DigitalSignalProcessing -> future?

[Polimorph]

quote: "Traditional feedback corrects some distortions, but
can add others in phase, tone, IMD, and overload recovery.
Perhaps feedback DSP could correct for this." (I don't now exactlly how to quote with the command)

I'm not sure about this. But I think DSP can't correct such distourbances just because de DSP device is placed before the power amplifier, and there would be a delay between the original signal and the feedback correctional sound from DSP having in mind that DSP needs too the signal from the output of the amplifier to compare it and correct it function of the original. If this delay isn't big enough than the correction is possible.
Anyhow I have in mind a DSP project witch will correct as much as possible the distortions caused by the ambient (walls and other kid of obstacles). Maybe with this we could get rid of all high expectation equalizing when u move with the system a lot in diffrent places. Maybe I'll start a thread on this sometime.

 

[Majestic]

quote: "I'm not sure about this. But I think DSP can't correct such distourbances just because de DSP device is placed before the power amplifier, and there would be a delay between the original signal and the feedback correctional sound from DSP...

That, of course assumes the theoreticaly 'ideal' way of doing things (ie dynamically and continuously). However by moving it to the set-up stage only, you could pass a sequence of complex (but domain-known) tones through the system, and have the result analysed by the DSP to produce a 'template' which then gets 'overlayed' onto all music that subsequently gets played through the system. This reduces the processing overhead + eliminates the above mentioned time-delay problem, whilst still correcting for many of the non-linearities that you could measure through the said, system.

I belive that there was/is (and probobly a lot more now..) a DSP equaliser component made by Behringer(?) which did something like this (I'm not sure what exactly it corrected over and above frequency response). It was well revewed by T.A.S. which, given their persinkityness about audio gear (and esp. digital) is at least a good sign.



On the other point about what to do/"add" to make a "tube sound":
It's a currently accepted maxim amongst sceptical engineers at the moment (eg Mr. Douglas Self) that listener preference for tube amplification must be due to it's added "euphonic distortions". Whilst this is difficult to refute straight away, and in fact, is possibly true in a technical sense; my gut instinct tells me that this is much less than the whole story. I think <tubetvr> hit the nail on the head by aluding to the also possiblity that this assumption might also be wrong. In which case we'll be looking in the wrong place for the solution.

Given the fact that despite increasing resolution of our measurments and equipment, we still have largely woeful sound (with exceptions) coming from modern equipemnt and appear to be no closer to understanding why, I feel the later case to be true.

There's definately some reason why tubes seem to provide a more huristically enjoyable musical expereience, and this reason(s) should be quantifyable and therefore physically measurable. But we seem yet to find all the pieces of the puzzle to put together.

 

[DrDeville]

Hey Majestic,

You took the keystrokes right out of my fingers!

have the result analysed by the DSP to produce a 'template' which then gets 'overlayed' onto all music that subsequently gets played through the system

Yes, there are all sorts of ways this could be done. When I worked for Lockheed we designed satellite optical systems on the same principle--they would send a test beam throught the optics *and* atmosphere, and then apply the corresponding correction to the signal so that it arrived on earth undistorted. As our understanding and processing power increases, this sort of thing becomes more feasible.

There's definately some reason why tubes seem to provide a more huristically enjoyable musical expereience, and this reason(s) should be quantifyable and therefore physically measurable. But we seem yet to find all the pieces of the puzzle to put together.

Absolutely. Although we are far from quantifying acoustic effects now, I believe that we will be able to do so, and reap huge rewards.

Best,

George Ferguson

PS What *is* that in your Avatar? My best guess is a tube and a Lotto ticket?

 

[fdegrove]

EUPHORUS TUBULUS...

Hi,

It's a currently accepted maxim amongst sceptical engineers at the moment (eg Mr. Douglas Self) that listener preference for tube amplification must be due to it's added "euphonic distortions".

Certainly a lot of SET amps produce a fair bit of this so called "euphonic" distortion...

What is meant by euphonic?
Basically, better than the real thing...But how's that?

Well, quite simple really:
The SET amps add a fair amount of even order distortion where the second harmonic predominates.
So you can have 4% of second H, 2% of third, 1% of 4 H and so on, with ever diminishing numbers.

The ear is much more forgiving towards even than it is towards odd order harmonic distortion.
Even order harmonic distortion adds an artificial warmth to the music that's not there in the first place.
Odd order harmonics are perceived as rather cold sounding and with sufficient dominance can make an amplifier sound subjectively cold, boring and uninvolving to the point you will unconsciouslly stop listening to it.

When the balance is correct you'll be wanting to play your favourite records one after the other, when not you'll be looking to start a conversation to avoid having to listen to the music.
At least that's what I've often observed.

Is this so called "euphonic" distortion typical of tubes?
Absolutely not. It just depends on the topology of the circuit, that's all.
Certainly a bad tube amp is just another bad sounding amp no matter how many tubes it has running.

But there's more to tubes than just THD figures, clipping characteristics for one...The ability to handle large voltage swings graciously, noise spectra is yet another.

Ever had a noisy SS phono preamp?
Boy, that hissing really bugs me....Less, much less so with a tube preamp. Why?
Looking at the way the noise is spread across the frequency band is one clue...

Just a little tip of the tubeberg,

 

[Majestic]

Re: EUPHORUS OR AGAINSTUS?

But there's more to tubes than just THD figures, clipping characteristics for one...The ability to handle large voltage swings graciously, noise spectra is yet another.

Quite right. And the 'clue' to that, is that those SE-deficient amongst us (<puts up own hand>) still prefer tubes even though we're listening to push-pull. Granted as you note, that a very badly designed tube amp, is still going to be a bad sounding amp. But the former points towards something inherent in tubes themselves (in addition to the way they're used {read: circuits}) beyond euphony (read: SE circuits) that's musically appealing to our human ears. At least when referenced to comparable silicon designs.


On the distortion spectra thing-o:
Apparently our own ears are actually highly inherently distored transducers themselves (how they measure something like this, I have no idea..). But our brain seems to employ some kind of post signal processing itself, so that our overall perception of what we hear is normalised.

This is apparently, one of the main reasons why so many listeners find the 300B so so sonically favourable, since it's distortion spectra (as least in its original WE incarnation) apparently mimics almost exactly the distortion quanta of our own hearing. -- Both in terms of the harmonic products produced and their magnitude. (Why such 'mirroring' should supposedly be prefered by humans, is another thing I'm unclear of... )

 

[Polimorph]

Majestic - quote: "But our brain seems to employ some kind of post signal processing itself, so that our overall perception of what we hear is normalised."

Is this a real fact?

fdegrove:

Nice piece of information there. I was waiting for that. If u know more about noise or anything that tube does with the sound that would be great!

Thanks all, respect.

 

[Polimorph]

even ordered harmonics

It can be possible to simulate those with an overdrive plugin?

 

[benny]

part one

hi all,

this is turning out to be a really interesting thread.

There's definately some reason why tubes seem to provide a more huristically enjoyable musical expereience, and this reason(s) should be quantifyable and therefore physically measurable. But we seem yet to find all the pieces of the puzzle to put together.

i think that firstly, all that matters is what the end product sounds like. it doesn't matter if we don't know exactly what went into it, as long as we like what comes out. one tube amp also sounds differnt from another one, and one might sound better. it is possible to design a tube amp so badly that even a mediocre SS amp sounds really good next to it... something like a change of operating class can alter the sound of the amp quite dramatically as i pointed out in my first post...

i think we will never find all the pieces of the puzzle though. we by now know about the colourization of the sound a tube amp has that makes it more aurally pleasurable, but to actually understand everything that goes on inside a tube amp fully is a huge task. for a little insight into the complexity of a tube amp, read something like the spanberg tube manual, which is a highly technical analysis of how a tube actually works. it is very technical, and moves a long way further from the basics, into some quite complex physics, but it is quite interesting if you want to take the time (it's not exactly light reading)... but there are things in there, that even if you raise just one point out of here, and if you consider it's implications, you realise the highly random, and imperfect nature of tubes. a good idea is read the chapter on electronic emissions. But i think one thing for sure, it is the imperfections in tubes that add the colour to our sound that makes it nicer to listen to. based on my listening, i think another thing tubes have over SS which is something that you can't move into the digital world if you plan to then use SS amplification is that tubes are more dynamic and hill preserve the dynamics of the program material better than SS.

There seems to be discusion now about using digital to compensate for imperfections and non linearities in the SS amplification... what a great idea. now this would really be of benifit to your tube model. despite the fact that it is still not something i recomend you call a tube model because of the reason that you can't really model a tube amp, i do think with the 'nice' colourisation of the sound your software gives, it would be made even better by compensating for imperfections in the system. there is one flaw in this scheme though that i think has been overlooked.. that is using this to compensate for distortions.

Yes, there are all sorts of ways this could be done. When I worked for Lockheed we designed satellite optical systems on the same principle--they would send a test beam throught the optics *and* atmosphere, and then apply the corresponding correction to the signal so that it arrived on earth undistorted.

yes, this is a great idea, but there is something about light (optical) distortion and sound distortion that is different for a start (the distortions we're talking about, you're on about atmospheric distortions, we're on about electrical distortions, not until sound leaves the speaker cone is it subject to atmospheric distortions). i am not an expert on light wave theory or anything as you youself probably are (if you're not i worry about what you're doing working @ lockheed doing what you said you did)... but atmospheric light distortion is due to things like refraction, difraction which are dependant on frequency/wavelenghth and the density/relative densities of matter that it is traveling through. light distortion is the scatering of different frequenies as refration and difraction takes place... but electronic audio distortion (within an amplifier), unlike atmospheric light distortion is mostly affected by amplitude... so the problem here is, everytime the user wants to adjust the volume, they are going to have to stop and run another test, or otherwise the distortion caractheristics of the amplifier have now changed, and therefore, colorization of the sound from the DSP will now no longer be fully canceled out by the amp, and therefore the sound would be colourised again... indeed it seems like a futile task in this case.

one way around this problem could be to create an amplifier with some sort of interface with the DSP so the DSP can automatically know where the volume control has been adjusted too. so when you run your test, you run it at one volume, and when done, turn volume up a little, and run again, and so on, untill you've tested for say maybe 12 points from almost no volume, all the way up to full. with these points, the DSP has a pretty good picture of the distortion spectrums/amounts with respect to volume, and with an interface so it know the volume the amp is at, it can automatically kick in the model for that volume point. and this could be further improved. say if you set the volume to somewhere inbetween two of your test points, you could get the DSP to calculate a new distortion image for this point, based on the reactive patern of distoriton V's volume your DSP has recorded in its test, and get it to use this pattern to calculate where the new distortion spectrum sits between the two previously tested points. all this sounds pretty complicated i guess, but rest asured, there is less to this than there would be to modeling a tube amp.

(continue to read on, i wrote more, but got told by the forum that message is too long, so i split it in two.)

 

[benny]

part II

Anyhow I have in mind a DSP project witch will correct as much as possible the distortions caused by the ambient (walls and other kid of obstacles).

here is another seeming DSP imposibility in my opinion. just like trying to model a tube amp, this is a very complex thing to try and do. to do this, you are going to have to do the same scheme as above for fixing the amp imperfections, only you are going to have to test your signal in the room with a mic. no problem here, a good electret condenser has a flat enough freq. response to be acurate enough... but the problem lies here, you need to test, and this would be from a fixed point. now, just one point (out of many, like before with tube amps), is dealing with standing wave cancelations from both reflective surfaces and the two (or more) speakers. When waves meet, wether they be reflected, or the waves from the two speakers, or all at once, the amplitude of the waves sum together and we get what is called either destructive interfearance, or constructive... destructive is when we have both a positive and a negative wave meet. basic math tells you when you add a negative to a positive, it is the same as a subtraction really. constructive then is obvioulsy when two positive waves meet, and their amplitudes sum like in adition. so, alright, you measure it, and you got it covered? well, now move to another point in the room and measure again, you now get completly different results. problem is, you set up the amp for one place in the room, but unless you sit in that place and don't move, as soon as you move, the test is once again a futile taks, as you are now colouring the sound unnecesarily. we could look at the flipside and say that it is the imperfections canceling out the aditional colour you add, so when imperfections change, colour is no longer cancled. chances are the colour added to cancel imperfections won't be 'nice' colour either, just colour that makes the sound crap if not in the right context. to get an idea of a very simple demonstration of this, play a test tone through both speakers in a stereo setup so both speakers playing the exact same sound in phase with each other. and do this test outside to eliminate room acoustics as this will demonstrate the complexity of just this issue before you even worry about the effects all factors have on each other... anyway, set up a stero speaker arangement outside playing same test tone in phase with each other. now, say two meters back from the speakers, start at one point, and move paralel across the speakers with one ear facing the speakers. move slowly, and you should notice that although your distance from the speakers is constant, the sound gets louder and softer. this is because of the sound either ariving in or out of phase, and is dependant on relative positions of speakers. you get a series of what are called nodes (min amplitude, out of phase so therefore destructive interfearance) and antinodes (sound in phase, constructive interfearance, max amplitude).

another point is that of difraction and refraction. just like what DrDeville was testing for at lockheed, atmospheric variations, and physical obstacles distort sound. refraction within a room is not really of concern as i doubt it would play enough in the overall colour to bother with... it is difration that is the problem though, as the amount of difration is not a linear constant, and is dependant on position and frequency. baisically, the lower the frequency, the more something is difracted, and higher frequencies result in less difraction. so, the amount of difraction is dependant on frequency, so if we measure, and notice a certain rolloff in highend from a certain point, we could increase amplitude of our HF so the amount of difracted HF's is percieved relative to LF's the same as original... one problem though, as soon as you move, once again conditions change, and amount of different frequencies in the difracted sound we hear changes.

so it would indeed be a very complex model to compensate for room acoustics... actually, i will go out on a limb and say impossible. Even using a big multi speaker arangement like all that doulby suround sound crap so you have more sound sources to play with to help even out the room, i think like this it's still pretty imposible.

anyway, i think i've said enough for now, maybe more later, but yeah, hope this helps. i still think as long as you do all the rigt things, your DSP should benifit computer sound a lot, and i wish you all the best with it. by the way, i'll add it again, that this is turning out to be an interesting discussion, and like DrDeville said, this is definite
flame bait, but it's good it's all been kept civilised, and none of us have got tempered at all, let alone to the very posible point on starting to call each other far-king see-you-en-tea's (check out that for creative way to get around auto parser )

cheers.

 

[Polimorph]

benny

quote: "When waves meet, wether they be reflected, or the waves from the two speakers, or all at once, the amplitude of the waves sum together and we get what is called either destructive interfearance, or constructive... "

Now this is a problem. But all the problems can be eventually solved.

Benny there would not be like u'll have to set DSP each time you change something (anything in ambient or in musical chain). The model that I'm trying to make here will have only mono perspective for now. There will be a mic. - here u're wright - that will capture the sound from ambient. The ambient could be even open space. Now my DSP model has 2 branches, and it can be made after it an analog discrete component device too, but is easier for me to work with the computer:

1.the source of the original sound (A);
2.the mic (electret, or one with flatest response) and a preamp with gain control (B) (the real output).

These two branches come togheter in a comparator-substractor that actually substracts the signals one from another, so it will have 2 outputs A-B & B-A. B-A (all that's in ambient and is not in the original) signal would be combined with the original in oposite phase and A-B (all that's in the original and not in ambient) signal who will be added in phase in the chain.

The plugin is always on and will correct in real-time all errors by the feedback u can easly see. The problem is that would be a delay, but I just hope from all my heart that would not be too big.
Of course the first time I'll set this plugin to work I'll have to make serious setting with professional tools.

So God help us! anyway if this won't work I have all the time in the wold because I'm still so young.

Respect.

 

[Polimorph]

...I'm such a forgetful figure

I wanted to add:

 

[Polimorph]

sorry

My latest 2 posts were a little bit off-topic I'm sorry, maybe we'll start a DSP thread, but let's try to keep it on the correct route (DSP tube model).

 

[benny]

There will be a mic. - here u're wright - that will capture the sound from ambient.

here i am right, but here is also the problem...

the mic only measures ambience in one place in the room, and what ambience it picks up is limited by it's position and pickup patern. it does not matter if your in stereo, 5.1 doulby suround sound... or, mono.

the position your mic is in will give dramatically different results. position closer to the sound source, and it will pickup more of the original sound relative to reflections in the room due to increased intensity. position further away, and sound mic picks up will be interfeared with more by relections in the room. now, even at same distance, relative position of mic to a relective surface will alter delay between original sound and reflection reaching the mic, and therefore give different results. the same is true for the compensating for refraction and difraction, relative positions of sound source, mic and reflective surfaces will give very different results.

now.. your scheme could work... if it is intended for use by one person, and one person only, who is sitting/standing in the same place constantly, or, has a mic strapped to their head or something rediculous so as their position changes, the mic does too and will therefore be able to compensate.

another flaw in the system you propose of it being monitored constantly. just say someone came in and said something to you, or there was some other sound external to the source. the mic would pick up this sound, and this would make the mic percieve this as part of the room acoustics. no matter how good a programer you are, you will never make a program smart enough to single out individual sounds/voices. so the problem here is the sound will the be altered to compensate thinking the person talking is the room acoustics, but because it's not, your sound will now sound crappy.

so i hope this helps. the scheme should work, but only in the situation i said above about either fixed position, or user with a mic straped on. and both of these only work for one person.

cheers

 

[benny]

My latest 2 posts were a little bit off-topic I'm sorry

and by the way, don't be sorry for posting something off the original topic... as all you have been saying is related to your topic, or what others have said, then it is good... that is how we get an even more interesting conversation.

cheers

 

[Polimorph]

benny

I'm sorry to say this but.... ummm you're perfectly right (Not that YOU are, but because is RIGHT). This could be done asuming that there is a pre-defined listener's position. And of course any sound that can appear from the exterior and has nothing to do with the musical (or less musical) signal will sum too to the original signal. Crap!
I should think more. This could be too hard 4 me. Some ideeas?

 

[Polimorph]

DSP thread started!!!

You're all invited on the "what we could do with DSP?" thread that's opened in Digital section. There we can (and I recomand to do it) move with the questions about anything else than tubes. This thread should stick on it's first ideea (DSP & tubes).

Thankyou for understanding and respect.

 

[thoriated]

There's nothing necessarily wrong with the idea of using dsp to alter the musical experience. However, it's incorrect to assume that 'tube' sound consists of unique distortions and colorations - it's more correct to look at the matter in the sense that tube sound is more characteristic of the *lack* of distortions and colorations that characterize solid state, for instance. With this understanding, the issue becomes more clear.

One problem is the limited actual resolution of today's even 24 bit converters - adding some arbitrary simulated 'tube' distortion to an already digitized source may not degrade it too much, but to convert a high quality analog signal to digital, *then* do this, then convert back to analog again will significantly degrade the sound, and all we're considering here is the resolution loss of the conversion process which is will result in little better than 16 bit resolution of the processed signal with existing real world converters.

 

[DrDeville]

Reduce Distortion, Not Add It

I find this thread interesting too. Heretical as it may be to discuss digital techniques for reducing tube distortion, the idea offers promise.

Thoriated's dancing digits tapped:

it's more correct to look at the matter in the sense that tube sound is more characteristic of the *lack* of distortions and colorations that characterize solid state

Abs-to-lutely!

I'm interested in ehancing tubes' exemplary fidelity, not adding euphonic distortions. The former is infinitely harder than the latter.

As I understand it, some of reasons that tubes offer lower and less annoying distortions than SS are:

• Tube distortion decreases as the signal decreases. This means that quiet passages--where most critical information lies--are blessedly clean. SS on the other hand tend to produce a constant amount of distortion, so the % distortion goes up as signal goes down. This gives quiet passages a veiled, gritty quality.
  
• As Frank stated so well, SS designs tend to produce more odd-order harmonic distortions, which are less euphonic than even-order (particularly 2nd-order) harmonics.
  
• SS devices tend to have higher feedback. Feedback can cause many problems, including poor recovery from overload. This doesn't sound like a big deal until one realizes that even quiet passages--particularly well-recorded ones--often have very brief overload transients. A low feedback design can basically ignore these, but a high-feedback design can choke on them and take a while to recover. This blurs details in even quiet passages.

I'm thinking of using DSP in a clever manner that leaves all these thermionic strengths intact, but reduces distortion even further.

To paraphrase the Wicked Witch of the West: "Matters like this must be handled with great care, to avoid damaging the magic."

More to come... (sorry )

Best,

George Ferguson

 

[DrDeville]

DSP Tube Distortion Reduction Techniques

Polimorph's dancing digits dictated:

I think DSP can't correct such distourbances just because de DSP device is placed before the power amplifier

Right, if the device is external. I'm proposing using DSP as in integral, internal part of amplification devices, to lower their distortion.

Benny's flicking philanges formulated:

i am not an expert on light wave theory or anything as you youself probably are (if you're not i worry about what you're doing working @ lockheed doing what you said you did)

Well, I did have them fooled.

Actually wave theory, whether optical or acoustic, is secondary. The more relevant concept is that of using servo ("closed loop") systems to correct errors. This is applicable to both fields, and many others.

There are basically two ways of doing this:

• Continuously comparing the device's output to its input. This is akin to a driver adjusting a car's steering to stay in her lane.
  
• Taking an intermittent "snapshot" of the device's transfer function, and pre-distorting the input signal to compensate. This is akin to a golfer putting at an angle to correct for a green's draw.

An example of the former would be intelligent feedback, that corrects distortions without doing the nasty things that feedback usually does, like freaking out on overloads.

An example of the latter would be an intelligent tube load, that takes a snapshot of the individual tube, and corrects its effective plate curves to eliminate odd-order harmonics.

I'm not saying that such things would be easy, or always desireable. An elegant engineering solution can often be simpler and more cost effective. Also, the processing power involved is not trivial.

But processors are getting cheaper, and I find the idea intriguing if thoughtfully applied.

And correcting room acoustics--in frequency, phase, and reverberation--is also feasible and promising. The most obvious method is the "snapshot" one, taken at the listening position. Of course it would not be perfect, just better. I believe there is a company (something like Behringer?) that offers a frequency/phase/reverberation corrector that does just that, and corrects for speaker and room distortions.

Again, that's not to say that clever open-loop techniques are not often more cost effective--like good speaker design, and simple room treatments. It amazes me how many people spend huge amounts of time and money tweaking components (tubes, capacitors, interconnects), when they could make an infinitely greater improvement with one well-placed wall hanging.

Thanks for listening to my pontifications.

Best,

George "Plays One On TV" Ferguson

 

[benny]

i think this DSP stuff is pretty good considering what it can do... as long as we can apreciate that we will never be able to model the real thing, so nothings going to replace a good tube amp and a well designed room, but we can at least apreciate the benifit it will gives us, and the improvement in our existing cheaper equipment by using it.

i think it's also a good idea of DrDeville's to use this technology to reduce distoriton in a tube amp. While i think it's a good idea, i think one disadvantage that may ocour from this, is maybe tube distortion can help cover up for digital distortion a bit... getting rid of tube distortion still leaves digital distortion. i don't know how you get rid of the distoriton of the AD/DA converter seing as the signal fedback must be processed by the AD converter before the computer can process it, and therefore, has had further distortion put on it from the ADC... so there's always going to be an element of distoriton that will be imposible to get rid of. other problem is, like i have been trying to get across, a tube amp does not act the same across all frequencies and dynamics... which you could compensate for in your initial set up tests, but you can't get the reactance the amp has with what it played before, and how it effects what it's now playing... such as overload recovery. also, if using a distortion reduction DSP, you wouldn't want to use it over the whole amp... such as the power stage of a SE amp... one of the things that had the first engineers who experimented with SE was that on the bench, they could create what was 'theoretically' a good amp through low THD. then when they listen to it, it sounded not so good... but when they made a theoretically bad amp with a high THD, they found it sounded far better. there are a number of theories about why this is so, but it is known that a SE with some THD sounds better than one with minimal THD.

so i wish you luck with your DSP creations polimorph. i hope they all turn out for you. btw, if it seems like what i'm saying is all negative as far as it's posibility of any of your ideas working, it's only because i point out the negative, because the positive has already been seen, that's why you thought the idea up... for some sort of benifit.

cheers