I’m not sure I’m exactly understanding his point about jitter: in my mind the process of upsampling (including its filter) is entirely independent from jitter. - Upsampling is a digital process and jitter is irrelevant in digital processes (otherwise they’d get the wrong result some of the time), also that upsampling digital process doesn’t do anything to lower RF (actually it generates some.) Now after things are done “perfectly” in the digital domain you need to go to analog and indeed different sample rates involve differing jitter and different RF.
If you are staring at lower frequencies each doubling of the sample rate allows you to get 3dB better S/N (at least if you are noise shaping.) On the other hand each doubling of the sample rate increases the relative jitter (assuming the same clock is used in both cases.) Jitter in time will cause noise in amplitude and as the sample rate goes up that noise grows as well. It turns out that somewhere around 5 or 6MHz is the sweet spot, if you go lower in frequency you get more noise because you are loosing space for the noise to go in the noise shaping process. If you go higher than about 5 or 6MHz the amplitude errors caused by the increasing relative jitter raises the noise.
Also FWIW above we assumed that we were keeping the same crystal and not dividing it as much. But also, as the frequency goes up when buying crystals the phase noise goes up, so arbitrarily using higher and higher frequency crystals adds even more noise.
I suspect that part of the improvement from the million tap filter is how he does his interpolation when upsampling (if I understood his talk correctly he’s getting 16 bit accuracy out of the million tap filter.) I do hear improvements with more taps and the DS uses a lot more taps than most (non Rob) DACs and (like Rob DACs) the DS uses much more precision during the filtering process than most digital filters. One of the things I want to experiment with in the next release is doubling the number of taps again and I’m sure it will make a different, but I’m still closer to thousands of taps than I am to millions of taps.
That increases resolution as explained above. But if you upsample it - the simplest way being just padding to zero the extra samples then what happens is you get these extra frequencies appearing (its called aliasing) so you need to filter it to get rid of them. The more accurate the filter the more samples you need to use in the filter - for historical reasons these are called taps. So the more taps you have the more accurate the filter. Thousands of taps is very high so Ted is to be congratulated - but a million is simply unbelievably high - amazing really. Rob Watts claims it was a revelation - a big difference - but huge processing power is required and he had to use a really fast processor - it has I think something like 740 cores ie 740 separate CPU’s. But processing power is increasing rapidly and the cost decreasing so eventually Ted will likely be able to use a million if he wants. Anyway the interesting thing is noise is evenly distributed over the frequency range so if you double the frequency you get a reduction in SNR in the audible band - 3db for each dounling. It may be higher to something like 6db if sophisticated dithering is used but we will be conservative. Ted up-samples 640 times so you get over 8 doubling’s which gives an increase in signal to noise ratio by at least 24db.
However every-time you double sample frequency there is an increase in clock jitter susceptibility:
The two work against each other and as Ted says the point where they balance out is abut 6 mhz - which Ted up-samples above - but down-samples by 5 in converting it to DSD before going to the output transformer - which is close to that 6mhz sweet spot he mentions.
As a consumer, when in the hands of an expert like Ted or Rob Watts, it means as processing power increases you aren’t seen nothing yet.
I am with you - I do not see how jitter can be reduced - susceptibility to jitter must increase as sampling rate increases. I think he is correct that increasing tap length is better - but a million - wow - still he claims it’s not just a bit better but appreciably so.
He also wants to use it as a test bed for MQA claims about the audibility of ringing in normal filters - he thinks that’s not the reason lower sampling rates sound worse - its the crappy down-sampling algorithms. Well he now has a million tap down-sampler and will put it to the test.
Interesting times ahead I think - what really leads to better sounding audio? I have the feeling there may be a bit of truth in both approaches. Personally I would do something like record at say 705k. Find the frequency where noise and music are about the same just as the MQA guys do. Dither the original to 16 bits then apply a high quality filter at that highest frequency. Transmit using the following compression method:
You can transmit at 705 if you like because anything above the cutoff frequency is zero and will take up virtually nothing using its adaptive compression scheme and you get perfect reconstruction because its lossless - plus you of course get the extra SNR from the up-sampling that results.
Why 16 bits instad of 20 or 24? Well according to Rob Watts if you use aggressive 7th order noise shaped dither at 705k then you cant tell the difference between 16 and 24 - at least he can’t. With that sort of dithering you get a extra effective bit for each doubling back to 705k giving effective 20-21 bit resolution. Don’t let anyone tell you you cant tell the difference between 16 and 24 - recording engineers and guys like Rob and probably Ted who do a lot of listening can easily do it. Its like the claim high bitrate MP3 is indistinguishable from the original. Those with experience can easily tell the difference.
I didn’t see any mention on the Playback Designs website or press releases but according to this review (via Google Translate) Andreas’ new ‘Dream’ series now up-samples to ~90Mhz. Which is double the Merlot DAC I believe.
I know you have great respect for Andreas so I don’t mention this for any other reason, other than I found it interesting. He’s obviously found doubling up-sampling rate improves things overall for his DACs (SNR etc).
PD. One thing I can say about them is they have for want of a better word a secretive culture to the point of silliness. I bought one and lent it to a friend to check out. You needed to download drivers but the process they had was as if it was so secret the world would come to an end if it was generally available - you had to supply all sorts of details to get it. Now the guy I lent it to designs DAC’s - its a boutique business by a guy just interested in audio. OMG - the blow-back I got, and I had paid my hard earned cash to own it, just for lending it to him and wanting the drivers. I rang the distributor I purchased it from and he was getting the same blow-back as well. I just couldn’t believe it - I own the bloody thing - I can lend it to who I like and complaining he wanted to download these supposedly ultra secret drivers was some kind of crime floored me. I wrote them a rather sternly worded email to stop wasting time on people just having fun lending things they own to others. BTW I sold it to get the DS - it was better IMHO. The PD for example had a very noisy output stage making digital volume controls in computer players pretty much useless. I complained about it and all I got was we have conducted listening tests that shows our output stage sounds best and needs to be used with a pre. I was glad to get rid of it - the DAC IMHO was reasonably good but their customer liaison, service, whatever you want to call it left a lot to be desired. Nothing like the first rate customer support you get here.
Thanks. We do pride ourselves on being responsive and if there’s a question we try to move to generous as the default. We get it wrong too, but heck, who doesn’t?
PDs stuff wasn’t bad - but PS was better. That happens all the time - and isn’t a concern - it was the attitude that got me - nothing like here which is first rate and one reason I own a number of PS products. I mean look at the technical response I got from Ted - fantastic. Although I personally am doing a bit of rationalization and will sell a few off. Not because they are not good - simply to rationalize. I was going to set up 2 systems but find myself listening to a lot of 2ch HT these days and have figured out how to combine the two without sacrificing quality. And the other thing is you do not mind discussing other gear. That Chord with enough power to drive a speaker by itself really intrigues me and I can see I will eventually get one. Not straight away though - it will be released in August and I have found from experience if you wait a bit you can generally get a better deal.
The issue isn’t how high you upsample, but how high the output data rate is, e.g. the DS outputs quad rate DSD. Going to octal rate DSD would increase the noise from jitter. The digital process of upsampling isn’t subject to jitter (in the data), it’s only at the moment that things become analog that jitter matters for audio and the amount of noise from jitter at that point is the issue.
To be clear I don’t know the sample rate at the digital/analog interface for Andreas’s products.
I was thinking. If the target for this is 20 grand I think MR Smith is going to run up against a brick wall. I know he can do a lot with a little. Nonetheless the DS SR. is already a world class product. Earning Stereophile class A+ 4 years in a row. I a don’t know much but I think like MSB Select is $130,000 for a reason. I think the DS is “already” a $20k DAC. To vastly improve upon it I think one is going to need a much greater budget. Unfortunately I see this getting scrapped as it is too big a risk. A DSII, possibly. With some subtle changes/upgrades. To make a DAC that walks all over the DS for 20 grand I am afraid is not doable. Although leave it to MR. Smith :) If anyone can pull it off I imagine he can. So I may stand to be very surprised when it is in my rack. I know with the DS there were compromises. Yet, the finished product is a force to contend with. You can really go compare it to anything at any price. Therein lies the problem. Clearly raising the bar I imagine shall not be easy and especially not given a “budget”. On the other hand if it were “cost no object” I think something can be done. Although I do not think MR. McGowan will be too keen on that. I do not by any means attempt to thwart your efforts, MR. Smith but I am just guessing you might not realize just how good the DS is. You did so good you don’t even know it lol.
MSB Select = $84.5K
2 x mono power base (optional) = + $20K
Femto 33 clock (optional) = + $10K.
So, more like $114.5K, depending on your upgrade options. But, no arguing that it is a big number.
In other words, the price point earmarked for the uber DS DAC ($20K) will only buy you the basic dual power base for the MSB Select, or an outright buy on the Femto 33 clock only.
Today I heard the fully optioned MSB Select DAC ($120K-ish). I like the modular design concept very much, and the build aesthetics. MSB, like Soulution, and Ted Smith, ascribe to the internal clock philosophy. Ted has talked about the technical reasons for this design point, and why that is better than external clocking. The MSB DAC had the Femto33 clock upgrade, which I think is one of the lowest phase noise audio clocks currently available.
The MSB sounded pretty good. If Ted can get his $20K uber DAC sounding (and functioning) as well as an MSB that costs 6 times as much, that would be quite a technical achievement.
When talking about meaningful price limits for digital equipment one might also consider the situation for people like me who also have a parallel high quality vinyl setup.
Due to the quality difference of available masterings (pro vinyl especially in case of a lot of Jazz, but also some part of Classical and Pop) such money (20k or more) spent on a DAC etc. would still mean inferior sound compared to vinyl for a considerable part of my music to me if listened in digital format. But for sure sound will still be improved by such high priced DAC’s for those fixed to just digital releases.
For me the DS is more than good enough to really enjoy everything where either no vinyl version is available or without any real advantage or where vinyl sucks due to mechanical limitations.
But I guess I’m in a minority cause most decide for an „either - or“ not both on a certain quality standard. And if one doesn’t know the better, he fortunately doesn’t care.
The second reason for me not to spend that much on digital equipment is, that technology and formats still move too quickly. But admittedly that’s a question of funds. One who can afford even changing 30k DAC‘s after a few years won’t care.
“The second reason for me not to spend that much on digital equipment is, that technology and formats still move too quickly. But admittedly that’s a question of funds. One who can afford even changing 30k DAC‘s after a few years won’t care”.
Exactly……. I am fortunate in the sense that I could afford to buy such DACs if I cared to. But I would never countenance such folly. Just dick waving. $120K on a product that becomes the norm after you lived with it for a couple of months when the novelty has worn off.
imo that’s right as long as it relates to some inherited differences between digital and vinyl as such, but I was talking about the differences between many masterings for digital and others (imo much better) available for vinyl. I’m sure they are objectively clearly better for everyone who hears, independent of and overruling differences between the digital/vinyl concept differences.
In case those masterings would be available for digital formats (i.e. if Grundman and others would make more of them for SACD), people who listen digital would prefer them to the masterings available so far in that format, I’m very sure.
So my argumentation here is not about vinyl/digital differences but mainly mastering differences.
Does anyone seriously doubt Ted’s ability to beat the MSB regardless of price? I don’t like to be “that guy” that calls out other products directly but I couldn’t resist. DirectStream is Ted’s best work within a budget. Just wait until you see/hear what he can manage when the financial shackles are removed and he gets to strut his stuff.
We’re designing a new chassis and look for the Ted Smith Signature DAC, TSS (for short), because it deserves it. I am in love with this new look - though it’ll immediately be obvious to you it’s part of the PS family. It’s going to be taller too. 5” height is needed to keep his boards centered and away from anything resembling noise and interference from the outside and to accommodate the bigger power transformers. And there will be two of these beautiful chassis: one dedicated for digital, the other only for analog. They are tied together with a wicked cool fiber optic cable allowing you to place the units wherever you wish. This will certainly be a stunning achievement, the likes of which the world of high end audio has never seen.
If anyone can pull it off I imagine he can. So I may stand to be very surprised when it is in my rack. I know with the DS there were compromises. Yet, the finished product is a force to contend with. You can really go compare it to anything at any price. Therein lies the problem. Clearly raising the bar I imagine shall not be easy and especially not given a “budget”. On the other hand if it were “cost no object” I think something can be done. Although I do not think MR. McGowan will be too keen on that. I do not by any means attempt to thwart your efforts, MR. Smith but I am just guessing you might not realize just how good the DS is. You did so good you don’t even know it lol.