then measuring what sample to take with about 5ps accuracy. So that is the limit of jitter. However, the digital servo loop that is used for that decision making needs to attenuate the clocks, otherwise jitter from those clocks will influence that decision. Analog provides the specs of that filter and one can see that it starts attenuating even below audible range. If that's what BB screwed up, changing filter coefficients should be enough to fix it (perhaps they've set it at 10-20kHz like they did for their DIR).

Isn't NJM4556 just a variant of 5532? By the way, why not use 5534, which is the improved version of 5532? I believe I replaced some in my Sony at some time.

 

[Iron_Dreamer]

Quote:

Originally Posted by pbirkett Or you could stick with the DT531, its magical (sorry ampgalore, I really couldnt resist)

LOL

As much as I like the DT531 (they have an especially nice synergy with my Archos), with this great source, their flaws and drawbacks are a bit more apparent, and I think I'd prefer some more resolution, extension, and power in some of the audible ranges (i.e. the lower bass is a bit weak, as is a certain aprt of the mids, which makes some drums sound a bit hollow). But I don't think I'm selling them immediately, since they are so nice, and hard to find, I want to make sure I have something better, and I have no use for the DT531 before I pass them on to another.

 

[pbirkett]

It was actually a joke mate

Hehe!

 

[jsiau]

Quote:

Originally Posted by Glassman why are you running the DAC at non-audio-standard clock? is that to prevent resampling to very close rates? like receiving 96.01kHz and resampling to 191.99kHz (just an illustration).. don't you think a good pll with long lock without async upsampler would be better for the transparency? what's your opinion on the new super fast opamps? aos, RA-1 uses NJM4556 found on front channels of Audigy2 cards too btw

We are frequency shifting the low-pass filters in the AD1853. This eliminates images (a form of aliasing) in the analog output of the DAC1. Virtually all DACs suffer from poor image suppression when the input frequency of the input audio is just below the Nyquist frequency. The DAC1 does not have this problem due to the (2^20)-1 interpolation ratio of the AD1896, and the frequncy shifting of the AD1853 low-pass filters. We are essentially replacing a 64 Fs interpolator with a 1048575 Fs interpolator.

BTW the AD1896 does not seem to be bothered by near integer conversion ratios.

John Siau
Director of Engineering
Benchmark Media Systems, Inc.
www.benchmarkmedia.com

 

[AndreYew]

John,

Welcome to Head-Fi! Ummm ... sorry about your wallet?

Could you explain why the DAC1's unbalanced outputs have such high impedance? This would seem to exacerbate the problems with unbalanced interconnects in the first place.

--Andre

 

[jsiau]

Quote:

Originally Posted by aos Analog's chip operates by (conceptually) interpollating at very high frequency and then measuring what sample to take with about 5ps accuracy. So that is the limit of jitter. However, the digital servo loop that is used for that decision making needs to attenuate the clocks, otherwise jitter from those clocks will influence that decision. Analog provides the specs of that filter and one can see that it starts attenuating even below audible range. If that's what BB screwed up, changing filter coefficients should be enough to fix it (perhaps they've set it at 10-20kHz like they did for their DIR).

Bingo! If I remeber corectly, they chose 5 kHz. Way too high! The corner frequency needs to be below 20 Hz, and preferably less than 1 Hz.

The existing part is cast in concrete. Too expensive to fix the problem (or too many parts already built).

 

[jsiau]

Quote:

Originally Posted by AndreYew John, Welcome to Head-Fi! Ummm ... sorry about your wallet? Could you explain why the DAC1's unbalanced outputs have such high impedance? This would seem to exacerbate the problems with unbalanced interconnects in the first place. --Andre

You are correct. As explained in the manual, we set them high enough to tolerate "Y" cord mono sums, and the inevitable misswiring of two outputs to each other. One of the design goals of the DAC1 was that it must tolerate a short on one ore two outputs without degrading the performance of other outputs.

The manual discusses these issues, and also suggests maximum cable lengths as a function of output impedance and cable capacitance.

BTW you don't have to be a DAC1 owner to download the 39 page manual. Lots of useful tidbits.

 

[aos]

I am sure glad the last thing I designed uses socketed ASRC so swapping between AD1896 and SRC4192 will be easier. While easy to fix, it will be a costly mistake for BB (retooling). Now I'm intrigued to try NE5532, I know I have a pair somewhere.

 

[AndreYew]

Thanks John. I downloaded the manual shortly after posting my question, and it's a fine manual with much more technical detail than one would expect. Would it be possible to make this an optional setting in future products, perhaps via a jumper? Default setting would be high Z so it acts the same, but people who primarily use unbalanced can open up the DAC1, and reset the jumper?

--Andre

edit: the manual says consumer outputs are around 15k. Most of my consumer audio gear has output Z much less than 100 Ohms, with most hovering from 50 to 75.