Given that You are making a "re-thinking" here, here it is a post that was doing it very much in your style in it's time, from Wildmonkeysects:
http://www.diyaudio.com/forums/digi...-reference-dac-8-channel-101.html#post1617282
http://www.diyaudio.com/forums/digi...-reference-dac-8-channel-101.html#post1617282
OK, is very possible that many peoples think in the same ways, in different places, in different time... I didn`knew about this post from 2008. Is a little bit difficult to read all in this forum... But anyway, I did like it was described in my posts, and I have the good result/improvements... I apply dis method since quite long time in another cases. I do think is a good way until will come out a better one...
I can not say that using shunt regulator is wrong, or lead to not so good results. But this one it could be a simple way with good result too... So, everyone choose at last what one want, and like, and think is better.
RE...
Sure it is all good, and experimentation is to be applauded and encouraged. I do think that it is important to point out, that incorporating the Trident regulators with the B-II/III does provide low impedance, isolated, supplies for all the relevant portions of the DAC: two for the analog sections with matched voltage too keep outputs balanced, 1 for the clock, and two for the: core and digital side of the chip-so with the Tridents, the goal of separate low impedance supplies to the different sections of the chip is met.
Coris, do the tants maintain their low z at all frequencies? My understanding is that the use of local shunts is beneficial because of the bandwidth over which low z is maintained.
One thing I am looking forward to trying is to clock the 9018 synchronously from my USB-I2S interface, defeating the ASRC and DPLL: my experience with other DACs is that the ASRCs and DPLLs usually result in some amount of unnatural sound. Doing so would disable some of the "jitter rejection", but I am planning on upgrading the clocks on my interface to Crystek CCHD series, so source jitter should be very low to begin with (async, XMOS interface).
I would love to try a low impedance in tube I/V for the B-II/III at some point, but I have no ability to design such a stage...
Sure it is all good, and experimentation is to be applauded and encouraged. I do think that it is important to point out, that incorporating the Trident regulators with the B-II/III does provide low impedance, isolated, supplies for all the relevant portions of the DAC: two for the analog sections with matched voltage too keep outputs balanced, 1 for the clock, and two for the: core and digital side of the chip-so with the Tridents, the goal of separate low impedance supplies to the different sections of the chip is met.
Coris, do the tants maintain their low z at all frequencies? My understanding is that the use of local shunts is beneficial because of the bandwidth over which low z is maintained.
One thing I am looking forward to trying is to clock the 9018 synchronously from my USB-I2S interface, defeating the ASRC and DPLL: my experience with other DACs is that the ASRCs and DPLLs usually result in some amount of unnatural sound. Doing so would disable some of the "jitter rejection", but I am planning on upgrading the clocks on my interface to Crystek CCHD series, so source jitter should be very low to begin with (async, XMOS interface).
I would love to try a low impedance in tube I/V for the B-II/III at some point, but I have no ability to design such a stage...
Yes, is right that a tantalum can have problems with the z and a large frequency spectrum. But the ceramics can do it better... Now is not exactly right to think about a z in "all frequencies", but a large enough spectrum I think is possible to cover with a right (type) cap...
I`ve already noticed that is a difference in using one or another of those types in different places (stages). This can very well be in connection with this impedance over a large (larger) bandwidth. In my arrangement of decoupling, I will not want to use tantalum for bypassing ES9018 (f. ex.), but I will use tantalum in decoupling the final (op amp) stage. Or a combination...
I`ve noticed that is a different sound (not exactly the best) if one do not customize quite accurate this bypassing on every single stage... But here is important too using the right regulator, and not at least a dedicated regulator. Using a bypass cap type not in the right place can have negative consequences in the resulting sound.
You right, the meaning of a shunt regulator is to keep low impedance in a quite large bandwidth. Experimenting in this field is not yet over for sure...
What do you think about to use an oven controlled oscillator? A such oscillator is possible to get it today for a quite reasonable price. I did... But I had not the time to try it yet. I intend to use it on Buffalo. A such device is very high quality when is about jitter, and so. Now my system works on a 0,5ppm/125Mhz clock and I`m very satisfied so far. The next step is to use an 122,88Mhz. I`m waiting it. It will be interesting for me after those last experiences, to go back to an 100Mhz, but with an OCXO...
Else, is of course a theme too about synchronous clock. Is on my list, but the time is the king so far...
Last edited:
Such devices as OCXOs are made and meant to be used in extremely stable systems, where phase noise levels and frequency stability is much more sensitive/important than in audio field...
The only problem with those type oscillators is that are quite expensive...
If you intend to state that a Crystek which costs 30$ or less is better than an f.ex. Abracon OCXO which is happen to costs more than 200$, in its phase noise level or/and about stability, than one have to doubt seriously about such logic...
(I`ve used here different companies names only by chance, but both those producers have all kind of oscillators for sale, cheap or expensive...)
Last edited:
Coris:
Please provide links to phase noise plots for the Abracon OCXO. Then we can compare with the (published by Crystek) phase noise plots for the Crystek CCHD Series parts at low frequencies (10 Hz). I would like to learn if very expensive OCXOs actually have real advantages for audio use. As qusp mentions, long term stability (ppm specs) are not really very relevant for audio performance. Just because the OCXO costs more does not mean it will produce better audio performance: but if it does prove to have significantly lower phase noise, it is fair to conclude that the OCXO will offer advantages that may be worth the costs for some.
Please provide links to phase noise plots for the Abracon OCXO. Then we can compare with the (published by Crystek) phase noise plots for the Crystek CCHD Series parts at low frequencies (10 Hz). I would like to learn if very expensive OCXOs actually have real advantages for audio use. As qusp mentions, long term stability (ppm specs) are not really very relevant for audio performance. Just because the OCXO costs more does not mean it will produce better audio performance: but if it does prove to have significantly lower phase noise, it is fair to conclude that the OCXO will offer advantages that may be worth the costs for some.
This is not Abracon, but a real good one. It runs loops around a stupid little CCHD 957 (-97db /10Hz) with it's -130db/-132db at 10Hz...
And it's not even the top of the line at Wenzel.
Only setback is that it never will be available for audio frequencies..
http://www.wenzel.com/pdffiles1/Oscillators/ULN_4_to_30.pdf
I only report it here to show what real good crystal performance means.
Ciao, George
Ok, for example this is more real, and probably not too bad:
http://depnerlabs.hu/wp-content/uploads/2011/07/oven_clock_datasheet2.pdf
Ciao, George
http://depnerlabs.hu/wp-content/uploads/2011/07/oven_clock_datasheet2.pdf
Ciao, George
Thanks again
George. It looks like the Depner is an entire clock/power supply which must be applied externally. I wonder how much additional phase noise will be caused by having to distribute the clock signal?
Nice thing about the Crystek CCHD-957/950 is that they can be placed directly adjacent to the DAC, and with good board layout, should not degrade much.
These (CCHD Series) clocks are still looking pretty good to me from a practical standpoint, consider the ability to implement them in existing products/boards in mostly non-compromising ways. In direct comparision to the Depner, the phase noise is pretty close, and considering the degradation which may be caused by distribution, in practice, the Crystek may perform equally as well.
Of course, we muct consider the power supply as well...
George. It looks like the Depner is an entire clock/power supply which must be applied externally. I wonder how much additional phase noise will be caused by having to distribute the clock signal?
Nice thing about the Crystek CCHD-957/950 is that they can be placed directly adjacent to the DAC, and with good board layout, should not degrade much.
These (CCHD Series) clocks are still looking pretty good to me from a practical standpoint, consider the ability to implement them in existing products/boards in mostly non-compromising ways. In direct comparision to the Depner, the phase noise is pretty close, and considering the degradation which may be caused by distribution, in practice, the Crystek may perform equally as well.
Of course, we muct consider the power supply as well...
TAD-D600 CD player
???????TAD-D600?TAD (in Japanese)
uses this oscillator. This is not OCXO. However, its technology is of OCXO and manufactured by NDK. (TAD people thought the need of "warm-up" in OCXO does not satisfy a requirement in a consumer product.)
Phase noise plot of the oscillator;
???????TAD-D600?TAD (in Japanese)
uses this oscillator. This is not OCXO. However, its technology is of OCXO and manufactured by NDK. (TAD people thought the need of "warm-up" in OCXO does not satisfy a requirement in a consumer product.)
Phase noise plot of the oscillator;
OK guys. I see that it started up a little discussion here about OCXO. I feel the need to come with some clarifications about my previous post on the same theme.
First I have to repeat that the producer names of those OXCO I used in that post was quite by chance. Just those names came in my mind at that moment... I will not say that one is better than another. It may be, it may be not... I did not made a close research about yet.... Else both companies produce very good and very expensive oscillators, as very standard ones...
When I think at an OCXO I have in mind an 100Mhz oscillator. As I know, is very hard to find an OCXO usable directly in audio. I mean about the standard clock audio frequencies. As I know, and got some informations about, those OCXO are made mainly for radio range control frequencies. That because the most known and sold OCXO frequencies have quite strange range for audio domain. Is quite by chance that an 100Mhz OCXO can meet today the audio field by clocking this ES9018. So, the phase noise is to talk about at this frequency. This I meant and had in my mind when I`ve posted may previous post. A phase noise level at this frequency is quite good, and have to be very good at an 100Mhz OCXO.
Maybe I`m wrong, but I have this logic in this case: a jitter/phase noise is directly connected to the frequency stability of that clock. An standard oscillator (for say 100Mhz) with a 100ppm stability, have to have quite big phase noise level. Another one with an 1ppm have to have a better phase noise/jitter. An 100Mhz OCXO clock with 50ppb (f. ex.) have to have a much lower phase noise. I actually think that is an very close correlation in between this two parameters of an oscillator: stability and phase noise/jitter. I in fact just noticed in few cases that one face a much lower jitter when is about less ppm, or in ppb range.
If I`m wrong, please correct me...
First I have to repeat that the producer names of those OXCO I used in that post was quite by chance. Just those names came in my mind at that moment... I will not say that one is better than another. It may be, it may be not... I did not made a close research about yet.... Else both companies produce very good and very expensive oscillators, as very standard ones...
When I think at an OCXO I have in mind an 100Mhz oscillator. As I know, is very hard to find an OCXO usable directly in audio. I mean about the standard clock audio frequencies. As I know, and got some informations about, those OCXO are made mainly for radio range control frequencies. That because the most known and sold OCXO frequencies have quite strange range for audio domain. Is quite by chance that an 100Mhz OCXO can meet today the audio field by clocking this ES9018. So, the phase noise is to talk about at this frequency. This I meant and had in my mind when I`ve posted may previous post. A phase noise level at this frequency is quite good, and have to be very good at an 100Mhz OCXO.
Maybe I`m wrong, but I have this logic in this case: a jitter/phase noise is directly connected to the frequency stability of that clock. An standard oscillator (for say 100Mhz) with a 100ppm stability, have to have quite big phase noise level. Another one with an 1ppm have to have a better phase noise/jitter. An 100Mhz OCXO clock with 50ppb (f. ex.) have to have a much lower phase noise. I actually think that is an very close correlation in between this two parameters of an oscillator: stability and phase noise/jitter. I in fact just noticed in few cases that one face a much lower jitter when is about less ppm, or in ppb range.
If I`m wrong, please correct me...
Last edited:
i don't know how, but you missed the point. they are built to be stable averaged over long time intervals, incredibly stable even, but that is of no real relevance to audio clock quality. average accuracy over long periods and a second could be considered long here does not automatically translate to high accuracy over very short periods. then we have the practical concerns, especially as most are physically large, requiring less than optimal routing and grounding of the clock output and many(most) require some sort of secondary clock division to be useable.
being expensive is meaningless, it can be very good for its intended purpose, but if that purpose is not audio it could very well be less ideal than a 10 dollar clock specified for audio
Last edited:
Quite right qusp. In fact, there are 'in principle' even better XO's around that lock onto the GPS network. Nevertheless, probably the crystal they put into the OCXO's is of good quality considering the cost of the external constant thermal 24/7 correction system. I have a suspicion that an XO package actually qualifies as an OCXO simply because of its thermal conduction characteristics.....
I have an OCXO of 100 MHz sine wave output at my hand. If you want to evaluate it, I can send it to you.
The model is 9325D(Fixed Communication)/Oven-Controlled Crystal Oscillator (OCXO)/NDK
Its list price is approximately 1,000 Euro. Phase noise measurement chart for the individual device is available.
As the output is sine wave of +-0.9 V(1.8 V p-p) amplitude, I applied an adjustable DC bias using batteries and fed the biased sine output directly to XI pin of ES9018. The idea of the direct input was given by Dustin.
You need +12V power for the device.
I felt the sonic result of this OCXO was definitely better than the standard Crystek oscillator used on Buffalo II. However, the cost performance was not so high enough. After I once knew the superior performance of synchronous master clocking scheme, I decommissioned it.
Coris, Bunpei, et al
I am more interested in synchronous clocking of the ESS 9018, hence my desire for more standard (audio) clock frequencies.
Bunpei: it is nice to hear that you find synchronous clocking to produce better sound than using the ASRC/DPLL-I assume this is as long as the master clock/bit clock provided are very low jitter. I am looking forward to trying synchronous clocking when I get my hands on the CCHD-957 oscillators.
The NDK which TAD is using-do you know if that is a custom part, or is there a series available from them with that level of performance, and if so, can they be purchased in single unit quantities somewhere?
I am more interested in synchronous clocking of the ESS 9018, hence my desire for more standard (audio) clock frequencies.
Bunpei: it is nice to hear that you find synchronous clocking to produce better sound than using the ASRC/DPLL-I assume this is as long as the master clock/bit clock provided are very low jitter. I am looking forward to trying synchronous clocking when I get my hands on the CCHD-957 oscillators.
The NDK which TAD is using-do you know if that is a custom part, or is there a series available from them with that level of performance, and if so, can they be purchased in single unit quantities somewhere?