Triode's USB 24/192 plug in - sound quality impressions

[AlexM]

Dear All,

I am temporarily using a Cambridge Audio DacMagic 100 as a loaner while my modded 840c is off for repairs. I installed Triodes's superb USB plug-in, and am happily running in USB 2.0 high speed async mode.

Playback seems very clean, and relatively free from HF hash that gives listening a tiring quality. The sound is quite 'analogue', although not the last word in bass power or fine detail. I'm quite impressed with this little DAC so far - seems like quite good VFM, and sonically on a par with the the HRT Music Streamer II+, which is the only other USB DAC I have tried.

I have also noticed that the CPU utilization on the SBT stays at around 20-24% when playing 24/88 FLAC files, so plenty of headroom for 24/192 FLAC playback, if not WAV (previously noticed this increased CPU utililization signifcantly, so have turned it off on the server).

What have you tried Triode's app with, DAC-wise, and what do you prefer in terms of SP/DIF or USB, both for red book audio, and 96/192 digital files?.

Regards,
Alex

[SBGK]

Think you will find that as the Touch is bit perfect that nothing can change the sound quality, have you tried room compensation software eg Inguz, moving your speakers or physical room improvements to stop sound reflections ? These are all better solutions than any physically impossible voodoo magic imaginary improvements oft discussed in these parts.

[magiccarpetride]

Think you will find that as the Touch is bit perfect that nothing can change the sound quality, have you tried room compensation software eg Inguz, moving your speakers or physical room improvements to stop sound reflections ? These are all better solutions than any physically impossible voodoo magic imaginary improvements oft discussed in these parts.

One person's improvement is another person's degradation.

[JohnSwenson]

I have a HRT Music StreamerII (not the +), it works very well with Triodes plugin and a hub. Adding TT3.0 significantly improves the sound, with this combination it's getting astonishingly good. (it's not the best, my homebuilt DAC still blows it away). The II with TT3.0 is sounding better than several other way more expensive S/PDIF input DACs I have. (except for my own design -- I'm not biased am I?)

I had to do a little work to get TT3.0 to work with it, I had to set the buffer to over 5000, and I had to change some of the priority settings. The esiest way to change the priorities was to just turn them off all together. I didn't have time to figure out which one caused a problem.

I tested the screen off part of TT3.0 and it didn't make any difference so I left the screen on. The part of TT3.0 that seemed to make the difference was the kernal settings, when I implemented just them I got the same sound improvement.


John S.

[Triode]

I have a HRT Music StreamerII (not the +), it works very well with Triodes plugin and a hub. Adding TT3.0 significantly improves the sound, with this combination it's getting astonishingly good. (it's not the best, my homebuilt DAC still blows it away). The II with TT3.0 is sounding better than several other way more expensive S/PDIF input DACs I have. (except for my own design -- I'm not biased am I?)

I had to do a little work to get TT3.0 to work with it, I had to set the buffer to over 5000, and I had to change some of the priority settings. The esiest way to change the priorities was to just turn them off all together. I didn't have time to figure out which one caused a problem.

I tested the screen off part of TT3.0 and it didn't make any difference so I left the screen on. The part of TT3.0 that seemed to make the difference was the kernal settings, when I implemented just them I got the same sound improvement.


John S.

Note that the app changes how the buffer setting is defined - so you will need to edit the EnhancedDigitalOutputMeta.lua rather than the standard file to change buffer setting. I'm interested if the kernel idle option makes any difference wrt to the kernel tuning?

[lake_eleven]

... The part of TT3.0 that seemed to make the difference was the kernal settings, when I implemented just them I got the same sound improvement.
John S.

Can you give the details on how implement the kernel settings alone in TT3.0

[guidof]

For the last few days, I've been listening to my main system (see sig below) with EDO+TT3.0+buffer=20000+SBGK's priorities of 03-25-12.

My impression is that there is an improvement in sound over the SBT+TT3.0 alone, very much along the lines of the improvements from plain SBT to TT2.0 and 3.0. More realistic tonal balance, better S/N ratio (blacker blacks), more detail more easily perceived, more three-dimensional and airy sound stage.

The improvements are fairly subtle, but easily perceived upon listening to familiar recordings of a wide variety of music -- at least to my ears and on my main system.

Of course, I have no way to attribute these impressions to any one of the above mods. But I'm grateful for the combined results and happy to post them in reply to the OP's request.

Guido F.

[Triode]

There has been some discussion on the Touch forum about the EDO app and the potential benfits to audio quality from the spdif output at 96k and below. This is not the primary purpose of the enhanced digital ouput app but people have debated whether any possible change in audio quality exists. I thought it would be useful to spell out what the changes are that are in EDO. As I may speculate about any impact on sound quality, I thought I'd do it here as its definately a discussion for the audiophile forum...

EDO essentially comprises three parts:

1) A custom linux kernel which is the standard Logitech kernel with the following changes:
- spdif - additional clock divider values to allow 176.4/192k playback
- usb scheduler - minor changes including change of default compilation options to allow most dacs to connect to usb
- usb audio - major changes to include most of the more recent linux kernel changes that support uac2
- experimental alternative kernel idle routine [disabled by default]

2) A squeezeplay applet which allows the kernel to be installed and provides menus to select the output device. It also starts up the audio output of squeezeplay and does a couple of things differently from standard squeezeplay:
- it does not start an effects jive_alsa process (it is assumed users don't want to hear the effects processing)
- it bypasses the alsa plug layer and uses the spdif output "TXRX" directly, simplifying the amout of processing required between jive_alsa and the spdif output driver
- increases the priority of the irq task associated with the spdif driver (primarily to improve performance for 192k playback)

3) A new jive_alsa binary which adds support for additional alsa formats for packing samples into alsa frames - this is necessary for the usb dac support, but there are no changes which I believe would impact spdif.

Now I don't claim sound quality improvements from any of these for spdif playback and I have not done any critical listening for this. I would also not want to make any claims without being able to measure the circuit in more detail. However you could speculate about the following:
- not sending output to the analog will avoid there being any data sent to the internal dac and avoid any audio related load on this part of the circuit/psu
- bypassing the alsa plug layer will reduce cpu load
- not having a second jive_alsa process running reduces cpu load slightly

However all these could be achieved with TT3 (I don't run it, but reading the code it looks like it will do this and much more) So its probably best to see EDO as doing the more obvious bits of TT3 to me.

There is one other component of the app which is dissabled by default and is a complete experiement - this is the kernel idle option. This can be turned on from Settings > Advanced > Digital Output. What it does is subsitute the default kernel idle task (which allows the cpu to enter low power mode) for an alternative one (which does not enter low power mode). I speculate that if cpu load has an impact on sound quality then this should have an impact as less than 100% cpu means the idle task is being run for a portion of the time. One could speculate that things like buffer and priority tuning change the frequency and duration at which tasks run and so also impact the frequency and duration the cpu spends in the idle task and hence the low power waiting for irq state. I would be very interested if people can tell the difference with it enabled or not....

[guidof]

There has been some discussion on the Touch forum about the EDO app and the potential benfits to audio quality from the spdif output at 96k and below. This is not the primary purpose of the enhanced digital ouput app but people have debated whether any possible change in audio quality exists. I thought it would be useful to spell out what the changes are that are in EDO. As I may speculate about any impact on sound quality, I thought I'd do it here as its definately a discussion for the audiophile forum...
<snip>
.

Triode:

Thanks very much for your thorough explanation. I'm not able to follow it in its entirety, but I think it's fascinating. Especially since your intended goal was clearly not to tweak the sound but to effect streaming of higher def files.

And also many thanks for all your work. Those of us who are non-technical end users are indebted to folks like you, who do the heavy lifting for the rest of us.

Guido F.

[Dave1972]

Think you will find that as the Touch is bit perfect that nothing can change the sound quality, have you tried room compensation software eg Inguz, moving your speakers or physical room improvements to stop sound reflections ? These are all better solutions than any physically impossible voodoo magic imaginary improvements oft discussed in these parts.

Agreed 100%! I do suspect that any sonic improvements using this app with lower samplerates are imaginary - but who knows, maybe I'm missing extra-golden ears. It's easy to imagine improvements when there are none - I've been caught out in the past (suckered in by silly speaker wire claims!) and have been extremely cautious ever since. A classic example is hearing an improvement with 24/192 over 24/96. With most cheaper-end or mid-end DACs (which many here, including me, will be using) 24/192 can actually be WORSE than 24/96 or 24/88.2 due to artefacts introduced into the audible range as a result of poor handling of the extra data in the inaudible range. And yet people will almost always report audible improvements. But show me a single double-blind test that proves anyone can genuinely hear an improvement in 24/192 - bet you can't!

The reason I installed this app is due to claims of reduced load on the processor. That is a measurable, verifiable improvement therefore worth taking, especially if playing through the SB Tiny server. I'll be experimenting with increasing the number of files the Touch can see (i.e. not hidden in folders starting with a ".") to see how far I can push it now.

[JohnSwenson]

The way EDO works when selecting the 192 digital out will use less memory and will result in less processing going on. (please note that this output will do all sample rates up to and including 192, not JUST 192). What Triode has done here COULD have been done with the original S/PDIF driver (and IS done in TT3.0). Triode has just chosen to bundle the changes in the way the driver is used with the new driver, as far as I can tell there is nothing different in the new driver other than support for 192. The differences are all in the way the driver is used.

Are these differences going to affect sound? Maybe. First off it depends very much on how sensitive a DAC is to whats going on with the input. Some are quite sensitive to jitter in the input, noise, reflections etc. Others are fairly insensitive to this. (I have yet to find a DAC that is COMPLETELY insensitive to what's happening on the input) (all this of course is assuming the the bits are all getting across without errors)

I don't think anybody has a complete definitive explanation of what's going on, but there are some speculations. My theory is that most of the changes to the S/PDIF signal are coming about because of ground plane noise. There is a perception amoungst a lot of people that a ground plane is a large equipotential (same voltage everywhere) system, this is far from the truth. The impedance of a ground plane is NOT zero, its not a lot, but it is still there. A digital system like the Touch has all kinds of high frequency currents flowing all over the place through the ground plane, the impedance of the plane is enough to cause voltage differences to develope on the plane due to those currents flowing through it. Everything on the board is going to be affected by this noise to some degree. Digital logic IS affected by it, but in most cases it's not great enough to cause any issues. Where it becomes an issue are things like the local oscillators, the recloking flop, DAC chip etc. For S/PDIF output, the two most critical parts are the local oscilators and the reclocking flop. This noise can cause increased jitter on the clock and is also directly injected as noise onto the output.

I have actually measured this ground plane noise. I built a simple device that detects and amplifies the noise and sends it to a spectrum analyzer. I can actually see changes in this ground plane noise with changes in what is going on in the Touch. I'm in the process of building a much more sensitive ground plane analyzer, sometime this summer I hope to have it up and running and can do some better tests.

Of course then you have to do correlations between this ground plane noise and what changes this makes in SOUND at the output of a DAC. That is a whole different kettle of fish. This is going to be hampered by the way that different DACs will deal differently with any changes.

So the upshot is that yes I have actually measured real electrical changes in the Touch due to these types of changes, these changes have the possibility that they may cause audible changes, whether they do has not been determined. If they due cause an affect, that affect will almost certainly be different or non-existent for different DACs.

Can I personally hear the difference? On some DACs yes and on some no. On the ones that I can hear the difference it's not a huge night and day difference.

John S.

[Dave1972]

Can I personally hear the difference? On some DACs yes and on some no. On the ones that I can hear the difference it's not a huge night and day difference.

John S.

At least that's a more sober assessment than the guy who stated that going from the incredible sound of the Touch +App back to the stock Touch was like listening to a transistor radio.

[jksbt]

FWIW--one of the more obvious differences in SQ (in addition to those I noticed with EDO) was soundcheck's advice to cause FLAC decoding to be done on the LMS instead of the sbt. It doesn't require the TT3.0. This improvement was very clear and required a simple setting change to the file types menu in LMS settings.

[Dave1972]

FWIW--one of the more obvious differences in SQ (in addition to those I noticed with EDO) was soundcheck's advice to cause FLAC decoding to be done on the LMS instead of the sbt. It doesn't require the TT3.0. This improvement was very clear and required a simple setting change to the file types menu in LMS settings.

Have you tested that double-blind?

Performance-wise, I am finding my Touch running the Tiny server is performing very well. It's currently loaded with just short of 5000 tracks and Memory Used is at 73% so still plenty of headroom. Before installing the app I would generally get stutters at the end and/or beginning of songs with more than about 4500 tracks loaded. I'll keep moving more tracks onto it to see how it copes. May be a coincidence as the Touch can be temperamental anyway.

[jksbt]

Double blind testing...nope. Maybe double ear testing though. ) Just kidding.

I have checked it many times and I feel confident I can hear the difference. I don't mean to claim that you or anyone else will. I believe I heard more detail and more articulated sound. But, again, this is on my system with my ears. YMMV.