In my experience -- for others it may be different -- so long as the sound emerging from the drivers is "correct", then the ear/brain is clever enough to adjust for the acoustics of the listening space. A real string quartet will always sound "right", as in real, whether they play in a bathroom or in the middle of a football field. So what you want to do is make sure your system works as well as that.
An experiment to try: if a recording doesn't sound "good" then physically move closer and closer to one of the speakers while listening. Most likely the music actually emerging from the drivers will sound worse and worse the closer you get, meaning that the sound, at the source, is not "correct". As simple as that ...
Frank
I presume you've only tested one side of this, and are making a comparison to an inferred side that deals with crosstalk and HRTF in a correct manner which you have not actually tested.
You can only draw a valid conclusion if you've compared to listening to binaural recordings through in-ear headphones, where said recordings were made with microphones in the ear canals of a human whose HRTF is similar enough to your own--or simulating a virtually-positioned sound source and convolving the signal with such an HRTF. I've done this experiment by trying various HRTFs from a couple of the online-accessible HRTF databases until I found one that gave me the best positional audio. The 3D accuracy is beyond anything you can get without doing this, and makes it trivial to tell not only position in the horizontal plane but even up and down. Recordings made with dummy heads don't even come close, and regular stereo or multi-speaker recordings are a joke in comparison.
People seem to forget that the direction-sensitive frequency filtering that the head-and-ear system is of very large magnitude--larger than a decent speaker's frequency response unevenness. Not accounting for this while worrying about smaller distortions is silly. Most people don't do it because it's easier to minimize speaker and electronic distortion, and the solutions to accurate positional audio don't scale well to multiple users (each one really needs their own HRTF; using a dummy head is not a very good approximation; you can a) use a selection from a database using much trial and error, 2) get your HRTF measured, 3) simulate you HRTF using a 3D scan of your head--there was a paper on this last option a few years ago).
Last edited:
might look at the Smyth SVS realizer virtualization sysem
www.smyth-research.com
but this is more EQ than Predistortion as I understood the thread topic to be
maybe "linear predistortion" covers room/loudspeaker EQ, virtualizers?
www.smyth-research.com
but this is more EQ than Predistortion as I understood the thread topic to be
maybe "linear predistortion" covers room/loudspeaker EQ, virtualizers?
They talk about emulating listening to loudspeakers. What we should be emulating is the sound of being present at the actual live performance. These are nowhere near being the same.
Certainly, when I go listen to the VSO perform Schubert, it sounds nothing like a Schubert recording I hear through speaks, be they even the B&W Nautilus demoed at the Hi Fi Centre in Vancouver together with all the fancy room treatment they add. The geometry of the configuration is all wrong. It could be done with speakers, but an inverse HRTF needs to be applied that depends on listener position and the speaker/room configuration.
Certainly, when I go listen to the VSO perform Schubert, it sounds nothing like a Schubert recording I hear through speaks, be they even the B&W Nautilus demoed at the Hi Fi Centre in Vancouver together with all the fancy room treatment they add. The geometry of the configuration is all wrong. It could be done with speakers, but an inverse HRTF needs to be applied that depends on listener position and the speaker/room configuration.
Last edited:
is this a "what if?" wish thread or discussion for practical application right now
you can wish for the moon - or we can discuss what improvements can be made in reproducing the source we can buy right now
yes the Smyth people take on a simpler virtualization problem - replicating room&loudspeaker multichannel sound in headphones
for accessing existing recordings, the immediate future the overwhelming majority of source we have is recorded/mastered with the intent to be listened to in acoustically "small" room domestic listening environment over a few discrete loudspeakers in a horizontal plane
with the SVS Realizer you can make personal hrtf calibrations in any real loudspeaker setup in a real room or any in studio you can afford ~1/2 hr access to - in Bob Ludwig's chair at his mixing board for instance
you can wish for the moon - or we can discuss what improvements can be made in reproducing the source we can buy right now
yes the Smyth people take on a simpler virtualization problem - replicating room&loudspeaker multichannel sound in headphones
for accessing existing recordings, the immediate future the overwhelming majority of source we have is recorded/mastered with the intent to be listened to in acoustically "small" room domestic listening environment over a few discrete loudspeakers in a horizontal plane
with the SVS Realizer you can make personal hrtf calibrations in any real loudspeaker setup in a real room or any in studio you can afford ~1/2 hr access to - in Bob Ludwig's chair at his mixing board for instance
Attachments
Last edited:
I'm not wishing for anything that is not currently achievable at moderate cost.
A $3000 laser scanner can capture a human head at sufficient resolution to allow computation of a person's specific HRTF, which would take a couple of hours on a modern GPU. No expensive measurement system in an anechoic chamber needed.
If ignorance about positional audio was not as prevalent, then it would be practical to provide this as a service at audiophile stores, for example.
A $3000 laser scanner can capture a human head at sufficient resolution to allow computation of a person's specific HRTF, which would take a couple of hours on a modern GPU. No expensive measurement system in an anechoic chamber needed.
If ignorance about positional audio was not as prevalent, then it would be practical to provide this as a service at audiophile stores, for example.
Last edited:
All of this I believe is equivalent to the way activity on an acting stage can be emphasised by strong lighting effects -- the acoustic markers which make it possible for the mind to decode the audible characteristics can be very strongly highlighted using various "tricks", spotlighting the key information, and so the mind understands what's going on, acoustically. Alternatively, if the detail is revealed as free from distortion as possible your hearing system can still work it all out -- equivalent to taking all the fancy directional lighting out of the equation for a stage performance, and just increasing the overall, uniform light level being directed upon the action.
Frank
Frank
How about making reproduced music sound like real instruments being played in the room. How do you do That? Nothing yet heard does sound real. Where is the blockage?
Despite the fact that feedback is fully understood* and causes no ill-effects*, it would be an interesting experirment to attempt to implement a genuinely feedback-less amplifier, say 1W or so into an 8R speaker using neural network-style (or SVM) pre-distortion system. Any suggestions as to the hardware I should use? As I mentioned before, running the output stage at a fixed temperature may make the problem more manageable.
* I defer to others on this!
* I defer to others on this!
I don't know the answer to that. Maybe if it's a simple 1D mapping problem, yes. Presumably your hardware would need to be reasonably monotonic e.g. no crossover 'dead bands'.
But if there's a '2D' error that is 'triggered' by a certain input sequence e.g. ringing, then it may not be possible. With a neural-style system you would have the option to avoid presenting that input sequence to the amplifier, or at least to modify it in such a way so as to minimise the overall error, as measured after thousands of trials. This would happen automatically during the training.
I listen to LP because I have LPs and do not wish to rebuy them in another format, even where they are available (not all are, or may have been 'remastered').
Audio amps have negligible delay, so feedback is fine when properly applied. It has some ill effects, but these can be minimised and predistortion is likely to be no better.
I think that output errors can all be corrected at the input provided that the amp is monotonic, in the sense that (apart from linear filters) an increase in input voltage causes an increase in output voltage (or decrease if the amp is inverting). That is, no hysteresis or clipping. I think the same conditions are required for feedback too.
There is an exception: predistortion cannot remove noise, while feedback can. Predistortion relies on accurately predicting what the amp will do, but you can't predict noise. Feedback can reduce noise in real time because it simply compares input and output and removes what should not be there. If you tried to make an amp with no feedback, just predistortion, then noise could be a problem. With BJTs you can't have a Vbe signal of more than a few mV because of the exponential characteristic distortion - so poor signal-noise. You would have to allow yourself local feedback, such as degeneration, to get round this but then you no longer have a feedback-free amp.
Audio amps have negligible delay, so feedback is fine when properly applied. It has some ill effects, but these can be minimised and predistortion is likely to be no better.
I think that output errors can all be corrected at the input provided that the amp is monotonic, in the sense that (apart from linear filters) an increase in input voltage causes an increase in output voltage (or decrease if the amp is inverting). That is, no hysteresis or clipping. I think the same conditions are required for feedback too.
There is an exception: predistortion cannot remove noise, while feedback can. Predistortion relies on accurately predicting what the amp will do, but you can't predict noise. Feedback can reduce noise in real time because it simply compares input and output and removes what should not be there. If you tried to make an amp with no feedback, just predistortion, then noise could be a problem. With BJTs you can't have a Vbe signal of more than a few mV because of the exponential characteristic distortion - so poor signal-noise. You would have to allow yourself local feedback, such as degeneration, to get round this but then you no longer have a feedback-free amp.
The point about FM is excellent . A system so fraught with problems as to look ridiculous . The CD players the broadcasters use are obviously better than mine despite the changes of horse the signal has taken . Then the brick wall filtering just to be sure it wont work . Work it does . My Quad FM3 although never considered very good sounds beyond my wildest dreams ( I have Kenwood and Armstrong which are excellent also , the Kenwood is state of the art KR750 ) .
Now one can say the FM tuner has performed alchemy on the signal ? I don't buy that . Recently I have been working on a SE valve amp using KT 88 . Mostly to complete my CV I would say . Having borrowed an amp using 211 valve I was hoping for a bit of less expensive magic . As far as I can tell what I have built is as good as it gets for the money spent . What alas it can not seem to do is add a bit of magic . A bad source signal is just as bad as it ever was . Then I remembered . I used to feed my cassette recorder into my Sobel valve radio when 12 years old . Whilst the radio was great the cassette was rubbish . In fact more obviously rubbish . Euphonic distortion ? If only .
Now one can say the FM tuner has performed alchemy on the signal ? I don't buy that . Recently I have been working on a SE valve amp using KT 88 . Mostly to complete my CV I would say . Having borrowed an amp using 211 valve I was hoping for a bit of less expensive magic . As far as I can tell what I have built is as good as it gets for the money spent . What alas it can not seem to do is add a bit of magic . A bad source signal is just as bad as it ever was . Then I remembered . I used to feed my cassette recorder into my Sobel valve radio when 12 years old . Whilst the radio was great the cassette was rubbish . In fact more obviously rubbish . Euphonic distortion ? If only .
A good FM setup will have lower distortion than LP, and about the same noise level, but narrower bandwidth. FM has a fairly sharp filter at 15kHz, but not as sharp as the brick-wall needed for CD. Some of the most realistic sounds I have heard from my system were live BBC Proms concerts in the 1980s, when their sound engineers knew what they were doing and were not under pressure from management to make everything loud.
Spot on . As said the signal has been through so many changes of horse . The 13 bit digital system BBC used ( use ) would be something I would like to try .
If FM is removed I will be so sad . If people hear of it being reality I will be very willing to lobby my local MP our Prime Minister . He wrote to me about the World Service so does care . We saved long-wave which I listen to when in France .
Yes, the BBC used 13-bit NICAM for programme distribution. This was different from the NICAM used for stereo sound on analogue TV. I don't know off-hand what sample rate they used, but I would guess 32kHz or 48kHz. So, yes, there probably would have been a sharp filter in there.
There is still a real danger of FM disappearing from the UK, at least for national stations. As soon as digital listening (DAB, DTV, Internet, satellite) reaches 50% they may switch off - even if FM is still the single most popular method and DAB the least popular. FM will of course be the cheapest and most portable method, and the most energy-friendly.
There is still a real danger of FM disappearing from the UK, at least for national stations. As soon as digital listening (DAB, DTV, Internet, satellite) reaches 50% they may switch off - even if FM is still the single most popular method and DAB the least popular. FM will of course be the cheapest and most portable method, and the most energy-friendly.
Where are the flaws in this article? - they're saying DAB is 20X greener than FM.
Major Moments, Minor Thoughts on Media, Technology and Travel: DAB 20 Times Greener Than FM
Major Moments, Minor Thoughts on Media, Technology and Travel: DAB 20 Times Greener Than FM
This looks worth a read sometime ?
BBC RD - Library - BBC Engineering
My feeling is 13 bit BBC digital would have been ideal for CD . It could have been upgraded a bit . I don't know enough to say why , just the way it sounded . As you say DF 1980's BBC broadcasts were outstanding . I now feel such a fool in denouncing digital as I did ( do ) . I had been enjoying it for years .
About additive distortions and similar . When Garrard closed down a friend asked me to design a Garrard 501 . Like most people I thought that idler drives were rubbish . I was called to visit this friend as he was listening to a Linn LP12 and his 401 . My first thought after accepting the Garrard to be better is that rumble might be causing vibrato . The 301 is about - 45 db , 401 - 57 dB and the 501 - 79 dB ( weighted ) . I can now say it isn't vibrato as 501 has the verve also . It might interest people to know I changed nothing on 501 . Just attention to detail . It measured 3 dB better than the Thorens reference on the same test rig . Wow and especailly flutter nearly non existent . We made our own motor . It was adapted medical motor parts ( blood centrifuge I think ) .
BBC RD - Library - BBC Engineering
My feeling is 13 bit BBC digital would have been ideal for CD . It could have been upgraded a bit . I don't know enough to say why , just the way it sounded . As you say DF 1980's BBC broadcasts were outstanding . I now feel such a fool in denouncing digital as I did ( do ) . I had been enjoying it for years .
About additive distortions and similar . When Garrard closed down a friend asked me to design a Garrard 501 . Like most people I thought that idler drives were rubbish . I was called to visit this friend as he was listening to a Linn LP12 and his 401 . My first thought after accepting the Garrard to be better is that rumble might be causing vibrato . The 301 is about - 45 db , 401 - 57 dB and the 501 - 79 dB ( weighted ) . I can now say it isn't vibrato as 501 has the verve also . It might interest people to know I changed nothing on 501 . Just attention to detail . It measured 3 dB better than the Thorens reference on the same test rig . Wow and especailly flutter nearly non existent . We made our own motor . It was adapted medical motor parts ( blood centrifuge I think ) .
Such honesty - kudos
Its rather odd that the people who don't like digital sound blame it on the theory, or the 'flawed' CD format rather than on the implementations they actually hear don't you think? What might be behind this misunderstanding in your view?