Hi Andrea.
Your observations are correct. But there are sonic differences among dacs & dacs.
For example the multibit chip dac like PCM63 is almost identical to PCM1702 but the latter doesnot sound at all as engaging and effortless as the former.
SABRE dac chips are not DA converter chips, they are converter systems, containing a lot of additional circuitry (SRC, MUX, DMUX, various DF's, Reclockers, switches aso).
By the time the clock signal reaches the converters, it's been heavily polluted.
It is slightly dependent on the DA architecture, but you are correct of course.
Well, the conversion process of a Delta Sigma modulator is slightly different and not that much affected by the x-axis precision of the clock flank. But i agree with you that a good clock can improve the sonics of 1-bit dacs as well.
Last edited:
Hi Damian.
I had an Entec DAC, (dont remember which one) and absolutely LOVED it.
Actually i might still have it somewhere!
I still think that for 16/44.1 the CS4328 is tough to bit - but that's just my opinion.
Use other RX chips if you can.
I remember Crystal's own CS8414 as being slightly better than the lowly CS8412.
If I recall correctly, I found the Crystal RX chips as being able to output better timing signals with some attention to the mode selection pins.
However, this is like 20 years ago for me and i don't remember everything
Good luck
Last edited:
Hi again Terry.
I dont know anything about the Transport, only the Forsell DAC, as the Transport was done by some other people.
Sorry to hear you didnt like them. In combination, they produce fantastic sonics.
I recall comparing the Forsell set up in a hifi show with a ML setup (dac+transport) and the Forsell front-end completely blew the ML rig out of the water.
You are right about the DAC: it has no provisions against jitter.
I recall talking to the designer back in 1994, and he told me that he had invested a lot of time & effort in reclockers & didn't really like the results.
I have done some extensive work on finding a jitter-reduction solution for the Forsell DAC unit (I own both Forsell and the later Bremen DAC) and I can reprt that it did improve the sound.
However, the best result was not with the Master clock driving the DAC chip directly and reclocking the data stream, but by placing the clock prior to the TX.
I have no (logical) explanations for this. Maybe TNT Johan has some additional info?
If I'm not mistaken Peter B. (the designer) runs his DAC with an expensive OCXO (Vectron) Mclk located in the transport unit.
I myself have a dedicated Transport for the Forsell Dac, useing a fairly good clock (Crystek - sorry Andrea
) which i am planning to upgrade to an Andrea-Clock soon.All the timing signals are driven from this Mclk, feeding the TX chip (old crystal).
Last edited:
The data sheet I have shows FFT's on P12. Here....
Attachments
Hey, it's no problem, this was many years ago. I was learning and trying to improve my understanding of design differences between various cutting edge Pro and Hi End digital systems. I was working with a former associate at that time who was more knowledgeable than myself WRT digital design but it was clear to both of us that the design approach used by F did not really make total sense from an engineering POV.
It is likely that they did all their development purely via listening and ended up with some kind of sonically pleasing jitter spectrum or synergy between the transport and DAC but from an engineering perspective I didn't really get it. In the Pro sector designers were starting to focus a lot of energy on jitter attenuation in the DAC.
That would not surprise me.
However, the best result was not with the Master clock driving the DAC chip directly and reclocking the data stream, but by placing the clock prior to the TX.
I have no (logical) explanations for this. Maybe TNT Johan has some additional info?
If I'm not mistaken Peter B. (the designer) runs his DAC with an expensive OCXO (Vectron) Mclk located in the transport unit.
I myself have a dedicated Transport for the Forsell Dac, useing a fairly good clock (Crystek - sorry Andrea
All the timing signals are driven from this Mclk, feeding the TX chip (old crystal).
If best results are coming from anything *other than parking the clock
next to DAC then this is indicating something else is not right.
This is and particularly used to be (before USB) a very common scenario in these types of systems and I often wondered if there was some kind of
euphonic jitter thing happening.
I moved on from all this, a $100 USB -> I2S board with isolation and reclocking is a beautiful thing.
TCD
One other solution, however not standard (I think Linn did it) is to feedback an isolated/buffered version of a local DAC clock by wire to the "drive"... why not via opto? This would be a much more simple and elegant solution to the problem. But maybe there is a Fs switch problem that I dot see...?
I cant recall any reclocking designs being made but memory is fading - but a lot of drive and PLL tweaking that did have impact. We where using Sony 50ES CD players modifyed to be drives. Very nice HW.
//
I cant recall any reclocking designs being made but memory is fading
- but a lot of drive and PLL tweaking that did have impact. We where using Sony 50ES CD players modifyed to be drives. Very nice HW.//
Hi Terry.
Lots of interesting information. Thanks for sharing.
In reference to the Forsell combo, I am inclined to believe that you might be on to something - that the combo was indeed designed with a strong focus on achieving a certain, sonic synergy.
I know that dr Forsell himself was a very opinionated, hardcore audiophile and had a very trained pair of ears.
I have tried (almost) everything with the F dac:
reclocking, active reclocking, passive reclocking, sync operation slaving the transport, async operation with 100 MHz mclk, distributed PLL (like Linn), SRC and so on, but the best result was TX --> RX --> I2S --> DAC chip with TX driven by a quality clock.
I still have no idea why....but Everytime I listen to the likes of Naim dacs, Benchmarks, Mark Levinsons, and so on and then switch back to the Forsell/Bremen dac, it just blows my socks off.
Lots of interesting information. Thanks for sharing.
In reference to the Forsell combo, I am inclined to believe that you might be on to something - that the combo was indeed designed with a strong focus on achieving a certain, sonic synergy.
I know that dr Forsell himself was a very opinionated, hardcore audiophile and had a very trained pair of ears.
I have tried (almost) everything with the F dac:
reclocking, active reclocking, passive reclocking, sync operation slaving the transport, async operation with 100 MHz mclk, distributed PLL (like Linn), SRC and so on, but the best result was TX --> RX --> I2S --> DAC chip with TX driven by a quality clock.
I still have no idea why....but Everytime I listen to the likes of Naim dacs, Benchmarks, Mark Levinsons, and so on and then switch back to the Forsell/Bremen dac, it just blows my socks off.
Last edited:
IME best sounding clocking varies according to how many other problems there are with a dac. Same effect with preferred choice of dac filters, depends on other remaining problems. Same effect with some other things too.
Thus people may be right when they say one thing sounds better than another in their particular listening situation. However, whatever is learned in that regard may not necessarily apply to other situations. In other words, the thing that sounds better may not really be better in an absolute sense
Thus people may be right when they say one thing sounds better than another in their particular listening situation. However, whatever is learned in that regard may not necessarily apply to other situations. In other words, the thing that sounds better may not really be better in an absolute sense
Very interesting discussion. I recent example of similar principle might be the transport/DAC combo by the dutch company EC Designs. The clocks are located in the transport, data is sent via their own protocol and electrooptical interface to the DAC unit, where the clock is recovered by the reciever PLL (DIR9001). No reclocking, still their combo sounds very good, by far the best I had in my system.
I intend to use Andrea's clocks to replace the transport clocks that have typical Crystek-level phase noise. I have had some doubts about how much this solution would benefit from the very best oscillators, but you have me thinking it might be more worthwile than I thought.
Andrea-Clocks vs other clocks:
Based on data & graphs submitted by Andrea on his new oscillators' PN, they are - in terms of close-in PN - very close to State Of The Art available today.
I have seen a Wenzel & one BVA that had better close-in performance but they were 5 MHz, and cost as much as a nice apartment in Lithuania.
I am not familiar with EC-Designs Mosaic dacs, but I know John Brown to be an excellent designer.
I am not sure the Master clock in his Mosaic dac has anywhere near the performance of Andrea-Clocks of the same freq.
So your updating the clock in the Mosaic dac is relevant.
I plan to upgrade the Master clock in my Transport I use exclusively with Forsell Air D/A (my Forsell Transport died a few years back due to laser pick-up issues).
In this Transport (SQB Duet) I use a Crystek 11.2896 MHz driving an old Crystal TX, where all timing is Crystek-driven and the result is VERY good indeed, but not as intoxicating as when driven by Forsell's own Transport.
I will use a 11.2896 Andrea-Clock, perform measurements and see what'll happen.
I would recommend doing some J-Dunn test on your set-up before & after clock modifications.
---------------------------------
May I ask where in Sweden you are located? You can send me a pm if you like.
Based on data & graphs submitted by Andrea on his new oscillators' PN, they are - in terms of close-in PN - very close to State Of The Art available today.
I have seen a Wenzel & one BVA that had better close-in performance but they were 5 MHz, and cost as much as a nice apartment in Lithuania.
I am not familiar with EC-Designs Mosaic dacs, but I know John Brown to be an excellent designer.
I am not sure the Master clock in his Mosaic dac has anywhere near the performance of Andrea-Clocks of the same freq.
So your updating the clock in the Mosaic dac is relevant.
I plan to upgrade the Master clock in my Transport I use exclusively with Forsell Air D/A (my Forsell Transport died a few years back due to laser pick-up issues).
In this Transport (SQB Duet) I use a Crystek 11.2896 MHz driving an old Crystal TX, where all timing is Crystek-driven and the result is VERY good indeed, but not as intoxicating as when driven by Forsell's own Transport.
I will use a 11.2896 Andrea-Clock, perform measurements and see what'll happen.
I would recommend doing some J-Dunn test on your set-up before & after clock modifications.
---------------------------------
May I ask where in Sweden you are located? You can send me a pm if you like.
Last edited:
Educated guesswork: unless the designers of the SABRE DAC chip are incredibly stupid, they will have made sure that the clock path to the DAC is as short and low in jitter as they could make it. The digital stuff doesn't need a low-jitter clock and presumably got a much longer clock path (probably with dozens of clock buffers inserted by an automatic placement, routing and timing analysis tool).
I don't agree with your claim that the clock flank precision matters less. Close-in phase noise just phase modulates the DAC output signal no matter what the DAC topology is. This phase modulation causes sidebands around the desired signal.
Far-off phase noise can be more critical in a sigma-delta DAC than in a multibit DAC because of all the out-of-band quantization noise that can be partly converted into the audio band when there is far-off phase noise on the clock. It also depends a lot on the exact type of sigma-delta DAC: for example, switched capacitor DACs are less sensitive to period jitter/far-off phase noise than NRZ or RTZ DACs that just switch a current on for one clock period or half a clock period. FIRDACs are less sensitive to far-off phase noise than single DAC stage.
And as shown in post 2850, https://www.diyaudio.com/forums/dig...itter-crystal-oscillator-285.html#post6373471 , some types of sigma-delta DAC are particularly sensitive to second subharmonic in the clock signal.
Measuring the effects of clock upgrade on Delta-Sigma and Multibit dacs:
Hi Marcel and thanks for the info:
Let me think a little bit more about what you wrote, and try to digest it. It makes sense, as upgrading the Mclk improved the perceived sound quality of many DS dac chips I have tested.
Although the J-Dunn test cannot differentiate between side-bands generated due to jitter in the d/a chip, or by the proceeding stages after the chip, I still find it to be a useful indicator of general problems in separate Transports/dac set-ups.
Please (if you have the possibility) perform this test before & after Andrea-Clock upgrade.
Also remember that the newer D-S dacs from TI, Cirrus (Crystal) or AKM are Multibit-Delta Sigma converters & not exactly one-bit devices anymore.
The TDA1547 or CS4328 are pure, one-bit, devices (correct me if I'm wrong).
CS4328 as opposed to TDA1547, features a very interesting d/a converter that in theory should be less sensitive to MCK jitter.
Some have stated that the excellent sonics of CS4328 is due to the output amplifier designed to sound euphonic.
I'm not sure I agree, as the output "amp" in that chip appears to be the usual mclk-scalable SCF, ie: nothing out of the ordinary.
Updating the Mclk (sorry Bclk) on my TDA1541a based dac some years ago MASSIVELY transformed the sound.
Going NOS was another eye-opener.
Then the signal conditioning ala Henk ten Pierick took the sonics to another level.
The dac now sounded so smooth and articulate, but a little less precise and less "airy" than the best DS dacs.
So I think (just my opinion) updating the clock in Multibit dacs appear to be more rewarding than in some DS dacs.
Cheers
Hi Marcel and thanks for the info:
Let me think a little bit more about what you wrote, and try to digest it. It makes sense, as upgrading the Mclk improved the perceived sound quality of many DS dac chips I have tested.
Although the J-Dunn test cannot differentiate between side-bands generated due to jitter in the d/a chip, or by the proceeding stages after the chip, I still find it to be a useful indicator of general problems in separate Transports/dac set-ups.
Please (if you have the possibility) perform this test before & after Andrea-Clock upgrade.
Also remember that the newer D-S dacs from TI, Cirrus (Crystal) or AKM are Multibit-Delta Sigma converters & not exactly one-bit devices anymore.
The TDA1547 or CS4328 are pure, one-bit, devices (correct me if I'm wrong).
CS4328 as opposed to TDA1547, features a very interesting d/a converter that in theory should be less sensitive to MCK jitter.
Some have stated that the excellent sonics of CS4328 is due to the output amplifier designed to sound euphonic.
I'm not sure I agree, as the output "amp" in that chip appears to be the usual mclk-scalable SCF, ie: nothing out of the ordinary.
Updating the Mclk (sorry Bclk) on my TDA1541a based dac some years ago MASSIVELY transformed the sound.
Going NOS was another eye-opener.
Then the signal conditioning ala Henk ten Pierick took the sonics to another level.
The dac now sounded so smooth and articulate, but a little less precise and less "airy" than the best DS dacs.
So I think (just my opinion) updating the clock in Multibit dacs appear to be more rewarding than in some DS dacs.
Cheers
Last edited:
One other solution, however not standard (I think Linn did it) is to feedback an isolated/buffered version of a local DAC clock by wire to the "drive"... why not via opto? This would be a much more simple and elegant solution to the problem. But maybe there is a Fs switch problem that I dot see...?
I cant recall any reclocking designs being made but memory is fading
//
I believe Spectral did this in one variation of their DAC/transport system. But that was long ago. Not that hard to implement, except fiber optic may introduce enough latency or jitter to cause clock errors.
On the matching transport I built using the CEC belt drive I could hear the disk rubbing in the transport on some occasions but no degradation on the audio. Left me wondering just how important the belt drive stuff was.
Probably the J-Test SPDIF jitter test, developed by Julian Dunn: http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.68.2092&rep=rep1&type=pdf
Yes Andrea's oscillators have much much lower PN than the low cost oscillators ECD use in the transports. My doubt was that the DAC combo would not benefit so much from a SOTA oscillator since I thought the receiver chip would add too much PN on its own. I will PM you some info.
Thanks, but how does measuring how well the jitter of an AES3 or S/PDIF interface is suppressed help in assessing the effect of the quality of a master clock?
Edit: I see that my question was answered in the thread that post 2919 links to: it doesn't.
Last edited:
- Status
- Not open for further replies.