Yes, that is what data sheet is indicating.
But exactly what happens to DSD data when it gets to OP DAC prior to and
during conversion can be varied. Generally the 1 bit data is circulated through the array of OP 1 bit DACs to form an averaging low pass filter, but there are varied ways to implement this.
I'm guessing they are also using some form of RTZ coding or an equivalent scheme (algorithm) to reduce ISI distortion as the -124dB THD spec holds for DSD (which is pretty amazing).
The take away from all of this though is that regardless of above it will still retain the transient characteristics of DSD with no ringing but reduce OOB noise significantly.
I don't think it is remodulated, see above. It would be great to get a better
understanding of *exactly how they do all this but obviously that's proprietary.
TCD
question on oversampling for Markw4
hi Mark,
I've been following this thread as well as others where you made some positive comments on oversampling/resampling files from standard CD resolution up to higher sample rates. Have you tried playing with any of the resamplers available for Foobar ? I've been playing around with higher sample rates in
Foobar using a recently completed AK4493 dac, there is a definite change in the presentation and sound, which seems different depending which of the filters is being used. For example using the super slow filter and upsampling to 352.8 seems to increase the upper frequency detail and clarity but moves the sound stage forward. doing the same with the brickwall filter seems to warm the mids a touch and increase the overall sense of clarity but it's not as noticeable as with the super slow. If you had used Foobar how did you find it compares to the software that you have been using.
Thanks,
Paul
hi Mark,
I've been following this thread as well as others where you made some positive comments on oversampling/resampling files from standard CD resolution up to higher sample rates. Have you tried playing with any of the resamplers available for Foobar ? I've been playing around with higher sample rates in
Foobar using a recently completed AK4493 dac, there is a definite change in the presentation and sound, which seems different depending which of the filters is being used. For example using the super slow filter and upsampling to 352.8 seems to increase the upper frequency detail and clarity but moves the sound stage forward. doing the same with the brickwall filter seems to warm the mids a touch and increase the overall sense of clarity but it's not as noticeable as with the super slow. If you had used Foobar how did you find it compares to the software that you have been using.
Thanks,
Paul
Paul,
Sound perceptually moving more forward sounds like a problem with filter quality. That effect seems to be caused by some low level distortion, IMO. 'Slow' filters often tend to roll off frequency response at the upper end of the audio band.
Haven't tried foobar myself for that particular use. Here, its mostly use AK4137 for hardware upsampling and or DSD conversion, or use HQ Player if in software.
If AK4137 has good, low-jitter clocks and good power supply regulation it could be very good with your dac.
Mark
Sound perceptually moving more forward sounds like a problem with filter quality. That effect seems to be caused by some low level distortion, IMO. 'Slow' filters often tend to roll off frequency response at the upper end of the audio band.
Haven't tried foobar myself for that particular use. Here, its mostly use AK4137 for hardware upsampling and or DSD conversion, or use HQ Player if in software.
If AK4137 has good, low-jitter clocks and good power supply regulation it could be very good with your dac.
Mark
Hi, I am also using the Allo USBridge Sig, yet into a SingXer SU-6 who's clocks and output are very accurate and clean. I feel (just feel) that the Topping D90 with its HDMI might just be a fantastic combination - I am struggling to find out if I can feel a direct signal from the SU-6 signal to the DAC (direct) as the SU-6 has better clocks etc. This might be a giant killer combo? Any thoughts?
Don't know why make that assumption. Many people seem to prefer the sound of NDK SDA clocks over that of Crystek.
Interested in 4499
I know a guy in the industry who regularly measures the phase noise of clocks. While chatting with him at RMAF this past fall, he mentioned that the NDK SDA series clocks consistently measure better for close in phase noise than Crystek's best ones. Of course, even better performance can be had if one sorts the NDKs. At their price, it would not be unreasonable if one used only the top 10% for really high end DAC builds. There is enough variation in these clocks, that the top 10% can be better by 3-5 dB or so at 2 Hz.
I am very interested to learn if the the direct DSD mode of the AKM 4499 is not going through a DSM modulator? If it is going straight to the switched resistor array, I guess we could consider each element as a separate bit, but the signal is not re-modulated to multi-bit: it would be a single bit stream going to the multi elements, kind of like a DSC-1 on a chip?
I know a guy in the industry who regularly measures the phase noise of clocks. While chatting with him at RMAF this past fall, he mentioned that the NDK SDA series clocks consistently measure better for close in phase noise than Crystek's best ones. Of course, even better performance can be had if one sorts the NDKs. At their price, it would not be unreasonable if one used only the top 10% for really high end DAC builds. There is enough variation in these clocks, that the top 10% can be better by 3-5 dB or so at 2 Hz.
I am very interested to learn if the the direct DSD mode of the AKM 4499 is not going through a DSM modulator? If it is going straight to the switched resistor array, I guess we could consider each element as a separate bit, but the signal is not re-modulated to multi-bit: it would be a single bit stream going to the multi elements, kind of like a DSC-1 on a chip?
"I am very interested to learn if the the direct DSD mode of the AKM 4499 is not going through a DSM modulator?"
Quoting myself above. I looked at the data sheet and it seems pretty clear that direct DSD bypasses the DS modulator in the 4499, good news to me. As I would prefer having the only modulator being run in software than on the chip.
Quoting myself above. I looked at the data sheet and it seems pretty clear that direct DSD bypasses the DS modulator in the 4499, good news to me. As I would prefer having the only modulator being run in software than on the chip.
There used to be folks selling tested and sorted NDK clocks. It was very time consuming, recall reading on their website it took two hours to scan one clock. Scan time takes longer at lower offset frequencies, IIRC. Also, the equipment required can be quite costly, JohnW said he has a rack of equipment to do it which would cost about $50,000 to buy today. More info at: http://www.ke5fx.com/phase_noise.pdf
The other thing I would say about NDK SDA verses Crystek's best is that they seem to differ audibly in their strengths and weaknesses. Neither one consistently sounds better to me overall. Could be jitter is not just an important factor at close in offsets. It turns out there are different jitter mechanisms that have been identified in different offset frequency ranges. More info at: https://www.crystek.com/documents/appnotes/ImpactUltralow.pdf ...there are some references at the end which in turn lead to other references going back many years. Too bad its so costly to measure the noise in detail. Here is an older publication: https://www.am1.us/wp-content/uploads/Documents/U11613_Phase_Noise_Measurement_Art-Scherer.pdf
Last edited:
yes.
I actually work for a company which has purchased sorted NDK clocks in the past, and am fairly aware of the process, the equipment required to make close in phase noise measurements, and who can provide this service.
I have also participated in listening tests which correlated listening results directly with phase noise plots of various XOs, those tests left me with the overall impression that the lower the close in phase noise, the better.
It is still possible for manufacturers to purchase sorted clocks, or they could purchase the necessary gear and do it themselves. Although few high end audio companies may have the engineering budget for such purchases.
Or, they could develop their own XOs as Ayre has done in partnership with Morion, they use a proprietary, SC cut crystal (typically used in an OCXO) without the oven.
Of course just having a good XO does not guarantee good performance at the point where it matters (the clock pin on the DAC chip, and re-clock flip flop if used). There are a myriad details which all matter in the end: PCB layout, clock power supplies and decoupling elements, RF levels in the vicinity of clock PCB traces.
Suffice it to say, for me, i like the NDK SDA series. One can do a little better for a LOT more money, but the increase in cost basically skyrockets with some of the ovenized, SC cut crystals.
I actually work for a company which has purchased sorted NDK clocks in the past, and am fairly aware of the process, the equipment required to make close in phase noise measurements, and who can provide this service.
I have also participated in listening tests which correlated listening results directly with phase noise plots of various XOs, those tests left me with the overall impression that the lower the close in phase noise, the better.
It is still possible for manufacturers to purchase sorted clocks, or they could purchase the necessary gear and do it themselves. Although few high end audio companies may have the engineering budget for such purchases.
Or, they could develop their own XOs as Ayre has done in partnership with Morion, they use a proprietary, SC cut crystal (typically used in an OCXO) without the oven.
Of course just having a good XO does not guarantee good performance at the point where it matters (the clock pin on the DAC chip, and re-clock flip flop if used). There are a myriad details which all matter in the end: PCB layout, clock power supplies and decoupling elements, RF levels in the vicinity of clock PCB traces.
Suffice it to say, for me, i like the NDK SDA series. One can do a little better for a LOT more money, but the increase in cost basically skyrockets with some of the ovenized, SC cut crystals.
Where it matters being the audible sound quality, if it doesn't make audible difference, there is no point spending more money.
Right...
And, as mentioned, i was involved in listening tests where we correlated lower levels of close in phase noise (measured) with a better subjective listening experience. Such that every time we were able to lower the close in phase noise, the listening experience improved.
My take away is that it is worth it, within limitations based on cost, to use the XO with the lowest close in phase noise one can afford to specify. Too bad the Pulsar clocks are no longer available, as they offered quite good performance at fairly reasonable pricing (compared to say, NDK DuColon). It appears that the company which manufactured the Pulsar Clocks does not make audio rate clocks as part of their normal offerings, so only if one is able to special order custom rates in quantities would appear possible to get these. The next best thing would be NDK SDA series, hand sorted if one can withstand the extra expense.
And, as mentioned, i was involved in listening tests where we correlated lower levels of close in phase noise (measured) with a better subjective listening experience. Such that every time we were able to lower the close in phase noise, the listening experience improved.
My take away is that it is worth it, within limitations based on cost, to use the XO with the lowest close in phase noise one can afford to specify. Too bad the Pulsar clocks are no longer available, as they offered quite good performance at fairly reasonable pricing (compared to say, NDK DuColon). It appears that the company which manufactured the Pulsar Clocks does not make audio rate clocks as part of their normal offerings, so only if one is able to special order custom rates in quantities would appear possible to get these. The next best thing would be NDK SDA series, hand sorted if one can withstand the extra expense.
If DACs available on the consumer market these days made audible difference in music listening due to phase noise, I would be surprised.
have you...
auditioned different clocks in the same DAC circuit? If not, I would expect you to be surprised then.
auditioned different clocks in the same DAC circuit? If not, I would expect you to be surprised then.
You can pick the best one that you've experienced and compare side by side music listening with store bought DAC (including built in disc player or receiver) with bias controlled, I would be surprised if you hear a difference.
As mentioned...
I have actually done the tests. Same DAC, different clocks, measured lower jitter results with the better clock(s) and confirmed better listening experience with the better clocks. It is a closed issue for me, lower close in phase noise is worth it for those seeking best performance. I have moved on to other issues, as worrying about what I have already determined for myself is a huge waste of my time.
You can choose not to believe my results, but until you do the same testing yourself you are just choosing ignorance.
I have actually done the tests. Same DAC, different clocks, measured lower jitter results with the better clock(s) and confirmed better listening experience with the better clocks. It is a closed issue for me, lower close in phase noise is worth it for those seeking best performance. I have moved on to other issues, as worrying about what I have already determined for myself is a huge waste of my time.
You can choose not to believe my results, but until you do the same testing yourself you are just choosing ignorance.
"You can choose not to believe my results,"
If you provided sufficient evidence to support your claims, then people would not need to "believe" anything.
"but until you do the same testing yourself"
If someone wanted to replicate your tests, how would they go about it. Can you provide details of your methodology?
"you are just choosing ignorance."
I don't think so!
If you provided sufficient evidence to support your claims, then people would not need to "believe" anything.
"but until you do the same testing yourself"
If someone wanted to replicate your tests, how would they go about it. Can you provide details of your methodology?
"you are just choosing ignorance."
I don't think so!
"If you provided sufficient evidence to support your claims, then people would not need to "believe" anything."
There is no way possible way to have "evidence" of subjective listening tests, so it is impossible to present anything more here. The point I am making is that the difference between a decent clock and a better clock is audible in the tests I have done, that is all there is to it. To anyone who has not done such tests, but just "believes" that the difference is inaudible, indeed they are just moving forward in ignorance, as their "belief" is not based on anything at all.
I really do not care if anyone believes me, I know what my experiences are and I shared them here, that is all I can do.
If others perform the same tests, and hear no difference, I have no problem with that, but just suggesting that there will be no difference, with no experiences whatsoever to back that up is nothing more than total speculation based on ignorance.
There is no way possible way to have "evidence" of subjective listening tests, so it is impossible to present anything more here. The point I am making is that the difference between a decent clock and a better clock is audible in the tests I have done, that is all there is to it. To anyone who has not done such tests, but just "believes" that the difference is inaudible, indeed they are just moving forward in ignorance, as their "belief" is not based on anything at all.
I really do not care if anyone believes me, I know what my experiences are and I shared them here, that is all I can do.
If others perform the same tests, and hear no difference, I have no problem with that, but just suggesting that there will be no difference, with no experiences whatsoever to back that up is nothing more than total speculation based on ignorance.