XXHighEnd

I've just discovered something really weird with the XX volume control...

For the last couple of weeks I've had my old Pass Labs X1 preamp present in my office system because I've been playing around with loads of different components. The messing around with components ended yesterday evening, but I was too lazy to take the preamp out of the chain. So I set XX to 0dB and had music in the background as I worked. I was quite shocked at the audible clipping in so much of the modern pop music I tried (the loudness wars really have gotten totally out of hand). In the past I would have switched PeakExtend on, but last night I just turned the XX volume to -1.5dB. Of course, the audible clipping stopped, but I was shocked in the change in the overall sound.

At 0dB, the sound is thick, full, and pretty boring. At -1.5dB (and any greater attenuation setting thereafter), the sound immediately comes to life. And I'm now talking about well-recorded stuff, well below clipping level.

Peter, there is something really strange going on. There simply shouldn't be such a big difference between 0dB and any other volume setting (assuming no clipping).

Mani.

Hi Mani,

You could try normal Arc Prediction, assumed that you use a Custom Filter setting.
Difference ?

You aren't using Volume Normalisation ?

Peter

Yes, I was using the Custom Filter (705600 Low). But I can hear the same effect with AP too - maybe slightly more difficult to hear because there is more 'sparkle' with AP anyway, so the 'dullness' at 0dB is somewhat compensated for.

I've never tried Volume Normalisation.

Mani.

On further listening...

On many tracks, I actually prefer the 0dB setting - there's way more energy in the upper-bass/lower-mid area, which gives a fuller, fatter sound. With 'thin' recordings, this is very welcome. But if the recording is already very 'thick' sounding, 0dB just makes it sound too boring.

But the point remains that there is a very big difference in sound between 0dB and any other attenuation setting, and I don't think there should be.

Mani.

Mani,

Maybe by now you have forgotton the quite some posts throughout time about sweetspots and such of the D/A chips. I'd even say that about everyone using a pre-amp (which are not many people in the first place) does not use the -0dBFS setting (but 6dB or so less). So, all as expected.
Btw, the real sweetspot is at -21dBFS or so.

As long as your mentioned "clipping" is not that at all, all is fine. Or "normal" if you want.

Btw, technically you shift the used bits in the D/A chips, assumed that you can not hear (perceive) the full dynamic range of 144dB. So say that you(r ears) use 96dB of dynamic range maximum, then at -0dBFS you use the upper 16 bits, and at -6dBFS the next batch of 16 bits (say 17 MSB's with the first one not in use). And so on.
Of course you can only really notice a difference when first the music utilizes the full scale (say continuously).

Peter

But music is dynamic, so you're never going to be in the sweet spot for very much of the time, irrespective of where you set the attenuation, right?

All I can say is that I have now tried a whole bunch of tracks with RMS levels ranging from -30dBFS to -8dBFS (according to TT-DR meter) and changing the XX attenuation from 0dB to -1.5dB has exactly the same effect on SQ on all of them.

Coincidence, or what?

Mani.

You talk like albums/tracks with an RMS level of -30dBFS play only up to that level. They will play into the MSB just the same.
Albums which never exceed -6dBFS can be counted on two fingers (and then I even make one up).

Peter

You can judge from the first few seconds of various tracks, some of which never go about -20dBFS.

I know what I'm hearing... 0dB sounds totally different to any other attenuation level.

Mani.

BTW, I've just compared a few tracks with quiet beginnings in 3 ways:

1. XX set to 0dB, preamp set to 11
2. XX set to -1.5dB, preamp set to 12
3. XX set to -21dB, preamp set to 23

There is a slight difference between 2 and 3. There is a massive difference between 1 and 2/3. Something is going on...

Mani.

I don't have a pre-amp. :(

What about the combination -0dBFS / 12 ?
(no, I am not joking)

Yep, the change in SQ is still easily audible:

-1.5dB/12 sounds like -1.5dB/11, just a little louder.

0dB/12 sounds like 0dB/11, just a little louder.

Do you want me to get my ADC out again???

Mani.

Mani,


Well, I have been thinking what we could do with the gear you have. I was thinking about something without microphone. And your ADC would be right that ...

Use this test tone :

1000Hz test file (http://www.stordiau.nl/downloads/Test1000-120-44100.rar)

and play at -0dBFS and at -1.5dBFS.

I am not sure how to calibrate the recording level but I think the only importance is that you don't record too loud. So be low 20dB to be on the safe side, is no problem.

The signal itself is at -3dBFS.

Peter

I forgot ! ...

... and send the result to me.
Recording of a few seconds only is sufficient. Make it 10 or so.

Peter

Will do later today...

But I was also thinking of simply making two 10 second recordings of a track where the effect is pronounced, one at 0dB the other at -1.5dB. You could match volumes at your end and take a listen - as long as you're not at 0dB, you should hear what I'm hearing, no?

But I'll start with your test.

Mani.

Mani, that could be a good idea. But assumed I hear the same as you do (which will just be the case) I next must "see" something anyway. The other way around (sort of) : If I see the test signal but don't see anything strange, I'm curious what you actually perceive from it (music, not the test signal).
So I guess both are as useful.

Regards,
Peter

Here they are...

The recording chain is:

Mach II -> NOS1a -> Prism ADC -> Tascam spdif input

1. Drum kit (XX at 0dB):

https://drive.google.com/open?id=0B0PU5LO5jVjfMVQyYjg2Tkp2NEE

2. Drum kit (XX at -1.5dB):

https://drive.google.com/open?id=0B0PU5LO5jVjfajJkN2N0SWl6Uk0

3. -14dB 1kHz sine wave (XX at 0dB):

https://drive.google.com/open?id=0B0PU5LO5jVjfREwwS0pwMkd2Ujg

4. -14kHz sine wave (XX at -1.5dB)

https://drive.google.com/open?id=0B0PU5LO5jVjfWURHS0N4RHg1N2c

I used 'my own' sine wave for the last two (ripped from a test CD) because it was already at -14dB and meant I didn't need to go through the preamp to attenuate further. But if these 2 files are not good enough for you, then I will redo them with the 1kHz file you linked to, but won't be able to sort the ICs from the preamp to the ADC until this evening.

Listening to files #1 and #2 here, the effect I described at the beginning of this thread is very clear to hear.

Mani.

Haha...

I can play tracks #1 and #2 at any XX volume settings (track #1 played back 1.5dB lower than track #2) and the difference in sound is obvious. BUT... when I play back at -1.5dB and 0dB respectively, they sound identical!

Mani.

Mani,

I can't use your test signal. Only my own. Sorry.

Peter

No problem Peter, I'll redo it, but it'll have to wait until this evening because I need to sort some extra balanced ICs out from the preamp to the ADC.

Mani.

I thought it might be useful to redo the files with an all-digital recording chain: Mach II -> AR-T Legato USB-to-spdif -> Tascam spdif input

Here they are...

1. Drum kit (XX at 0dB):

https://drive.google.com/open?id=0B0PU5LO5jVjfWnF3QkFWRVpkd3M

2. Drum kit (XX at -1.5dB):

https://drive.google.com/open?id=0B0PU5LO5jVjfLUw3VEhGUTc1enM

(On this track, you may notice that the L/R channels are reversed for the first second or so - you're not going mad, it's just a bug in the Tascam's 'auto record start' function.)

3. -14dB 1kHz sine wave (XX at 0dB):

https://drive.google.com/open?id=0B0PU5LO5jVjfVUt6NE4xZ0t5N2c

4. -14kHz sine wave (XX at -1.5dB)

https://drive.google.com/open?id=0B0PU5LO5jVjfaFljcGJCRTBfdUk

(Again, I'm using 'my own' sine wave for the last two.)

Mani.

I'm hearing the effect on these digital recordings too. So, it has nothing to do with being in the DAC's sweet spot.

Mani.

What is the upsampling rate you use for the AR-T ?
IsNOS1 should be inactive, but please confirm whether you really had it inactive during the recording (and the audible judgement).
Filter settings ?

These questions are only for me setting the exact same when I'm diving into the code to look for the unknown.

I think that the testsignal recording (of my own file) will be useless because if the origine is in XXHighEnd, it can only be that I see the same as always. Am I right ? So maybe don't put too much effort in getting the interlinks etc. Maybe later.

What I do wonder is whether the Peak Extension implies the same effect. I am not sure yet why I ask, but I recall some combination / oddity with something like the "base" volume of "no attenuation" and Peak Extension, me never being able to find that culprit and made a work around instead. So possibly all is related to this work around. *If* I am able to find back this particular piece of code ...

Peter

With the AR-T I had the upsampling slider set to '1', 'Is NOS1' = inactive, and no filtering (both Custom and AP off).

What's interesting for me is whether 0dB actually sounds better/more accurate than any >0dB attenuation. Your large SFS and other Q settings seem to work well for >0dB attenuation, but not for 0dB - they make it sound too 'thick'.

Mani.

OK, I'll just assume that no-one else is hearing what I'm hearing... :(

Mani.

Hi Mani,

I once communicated with Peter (maybe last year or somewhere in the beginning of this year - can't recall accurately) about the XXHighEnd digital volume. I use a preamp because I find the double output practical to feed 2 active subwoofers.

At the time and before I got my NOS1, I was not using the digital volume, so it was always at 0db. After some time (it took me time to realize this) I felt that sometimes the sound was harsh and it was without a doubt due to the level of compression that many of my CDs had.

So I experienced different levels that would be convenient to overcome this situation. I started using -6db, then -12db. At some point I felt that the music was "leaner" than usual and communicated with Peter by PM to ask if there were recommendations about this. Peter told me about a "sweet spot" that could maybe help. I finally managed for -18db since -21 was a little too low.

From time to time I was coming back to -3 or -6db because I felt that the music was fuller. In all honesty I wondered if I was imagining things. I am still unsure nowadays.

The reason why I hesitated to chime in was because I am not sensitive to many aspects of the settings dials. I try to follow the suggested adjustments but at an early point I fail to hear differences. What is obvious to some of you is beyond my ability.

Do I still hear this "fuller" sound near 0db ? I am not sure. With the NOS1 now a NOS1a, there are differences and for the better. I could do some tests but I feel quite limited even in testing.

So the few tests I did then are beginning to "dilute" as time goes by, but at the time I really felt there were differences. Near 0db some music seemed to have more body, but I never really used a meter to adjust the volume more accurately, so I can't say if my approximate adjustments in gain were ok or not...

Regards,

Alain

Hi Alain, thanks for sharing.


I just want to make it really clear that what I'm hearing is not a gradual build up to a fuller sound as 0dB is approached. It's a definite ON/OFF. At -1.5dB, to my ears the sound is pretty much the same as at -9dB, or -15dB, provided there is no clipping in the material. But at 0dB, it fundamentally changes.

Anyone should be able to hear this with the files I've linked to. Play the 0dB file 1.5dB lower than the -1.5dB file. The level you play back on doesn't matter, as long as it's not -1.5dB and 0dB. For example, playing back at -19.5dB and -18dB are fine. If you're hearing what I'm hearing, there will be a quite big difference in the way they sound, at yet the signal level reaching the DAC will be identical.

Mani.

Hi Mani,

Any news on this ?
Please keep in mind that I simply can-not test such things myself because it requires a pre-amp and I don't have that.
If you don't feel like testing it yourself, you can always just record it at -0 and -1.5dBFS output level. Then I can test it myself (I just attenuate both situations and the difference (if any) will be preserved).

Thanks,
Peter

Peter, I've moved all my recording equipment back into the basement and won't be able to do anything now until the end of next week at the earliest.

However, I've just taken a quick listen and it's interesting...

Peak Extend OFF:
0dB - thick, full sound
-1.5dB (onwards) - much more sparkle and life, but maybe too thin for some material

Peak Extend ON:
0dB - thick, full sound
-1.5dB (onwards) - thick, full sound (exactly the same as 0dB, but slightly quieter, obviously)

I think your thoughts about PeakExtend playing a part might be right.

Mani.

Okay Mani !
I will take your observations for granted (read : no further recording necessary) which (P.E. thing) is SUPER useful. Tis really gives me something to alost physically (text)search for.

Thanks a lot,
Peter

Perhaps my most controversial post ever...

There really is something weird going on, but it's not exclusive to XX - HQPlayer displays exactly the same behaviour. If the volume is attenuated anything below 0dB, the sound changes in the way I described in my first post - it appears to get clearer and more detailed. But I'm now convinced that 0dB (for both XX and HQP) gives the 'correct sound', and that with attenuation >0dB, the sound actually gets 'thinned out', and that other weird effects (such as a very slight 'echo' effect) come into play, giving the sound a 'lively' character. This seems to be counteracted to some extent by using the Custom filter, or indeed the Phisolator, but I think these are band aids to a certain extent.

But there does seem to be a 'fix' for those who refuse to go the preamp route (everyone here?):


In XX, using PE seems to keep the sound character exactly the same, irrespective of XX's attenuation level. What on earth is going on here, I have no idea.

[Interestingly, I've only started really noticing these effects since moving to single-ended tube amps and horns (in both my systems). Not sure if there's any connection. Perhaps for those with other system setups, this whole thing is simply a non-issue.]

Mani.

Mani,

I think with Peak Extension the volume is 3dBFS down internally to begin with. So set the slider at -0dBFS but in reality it is at -3dBFS. So the attenuation already took place. Now going down with -1.5dBFS is not changing from -0dBFS to -1.5dBFS but from -3dBFS to -4.5dBFS.

So I think what you see in the end, is expected.
The thing as a whole is still a mystery of course. But maybe something is happening like a kind of overshoot which crosses the voltage rails (we know that the margin is not all that large, right ?) and which won't happen with a small digital attenuation beforehand. If so, this should be happening at the higher frequencies and not with the lower we test with (1KHz). AND :

... I suppose I dreamt something about you telling that "it is since the SE tube amp and speaker" ...
(but I'd rather think you just removed that text)

It can just as well be caused by that amp etc. Same thing about the voltage rails, BUT you'd need to play at maximum "power". Read : your amp could be clipping.
Not sure !

Peter

If I remember correctly, Mani was hoping for more output from the B'ass I/V so that he could drive his amplifiers louder.  This would indicate that the amp is not clipping with anything the G3 can throw at it (nominal 1.5vrms I think?).  Of course that is no guarantee that the amp in not clipping, but is an indication that it probably needs more than 1.5vrms to achieve full power, and therefore induce clipping, so 0dB attenuation in XXHE should be no different to -1.5dB.

I am determined to find an active preamp that does justice to the G3 but that is proving difficult.  My most basic needs are instant muting and source selection...neither of which the G3 can do.  Analogue volume control that does not mess with the sound would also be nice...

XXHighEnd 2.08 can mute instantly ...

Nice!

Hi Anthony,

Hmm ... that makes me think twice. But I think that only for some sense of "logic" you may be right. This does not mean that technically you need to be right per se. So I think the other way around :
With exactly the same thought (there's lack of power in general), you thus feed the amps with full voltage (whatever that is for the moment) and thus for that situation it requires the most power. So of course you can think that a DAC with e.g. 2x the output voltage should be able to work too for any situation (because "an amp is an amp" ?) but it doesn't really work like that. So I think the other way around : we have some amps that suffice for 109dB sensitive speakers (at say 3 Ohms) and that thus siffices the same for 115dB sensitive speakers (at 3 Ohms) ? I don't think so. And thus : if the amp suffices "just" for the 115dB speakers it lacks a factor of two for power on the 109dB speakers.

So all what's required to let the amps clip is :
a. have underpowered amps;
b. feed them with a voltage which is just more than they can handle.

Ad b.
Make that a little less (digital attanuation or by other attenuation means) and they are fine.

Right.
And now I say that all is moot because I said that I have dreamt that I saw a text from Mani telling about "since this tube amp and speakers" but I don't think I dreamt and Mani rather pulled that text (because ... and I didn't look it up - Mani had the problem prior to that already).

I already said "not sure !" and it is far more easy to think that the NOS1a clips. But then "unnoticed" for me and it should be related to a frequency because I coincidentally checked it for exactly this phenomenon (clipping) and it does not (but with 3dB more IIRC it does). It does not at 1KHz.

While this would be some technical explanation from my side, I only present you this information because it is an understandable explanation (I hope). But what I really think and should say is that each NOS1 has its sweet spot for attenuation which is related to the PCM1704 chips which has the sweetspot and which is not equal at al for each chip. But as you can understand in itself, this is not any explanation to bring forward because nobody can deal with it (I wouldn't be able to either). From there, another explanation is more feasible and this is that the MSB (Most Significant Bit) from the chips (or one of them etc.) implies a "not the best" sound. So remember, when we attenuate digitally even the slightest, that bit is turned off and never comes up throughout and piece of music (no matter how loud). Or an even better explanation for those understanding the PCM1704 : that one half of the chip (forming the 24th bit (MSB)) is shut off and which is special (regarding to all of the other bits).

... And of course it is already known that about each person trying to digitally attenuate with a preamp (or analoge volume of other means) present, always comes up with this 6dB. I never tried it, but this may not be a coincidence.
So in the end it is nothing new ?

blablabla ... :)
Peter

A very quick post for now, with a few points...

1. I don't think it has anything to do with the amp - the effect is the same in both my systems (totally different amps)

2. I noticed this quite a while ago, when I still had the Orelo speakers (though I may not have posted immediately)

3. The 'character' of the sound at -3dB with PE is very different to that at -6dB without PE - levels identical, so this has nothing to do with the 1704 'sweetspot'

4. The 'character' of the sound at 0dB without PE is similar to that with PE at any attenuation level

More later (gym first!).

Mani.

Mani, are you talking about this happening in two different systems with two different NOS1 G3, both systems tubes/horns, regardless of HQPlayer or XXHE doing the digital attenuation?  If so, then I assume that each system has different tube amps which leaves the G3 as the common denominator.

Perhaps what Peter says above regarding the PCM1704K chips is relevant, but you have two G3's and it is probably unlikely that they would display the same behaviour if it was a quirk between different individual 1704's.

A lot of tube amps need 6VRMS or more before clipping, it just depends on the first stage tube and to a lesser degree the circuit, but the winding of the output transformers and the speaker load, as Peter says, can also induce clipping in the amplifier.  That both of your systems display the same trait still leads me to think there is something about the B'ass driving those amps, but even that is unlikely in my view.

Do you have a spec for the input sensitivity of the amplifiers?  What about impedance plots of the speaker load that they are driving?  Probably not to the latter, but it can be measured, and you may even have the gear there to do it.

LATER:  Reading your post above Mani, where you mention the Orelo's, perhaps it is something that others with that "known" system can try to emulate?  Surely you must have some other way of attenuating full power with 20w/30w amps and 118dB/m/v sensitive horns?  How do you even test that?

Thanks for this post Peter, it is most informative, especially the bit I have made bold.  In the past I have noticed that no attenuation does not sound as good as just a little bit, but I always put it down to the untreated room I was in and thought that the "problem" was a result of  non-optimal acoustics as more sound power was added to the room.

I have not bothered to find the attenuation sweet spot with the preamps I have here but maybe I should one day.  Have been waiting to properly treat the room first before I bothered with such relatively minor tweaking, but I also have other more pressing audio projects on the go...

Anthony, by now I think you saw it yourself (reading Mani's post later than you wrote yours) but this was already happening prior to the G3/B'ASS upgrade.

I also dare say that Mani has "visions" (don't want to call it issues) on this from what ? ... 6 years back ? or was it 8 years because even prior to the NOS1 ? I think so ...

Peter

Yes, I have read that now.  I am not sure how Mani tested it with the Orelos though...it must have been far to loud to listen with no attenuation.

Oh yeah, I've been playing around with active, passive and digital attenuation for quite a while now. For a long time, I refused to use digital attenuation because it just sounded wrong to me, and used instead an Audio Synthesis balanced discrete resistor passive or a Pass Labs X1 active. I switched to using digital attenuation when I received the Orelos.


With the Orelos, I just compared -21dB_no_PE with -18dB_with_PE - the difference in sound was obvious. I assumed that PE was messing the sound up in some way, but I now wonder if attenuation without PE is the issue...

Mani.

Yes but -6dB without PE still sounds different to -3dB with PE. You may feel that this is obvious - I mean, PE should be doing something to the sound, right? But I'm interested in why PE at any level sounds similar in 'character' to 0dB.

Any attenuation level without PE now sounds wrong to my ears. There's an edge to the sound. A 'hollowness'. The full-bodiedness of instruments and voices is gone. But if I switch to a Custom filter or put the Phisolator in the chain, PE simply sounds too dull.

Mani.

OK Mani, so I must start listening to everything just with PE engaged, right ?
So I will !

Peter

PS: As you may recall I myself am quite confident that something *is* going on, just per your reports about this all so often, combined with the fact that what all happens is complicated so prone to error. But then you may also recall that I can't find any bugs in the code and I looked so often (in the past) ...

Great! I'd recommend everyone who currently uses XX's digital attenuation to try this - takes 5 seconds to click on PE and decrease attenuation (i.e. increase volume level) by 3dB.

BUT... I almost guarantee that the sound will become duller. My feeling is that you should stick with it, and adjust other software parameters and/or hardware configs until the balance is restored.

Interested in hearing others' thoughts...

Mani.

Yep, PE+3dB sounds different to no PE.  I prefer no PE.

How would you describe the difference?

Mani.

Well, with PE is more "congested", but that sounds bad which is not necessarily so because it still sounds very nice.  There is more substance, a smoother, thicker sound but what is lost is some dynamics, transients and detail which is replaced by a certain monotony, or lack of diversity of sound. 

I've not spent a great deal of time comparing the two but I have found the difference easiest to hear when listening to electric guitars...think Pink Floyds "Coming Back to Life" from Division Bell with the variation in guitar techniques...I find the difference relatively easy to spot at the start of that track.

Non-PE by contrast is livelier thanks to the dynamics and transients and seems to have more time or space between the notes.  It sounds full to me but not with that hint of bloat and monotony.

Who knows which , if any, is correct?  I just know that in my system as it is at the moment that I prefer the normal VC without PE activated.

Hmm ... That would be about 100% my description. One quite important thing :

I am judging other things these days, and at trying PE I right away felt disturbed because it influenced my original listening. So really after 10 seconds I thought I shouldn't do this at this time (and switched it off indeed), but was left with something like the above description anyway. The "congested" would nicely summ it up because for me that was the most jumping at me.
I didn't report it because obviously this hasn't been "listening" at all. But now Anthony tells about the same ... why not admit I heard that too.

Peter

Well, it seems we're all pretty much agreeing:


My observation remains though that both XX and HQP sound very different at 0dB vs. any level of attenuation... and that this has nothing to do with clipping or the 'sweet spot' of the the DAC. And also that in XX, PE has a similar character to 0dB.


Yep, kind of like the vase vs. the face - does PE make the sound 'too congested', or is the sound 'artificially etched' without it? Take your pick and see (hear) what you want to.

It's going to take me a few more weeks to get my little 'vinyl studio' back up and running. Once it is, I intend to compare some needle drops to the original vinyl. I'm hoping that this will give me a clue as to which digital volume setting is the 'most correct'. But even then, I doubt I'll be able to come up with a definitive answer because of so many variables involved (e.g. the transparency of the ADC, etc.).

Mani.

You are in a unique position to test this Mani given that it sounds like you are wringing every volt from the B'ass by using no attenuation.  Only Peter could tell you how PE changes, or does not change things and if it behaves differently to the normal digital volume control.

I can also see now why you think the more congested sound may be the correct one given that you observe that it has the same sound characteristics as -0dbB.  But that begs the question "does -0dB have the correct sound."

Hi,

I recall that I even compared the byte output (after the different means of attenuation) and there was no difference. But maybe this depends on the actualy music (file).

Peter

Haha... I had some 'good ears' over here for the weekend, and although we didn't do a detailed listening test, everyone heard more 'sparkle' when I went from 0dB to -1.5dB. One of them thought I had actually increased the volume!

All-in-all, I think my original findings (that 0dB sounds dull and that >0dB attenuation sounds 'better') now stand. Why PE destroys this better sound, I have no idea. In any event, I'm back to no PE.

Mani.