XXHighEnd

Hi there,

I am still overwhelmed by the sound of your new release, Peter, just amazing what a decent PC-DAC combination can produce in terms of touching musicality!

Esp. the PA and the Play from Drive option makes all the difference!

It all played well on my system and with the settings I am currently running (see my sig) with one exeption and I couldn't really solve it by following other posts here regarding plops, cracks etc (hope to depict the sound I am hearing correctly).

Besides the little ticks after every song (woulnd't even bother me at all), I get a very loud "crack" when the playlist stops or I hit "stop", either in unattended or in attended mode. The crack can e compared to putting a needle on a vinyl disc at full volume, really giving a blue burst in my poweramp valves. If I don't turn the volume down before I even fear to fry my speakers...

As recommended I have tried to go as high as possible with Q1 and xQ1 settings, the system plays music (although receiving the warning message that the combination of... is larger that 122880 and only one track might play in attended mode) at Q1=15 an xQ1=40. But all with the same crack at stop. Same with higher Q1 values, but music only plays with low xQ1.
At Q1=15, I have tried xQ1=1,2,3,4,5, the all give the crack.

Any idea how to proceed?

Other than that I will live happily ever after (...still have to figure out how the Gallery feature works, obviously I always load the pics from my HDD music folders, not from the SSD...)

Thanks again for this milestone in computer audio

cheers Carsten
 

Hi there Carsten,

With a big thank you for the nice words, to me it looks like we have a new phenomenon at our hands. Not sure, but it seems so;

In order to me being on the right track, you must observe your Weiss without PA. Now, I am fairly confident you hear ticks all the time (not really always but more than half of the time), especially when you don't play at the set sample rate in the driver (I think this can be set although I never managed; anyway, it *is* set at some rate).
True ?

If this is true indeed, especially at the *start* of a track, you can try to recognize whether this is actually the same. Btw, I wouldn't describe what happens at the start as a crack, but merely as a (way too loud) tick. Remember, just without PA.
*If* you have these ticks at the start at all, I think we should try to find the culprit in the Weiss driver. Not that we must solve it in there (via Weiss etc.), but it would be good to know it's actually the same problem.

If the above is not it/related, there's the other chance of the filtering in the Weiss bugging us. So, (digital !) filtering as such works upon the sequence of samples and roughly put, it will make a nice filtered sequence out of that. But, this anticipates normal music behaviour, and what happens at stop is not that at all. So, only listen to what happens (if you could without "cracks" or whatever it is) and you'd hear a buffer repeating. On that matter, the higher Q1xQ1F the longer the repeat lasts but also the less repeats. Anyway, this is about the end and beginning of the buffer -when connected like what happens- not being music. So, there can be a larger voltage jump, depending on the last sample of the buffer and the beginning of the same buffer (which thus is played after its own end). This voltage jump can be audible within itself, and I know -using the NOS1 which doesn't do a thing for filtering- that this can be audible. But, this is coincidental and when I hear something like a small tick once out of 50 times, it will be much. What I'm saying is : the voltage jump itself is not doing it, or otherwise I would be hearing it just the same. But now the filtering ...

I know from your Weiss that it can produce totally distorted music if only the music is too squary. So, here too the higher voltage jumps occur, but now we call them high transients. This is not what the Weiss can deal with very much and remember, this is about the filtering. So, the algorithm in there can not make sense out of too high trancients, also explicitly referring to filtering generally being *about* eliminating high transients, because that's what's in Redbook unfiltered (like a 20Khz sine implies a square because only two samples can form a 20KHz sine).

Still there ? Then now you can maybe see better how ticks in between tracks emerge if they only are with gaps. Here too those filtering anomalies happen, but now in XXHighEnd. It is just difficult to solve.

My last option for now would be the sort of strange habit of Weiss (but others sure too !) that it's actually advised that playback software should start with 2 seconds or so of silence (ask Daniel). Although this will help the Weiss to avoid the loud tick in the beginning, it is a backdoor solution. I mean, yes, *I* could solve that for Weiss and the others, but actually it is a DAC problem and I don't say it is easy to solve in-DAC. Now watch this :

What I suspect is that at least the Firewire drivers in general (from DICEII which your Weiss will be based upon) are clunky (well, they are and always have been) and that something happens like the playback software tells to start, but the driver right at that time starts streaming. Wrong, because it didn't receive "sample rate switch data" yet. This is where the tick at the start happens. So, the first couple of samples are played at the wrong sample rate and such, and this is where the tick comes from; actually it's a very brief moment of static. Now the stopping ...
When the software plays the last buffer at some point in time it will tell the interface to stop playing. This is when the last buffer is empty. However, this is when the software is TOLD that the last buffer is empty. And this can easily be wrongly programmed. So, playback software hears that last buffer has been played, it stops the device, but in the mean time the driver isn't playing out the last buffer - it just stops right away. It is similar to the start problem, though different. But made by the same programmer who didn't know all about it ...
So, this last buffer needs to be played, or otherwise there's an abrupt cut of the sound, and that means "depending on" again, a large voltage drop (a super transient downwards).
Without Phase Alignment this can be audible already, but with for sure it will.

Enough blahblah. But with this "information" you might be able to see better what could be going on.
Notice that I could provide another parameter which anticipates the early stop like described, but it gets more and more complicated. Not only for me, but for sure for people like you. But if it helps ...
Btw, I'd think that shortening the buffer in the Weiss control panel could help here. So, if you have it siginficantly larger than 512, try 512 which is doable (I know). Maybe you can reach a little lower (like 384) but not much more I think.

At testing these things, try to always use a same "ending" so you're not depending on the audio data; drag the time slider towards the end (try until where it will play, because something like only 15 seconds before the end may not work for XXHighEnd) so then it will always end at that end of the track with the same audio data (could be silence by that time). Use short tracks because I think it will be more easy to go towards the end with the slider (so that it will work for XXHE).

Apologies for the long post ...
Peter

Peter,

thanks for sharing your insights here, I have tried all your suggested routes. With or without PA, I never get ticks or any other noises at the start of a track. No anomalies whatsoever.

It only happens at the end of the playlist or at stopping, and I guess your assumption with DAC playing the last buffer is right, the it gets this transient downwards and this makes the voltage jump. At my settings, the last buffer repeats once, then the crack comes. Changing the Q1/xQ1 settings or the buffer lenghts of the DAC and in XXhe just changes the number of buffer repeats, but the crack still occurs, in some settings even worse. But it never goes away, even with very small buffer sizes.

I will check for a driver update (however the last driver update did not work with XX9.6, don't know if it would work now and if it would change this stopping behaviour), but if you could make provisions within XX in one of the next releases, would be highly appreciated!

Thanks for all your help

Carsten

Carsten,


This suggests that you have the "crack" at the end of the track just the same without PA ?

Peter

No, just with PA engaged it gives the crack, sorry for confusion / misunderstanding.

Best Carsten

Carsten,

But is there any change noticeable when you change the Weiss Control Panel buffer size (to larger or smaller) ?

Say you have Q1 at 30, Q1F at 10, SFS at 200 and then change the Weiss settings to extremes (but it must keep in playing of course);
Does this change things audibly ?

Peter

I can play at Q1=30, Q1F=10, SFS=200, PA on, I set the Weiss buffer length to 256 (smallest possible value in the Weiss control panel), only change is the length of the music that is repeated once (seems like playing on buffer length), but te crack noise always stays the same (same tone, same length, same volume).  My valves show a blue burst at higher listening volume, I almost fear damaging the equipment with this burst sound.

Same behavior with XX matching the Weiss buffer settings or not. With small buffer length I receive a messag: "Asked for 15ms, received 1ms", my original settings it seemed to play 15ms.

What I notices is that when I hit stop, the music plays a second or two longer, than it repeats a second, but not the full part that I had heard after hitting stop, so it stops somewhere inbetween!?

Does that help?

Many thanks for taking care of this little issue

Carsten

Carsten,

You mentioned that blue flame in the tube(s) before, but then I thought it was merely a manner of speaking. But if this is real ...
You better quit this route ?


That you don't perceive the very last part of a track (or what you heard before at pressing Stop) is normal (and is described in a ToolTip somewhere I think).

As I said in my first response, this all doesn't seem normal;
Did you ever measure DC at the outputs of the amplifier ? You should. No, you HAVE to. But whether you did it or not, you should do it again, but now focused on what happens at the end (Stop). I am starting to suspect some kind of reversed effect, possibly created by the amplifier. Maybe some auto-bias which can cick in quickly, but which can't "cick out" so fast. So, a tube amp may be inherently protected against DC at the outputs, but to me that doesn't mean it can't choke in further processes in there.

Your "mod" of the Weiss can be about the removal of DC protecting capacitors. So, good for sound, but now it will let pass DC. Together with an amp which isn't protected against it either, things may go wrong. In your case possibly the output transformers (??) get saturated and now *that* is doing strange things when suddenly the DC stops.

I say it again, you should stop this. It looks too strange to me. And you know that I won't be responsible ...
This doesn't prevent that I will keep on helping.

Btw, I saw that your Ph Strength is at 2. Well, if I were you this would be the first I'd set to 0 before proceeding. Also use attenuation settings in XXHighEnd of over 20dBFS. Possibly it is gone then and if so, I think I can work out a solution for you. But I wonder, because so far nothing I suggested helped, and it should have.

Kind regards,
Peter

Hi Carsten,

I think you must follow Peter's advice on measuring the DC output fed to your speakers when engaging PA.

It seems that output transformer in all tube amps should block DC. But in my case, I always have a DC reading of +/- 1.5 V (phase strength =0) at amp binding post when hitting STOP. My tube amp has 10 uF upfront on the input signal path too. Like Peter said there must be something going on with our tube amps which does not happen in SS counterpart.

You may like to set your DC volt meter to null mode where the needle is in the center of the scale so you can read both negative and positive DC pulse. Have someone click STOP for you while you stick meter probes to the binding posts. See if you have a DC pulse. Well it looks more like very low frequency AC with high amplitude. In my case, the reading can be as high as +/- 1.9V at phase strength of 3. If you have this phenomenon, look at your woofer diaphragm. The moment you hit STOP, the woofer excursion you see can be worrisome. Mine is about 10 mm travel.

BTW, in my case, I can manage to get rid of "crackkk", "tick", "plop" at either start or stop or next. Still large excursion of woofer (at hitting STOP only) makes me feel uncomfortable. Folks in here said this is not unhealthy but it is unusual though. If I cannot find a solution to get rid of this DC or AC pulse, I will refrain from engaging PA.

Best regards,
Praphan

I had the crack too at the end of playing.
Not a crack but a couple of explosions! :NY01:
Just wondering if it was my transformer in the HongKong-Dac, or the tube in my preamp, or the capacitor in there or the Gainclones...
Then I changed Q1 from 1 to 15 and what remains is a short crack like removing the needle from a record.
Before it was more like pushing the needled over the record and turning the volume up.
So maybe not that much usefull, but I think it is in XXHE the solution has to be found.
And at least now I can enjoy XXHE in company of my family/friends.

Before engaging PA, I had contacted the german distributor of Jadis, asking about my amp blocking DC. He confirmed, that the preamp has a coupling capacity to block DC and that the output transformers by definition do not pass DC.

I had also measured as indicated DC at the speaker output, no DC during playback with PA.

When the crack at stop occurs, I get a huge voltage peak and the speakers show huge travel. The peak is at about 1,5Volts, when normal music is playing at 200mV.

The problem is I am already addicted to the pure PA sound and cannot simply go back...

I need to find a solution and Peter, I still believe in your theory of a transient jump at the end.

By the way, PA strength always was and is set at 1 (not 2, have to correct this), the crack doesn't change with strength=0. And it does not change with lowering the Volume attenuator to -20 or -25.

I hope that we find a solution somewhere in the XXHE-DAC route...

Cheers Carsten

While typing a response, I saw this inconsistency :


and


So I don't understand what you realkly wanted to say here !

Hi Peter,

I guess Carsten is trying to explain the similar experience as mine. He is trying to report his findings on DC readings focusing on the moment of clicking stop or next button.

I am not quite sure whether it is an AC or DC pulse. But I used DC null mode to measure voltage at amp binding post at the moment of hitting STOP .  I guess it is a high amplitude very low frequency AC which slowly moved my woofer diaphragm for about 10 mm.   

However, I still don't get his contradictory message about DC reading at normal playback. In my situation I always read slight mild DC swing (+/- tenth of mV) during either with or without PA playback. Never zero DC. Then at the hit of STOP the large LF AC pulse kicks in.

Hope this might help.

Kind regards,
Praphan

Hey Praphan, thanks.


Must I read this like : also without PA this happens ?
If Yes, I almost start to believe there is a general problem elsewhere. So, I already started thinking like this, but when you answer Yes to this question I can be more confident about it. Btw, not a software problem, but something VERY different ... (not that I know what, but I may theorize about it).

Regards,
Peter

Thank you for all the feedback Carsten - and don't think I gave up !

On a side note and possibly important :
By coincidence I noticed myself that when the product of DevBufSize x Q1 x Q1F is too low, there will be an unavoidable loud tick at stop. And while I know you already tried a few things, for me the product is like this :

4096 x 1 x 1 = 4096 at SFS=60. So, here I have the unavoidable (loud !) tick (thump); this is WAY too low. But this works for me :
4096 x 14 x 1 at SFS=60

The message here is not that my latter setting should work for you, but merely is about the too short "buffers" theoretically *never* working. So, whatever you are going to try further, have the result of that product to be 4096 x 14 (x 1) = 57344 *at least*.
In my case this is in conjunction with a physical device buffer of 4ms. This is a dangerous one, because for me this is about 32/705600 while for you it might (will ?) be about 32/176400 (for 4x upsampled Redbook). Here is the math for this :

1 sample in my 705600 case takes 0.0000014 seconds. Two samples for stereo take 0.0000028.
My setting is 4ms;
In samples this is 357.

This very theoretically means that you should set the buffer length in the Weiss Control Panel in samples to 256 or 384 and then that part should work out the same. Together with my other settings (above) all should work out the same. And btw my Strength is 1.

So far for this little part ...
Peter

Carsten, besides my other question about the consistency of your earlier post, I find it hard to believe that this is 1,5V (DC !?) only. So, connect a nornal 1.5V battery to the input terminals of your speaker and watch (excursion) / listen (small tick) how harmless this is.

Maybe your Volt meter is "slow", but maybe this is another indication that something is reacting (?) strange.

Btw, don't switch on Ph + with Normal Phase. Same for Ph - and Inverted Phase. *That* would explain what you see I guess ...

Yea, uhm, but ...
When your Absolute Phase is inverted somewhere along the lines all works the other way around MAYBE. So when you perform this correctly (see above) you just as well may try doing it the wrong way. So, switch on Ph + while using Normal Phase in XXHighEnd ... TRY THIS CAREFULLY because when I am wrong here (and probably I am), the "crack" will only be worse.

Peter

Hello Peter,

The answer is "no". The strong LF pulse kicks in only with PA engaged. Without PA, XXHE on my system works perfectly well. No noise , no stall, smooth as silk though but not as pure as PA mode.

Regards,
Praphan

Sorry for expressing myself in a misunderstandable way, I still have to sort out what is happening here.

I have done some measurements at the DAC, without the Amp running.

The whole problem already lies here, so we should not suspect the amp. The DAC has a headphone outout. When playing with or without PA engaged, I measure a constant DC-level of -1.8mV, so I guess a normal behaviour and not related to XX, nor to XX with PA engaged. During normal music playing with maximum volume, I get AC levels of max 100mV. When I hit stop, I get a peak of DC (!) of 1.5 Volts. A little more accentuated on the left channel, but I guess that depends on the music playing.

The same thing happens on the line outputs. No DC during normal play, when hitting stop, I get a huge DC burst of 1.5Volts. Why this is getting through the amp I have to ask the manufacturer. But it would be better if it would not happen at all. So it must be somewhere in the XXHE-DAC route.

Sorry to bother , I would love to dinf a solutin for this.

Best Carsten

Sorry Peter, I had not read your suggestions on the device buffer length, will try asap

ok, things change somewhat with those parameters.
I have put the buffer length in the Weiss panel to 256, which is the minimum. When I change the setting in XXHE also to 256, things get worse. But if I leave the setting in XXHE at 4096, while putting the Weiss control setting to 256, the crack gets a lot softer. All this with Q1=15, xQ1=1. Lower Q1 values make things worse. Changing clock resolution does not make a difference. Higher Q1 values up to 30 do not change it much.
Hope that helps best Carsten

Hi All ,

I did some work around this PLOP at Stop issue .
Since I did read a lot of things about amplifiers and such ( some of them rather displeasing ) , I want to put a couple of pictures that hopefully will clarify a bit what is happening at the stop (or when we push the stop button in XXHE) of the song with PA working . Notice that the following pictures do show a *good* case when the bad cracks are not happening . What we see , are some loops ( I think many people here know that already ) and the stop itself , that we *perceive* as a bad plop .

Set UP : In this test I didn't use a real DAC at all . Infact I did use the Virtual Cable ( which can be configuarble up to 16X sampling frequency/32 bit operation and for KS) .
Then I just recorded the sound directly from the software via windows . So: no real USB to I2S is involved , no real USB DAC is involved .

Then what we see in the little sample representation is 2 channels of the stereo song playing at around -30db , then we see the samples repeated at the end when Stop is Pushed in XXHE .

Again , I did this to have an idea of what is going on . But I think is close to what happens in our USB DAC setup .

In the first figure we see the meter ( in ROSE color evidence) and the song position triangle ( in ROSE color evidence) during playing music .

In the second figure we see the meter ( in RED color evidence) and the song position triangle ( in RED color evidence )just about at the end of the whole thing.

What I do note , is that *during* the 2 sample repeats , the level of the sample drifts from 0 with respect to its "symmetry" . The METER of the player ( or editor ) shows incresing volume ( (see the RED Evidence ) but of course we do not hear this increase at that moment ) that then stops as a form of plop .

I have the recordings made at 88200 , and of course the plop is there as a reproduction for every player , and yes it is audible just the same .

stefano

PS : volume in XXHE was set at -9 db . PH - and strenght 0
      PE + Arc Predict (2X).

Hi Stefano,

I am sitting here with a big smile looking at your analysis. Great. And sure it shows what some perceive. But one thing :


W r o n g ...
And no, not your fault. Mine. Nobody asked or noticed thus far, but I did last week myself ... So, in the version I use, PA is not possible outside of Arc Prediction ...

I honestly can't even predict what happens with AI, but your graphs show a sort of reversed effect.

Small issue : This will not be Carsten's problem.

What I (seem to) learn from this, is that I possibly can build in protections against this "reversed" work out ...

Thank you Stefano,
Peter

Carsten, maybe you can get there after all ...


Not much ? to me that seems "a little", or anyway : something.
Now, you must realize that Q1=30 is only twice as long as Q1 = 15. This is not all *that* much. It becomes much when you now start dialing Q1Factor. So, set that to 2 and again it is 2 times larger. And so, 40 is 40 times larger than when set to 1.
It would be my idea to slowly increase it and watch what happens.

Regards,
Peter

Carsten,


Don't you mean DC here ?


Don't you mean "speaker outputs" here ?


Maybe not.
(*remember*, this burst of 1.5V at the end shouldn't be there in the first place !)
I don't think this DC can be blocked, because it is not DC. On this matter we may just as well look at Stefano's graphs, because no matter it is a wrong situation incurred for there, it sure is representative for what happens. So, look at the right side of the graph ... that is no DC. So, supposed that "end period" takes 2.8 seconds or so (looking at Stefano's graph), there's a change of voltage from 0 to the 1.5V in 2.8 seconds. Call that a(n AC) frequency of 0.075Hz  (4 times the length of what you see for a full cycle), and I don't think that can be blocked. Mind you, this is this end period.


And ?

no prob Peter

I did noticed this morning , that into Anti Image filtering there wasnt sample repeats and other things for the first time , but ... it was late .. late and more over engine3 would crash in that situation ( ... ) .

Well . Not bad for now

:-)

stefano

no there isn't a reverse effect , but just a conga playing in the music just when the stopping  sequence commence ( PE/AP) .

So what I did call " first loop " effectlively will be repeated after  one time ( second loop ) and a third time just in part .

S

I can switch Phase at the DAC and at XXHE. Original was DAC=+, XXHE=-. When I keep the DAC at + and switch XXHE to +, the cracking sound varies a little bit, but not much. Just a little crack more before repeating the last buffer. The crack at the end might be a tad softer, but not really noticeable. Keeping XX at + and switching DAC to - gives the same pattern like the beginning (i.e. DAC +, XX -). No big changes here.

I have also increased xQ1 now from 1 in little steps all the way up to 40. What changes (clearly) is the length of the repeated buffer, but the crack at the end (and a littel one before repeating the last buffer) stay almost the same, might again get a little softer all the way down to 40, but not really considerably. Still sounds frightening at higher volume levels. The biggest change was really setting the DAC control panel buffer length to 256 and keeping the XXHE setting at 4096.

Thanks for all the help

Carsten

Hey Carsten,

Regarding your Absolute Phase tests, I have the feeling that you did not understand what I wanted you to try. But maybe my text was too much out of context. This is what I meant to say :

In XXHighEnd you can have Normal (Absloute) Phase or Inverted (Absolute) Phase. This is what the IPhase button is for. When not active it is Normal Phase.
Now, when Normal Phase is active, Phase Alignment should be used in its "Minus" form. So, think like something needs to fight back against plus. So, Minus fights Plus;

You can just as well use Inverted Phase because your chain inverts it somewhere (often DACs do). So, now in XXHighEnd Inverted Phase is output, and e.g. your DAC makes it normal again (because that Inverts too). This is what the IPhase button is for and always has been for. However, thinking about Minus wants to (intendedly) fight Plus, now you logically need to use the Ph (+) Plus button.

The above is the normal situation of usage.

Now :
a. you may be using this wrongly;
b. you may NEED to use this "wrongly".

My suggestion in the earlier post(s) was to try b. But of course when not first a. applies. :swoon:
So, when a. does not apply, it was my suggestion to use Normal Phase with Ph+ OR Inverted Phase with Ph-.

(and what you did was just playing with the Absolute Phase settings on either end of the chain (XXHE vs DAC) without ever telling about the Ph+/- settings which already make me believe that you can have it "wrong" in the first place ...)

Ok, so what I want you to try out is pretty easy to follow now, I assume. And so here I should stop talking ...

But I don't (so watch out not to get confused again) because what further can be in order I can't get myself. So, it is for a reason that you can set Ph+ together with Normal Phase. Thus, as explained above, this would be an "illegal" situation, but sadly *now* I don't know what happens when this illegal situation is fed to a DAC which inverts the Phase. And this is just an example, because preamps and amps can invert the Phase as well. It is just a so crazy difficult thing (and possibly the most difficult I ever made in my life) and I can't reason it out. But what SEEMS to be happening is that the workout in your chain incurs for that "reversed" effect just because something inverts along the lines, *OR* a. above is in order in the first place.

Is it more clear now ? (:nea:)
Peter

ok, I got it, I had read in the IPhase description that it would only work for Engine#3, but I was wrng, I probably need to wear glasses...

I have worked out the following combinations:

Iphase   Ph+   Ph-   DAC_Phase    Result
------------------------------------------------------
  off         off     on          +               crack
  off         on     off          +               softer crack
  on         on     off          +               softer crack
  on         off     on          +               crack
  off         off     on           -               softer crack
  off         on     off           -               crack
  on         on     off           -               crack
  on         off     on           -               softer crack
       
softer crack means that the last buffer gets repeat with alittle lower volume, but very distorted and the crack is still audible, but a little softer, but nothing compared to the changes that I got with setting the DAC panel to 256.

All tests were made at Q1=14, xQ1=2, Buffer length in XXHE=4096, Buffer length setting in DAC panel =256.
Then I have left the DAC setting at 256, set xxHE buffer also to 256 and put Q1=30 and xQ1=16 (system would still play ant unattended!) and that gives again a little relief to the crack, but not letting it go away. But I get an error message about clock resolution (15ms asked, received 1ms). Seems to be controllable.

But what I have noticed is the following, not knowing if this is normal or not: XXHE choses enginge #4 for the DAC, but when I look in the TaskManager, a process (or task) is active called XXEngine3 ??? Normal?

Do you call all this case a) or b) ;)?

Cheers Carsten

Welp

After having everything go well, I decided to pull out the volt meter and read again about Phase Alignment to check it out.  I have two Bryston mono amps that work the bottom end of my speakers and one Moscode 401HR  stereo amp that works the top end of the speakers.  (I also have 3 Sunfire subwoofers that get their signal from the Bryston amps via speaker cable; they have internal amps...hmm never checked them).  The Bryston don't pass DC and the Moscode swung from 7mv to 20mv while playing with speaker cables off and -30db.  Ok try with speaker cables.  Everything good to go...
until I hit Stop or album ended.
very loud crack. and at normal volumes just way too loud and scary.  I can't afford to damage the system so...
So I'm off the PA until I/we figure this out.  I suppose I could try turning down the vol right before 'stop', if I really want to listen and evaluate...

no cracks, pops ticks at beginning .middle of songs... but man, when I push stop or it comes to the end of an album all hell breaks loose.
I"ll change my sig soon so you can see where I'm at.  I was  even able to use special mode this time and everything ran well.  I'm using  both special or adaptive mode at the moment.  Everything runs great.... just not PA.

oh ya, this crack at stop happens whether I have the Brystons shut down and the Moscode on, or vice versa.

Carsten,

Quick question :
Must I assume that with all your previous attempts you used Engine#3 ?
And that your report in the last post is about Engine#4 ?

Please answer this precisely.
And yes, sorry about that XXEngine3.exe which looks confusing. The sound engine is just called like that (always).

One more thing :


It seems logical that you always use the same track for this testing. Can you try two others and tell about that "very distorted" ? I mean, nothing should be distorted at all ... (unless it is about the end-start connecting of that last buffer and e.g 100 very fast repeats make that "distorted sound" indeed)

Hey, great list. I will try to make sense out of if after knowing the answers to above questions ...

Peter

So from the above I should read that if Arc Prediction is engaged but the reproduction is set to say 32 bit and 44100Hz PA is out of Arc Prediction anyway ...

stefano

Better out of PA for now , yes . IF it is also a crack and not only a plop it is to be considered a bit demaging in my view .

I am building amplifiers from ten years now , and somehow I do know a bit of them . But also your dacs may be at risk .

So , being it a crack or not it is to be understood that is not only DC . That fenomenal generates a spread of frequencies up to 40KHZ , which can also be bad karma for your Tweeters .


Below find the picture of the spectrum plot referred to the previous (simulation Peter ? ) example posted by me .
The analysis of the spectrum is made just when the thing happens , during the drifting .

Stefano
Stefano

For example: the TDA 1543 invert the phase as well as a TDA1541 . That was made from project I believe since Philips would always engage an OP amp in the place of the I/V converter wich would invert the phase itself (for amplification topology reasons ) .So in that case I know where the chain inverts . The case of TDA1543 and 1541 with OP amp I/V conversion would be 2 times phase invertion .

But lets say I build a balanced dac out of a I2S signal ... and I know where it inverts ....  I would have 2 polarity signals ... right?

Maybe I am missing something here about the difference between Polarity and Phase .

Stefano

Hey Stefano,

[this responds to the picture with the FFT graph; not to your last post]

Yesterday I wrote you a post which became more and more complicated and I decided not to post it.

Today, with your new graph, I must really try to let you refrain from judging (and posting) based up wrong "data". So, I already told that your graph exactly shows what Carsten will have on his hands, and this is "wrong" hence not intended. So, now we can try to proceed on what's wrong in the base anyway, but to me this seems useless *and* confusing.

That it happens like that is another story and it is to find out why. But "explaining" like you do what happens with Phase Alignment ... please don't.

Btw, that 40KHz will be about the end-front connecting of the buffer and the steeper transient occurring because of it. And otherwise it is about the upsampling or whatever (like DSD origine).


So, don't please.

Regards,
Peter

It is perfectly ok with me Peter , really . Feel free to delete the post . I am looking elsewere right now , and I made 2 other post to try to have some explanations from you .

stefano

PS . I understand it is confusing .

Well, maybe I don't even see where it is related, or what you want to pose with it. So :

Polarity, in my view, can be seen as plus and minus and a swap between them. Swap it, and the phase will be in reverse order (180 degrees). Plus becomes minus and minus becomes plus.

When you have a differentially setup device (which could be called "balanced") then one channel "contains" or outputs two signals of opposite phase. So, here too a 180 degree difference. If you swap the "input polarity" to this (so, plus and minus), both signals will change phase with 180 degrees with a net effect of nothing. But, this is a moot thing, because nobody is doing this. So, we could swap plus and minus on a loudspeaker terminal, but nobody is going to swap the "hot" and "ground" of an input signal to the DAC (or from DAC to amp). But it is not about that, in the context we talk here;

When we'd look at one of those "balanced" signals, its has phase characteristics within itself. So, for example, when we hit a drum, the signal should get positive (at first) for the normal signal (wire). This is the "absolute phase" when it is correct (the story is miles longer but never mind). At the same time, the other signal wire also gets positive at this same drum hit, although we tend to say it is "negative" because the voltage is. Anyway, the wave expands to the opposite direction of the normal ("plus") wire;
When we'd change the input polarity, all what happens is that what happened in the positive wire now happens in the negative wire and the other way around ...

... But when we change the Absolute Phase of the input, per wire the phase changes. And now a hit on the drum extends to the "negative side".

When all is wired correctly, and assumed the recording is correct as well, the DC level - which represents the average AC level ! - will be positive. This is because a transient rises fast while a decay never is fast. So, net = positive.
Changing the input polarity to a differential system doesn't change a thing to that. Changing Absolute Phase does ...

Something like that ? :)
Peter

I just finished the ToolTip text to a new little feature regarding PA and I had to do that three times over. Just because I got it the wrong way around. That's why I didn't post that "too confusing" post either, because it is so easy to make mistakes here. And really, when you look at the table from Carsten, assumed he has all right like using Engine#4, it is totally clear how things relate; it is consistent, and no way Carsten will have made mistakes there. But don't ask me to explain it, no matter how consistent it is. It is too damn difficult, even for me.

... Bit because I can't explain it, something will be going on what I miss. Should be your situation too, because your first screenshot IS NOT RIGHT. So ... *that* is important because it must be worked out. And when it is correct you can have all the screenshots you want with the comments you want.

Well, that is what I was trying to say from the start to you ... :yes:

Regards,
Peter

Hahaha , ... good .  ;)

Forgot this one :


Well, that's a good one. So, for the upcoming 0.9z-7-3 a few things changed, but this one stayed (active). Just because I wasn't sure and didn't have time at that particular moment to check. But what happens ? (I now checked) PA is NOT engaged, but because of that "mechanical" thing I talked about earlier, it engages at the end, and when it's finished with that, we're out of data. TICK.
... No, I did not talk about that "mechanical" thing, because that was in the post I did not post. So, at the end things are "balanced out" as should, but since nothing really happened, the balancing out goes in the wrong direction; it engages while it should ... do nothing here.
Anyway, we can call this a bug for sure on my side. Similar as that Engine#3 just doesn't do a thing with it FOR THE END OF IT, while it was (is in pre 0.9z-7-3) allowed to activate PA for Engine#3. So, in this case it will engage it, but here no "balancing out" happens. So, a bug, this time for allowing it to (set by you and) activate.

And there can be more of these bugs.

But since I now have the example of 44100, which again would just resemble what you see in your first screenshot, I can happily build in a check for such a situation. I mean, since I can copy the behaviour now, I can counter attack it.

This is how your posts help ...
Thanks,
Peter

I found the whole damn culprit !!

Okay, don't hit me. The whole stupid thing only works at 44100 into 352800 or 705600. Didn't check 88200 and other bases, but it is to be regarded that it just doesn't work for most of you. Never mind you thought it sounded better ...

Djeezz ...

So, here's that "mechanical" thing at play, I talked about;
It does not engage, tries to "balance out" towards the end, but there it goes the wrong way. And when done ... TICK. Nothing more and nothing less than your picture shows Stefano.

Well, can I dedicate the solution to you ?

Carsten, obviously - hold all your horses.

Oh boy.
:tomatoes:

Of course you can , hahaha . Big Smile .

Stefano

I don't get it, but never mind, as long as you know what to do, Peter. I can live with the status quo without damaging my equipment.

Btw, I had always engine#4 on, I do KS, so I cannot even chose, XXHE always puts it automatically at engine#4.

And yes, with another piece of music it shows just the repeat of the buffer and then the soft crack (or is this a tick by now?) at the end, no distortion. I played CD resolution 4xupsampled in XXHE to 176kHz, never tried highres material up to now. Should I or are you done?

Thanks Carsten

No, I'm done Carsten; It's all solved. Thanks for all of your time ...

Are you saying this is solved for my situation as well?  This is no tick, or soft crack (hehe at the levels I play) but a sharp, harsh, loud, speaker-damaging CRACK (although just checked all drivers and they all miraculously work, even the super tweeters... I may not hit you Peter love, but my Drivers want to kick your ass hehe)
I also play mostly 44.1 material going to 176.4 @ 2048

Hope you're talking about a fix for the above situation. I hate to be out of the innovation loop.


edit, and wow that was quick with the new version just while I was posting,,, must have been a simple fix hehe ya right

ALSO
Just to be clear to the noobs out there.  I've been in this experiment with Peter for a long time.  It has always been a Try At Your Own Risk, and I'll take that risk.  If it does go horribly wrong I (and you) will have to deal.  I'll just start over buying what I can.  There were more warnings than usual with PA and I knew the risks.  If you really can't risk, wait for lots of time to go by before trying a new version or want to be perfectly safe go try foobar.  From the beginning this has been Peter's player he is developing for himself, he's just using us for further development and confirmation (the more ears the better to help him make a the best for himself).  We are lucky he decided to let us in on this sometimes risky game.

Hello Peter
Thougt I had to dare using the PA now, with the version 3, but I get the same crack/beplop sound at end, and while stopping playback as the others did before. did I misunderstand anything? thought it was safe now. tried with the fade out step as well.

Regards Arnt

For me everything is running smooth with 9z-7-3 now! No more crack at the end, one buffer repeat, thats it. Thanks Peter!

Just one question: Shall I keep all the changes in the settings? I.e. is it advisable to keep buffer length at 256 or shall I go back to 4096? Whats the advantage if both are running ok? Q1 and xQ1 (now at 30, 10) shall I go back to (14,1)?

Many thanks for the improvements,

Carsten

You can do what you want Carsten. Those settings were just mine, so it was to compare with settings which had to work by "guarantee". That's all.

Regards,
Peter

hmm 9z-7-3
Tried with my regular settings, but not minimized and not unattended.  Set fade out to 0.  Very low volume for me @-36
didn't hear the crack when pushed stop.  Turned up the volume to close to listening levels... pushed stop and heard the crack :/  It wasn't too loud compared to what I heard with 9z-7-2... but the volume was a little lower.  Tried with Fade Out set to default 3 and that seems to work at not hearing the crack... but does that mean PA is still not working properly?  Very late here... need sleepy time.  Will try more tomorrow.

Okay ...
What people (Arnt too) now must know is that now all applies what applied before this topic emerged. So, buffer settings stuff etc. For example, when I set my Q1 and Q1F to 1 (DevBuffer at 4096) I sure have that louder tick at the end too. This is just the basics of all. Too bad of course that before 0.9z-7-3 things didn't work out for other reasons. But I think it is a good idea to now think *this* part just works, so now it's back to those basics again.

Note :
What I said before is that I now would be able to check whether it all didn't work out in reversed mode. But I can't. That is, looking at the code-changes I "forgot" why it worked with 8x or 16x upsampling only - such a reason can not be checked for (sadly !).
But when everything is followed in normal fashion (hence the protections are normally obayed), all should be working. For example, for Engine#3 it is not allowed and this is obvious when you try it. However, I think that me myself and I would be able to activate it for Engine#3 anyway, but it is obvious that should not be attempted (because it really doesn't work).

Peter

Last night I had the pleasure to finally 'listen' to PA . IF you did note , I didn't tell about anything for the 0.9z7 sound in general .
That was because I think , PA wasn't working or engaging correctly at the famous 88200Hz , which is the only upsample frequency I can afford now ( but soon will be more than that ) .

It worked all for me at : Q1 from 14 to 30 , XQ1=1 Dev Buff size at 4096 and its value on the DAC the same .

It did set perfectly to fade out at a Step of 4 .

Changing the DAC buffer size in front of a fixed XXHE buffer size ( 4096) didn't change anything for the fade out Step .

So from what I hear , I say : IT IS WORKING !!!!   :)


For the sound ... congratulations . It seems to me a wonderful feature . I can hear basycally a very RIAA like dynamic sound , much more musical all over , with a good notice for the bass zone which is delicate , round , and .... dynamic .

Also all the cymbals and drums details are better ;
It is another sound , it is audible pretty soon turning on and off PA , but of course the enjoyment comes listening to all the album . More there is tons of stereo information of the rooms , keyboards and fingers to the bassguitars . Mmmmm . Sweet .


Stefano

CaOd & Stefano I'm assuming you both are using the Fade Out feature to avoid the crack?  If the crack is still there but being hidden by the Fade Out, does that mean that PA is not engaging or not engaging properly?  The player sure sounded wonderful for the small amount of time I got to listen to it last night.  I had it running in Minimized/unattended mode and my "good" settings and it has been almost 3 weeks since I've gotten to listen, so it could be I was just starved for good 'ol XXHE, but it sure sounded good with PA knowing I could relax with Fade Out engaged.

I do not have severe cracks at stop , only that plop thing . But to me is clear that now PA engages , and I maybe know why .

Engaging properly ? How could I know , since the feature is considered a secret?  ;)


stefano

I hope you are just kidding with this .  :o