To SAM or not to SAM and DRC

[yabaVR]

So I'm on the fence and start a new thread on digital alteration of a music signal. Is it nonsense or not? If it's not who benefits and who undergoes different experiences?

At first...I love the sound of my D200 but I don't use all the features Devialet offers me. You can call me a purist on sound and...yes I am.

I will c/p some parts of my other postings where I went a 'little' bit too far forward based on the original post   .
Ok, here it comes:

Most speakers have wildly turning phaseshifts in their low frequencies. Phaseshift, in short, means that these low frequencies (who once were time aligned to a tone that carries many frequencies) now seperate from this original tone in time. These low frequencies run beforehand or behind the original tone they belonged to (on the timeline).
This seperation causes fatigue/disruption because your brain, while listening, trys to assemble these frequencies to the original tone all the time. That's why when there is no (or less) phaseshifts in the music you can say at once "Hey, this sounds very natural". It's an ease for the brain. It notices right away the less (or not appearing) phaseshifts and doesn't have to work (thus calms done).
Got it? Difficult for me as english is not my native language.

For your brain fixing things on the timeline is much more difficult than e.g. fixing different volumes in the frequency response of speakers.
Therefor Jitter (digital Phaseshifts) bothers us so much. That's the 'digital sound' we all complain about. The brain has to work in large amounts on these digital phaseshifts.
Now with SAM it's a mixture of both analogue and digital phaseshift. Very complex.

Devialet with SAM measures these (analogue) phaseshifts for each speaker model and 'bends' them digitally inside the music signal (process them on the timeline/volume) to the position THEY think it would be optimal.
However, it's their 2D(imensional) approach on a 3D job. You loose Bit-Perfection of the recording with the first touch on a bit with this.

For many here SAM works and they are happy. For me and some other D owners I know it only works at the expense of loosing ALL the fine details (especially room mapping of recording and transients).
Maybe because of loosing these details you also loose the Jitter of these details thus your brain has to work less and you're experiencing the result as sounding more analogue? I don't know.
It would be better though to work on the problem another way and maintain the details of music.

Test:
To experience for yourself what I'm hearing listen to a recording with no bass but high frequencies (single instrument, woman's voice) and many of room information (best is live recordings). Listen - switch off SAM on RC - listen again. Then you know if SAM alters something in your music...may sound better or worse...your comments.

again c/p

As far as SAM 'only' deals with low freqeuencies.

Have you ever listened to a system with a perfectly speaker matched subwoofer with only high frequency music like lets say a Xylophon or a solo woman's voice.
You would say there is no bass in the recording BUT you will hear a dramatic collapse of stage/room mapping when you turn off the subwoofer.

How to explain this? I'll try but I'm not safe on this one.
Even tones of high frequencies resonate in a room and are alocated to (generate?) low frequencies of very low volume level.
These low frequencies belong to the original high-frequency-tone and imprinted is the precise 3D-Location of them in 'time-volume-freq' because these values are unique for the room they were recorded in.

When listening with a perfectly speaker matching sub the increase in holographic room mapping is huge when the sub is switched on.
So, if you think the reverse way, it would be most likely, if you swirl around with the bits and bytes of low freq (as SAM does) you loose all the informations of room mapping because these altered low freq get displaced out of phase/volume and hence can not be allocated to the original tone by the brain. These 'Out of Phase informations' will just add upp to a fidgety background for which reason some call it 'the blackness of background increases' when such freq are back in phase. Almost for me this is quiet coherent.

..ummpf...weird stuff though when I reread it



This does not mean that SAM or Digital Room Correction is ever to be avoided but possibly CAN be avoided when refining listening room, speaker/listening position etc. is an option for you. Though the whole procedure depends on wether or not you hear a difference with the SAM On/Off test because elsewhere it gets obsolete.

I would appreciate repliers to do this quick test mentioned above with SAM On/Off so we speak of similar experiences. 

Big point. If I'm wrong with my theory, can someone try to explain the precise function of digital altering of a music signal to me?

gui

[Antoine]

'Confused' already responded and did the test which he described here: http://devialetchat.com/showthread.php?t...5#pid53035

This is what he wrote:

OK - I have just been trying the 'Gui' test. (Per #53 above)

Very interesting!  I used the Devialet remote pseudo blind test, that is you repeatedly press the SAM on / off button to the point that you do not know if it is on or off, sat in the listening position I cannot see the display on the Devialet.  I did manage to find a number of recordings with more or less zero bass as suggested, all 24 bit and decent recordings.  Toggling SAM on and off I, I could tell a difference between each one.  In terms of details, I could notice a very slight improvement / reduction in the fine details.  The first couple of time I checked, the slightly better details were obtained with SAM off.  I was able to repeat this a number of times, there was one occasion where I preferred the SAM on version, but other than that, each time SAM off prevailed.  I recall YabaVR (Gui) posting that SAM 'looses all the fine details'.  I would not put it as starkly as this, for me it was a very small reduction in fine details, very small but clearly discernible.  As for soundstage, this was a bit hit and miss for me, so I reckon this aspect may be more system / listener specific.  Sometime I preferred the soundstage with SAM on, but I think a fairly consistent result was a narrowing of the soundstage, which in same cases I think I may have preferred as it came across as a more accurate sound stage, possible because it was actually less accurate!  So yes, definitely a ***** (replace ***** with insignificant, small, major / unacceptable, depending on your perspective) loss of details. I think the sound stage effect could be system dependant, but in my case it was a narrower sound stage that perhaps seamed more precise and more detached from the speakers, which could mean more mono!

Well done to Gui for spotting this and posting such a clear method of trying it for yourself.

Being slightly less serious (although there may be a serious point lurking here), I can now propose the Confused SAM test.  This is for the very brave only.  Play the first 30 seconds of Dizzee Rascal's 'Dirtee Disco' on a reasonable volume. Please note that it is neither safe nor sensible to play more than 30 seconds of this track.  I suspect that Confused SAM test may also be very system dependant.

[Confused]

Thanks' Antoine, you have saved me a job!

This is definitely an interesting test, and quite simple to do so I would recommend this as something to try. If you want to know how much effect SAM has in the fine details in your music, there is nothing like listening for yourself. @yabaVR does have a valid point here I think.

Although I have to say that in general listening I am someone who prefers SAM on, for me it does improve the accuracy of the bass and improves the coherence of the sound stage. Also it does seam to clear up the mid-range with a lot of the music I listen too, quite what the mechanism is for this I am not quite sure. My understanding is SAM is designed to mitigate phase errors which are a fundamental part of the mechanical restrictions and compromises in speaker design, after all, a 8" bass driver cannot react as quickly as tweeter, thanks in no small part to the laws of physics.

To quote Devialet: "In its current implementation, SAM is effective on lower frequencies, up to 150Hz; yet its effects can be heard throughout the audio range.
By its principle, SAM ensures that the acoustic response is at every instant an extremely faithful image of the recorded signal ; this is true on transient signals (notes attacks, drums, …) as well as on continuous signals (notes sustain, organ…). In particular, SAM prevents any delay between bass and the rest of the spectrum, avoiding the effect of the phase rotation at the vicinity of the natural low-frequency cut-off of the loudspeaker."

http://en.devialet.com/assets/Systems/De...per-EN.pdf

I suppose the finest aspect of SAM is that it does have an on / off toggle on the remote, and we can all make our mind up what to do with that depending on the mood of the day. I would urge others to try vabaVR's test though, fascinating stuff.

[ogs]

Hi gui
there are weaknesses in your reasoning here; you assume that the loudspeakers in use are transparent and 'correct' sounding. Very few speakers have linear phase response. Example: the Kef LS50 has the tweeter inverted, many (most?) Wilson models invert the midrange driver, many High End speakers do not even try to have a decent time response. If a dome tweeter is mounted on the same baffle as the bass/mid, the treble arrives too early. You'll have to account for this when you develop your theory. Your ML speakers is very good in this respect, I am sure.
You may be correct when you say that digital room correction reduces information density, but this is not my experience. I use Audiolense with long filters (132K) and can not say I miss something. If I use short filters (8K, 16K) I can hear the 'coarseness', but not with the long ones. Your test with SAM on/off and a female voice is interesting. I will try this as soon as I get my amp back from upgrade. Your binding post mod will be hard for me to resist, however! And what about this 'secret mod' you have mentioned?

[Soniclife]

OK - I have just been trying the 'Gui' test. (Per #53 above)

Very interesting!  I used the Devialet remote pseudo blind test, that is you repeatedly press the SAM on / off button to the point that you do not know if it is on or off, sat in the listening position I cannot see the display on the Devialet.  I did manage to find a number of recordings with more or less zero bass as suggested, all 24 bit and decent recordings.  Toggling SAM on and off I, I could tell a difference between each one.  In terms of details, I could notice a very slight improvement / reduction in the fine details.  The first couple of time I checked, the slightly better details were obtained with SAM off.  I was able to repeat this a number of times, there was one occasion where I preferred the SAM on version, but other than that, each time SAM off prevailed.  I recall YabaVR (Gui) posting that SAM 'looses all the fine details'.  I would not put it as starkly as this, for me it was a very small reduction in fine details, very small but clearly discernible.  As for soundstage, this was a bit hit and miss for me, so I reckon this aspect may be more system / listener specific.  Sometime I preferred the soundstage with SAM on, but I think a fairly consistent result was a narrowing of the soundstage, which in same cases I think I may have preferred as it came across as a more accurate sound stage, possible because it was actually less accurate!  So yes, definitely a ***** (replace ***** with insignificant, small, major / unacceptable, depending on your perspective) loss of details. I think the sound stage effect could be system dependant, but in my case it was a narrower sound stage that perhaps seamed more precise and more detached from the speakers, which could mean more mono!

Well done to Gui for spotting this and posting such a clear method of trying it for yourself.

Being slightly less serious (although there may be a serious point lurking here), I can now propose the Confused SAM test.  This is for the very brave only.  Play the first 30 seconds of Dizzee Rascal's 'Dirtee Disco' on a reasonable volume. Please note that it is neither safe nor sensible to play more than 30 seconds of this track.  I suspect that Confused SAM test may also be very system dependant.

I just gave this a try, I didn't find it easy finding recordings that fit the requirement, most solo female vocals recorded in a big acoustic have ambient grunge going on low down that SAM reproduces more accurately, and expands the soundstage as I would expect, studio recordings don't have much acoustic space present, and it's likely to have been through so much processing as to be unreliable.  However I never found less soundstage or detail in any track with SAM on, often the opposite, and I always prefered the with SAM version overall.  This makes sense to me, part of the female vocal range is within the operating frequencies of SAM.

PS Maybe all of what I tested can be discarded because I did the Confused test first.

[yabaVR]

Thanx Confused for testing and having the kind words.

Ok.
The most difficult thing in this test is the transparency of your system chain...I know, I know, everybody in this forum would claim his system for being very transparent   .
But may be there is a here and then of someones system resolution being a little bit better than the average   ...you know what I mean   .
For those having a 'highly' transparent system the effect the test discovers will be of a large amount, I promise. Saying that we did this test with 'highly' tuned rooms and systems here.

Does anybody take a thought discourse on the Bit-Perfection? Am I right that Bit-Perfection is out the window if the music signal is touched by e.g. SAM, etc.?

gui

[yabaVR]

Hi gui
there are weaknesses in your reasoning here; you assume that the loudspeakers in use are transparent and 'correct' sounding. Very few speakers have linear phase response. Example: the Kef LS50 has the tweeter inverted, many (most?) Wilson models invert the midrange driver, many High End speakers do not even try to have a decent time response. If a dome tweeter is mounted on the same baffle as the bass/mid, the treble arrives too early. You'll have to account for this when you develop your theory. Your ML speakers is very good in this respect, I am sure.
You may be correct when you say that digital room correction reduces information density, but this is not my experience. I use Audiolense with long filters (132K) and can not say I miss something. If I use short filters (8K, 16K) I can hear the 'coarseness', but not with the long ones. Your test with SAM on/off and a female voice is interesting. I will try this as soon as I get my amp back from upgrade. Your binding post mod will be hard for me to resist, however! And what about this 'secret mod' you have mentioned?

No, you got me wrong. I know loudspeakers are not correct sounding when it comes to phases. And the ML are by far no reference for me.

But why does it matter anyway. For me this is a good distinction for good and bad loudspeakers. The one which is more neutral on the phase is mostly the better sounding for me.
The music signal is not meant to compensate on this. It's a 3D-Matrix. You can not cut in and take one 'thing' (freq/vol) out of it. If you do you cut this 'freq/vol thing' out of every (sound) object in the 3D-Matrix. You can only do harm to the signal when you try to solve problems here though their cause is further down the system chain.

It might not be easy to imagine what I mean about this 3D-Matrix (freq/vol/phase) of the music signal. 'freq/vol/phase' of a sound in a room is unique at every point inside this room and these points are dependent on each other. Think of reconstructing a single moment with all sounds in this room. The values of 'freq/vol/phase' have to be variable to describe every point in the room. You can not use a set value for a unique freq to be assigned to every point of its occurrence because this unique freq might belong to a saxophone AND also to a voice AND a guitar AND...it has to be different in 'vol/phase' at every of this points.
Man, it's hard to get it down to words  
Think of it again and it might do a 'klick' in your head    .

[Jean-Marie]

Thanx Confused for testing and having the kind words.

Ok.
The most difficult thing in this test is the transparency of your system chain...I know, I know, everybody in this forum would claim his system for being very transparent   .
But may be there is a here and then of someones system resolution being a little bit better than the average   ...you know what I mean   .
For those having a 'highly' transparent system the effect the test discovers will be of a large amount, I promise. Saying that we did this test with 'highly' tuned rooms and systems here.

Does anybody take a thought discourse on the Bit-Perfection? Am I right that Bit-Perfection is out the window if the music signal is touched by e.g. SAM, etc.?

gui

I will bite the bullet on this one. 

I don't think that bit perfect is what really matters. The ideal would be a system that reproduces at the listening location as the output of the speakers the exact sound Field that the microphone was attempting to capture. 

Before continuing I would like to express a number of postulates that underpin my reasoning:

1) 24bits PCM is inducing a quantization noise that is well below the audibility threshold
2) The dominant effect of an ADC DAC chain is the low pass filter that is required to fulfill the Nyquist condition, and is way beyond the audibility and even most speakers capability if you take a 96kHz sampling rate, so not even speaking of 192. 

We can debate those statements, in which case the following will have to wait... 

So the notion of 'bit perfect' is only valid if the rest of the chain is perfect. If it is not, then it does not really matters whether you correct the defects in the digital domain, the analog domain or the accoustic domain. 

The only problem is that it is always a matter of compromise and cost. The acoustic correction of a room mode can be much more costly and requiring much more real estate than a digital correction.

The digital correction will have other compromises. The biggest is that to compensate cancellation you need a lot of power that can quickly reach the limits of the speakers, meaning that most of the time room correction will only allow you to play at a lower level than the same system would have been able to play with an accoustic treatment. 

An other consideration is that a phase shift is a phase shift and an attenuation is an attenuation and it does not matter if it is done in the digital, the analog or the acoustic domain. Similarly, if I can insert somewhere in the chain the exact inverse function the end result will be much more accurate, and it does not matter in which domain the correction is done. 
It happens that digital signal processing is the most cost effective way to achieve it. So yes the digital chain is no longer bit perfect but the overall chain becomes much closer to the identity function. 

One last consideration is that our brain has evolved to discover changes and patterns. That was a matter of survival, therefore the hardest situation for us is to listen to two identical things and find them identical. 

My post is probably way too long already but let's have the ball rolling. 

Jean-Marie

[GuillaumeB]

Thanx Confused for testing and having the kind words.

Ok.
The most difficult thing in this test is the transparency of your system chain...I know, I know, everybody in this forum would claim his system for being very transparent   .
But may be there is a here and then of someones system resolution being a little bit better than the average   ...you know what I mean   .
For those having a 'highly' transparent system the effect the test discovers will be of a large amount, I promise. Saying that we did this test with 'highly' tuned rooms and systems here.

Does anybody take a thought discourse on the Bit-Perfection? Am I right that Bit-Perfection is out the window if the music signal is touched by e.g. SAM, etc.?

gui

I will bite the bullet on this one. 

I don't think that bit perfect is what really matters. The ideal would be a system that reproduces at the listening location as the output of the speakers the exact sound Field that the microphone was attempting to capture. 

Before continuing I would like to express a number of postulates that underpin my reasoning:

1) 24bits PCM is inducing a quantization noise that is well below the audibility threshold
2) The dominant effect of an ADC DAC chain is the low pass filter that is required to fulfill the Nyquist condition, and is way beyond the audibility and even most speakers capability if you take a 96kHz sampling rate, so not even speaking of 192. 

We can debate those statements, in which case the following will have to wait... 

So the notion of 'bit perfect' is only valid if the rest of the chain is perfect. If it is not, then it does not really matters whether you correct the defects in the digital domain, the analog domain or the accoustic domain. 

The only problem is that it is always a matter of compromise and cost. The acoustic correction of a room mode can be much more costly and requiring much more real estate than a digital correction.

The digital correction will have other compromises. The biggest is that to compensate cancellation you need a lot of power that can quickly reach the limits of the speakers, meaning that most of the time room correction will only allow you to play at a lower level than the same system would have been able to play with an accoustic treatment. 

An other consideration is that a phase shift is a phase shift and an attenuation is an attenuation and it does not matter if it is done in the digital, the analog or the acoustic domain. Similarly, if I can insert somewhere in the chain the exact inverse function the end result will be much more accurate, and it does not matter in which domain the correction is done. 
It happens that digital signal processing is the most cost effective way to achieve it. So yes the digital chain is no longer bit perfect but the overall chain becomes much closer to the identity function. 

One last consideration is that our brain has evolved to discover changes and patterns. That was a matter of survival, therefore the hardest situation for us is to listen to two identical things and find them identical. 

My post is probably way too long already but let's have the ball rolling. 

Jean-Marie

Great post Jean-Marie 

Guillaume

[Hifi_swlon]

I much prefer my setup with Dirac live DRC.

SAM does very little here. Yes it makes the facts go deeper, but i always feel it comes with a narrowing of soundstage, and in my average shared use lounge this extra bass doesn't really sound right anyway. I also sense a slight 'softening' of the sound with SAM.

With Dirac it just sounds 'right', and makes for a really nice fuller bodied sound, takes away the brightness, and equalises the LR volume imbalance and timing all in one hit. Plus it's flexible and configurable.

The only downside to me is the complexity (one extra piece of software to get in the chain) and taking the measurements is a bit of a faff and doesn't go down well in the household! Oh, and the other downside is I can't use it at the moment because of Roon. As soon as that changes - even if not Dirac - I'll re-enable it straight away.

The bit perfect thing used to worry me, but to be honest pretty much all modern DACs are doing something to those bits, and the Devialet certainly is too. i understand why we want it, and we certainly don't want stuff going on that we don't know about, but I'm less worried about processing with modern digital audio techniques.

Of course I'm also coming from the perspective of someone with an OK hifi in a shared living room soace where speaker placement is a secondary priority for the room and sound treatments are out of the question. For those using a dedicated room things are probably very different. I'd love to be able to do it without Dirac, but that's just not the reality. So better to just focus on getting the most enjoyable sound.