Re: Help with Kernel Streaming 0.9y-5-01

[manisandher]

Peter, the USB boys can wait... focus on getting the Weiss firewire interfaces working first

Mani.

[PeterSt]

Haha Mani - From your description I think I can see the problem with the Weiss may be similar, hence requires a smilar solution.

That solution though requires a quite different setup of, well, all, and I'm not sure it can be done fast (everything and all can fail again on the same subjects which were just solved here). But ... I think I will do as follows :

It is relatively easy to create a kind of laboratory setup, which will allow you to play a first "file part" which is allowed to be relatively large (think of a setting of 200MB). If that works (my own USB ahead), you (all) can test that. If that works I'll know what to do, plus I'll know it's worth the effort.

I estimate this laboratory setup won't be ready today, but it will tomorrow *if* it solves things for my USB here (but I think it will).
So, more later.
Peter

[manisandher]

Hey Peter, as always, your efforts are much appreciated.

But quite honestly, I love the sound I'm currently getting from 0.9y-5 with Engine#3. However, this is only the case when I'm running my laptop on battery power. The difference between batttery and AC is just huge... the former I love, the latter I just can't listen to. The master clock is in the DAC, the firewire interface has it's own power and yet how the laptop is powered still matters...

On another point, I think you may have over-done the attenuation with AP. Playing a 0dB FS 1KHz sine wave, I'm getting a reading of -3dB on my machine. I know this won't have any reconstucted peaks, but I'm not getting above -3dB with any type of music either. Of course, maybe my meters aren't 'sensitive' enough to register a limited number of samples above -3dB...

Are you sure you're attenuating by 1.5dB and not 3dB?

In any event, QAP sounds phenomenal, so my worries about digital attenuation in this case seem to have proved wrong.

Mani.

[jkeny]

Hey Peter, as always, your efforts are much appreciated.

But quite honestly, I love the sound I'm currently getting from 0.9y-5 with Engine#3. However, this is only the case when I'm running my laptop on battery power. The difference between batttery and AC is just huge... the former I love, the latter I just can't listen to. The master clock is in the DAC, the firewire interface has it's own power and yet how the laptop is powered still matters...
..........................
Mani.

Mani, what you are describing sounds like dirty ground from the laptop contaminating the ground of your DACs. Three ways to deal with this
- You could disconnect the earth pin in your laptop's plug so no earth connection when connected to grid. This would probably be frowned upon by electrical engineers, but it works - I have been using this set-up for a long time. This is not a complete solution - for that you need galvanic isolation (see below)
- Always run your laptop on batteries
- use some form of galvanic isolation - unfortunately this probably requires some DIY. This is where your signal lines from your laptop to DAC are physically interrupted by an device which allows the AC signal through but stops any DC signal.

[boleary]

Guess I spoke too soon. With both 9y-5.01 and 9y-5.02, kernel streaming, I have had the horrific hiss occur when one track ended and the next began. Both times I had 2x arc prediction selected. Attached is the log file for 9y-5.02.

Regarding the Hiface, John and Peter, the two times that it "worked" for me, xx played but there was no sound, I had "24 bit only" selected. This was when I renamed a device with the initial release of 9y-05. I couldn't get it to repeat this with 9y-5.01, just got crashes. I haven't tried the hiface with 9y-5.02.

Here's a proposal, Peter. Email me your address and I'll ship my hiface to you so you have a 24 bit only device to sort these issues out. I'll pay shipping to and fro, and I won't complain (out loud   )regarding how long it takes. Once its sorted out you can devote all your time to Mani and the Weiss issues he's having!  

I'm serious, Peter, but I also understand if thats not a route you want to take.

[PeterSt]

Dear Mani,

It wouldn't be the first time that you's prooven me something is wrong by means of some simple measurement, so I guess the last time is yet to come. It would proove why at least one other person (sorry I forgot who it was) complained about the huge difference in level between Arc Predition and without. So yes, it would be very easy for me to make the mistake which is because by now everything runs through the "replay gain" and the things you see are not what they are (in the program that is).

It is obvious that right now I'm not near the program, so I can't look. But, when you switch off QAP and use Anti Imaging instead, is the level 3dB higher then ? If so, you *must* be right, but if 1.5dB only something else is going on. And to keep in mind (and you implied it yourself) : I don't think peaks at -1,5dB will be measured so. Not only because of too few samples forming the peaks, but merely because you can be measuring SPL-like (hence/or RMS).

But never mind, it's fairly easy to find out once I know it can be wrong. Will let you know.
Peter

[boleary]

This morning, with 2x arc prediction and kernel streaming, the first time one track ended and the next was to begin, I had the hissing explosion occur. I have switched off 2x and have played eight tracks in a row with no problem. Is it possible that kernel streaming doesn't like arc prediction?

[manisandher]

... when you switch off QAP and use Anti Imaging instead, is the level 3dB higher then ?

Things are a bit stranger than I originally thought.

I'm measuring -1dB for the 0dB signal! This is true for XXHE no upsample and with QAA. (Just to make sure, I checked with Foobar, and am getting the same result.)

So, three possibilities:
1. the test tone is not 0dB (I doubt it)
2. there's some attenuation happening in my computer (I doubt it - HDCD would not be recognized by my DAC if this were the case)
3. the meters on my PM Model Two need re-setting/re-calibrating

In any event, for the 1KHz 0dB sine wave; with AA, I'm getting -1dB; with AP, I'm getting -3dB.

With music, the difference is greater. Taking something like Whitney Houston's 'Your Love is My Love'; with AA I'm getting 'overs' from the first note; with AP, I'm never getting above -2dB.

What I think we can surmise is that the attenuation is greater than it needs to be. I would say by 2dB.

I can change the sensitivity of my 'over' meters to measure between 0 and 7 successive full scale data words. I'm happy to play around with this if you were to decrease the attenuation.

Mani.

[PeterSt]

Boleary - This is about this I guess (from your log file) :

FL1 : 100000222
FL2 : 100000216

and my earlier remark about "if not currently static emerges, it will after I solve this bug" and "I put it up anyway" (5-02).
So, firstly this never happened to me (static) in two months of time, but what shows is wrong. Ok, I found how to solve it, and this will be in 5-03 and I never put up this solved bug because the static never happened. However, if you look at those two numbers, you'll see that in your case the second number is 6 smaller than the first. And THIS is dead wrong, because it implies a "no sample boundary" which whould be 8. UNLESS you used a "DAC Needs" of 24 (which I don't think you do -> nope, your log file says not), then it would be just right (and the hiss remains unexplained).
Notice that this can be caused by using ripping software which is other than what I use and which ... wait ... yep, it's in the log file as well. -> Which uses another data offset in the file opposed to what I always use. For me this would be 44, while for you this is 216 for this particular track.
Not using Double indeed makes a difference, unless you used the digital volume to attenuate, then it won't matter.

Allright, no matter this bug has been solved, I now see that I can build in an additional check for this, and without harm(ing SQ). So I will do that.

About sending me the HiFace ... I am trying to get one from here. I hope to let you know in a couple of hours whether this will work.
Peter

[PeterSt]

Haha Mani - From your description I think I can see the problem with the Weiss may be similar, hence requires a smilar solution.

Mani - What slipped into my mind ... how strange that your AFI1 behaves differently from the DAC2 (which just works). On the other hand, you slave the AFI2 to the PMm2 which may incur for such a strange (unknown) situation that I might not be able to deal with it from theories only.

So come on, send all your soundcards to me, and I will sort it out.


Btw, the whole point is : I am not performing any normal means of Kernel Streaming here. If I had only done that ...
then the sound would have been different, haha

[manisandher]

I thought bernharddiner (with a DAC2) was experiencing the same problem as me, no?

In any event, I've tried all the clock variations, to no avail. I'm pretty certain it's the DICE firewire driver that's to 'blame'. Hopefully, you'll find a solution...

Mani.

[boleary]

Thanks Peter, I very, very much appreciate your efforts.

[PeterSt]

I thought bernharddiner (with a DAC2) was experiencing the same problem as me, no?

You are correct. Sorry.
I guess he got stuck in my mind because of the 6 seonds of music and "I like it way better" (similar). Had to laugh about that really.

[manisandher]

Mani, what you are describing sounds like dirty ground from the laptop contaminating the ground of your DACs.

jkeny,

Thanks for the suggestions - I will try the first, and continue with the second if that fails.

But just to be clear, I cannot hear any obvious noise with AC power, i.e. no buzzing or increased static/hiss. What I do hear though is an added 'shrill' or 'edginess' to virtually everything. It makes it difficult to listen to - almost bringing too much attention to the sound.

FWIW, I really can't imagine how KS could improve upon what I already have (when I'm running from battery). If what I have now was the end-game, I'd be happy!

[EDIT: Is there such a thing as an 'end-game' in this hobby-cum-obsession of ours?]

Mani.

[pedal]

But quite honestly, I love the sound I'm currently getting from 0.9y-5 with Engine#3. However, this is only the case when I'm running my laptop on battery power. The difference between batttery and AC is just huge... the former I love, the latter I just can't listen to. The master clock is in the DAC, the firewire interface has it's own power and yet how the laptop is powered still matters...

John Atkinson/Stereophile reported some so-so measurements from a Benchmark DAC (a 2009 review I think). In respons, in a letter from Benchmark, they showed measurements how the jitter increases 10-fold when a typical laptop is run by it's standard powersupply. Most such power supplies are el-cheapo switchmode types, which generates a hell of a lot distortion.