Those were pretty much what I had settled on as well, but in the end they sounded more artificial to me than solo Dave.
This is new information for me, do you have some links where I can find some more insight as to why 18 bits ?
CHORD ELECTRONICS DAVE
- Thread starter magiccabbage
- Start date
-
- Tags
- chord-dave
Because the DAVE has a THD+N of 115dB which is 18.8 bits. So you'd take the number of full bits which is 18. ( https://goldensound.audio/2022/03/14/chord-dave-measurements-with-mscaler/ )
This ensures that the noise shaper is running entirely within the linear range of the DAC and is working to full effect. Setting it higher will actually cause a degradation in quality of low level information.
You could arguably go by dynamic range which is 117.6dB / 19.2 bits and set it to 19 instead. But it's better to go too low instead of too high.
As an example here's a Holo spring 3, playing a very low signal (-120dB 1khz sine). With LNS15 running at 24 bit:
But because the spring isn't linear to 24 bit (no real DAC is), there is some distortion present.
The spring 3 dynamic range is 129dB. Or about 21 bits. If we set LNS15 to run at 20 bits to ensure it's operating within the linear range of the DAC:
The DAC now performs demonstrably better.
The DAC bits should be set to the correct range for the DAC you're using, otherwise it is effectively wasted and you won't get the full benefit.
It's counterintuitively better to go a bit too low than even just 1 bit too high. But for the DAVE 18 bits is correct.
Thanks for the explanation, I'd love to see such straightforward explanations for all the mojo surrounding Dave that actually make a difference.Because the DAVE has a THD+N of 115dB which is 18.8 bits. So you'd take the number of full bits which is 18. ( https://goldensound.audio/2022/03/14/chord-dave-measurements-with-mscaler/ )
This ensures that the noise shaper is running entirely within the linear range of the DAC and is working to full effect. Setting it higher will actually cause a degradation in quality of low level information.
You could arguably go by dynamic range which is 117.6dB / 19.2 bits and set it to 19 instead. But it's better to go too low instead of too high.
As an example here's a Holo spring 3, playing a very low signal (-120dB 1khz sine). With LNS15 running at 24 bit:
But because the spring isn't linear to 24 bit (no real DAC is), there is some distortion present.
The spring 3 dynamic range is 129dB. Or about 21 bits. If we set LNS15 to run at 20 bits to ensure it's operating within the linear range of the DAC:
The DAC now performs demonstrably better.
The DAC bits should be set to the correct range for the DAC you're using, otherwise it is effectively wasted and you won't get the full benefit.
It's counterintuitively better to go a bit too low than even just 1 bit too high. But for the DAVE 18 bits is correct.
I am curious If you had a chance to listen to Dave and the M-scaler with the optimizations that a lot of us found make a big difference. I took it for snake oil and fought it every step of the way but, 2 years later, my tin foil hat just came in the mail. Stuff like :
- The M-scaler needs ferrited cables/Opto-dx or a battery + an optical connection to the streamer. I absolutely hate my M-scaler when plugged in. I'd almost go without it. There's clearly a much higher noise floor. Every time I forget to unplug my battery when listening, I can tell something is off. (The reason why is known, but I'm curious how you've experienced it for your reviews)
- Somehow, even via an optical connection there is a difference between streamers. I've stuck to my guns and my gaming pc-s optical out for as long as I could, but there's no denying there is a o considerable difference, perhaps greater than there is going direct to Dave. Maybe the M-scaler isn't as resistant to Jitter ? Maybe It's due to the up sampling itself? no idea.
-A mains filter is pretty much a necessity in my experience, otherwise, in my case (terrible power quality) the bass is just...flabby and diffuse and the noise floor seems higher. Highs are also irritating and stage collapses a bit.
- The latest FOTM, DC4/Farad power supplies. What makes such a big difference? Especially with headphones (some people claim it can drive Susvara/Abyss much better). Is it the quality of the power supplied to the amp? If so, why do people swear by it in dac only mode as well? Are they essentially doing what the filter is doing, or is there more to it?
I have some original Wave cables here atm and have some Wave Storms and an SRC-DX on the way so will give feedback once I've spent some time with them.
I have noticed some odd electrical behaviour with the MScaler.
Gonna look into it a bit more but there is notable leakage current in some situations and it seems that the ground level of the MScaler is at a different level to some connected devices.
The MScaler's power supply is ungrounded and the USB connection is galvanically isolated (so no ground) so it's entirely possible that grounding setups may be a bit finnicky with it.
Will look into this more soon.
The MScaler does have quite vastly different performance depending on the source. It does seem so far that using USB is the lowest jitter option by far, though that's ignoring any potential aforementioned changes in electrical behaviour which I'll look into.
Below are measurements of the jitter level and spectrum on the MScaler with USB vs SPDIF from the analyzer.
I'm using an audioquest niagara 5000 currently. I've not actually tried it without.
I'd really like to try the DC4.
Originally Sean Jacobs was going to be sending one over actually to try on the DAVE which I was looking forward to but I've not had any response from him in a while
Hopefully that's still going ahead but idk
Thanks! Any chance you can throw optical in there for good measure?
I'm sure @MvRBE10 and Farad might be more amenableI'd really like to try the DC4.
Originally Sean Jacobs was going to be sending one over actually to try on the DAVE which I was looking forward to but I've not had any response from him in a while
Hopefully that's still going ahead but idk
I bet everyone would like an objective comparison. I'm definitely looking forward to more of your impressions.Now I'm getting curious about how a Singxer SU-6 would perform with the M-scaler instead of my streamer. Not to mention Innuos and Taiko level stuff.
I had no idea it existed before I came across your website. I can see why the endpoint makes a difference, but I attribute that to the electrical "contamination" and the clock that's used. Beyond that, I can't see what special sauce a (Roon) server could have that would impact the performance out of the same aforementioned streamer, as some people claim.
MvRBE10
Member of the Trade: Farad Power Supplies
I could throw in a good word and ask if he wants to get a kind of demo model that can travel the world. But mattijs knowing he always want some background info. I would also advise you to just mail him and throw all your technical knowledge to him and the why and hows he loves and appriciates that.Thanks! Any chance you can throw optical in there for good measure?
I'm sure @MvRBE10 and Farad might be more amenable
So far I've not managed to get the same level of jitter performance with the MScaler with any SPDIF DDC as I have with USB. Including the SU6.
It seems that it just will always have poorer performance with SPDIF input. Though this isn't entirely unexpected. Buffering and processing synchronous signals like SPDIF is trickier as the receiving device has no control over the speed with which data is sent, and so it must adjust it's own internal clock speed slightly to prevent long term clock drift issues which can eventually cause buffer overruns/underruns.
Other DDCs even without resampling such as the DI20HE have the same thing, where you get better performance with USB because you don't have as many challenges and the device itself can be entirely in control of timing.
I would recommend from what I've seen so far using USB for the MScaler.
The inputs of the MScaler are galvanically isolated anyway so in theory there should be no mechanism for noise to be transferred and therefore a taiko/innuous etc are not going to make a difference unless they were changing the audio itself.
Though it depends on exactly how the galvanic isolation is done as different companies use the term for different methods of isolation, not all of which actually eliminate all noise. Schiit for example uses transformer based galvanic isolation which does prevent DC offset and ground loops but can still pass higher frequency noise.
I generally would VERY rarely recommend someone to spend large amounts of money on a server. The server itself does not make a difference to sound quality, the endpoint or device connecting to the DAC does.
But even then if you were planning to get a high end server and connect your DAC to it, devices like an Intona 7055-C will provide full galvanic isolation to eliminate any noise/contamination from the source (The 7055 has upto 5kv isolation. You could literally set off a defibrillator on the other side and the DAC wouldn't see a thing) and have demonstrably ridiculously low noise on the output.
The Innuos devices are pretty much custom designed from the ground up with some genuinely great isolation and design aspects, and I'm sure will provide excellent performance too but you're paying a huuuge amount for the convenience of having that in with the server itself compared to getting a network endpoint and decent DDC or USB isolator.
The taiko is actually just a standard Intel based server (on an Asus WS C621E SAGE board) with a fancy PSU, and there is pretty much no way that it is going to have noise as low as an innous or intona or even a raspberry pi given as it's just using the regular motherboard USB outputs.
You can pay £1700 for the extreme USB card ontop of the ~£28k the device costs stock, but it seems frankly ridiculous to me. And you can get truly excellent USB sources for fractions of the price.
Either standalone streamers like an SMS200 Ultra, or high performance isolators like the Intona 7055-C if you want to use a PC/direct source
Last edited:
I'm sure @MvRBE10 and Farad might be more amenable
I did actually have a Farad PSU here a while ago though it was used with a modified Singxer SU-1 DDC.
It did have excellent performance and I'd definitely be interested in listening to one with some other gear!
Unfortunately the SU-1 itself in that situation was a bit of a bottleneck but the Farad was lovely
Well @MvRBE10 and Mattijs are working on a Frankenstein solution to use 3 of them with the Dave for less than half the price of the DC4. Definitely good to see some competition.
Agreed it's ridiculous. I have a top tier gaming pc that I upgrade every 2 years, for a Roon/Hqplayer core. It will always be faster than whatever they're pushing.
For some reason, I've always preferred SPDIF to USB and coax (Most likely the noise the M-scaler is injecting into Dave). Differences between them aside, once I've introduced a battery for the M-scaler, SPDIF seems to win every time for me. If I plug in the M-scaler differences are debatable, but I almost want to go back to solo Dave and call it a day anyway.
With Dave I'd say the Jitter performance is almost inconsequential, but It doesn't seem to be the case with the M-scaler.
chesebert
18 Years An Extra-Hardcore Head-Fi'er
- Joined
- May 17, 2004
- Posts
- 8,969
- Likes
- 4,135
I am confused about Taiko. What are they trying to do? Is there any reason why you need dual Xeon with possibly 128gb or 192gb of ram for music server?So far I've not managed to get the same level of jitter performance with the MScaler with any SPDIF DDC as I have with USB. Including the SU6.
It seems that it just will always have poorer performance with SPDIF input. Though this isn't entirely unexpected. Buffering and processing synchronous signals like SPDIF is trickier as the receiving device has no control over the speed with which data is sent, and so it must adjust it's own internal clock speed slightly to prevent long term clock drift issues which can eventually cause buffer overruns/underruns.
Other DDCs even without resampling such as the DI20HE have the same thing, where you get better performance with USB because you don't have as many challenges and the device itself can be entirely in control of timing.
I would recommend from what I've seen so far using USB for the MScaler.
The inputs of the MScaler are galvanically isolated anyway so in theory there should be no mechanism for noise to be transferred and therefore a taiko/innuous etc are not going to make a difference unless they were changing the audio itself.
Though it depends on exactly how the galvanic isolation is done as different companies use the term for different methods of isolation, not all of which actually eliminate all noise. Schiit for example uses transformer based galvanic isolation which does prevent DC offset and ground loops but can still pass higher frequency noise.
I generally would VERY rarely recommend someone to spend large amounts of money on a server. The server itself does not make a difference to sound quality, the endpoint or device connecting to the DAC does.
But even then if you were planning to get a high end server and connect your DAC to it, devices like an Intona 7055-C will provide full galvanic isolation to eliminate any noise/contamination from the source (The 7055 has upto 5kv isolation. You could literally set off a defibrillator on the other side and the DAC wouldn't see a thing) and have demonstrably ridiculously low noise on the output.
The Innuos devices are pretty much custom designed from the ground up with some genuinely great isolation and design aspects, and I'm sure will provide excellent performance too but you're paying a huuuge amount for the convenience of having that in with the server itself compared to getting a network endpoint and decent DDC or USB isolator.
The taiko is actually just a standard Intel based server (on an Asus WS C621E SAGE board) with a fancy PSU, and there is pretty much no way that it is going to have noise as low as an innous or intona or even a raspberry pi given as it's just using the regular motherboard USB outputs.
You can pay £1700 for the extreme USB card ontop of the ~£28k the device costs stock, but it seems frankly ridiculous to me. And you can get truly excellent USB sources for fractions of the price.
Either standalone streamers like an SMS200 Ultra, or high performance isolators like the Intona 7055-C if you want to use a PC/direct source
Last edited:
Ragnar-BY
1000+ Head-Fier
They can simply rely on the idea that a person who has spent 24 thousand euros on a music server will never try and compare any "cheap" stuff from other companies
Last edited:
MarkusBarkus
Head-Fier
...interesting "fact" you note there @Ragnar-BY
Where did you find it?
I have an Aurender ACS10 and a RaspberryPi 3b, I think it is. And a Taiko Extreme.
If you optimize the network and perhaps add some ancillary devices (switches, LPSs, USB stuff) the less expensive gear can sound very good. Very good.
The Taiko Extreme sounds better. In my house, that's a fact.
Now you ask: Have I ever done a double-blind, climate controlled, fasting, A/B shoot-out with those components, or fill-in-the-blank streamer. And I say: I have not.
Then I say: If you listen to a high-school orchestra on Monday, and go to the London Phil on Tuesday, do you think you could determine which one sounded better? Or would you need to get them in the same hall to shoot it out?
24K is a little low. You definitely want the USB card and add some storage. Strongly recommended!
PS: no offense meant to the LPO or any high-school musicians! Love you all.
Where did you find it?
I have an Aurender ACS10 and a RaspberryPi 3b, I think it is. And a Taiko Extreme.
If you optimize the network and perhaps add some ancillary devices (switches, LPSs, USB stuff) the less expensive gear can sound very good. Very good.
The Taiko Extreme sounds better. In my house, that's a fact.
Now you ask: Have I ever done a double-blind, climate controlled, fasting, A/B shoot-out with those components, or fill-in-the-blank streamer. And I say: I have not.
Then I say: If you listen to a high-school orchestra on Monday, and go to the London Phil on Tuesday, do you think you could determine which one sounded better? Or would you need to get them in the same hall to shoot it out?
24K is a little low. You definitely want the USB card and add some storage. Strongly recommended!
PS: no offense meant to the LPO or any high-school musicians! Love you all.
Ragnar-BY
1000+ Head-Fier
@MarkusBarkus English is not my native, already corrected my message and wrote “idea” instead of “fact”.
As for the rest… I’ve compared a lot of transports. Sometimes there are things which are hard to explain, like clear difference between bit-perfect software (Roon and LMS in my case). But after I isolated my network transport with optical modules I can’t hear any difference between server hardware.
Maybe my hearing is not sharp as yours, or maybe you just heard a difference between software. As far I know, Aurender use their own soft which is not possible to install on Taiko server
Considering the fact that we are discussing this in DAVE’s thread, I also want to remind you about electrical noise, which can (as explained by Rob million times) influence DAC’s sound and some people subjectively like this effect. Due to higher processing power it is very likely that Taiko’s RFI level is higher than Aurender’s.
Long story short: I don’t believe in Taiko’s server, but my opinion has nothing in common with ASR’s pseudoscience, blind-testing and other ignorant theories.
P.S. Just in case if my English skills let me down again, I’m not arguing or trying to troll you. Only explaining my point of view.
As for the rest… I’ve compared a lot of transports. Sometimes there are things which are hard to explain, like clear difference between bit-perfect software (Roon and LMS in my case). But after I isolated my network transport with optical modules I can’t hear any difference between server hardware.
Maybe my hearing is not sharp as yours, or maybe you just heard a difference between software. As far I know, Aurender use their own soft which is not possible to install on Taiko server
Considering the fact that we are discussing this in DAVE’s thread, I also want to remind you about electrical noise, which can (as explained by Rob million times) influence DAC’s sound and some people subjectively like this effect. Due to higher processing power it is very likely that Taiko’s RFI level is higher than Aurender’s.
Long story short: I don’t believe in Taiko’s server, but my opinion has nothing in common with ASR’s pseudoscience, blind-testing and other ignorant theories.
P.S. Just in case if my English skills let me down again, I’m not arguing or trying to troll you. Only explaining my point of view.
Last edited:
In case folks have an interest in this, I have about 30k words to say about it a couple years ago (I had the same question going in) (and to stay on topic, this was paired with a DAVE for the listening tests)
https://audiophilestyle.com/ca/revi...the-taiko-audio-sgm-extreme-part-1-of-5-r907/
(I specifically answered this question in part 2...it is all about balancing memory to cores, so that process isolation to cores isolates memory I/O as well)
Last edited:
Users who are viewing this thread
