Hi,
THD is completely meaningless if we want to assess distortion audibility.
0.05% THD may already be way too much and objectionable, while 3% THD or even more may go unnoticed with music. For any funloving number of THD you care to choose* I can illustrate that it is the "wrong number". Because THD is an artifical measurement with no relation to distortion audibility.
* except around -120dB re 111dB SPL for all harmonics (that is < 0.0001% for any given harmonic), at all levels and frequencies
In order to assess distortion audibility for simple (HD and resultant IMD) Distortion we need to know:
1) SPL Level of the signal
2) Frequency of the signal
3) Level of harmonic and order
While Earl Geddes has provided more precise and evidence based weighting (and more stringent on higher orders than all earlier suggestions) for this type of distortion, the scheme D.E.L. Shorter of the BBC suggested in the 1950's# should suffice for a very decent quick approximation.
# D.E.L. Shorter suggested that the level of each harmonic shall be multiplied by N^2/4 where N is harmonic order. So for example H2 needs to be mulplied by a factor of 1, H3 by a factor of 2.25 and H7 levels need to be multiplied by 12.25 to get the equivalent audibility.
In other words, if we give 0.1% H2 an audibility score of 0dB (that is it is at the edge of being detectable), than 0.044% H3 and 0.008% H7 have a similar level of audibility.
D.E.L Shorters scheme does not account for SPL though
I would also submit a further thought, which is pure speculation of course...
We know that the human hearing mechanism is not a microphone, the hair-cells are not linear piezoelectric transducers and the analogue signal processing that happens in the grey matter that most humans are alleged to have between their ears (reading this board often makes one doubt the veracity of such an allegation) is not an Fast Fourier Transform of a limited windowed signal and so on etc. et al.
In his Heyser Memorial Lecture John Atkinson proposed that when listening to music (real or recorded) we effectively use an internalised model and judge and react on the basis of differences to this model.
What I would propose is that we as such do not directly react to the individual harmonics in themselves (in the way an FFT Analyser does), but rather we react to, for lack of a better term the whole transfer function, with no specific "audibility" of a harmonic as such.
As said, speculation, purely, lest someone knows research that supports my deductive/inductive leap of intuition...
I find it generally surprising how little the extensive work of a range of professional and industry bodies on many subjects closely impacting audio is among those who practice the craft.
Luckily enough we can take a a rather sturdy swordhilt to this gordion knot by looking closely at the EBU (adopted from IRT recommendations) for maximum SPL we come to 108dB continuous (or 111dB peak) if we look at THX we get 105/108dB so both sets of numbers are quite close.
Given that a stereo pair of speakers at 3m listening distance each fed with equal power gives around 3dB less SPL than the individual speaker at 1m we can easily work back to SPL requirements.
Basically we need between 111dB and 114dB peaks.
So for a 82dB/2.83V/1m Speakers we need 2.83V + 32dB = 113V peak (800W/8R).
On the other hand if we have (say) a speaker with 98dB/2.83V we only need 2.83V +16dB = 17V peak (18W/8R).
I suspect our 82dB/2.83V/1m speaker has long vaporised the voice coils should we use a sustained tone for this test.
Ciao T
If that is a speculation, then we share speculations, Thorsten.
I maintain that people are discovering vintage sound for precisely the same basic reason - it does not concentrate on this or that, but rather goes for a wholesome presentation of everything, top to bottom. Not as easy as it may appear to be.
This is my dictum, if I ever had one. This is the first and foremost thing I look (listen) for in a piece of audio gear, a balanced top to bottom sound, nothing pushes, nothing pulls, no dips in between.
I love my vintage Marantz gear sound precisely because it manages to pull off that stunt so well - I do not hear a group of individuals trying to get it together, rather I hear a band performing a piece in relative harmony. Not perfect, but very, very good.
It's still done today as well, or even better than it was in the old days - my Karan Acoustics KA-i180 is living proof, and I'm sure there are other worthy units out there (obviously, I can't audition them all, so I can't name too many of them).
Regarding your calculations of power, I agree wholeheartedly. For a big sound, you need a big amplifier, tubes can sound very pleasing, but inevitably their typical lowish power works against them.
And since a good preacher does as he preaches, I run my 92 dB speakers from a 22 dBW amp, giving me a nominal dynamic reach of 114 dB. In a room, and with a pair of speakers, that might be more, but then, the power/SPL equation is not really completely linear, is it? No such thing as a perfect speaker, crossover, etc.
Last edited:
TL, DVV,
I share TL views on this , I don't see why it can't be done , options can be small
In how much they vary from designer optimum. Bias already moves around due to temperature tracking , feedback could be applied as on/off ( degeneration only ) designer optimum , +/- 12 db for eg ...
Just saying ...
I share TL views on this , I don't see why it can't be done , options can be small
In how much they vary from designer optimum. Bias already moves around due to temperature tracking , feedback could be applied as on/off ( degeneration only ) designer optimum , +/- 12 db for eg ...
Just saying ...
Hi,
Don't hold your breath.
Even now switching at 1MHz with appreciable power is a serious challenge.
Even if we attempt to interleave (say 16pcs) of lower power very high switching frequency amp-chips (not that they exist anyway, but we can theorise) we would still be a few orders of magnitude short of 16 Bit true (unaveraged) resolution...
I do not think that this road leads anywhere, but feel free to travel it and keep us posted.
For most commercial applications the current generation of 0.3MHz switching Amp's suffices, they do not measure terrible and I have worst Class AB Amp's than these (especially big expensive American Muscle Amp's), even though in absolute terms these Amp's are pretty poor, but so is 90% of the all the A and V gear out there...
Ciao T
I'm banking on them getting there, eventually.
Don't hold your breath.
Even now switching at 1MHz with appreciable power is a serious challenge.
Even if we attempt to interleave (say 16pcs) of lower power very high switching frequency amp-chips (not that they exist anyway, but we can theorise) we would still be a few orders of magnitude short of 16 Bit true (unaveraged) resolution...
I do not think that this road leads anywhere, but feel free to travel it and keep us posted.
For most commercial applications the current generation of 0.3MHz switching Amp's suffices, they do not measure terrible and I have worst Class AB Amp's than these (especially big expensive American Muscle Amp's), even though in absolute terms these Amp's are pretty poor, but so is 90% of the all the A and V gear out there...
Ciao T
Hi,
In my case it is 90dB/2.83V and 16dBW, but my watt are tube watts (and every audiophile knows that TUBE WATTS are much bigger than transistor watts ) from a "no-looped-feedback" Amp (EL34 in "super triode" push-pull).
So the Amp compresses more than it clips, so subjectively it handles peaks as much as 6dB past it's nominal 40W power rather gracefully. Not undistorted, but without what most including myself would call "bad distortion", I guess some might call it "soft" on high peaks.
So I can also get the average levels very high, at the cost of "compressed" peaks... Plus, at the high levels the Aphax FX from the increased distortion makes the system sound louder than it really is...
Ciao T
PS,
The whole system lacks any looped feedback, all feedback is local
DAC Chip without on board analogue stage circuitry ->
open loop tube stage & buffer->
Volume Control ->
open loop buffer ->
open loop Tube Amp ->
open loop speaker
Three stages are buffers (unity gain or 1+1) the other three have voltage gain.
And since a good preacher does as he preaches, I run my 92 dB speakers from a 22 dBW amp, giving me a nominal dynamic reach of 114 dB.
In my case it is 90dB/2.83V and 16dBW, but my watt are tube watts (and every audiophile knows that TUBE WATTS are much bigger than transistor watts ) from a "no-looped-feedback" Amp (EL34 in "super triode" push-pull).
So the Amp compresses more than it clips, so subjectively it handles peaks as much as 6dB past it's nominal 40W power rather gracefully. Not undistorted, but without what most including myself would call "bad distortion", I guess some might call it "soft" on high peaks.
So I can also get the average levels very high, at the cost of "compressed" peaks... Plus, at the high levels the Aphax FX from the increased distortion makes the system sound louder than it really is...
Ciao T
PS,
The whole system lacks any looped feedback, all feedback is local
DAC Chip without on board analogue stage circuitry ->
open loop tube stage & buffer->
Volume Control ->
open loop buffer ->
open loop Tube Amp ->
open loop speaker
Three stages are buffers (unity gain or 1+1) the other three have voltage gain.
Hi,
Amplifiers with perfectly adequate distortion performance (that is not audible) AND damping factor (that is no appreciable differences in the actual damping factor at the woofer cone) can be made without global loop feedback. Local degeneration, which is generally considered negative feedback by some will be needed though.
That again is debatable. In the end it boils down to how we define "optimum" in technical terms and how we verify "optimum".
There can be only one optimal point if there is a wide agreement what "optimal" is, in respect to a given quantifiable quality.
If there is no such agreement, there will different "optimal points", especially when we need weigh up multiple and conceivably interactive qualities (as is the case for audio).
This is debatable.
What can be substantiated is the existence of a minimum THD bias point.
This Bias Points actually varies with signal levels (there is thermal memory component involved).
This bias point actually minimises THD and low order HD, but it leads to higher levels of high order distortion than could be attained using a different (higher) bias.
So the optimum differs if we want to minimise audible distortion or measured THD. Which then is the optimum?
I would argue that it is at a level of quiescent current where the given Amplifier no longer shows any appreciable reduction in higher order distortion components, EVEN IF THD IS INCREASED over the THD minimum.
D.Self would argue that it is THD minimum, though he'd probably start to waffle and wobble a little if you asked him to substantiate at what power level the THD minimum should be set and why it does not matter that this optimum point is signal dependent...
For the accountants D. Self Optimum is preferred, measured at the lowest possible power, as this minimises the need to spend money on heatsinks and power supplies...
Ciao T
Thorsten, in absolute terms, EVERYTHING is debatable. Some like the preacher, others prefer the preacher's wife. Even if one were to find a point at which most would agree he is at a tchnical optimum point, that may well turn out to be strictly second place to what our ears may tell us.
Regarding bias, the lowest point of THD is, in my experience, NEVER the same as the point of what most would say is good sound; as I see it, in very GENERAL terms, while a lower value of bias current may be technically preferable according to D. Self, in my view most amps do best with bias currents of 80...130 mA per output transistor. Just add bias for as long as you can hear a positive difference (in your view), and when you don't hear it any more, go back that one step. And pray your heat sinks hold out.
When working for myself, I have no such problems because I have no bean counter peering over my shoulder (my wife never even tried it, she understands I need a vice, since I don't drink, don't socialize in pubs, don't gamble and don't womenize, I am your boring home & family type of guy). It will cost what it has to, period, but it will done right. Need bigger heat sinks? Fine, I buy them. Need more output devices? Fine, I buy a population I am reasonably certain will be able to yield as many matched pairs as I need. It's great when there are no price constraints. And in the end, nothing is wasted, everything finds its place sooner or later.
Again, sounds nice in theory, but I very much doubt that any such practice will follow.
Already in progress
You and a thousand Chinese manufacturers have the same idea. Back to square one.
Then its abundantly clear you misunderstood my idea. Having been in close contact with Chinese for seven-odd years when some do take the time to listen to what I have to say they're surprised and realize what I'm beginning doing hasn't been done before in China (to their knowledge).
I am saddened to see fewer and fewer people pay attention to the actual sound.
That's fundamentally a choice - you choose to lament the current situation, I choose to leverage it. Enjoy your lamentations
So, my dear fellow - where are those customers you expect?
I have no expectations about customers, rather I have intentions. The customers I'm intending are already coming along do not fret
Some will be there, of course, but so will a hell of a lot of your competitors, many of which will offer more external machining than you, at a quarter of your price.
Argument from ignorance (again). I have no 'competitors' - partly because they're all bozos who can't understand the scheme I'm plotting. And partly because they're not patient enough to implement it even if they could grasp it. Generally nowadays the focus is on the 'fast buck' whereas I'm playing the long game. Tortoise vs hare.
Don't hold your breath.
Sure, I'm quite happily breathing class AB for the foreseeable future.
Even now switching at 1MHz with appreciable power is a serious challenge.
But the mass-market does not need 'appreciable power' so that's beside the point.
Even if we attempt to interleave (say 16pcs) of lower power very high switching frequency amp-chips (not that they exist anyway, but we can theorise) we would still be a few orders of magnitude short of 16 Bit true (unaveraged) resolution...
I do not think that this road leads anywhere, but feel free to travel it and keep us posted.
Can't see the point myself either when class AB does the job admirably well for most people right now. And when it doesn't I have other ideas up my sleeve which don't quite match up with your notions
Hi,
Actually, my claiming that TUBE WATTS ARE BIGGER and transistor watts are smaller was a bit tongue in cheek...
I gave the solution after that.
If you look at music, you can see a fairly low average signal and a very crest factor with comparably short peaks of substantial magnitude.
When most transistor amplifiers clip on such a signal it sounds really bad, not because of the clipped transient, but because they tend to saturate during clipping and do not recover instantly from clipping.
The more NFB is applied the greater the problem, lest measures are applied to improve the situation.
Meanwhile competently designed tube amplifiers, especially those with very low or no looped feedback can accept significant levels of instantaneous overdrive (what we would call clipping in SS Amp's) without sounding objectionable.
So if we have (say) a 70W SS Amp that sounds nasty at 0.1dB above 70W and a 35W Tube Amp that sounds fine at 3dB above 35W, then the lower powered tube amp, though technically "clipping" or "overloading" will play as loud or louder than the higher powered solid state amp.
So the 35WATT TUBE POWER ARE BIGGER than the 70watt solid state power...
![Wink ;) ;)](data:image/gif;base64,R0lGODlhAQABAIAAAAAAAP///yH5BAEAAAAALAAAAAABAAEAAAIBRAA7)
Ciao T
Tubes watts bigger ? puffy , swells up like a wet sponge...
Actually, my claiming that TUBE WATTS ARE BIGGER and transistor watts are smaller was a bit tongue in cheek...
I gave the solution after that.
If you look at music, you can see a fairly low average signal and a very crest factor with comparably short peaks of substantial magnitude.
When most transistor amplifiers clip on such a signal it sounds really bad, not because of the clipped transient, but because they tend to saturate during clipping and do not recover instantly from clipping.
The more NFB is applied the greater the problem, lest measures are applied to improve the situation.
Meanwhile competently designed tube amplifiers, especially those with very low or no looped feedback can accept significant levels of instantaneous overdrive (what we would call clipping in SS Amp's) without sounding objectionable.
So if we have (say) a 70W SS Amp that sounds nasty at 0.1dB above 70W and a 35W Tube Amp that sounds fine at 3dB above 35W, then the lower powered tube amp, though technically "clipping" or "overloading" will play as loud or louder than the higher powered solid state amp.
So the 35WATT TUBE POWER ARE BIGGER than the 70watt solid state power...
Ciao T
Already in progress
Then its abundantly clear you misunderstood my idea. Having been in close contact with Chinese for seven-odd years when some do take the time to listen to what I have to say they're surprised and realize what I'm beginning doing hasn't been done before in China (to their knowledge).
That's fundamentally a choice - you choose to lament the current situation, I choose to leverage it. Enjoy your lamentations
I have no expectations about customers, rather I have intentions. The customers I'm intending are already coming along do not fret
Argument from ignorance (again). I have no 'competitors' - partly because they're all bozos who can't understand the scheme I'm plotting. And partly because they're not patient enough to implement it even if they could grasp it. Generally nowadays the focus is on the 'fast buck' whereas I'm playing the long game. Tortoise vs hare.
Perhaps you didn't explain your idea well enough, the explanaition being as clear as mud.
But either way, I see you have it all figured out, so I'm not worried. Just impatient to read in the papers how you just bought half of the Bahamas and never felt the sting.
Playing the long game is inherently uncertain because much can hapen in a short time, let alone a long one. At the speed technology is developing, it's hard guessing what comes in two years' time, let alone 5 or more.
I agree with you on these being the time sof the fast buck, hit and run whizz kids.
Good luck with leveraging, being innovative, fully understanding and not being ignorant and the rest of it. You sure talk the talk, let's see if you can walk the walk. Keep us posted.
Last edited:
Perhaps you didn't explain your idea well enough, the explanaition being as clear as mud.
When you were over in UK, they didn't teach you Rudyard Kipling? The particular one I'm thinking of begins 'I keep 6 honest serving-men...'.
Playing the long game is inherently uncertain because much can hapen in a short time, let alone a long one.
Nope, its inherently more certain than playing the short one. Because in the long game, I create the future don't just try to predict it.
At the speed technology is developing, it's hard guessing what comes in two years' time, let alone 5 or more.
Ah but has it escaped your attention that in many respects (electronics in particular) things are actually slowing down quite nicely? Not only that, but revolving as much as they're evolving?
Hi,
While looking for some other stuff, I stumbled across a quite lucid and accessible treatment of the GedLee distortion metric, worth a read...
http://www.gedlee.com/downloads/THD_.pdf
Ciao T
While Earl Geddes has provided more precise and evidence based weighting (and more stringent on higher orders than all earlier suggestions) for this type of distortion, the scheme D.E.L. Shorter of the BBC suggested in the 1950's# should suffice for a very decent quick approximation.
# D.E.L. Shorter suggested that the level of each harmonic shall be multiplied by N^2/4 where N is harmonic order. So for example H2 needs to be mulplied by a factor of 1, H3 by a factor of 2.25 and H7 levels need to be multiplied by 12.25 to get the equivalent audibility.
In other words, if we give 0.1% H2 an audibility score of 0dB (that is it is at the edge of being detectable), than 0.044% H3 and 0.008% H7 have a similar level of audibility.
D.E.L Shorters scheme does not account for SPL though
While looking for some other stuff, I stumbled across a quite lucid and accessible treatment of the GedLee distortion metric, worth a read...
http://www.gedlee.com/downloads/THD_.pdf
Ciao T
Abraxalito is right to this extent:
Supply creates demand.
If the product does not exist, then there is no demand.
So if you're going to market you will have convictions and will take risk.![smash :smash: :smash:](https://files.diyaudio.com/forums/images/smilies/smash.gif)
Supply creates demand.
If the product does not exist, then there is no demand.
So if you're going to market you will have convictions and will take risk.
![smash :smash: :smash:](https://files.diyaudio.com/forums/images/smilies/smash.gif)
I have no expectations about customers, rather I have intentions. The customers I'm intending are already coming along do not fret![]()
Hi,
Yes, it seems since Apple everyone seems to believe this.
The fun part, non of Apple's major successes where Apple ideas or inventions.
What Apple was good at was to spot trends and to take the ideas smaller/other companies had created, repackage them and relentlessly promote them as fashion accessories.
Examples:
Graphical OS's
All In One PC's
Portable Flash/HDD based music players
Smart Phones
Tablet Computers
Supply existed before Apple, but no demand.
Advertising creates demand, not supply.
Ciao T
Abraxalito is right to this extent:
Supply creates demand.
If the product does not exist, then there is no demand.
Steve Jobs put it like this: 'the customers don't know what they want until we show them'.
Yes, it seems since Apple everyone seems to believe this.
The fun part, non of Apple's major successes where Apple ideas or inventions.
What Apple was good at was to spot trends and to take the ideas smaller/other companies had created, repackage them and relentlessly promote them as fashion accessories.
Examples:
Graphical OS's
All In One PC's
Portable Flash/HDD based music players
Smart Phones
Tablet Computers
Supply existed before Apple, but no demand.
Advertising creates demand, not supply.
Ciao T
Hi,
Actually, my claiming that TUBE WATTS ARE BIGGER and transistor watts are smaller was a bit tongue in cheek...
I gave the solution after that.
If you look at music, you can see a fairly low average signal and a very crest factor with comparably short peaks of substantial magnitude.
When most transistor amplifiers clip on such a signal it sounds really bad, not because of the clipped transient, but because they tend to saturate during clipping and do not recover instantly from clipping.
The more NFB is applied the greater the problem, lest measures are applied to improve the situation.
Meanwhile competently designed tube amplifiers, especially those with very low or no looped feedback can accept significant levels of instantaneous overdrive (what we would call clipping in SS Amp's) without sounding objectionable.
So if we have (say) a 70W SS Amp that sounds nasty at 0.1dB above 70W and a 35W Tube Amp that sounds fine at 3dB above 35W, then the lower powered tube amp, though technically "clipping" or "overloading" will play as loud or louder than the higher powered solid state amp.
So the 35WATT TUBE POWER ARE BIGGER than the 70watt solid state power...![]()
Ciao T
Well i guess Hollowstate has an advantage if there were only 70 watts SS amps available...
It does appear Mr Carver has decided that his new line of amplifiers will be Hollowstate and with real watts, as much as 300watts RMS/ch Attempting to give one a listen ..........
![radar :radar: :radar:](https://files.diyaudio.com/forums/images/smilies/radar.gif)
Yes, it seems since Apple everyone seems to believe this.
I believed it earlier - from the days of Akio Morita's Walkman, prior to Apple.
Supply existed before Apple, but no demand.
Where was the supply of graphical interfaces prior to Apple? Hardly within Xerox PARC, they hadn't commercialized it.
Hi,
The numbers where more or less based on the Dynaco ST-70 and Dynaco ST-120.
I could also have used the comparison between a 200W CAT Monoblock and 400W Krell, but the story goes back to ST-70 & ST-120, so why not use them?
High Power Tube Amp's are no news.
Ciao T
Well i guess Hollowstate has an advantage if there were only 70 watts SS amps available...![]()
The numbers where more or less based on the Dynaco ST-70 and Dynaco ST-120.
I could also have used the comparison between a 200W CAT Monoblock and 400W Krell, but the story goes back to ST-70 & ST-120, so why not use them?
It does appear Mr Carver has decided that his new line of amplifiers will be Hollowstate and with real watts, as much as 300watts RMS/ch Attempting to give one a listen ..........![]()
High Power Tube Amp's are no news.
Ciao T
Hi,
Xerox commercialised GUI's in 81. Apple Lisa appeared 2 Years later.
Ciao T
Where was the supply of graphical interfaces prior to Apple? Hardly within Xerox PARC, they hadn't commercialized it.
Xerox commercialised GUI's in 81. Apple Lisa appeared 2 Years later.
Ciao T
Xerox commercialised GUI's in 81.
In what market?
- Status
- Not open for further replies.
- Home
- Member Areas
- The Lounge
- Sound Quality Vs. Measurements