Preference vs. audibility - please keep them separate.

[andy_c]

Andy,
when did Bob come out with the book, was it after 2008?

Bob's book came out in 2011.

Coming back to distortion and patterns/structure
Andy, did you try the Keith Howard add distortion software?

I tried it briefly but did not spend much time with it. Keith's listening tests in that article were sighted. Using the foobar2k ABX comparator with it could be interesting.

After measuring the bass response in my room, I've stopped worrying about amplifiers for now.

 

[Orb]

LOL
Thanks for the heads up on Bob's book, missed it.
What I like about Keith is that he pretty much sumarises the caveats between a controlled scientific environment, sometimes real-world requirements,limitations; probably because his background was a strong scientific degree followed up with masters with research lab knowledge/work.

TBH with this type of testing the process done by Harman on distortion % perception makes more sense than the traditional ABX selection; not a fan of ABX selection for subtle JND like this.
But sighted it is ok if one approaches this from preference and anecdotal, after all a fair few seem to think some distortion/specific pattern makes it sound better (which is two of the patterns included).

Cheers
Orb

 

[j_j]

Ok, does anyone want, perhaps, a script in octave that they can process a wave file from, add various distortions as they like, and then listen to the results?

It's not hard, you know. Just have to build in enough oversampling and downsampling that there is no aliasing of distortion.

 

[Orb]

JJ, whats wrong with the one done by Keith Howard linked earlier if people just want experience with distortion patterns?
Can do a whole album in each pattern, that way can see how long they last with preferences.
Appreciate his is non-linear distortion patterns, but then in a way that is more like the real world rather than linear lab controlled JND % distortion.
Can you remember if the Sean Olive/Harman group test was linear or non-linear with regards to their testing perception of % value/threshold as I cannot?

Cheers
Orb

 

[Groucho]

JJ, whats wrong with the one done by Keith Howard linked earlier if people just want experience with distortion patterns?
Can do a whole album in each pattern, that way can see how long they last with preferences.
Appreciate his is non-linear distortion patterns, but then in a way that is more like the real world rather than linear lab controlled JND % distortion.
Can you remember if the Sean Olive/Harman group test was linear or non-linear with regards to their testing perception of % value/threshold as I cannot?

Hello Orb. What do we mean by "distortion patterns"? Is this a time domain lookup table where we feed instantaneous voltage in, and get a modified voltage out?

 

[Orb]

Hiya Groucho,
if you look at post 289 in this thread you can follow what was done by reading the Stereophile article link - please appreciate there were followups included in that article and they are important to read as well.
Basically distortion most of us are talking about is not random but has a harmonic structure or pattern, same way you can have with jitter (random "white noise" type jiiter but there is also critically correlated jitter that has a specific structure-pattern and can then be linked to the data/music played).
The simulation software is the other link in that thread is closer to the concept of correlated jitter rather than white noise; this is a loose comparison before any technical semantics are argued thanks
Cheers
Orb

 

[Groucho]

Hiya Groucho,
if you look at post 289 in this thread you can follow what was done by reading the Stereophile article link - please appreciate there were followups included in that article and they are important to read as well.

Thanks. All is clear: in the software's readme file it describes how the software produces transfer function files - which is what I was going to say would be very interesting to look at, possibly more so than the nominal harmonic amplitudes. (So does a 'saturating' transfer function sound better than a 'peaking' transfer function for example?) And the readme also makes another point I was going to mention: that real distortion mechanisms may be much more complicated than static lookup tables.

 

[Orb]

Ah ok I see where your coming from.
And I agree, but also please read (if you have not yet) the Stereophile article by Keith Howard as it also explains a few things, such as subtle changes to the transfer functions, their basis and validation,etc.
And yeah, you can only go so far doing a simulation software without spending years on the subject with the maths and critically if using it in the real world outside lab-scientific research conditions, but what he has done is pretty handy as it covers some major harmonic structures/patterns.
Hmm, there was one who did look at more complex harmonic distortion patterns relating to subtle tube differences and looked to correlate the harmonic distortion pattern to listening...
Will try to remember who that was and see if I can find it on the web, the guy is a bit controversial with some in the audio world
Cheers
Orb

 

[j_j]

It's free?

 

[NorthStar]

I don't know; is there a link?

 

[j_j]

I don't know; is there a link?

You have to get octave, with the signal processing and audio libraries, from sourceforge. Get the stable version.

I'll just PM you the script.

 

[Orb]

It's free?

From Keith's software:

AddDistortion should run under any Windows 32-bit operating system, from Windows 95 onwards.
It has been tested on Windows 95, Windows ME and Windows NT4.
To run, AddDistortion requires that the supplied CT_Pro.dll be installed in the same directory as the executable. This file contains the runtime files of Perfect Sync Inc’s Console Tools Pro
(http://perfectsync.com) which provides enhanced control over the console window’s appearance and function.
The AddDistortion console window is sized for use with screen resolutions of 1024×768 or greater.
If you attempt to use it with lower screen resolution you will have to navigate the console window via scroll bars.
AddDistortion is inherently a fast-running program. Its speed in practice is determined principally by disc reading/writing and memory operations, particularly when processing large Wave files.
If there is intensive hard disk activity as AddDistortion reads the input file this indicates that your computer is having to use virtual (hard disk) memory.
Operation will be significantly faster if all the memory operations can be achieved within RAM.
To facilitate this you may need to close other applications. If virtual memory is required even when AddDistortion alone is running, a RAM


Licence
This software may be freely distributed provided that it is unaltered and distributed in its entirety,
including the supplied DLL, Wave file and this information file. It may not be used for any commercial purpose.
No support or warranty is implied or given but if you encounter any bugs or have suggestions for improvements you are invited to post them via.....

Just scroll down and look for software title: Nonlinear distortion simulation
http://www.audiosignal.co.uk/freeware.html

But I do need to stress it is very useful to read the Stereophile article written by Keith Howard that is relevant to this software: http://www.stereophile.com/reference/406howard/index.html
As the follow-up mentions in the Stereophile article for 2.1

Meanwhile, readers interested in using my AddDistortion utility for their own listening experiments should know that I have recently posted a revised version (v2.1) on my freeware page.
This new version normalizes the transfer characteristic coefficients such that the output signal contains the input signal at precisely the same amplitude, which previously was not the case.
As it is not possible to arrange that the input signal and (distorted) output signal have the same loudness at all signal levels, this strikes me as probably the best compromise.—Keith Howard

Appreciate some of his software tools/simulations can require other audio utilities but this one is complete.
Bear in mind I think last updated in 2007, but Keith is pretty responsive (well has been to me in the past but I guess it depends if he is swamped with emails and the context).
Anyway some interesting software there (some really nice ones though are not available to public such as his one to analyse hirez download music due to the process and software involved)
Thanks
Orb

 

[Atmasphere]

Vinyl is pressed to reliably do 30-18K more than once or twice, but it sounds good. Tape can do 20-20K, but only with very reduced levels at high frequencies. It sounds good. (yes, I know about the various 4-channel audio systems, and, thank you, let's look at how they survive many playbacks.)

So ...

Vinyl can do 50KHz reliably, if the cartridge is up to it, the old CD-4 LPs of the 1970s are an example. We can cut 30KHz easily enough on our mastering lathe but we don't see much above 18KHz or so from any of our customer's source files or analog tapes.

Uh, not really.

Wideband amps have a higher unity gain bandwidth in general, and have much better performance at in-band frequencies, because the feedback has not disappeared at those lower frequencies.

That's the biggest mechanism.

A "fast" amp does not have THD start to roll off at 1kHz. A "slow" amp may have THD rolling off at 100Hz.

And the idea that transient speed matters is bogus as well, as long as you don't get into slope overload.

I am well aware of more than a few mistaken ideas about this issue, but for the most part, they are indeed mistaken.

What about amps that have no feedback and yet are wide bandwidth?

Err - no. Just no.

Bear in mind a cathode follower tube amp has enormous negative feedback, it's just local feedback.

This type of feedback is more accurately described as degenerative feedback. If described as 'local' it can still be loop feedback. Degenerative feedback works in real time, loop feedback does not, which is to say that it arrives at the input of the loop slightly delayed (due to propagation delay) from the original signal.

completely possible for my part, micro, I got that line from Wikipedia. A wonderful resource most of the time, but not always accurate. It appears, however, that if he is being misrepresented here, we are far from alone. He is being widely misrepresented to have said exactly the same thing in every misrepresentation. Seems odd...

https://www.google.com/#q=Peter+Walker+all+amplifiers+sound+the+same

Tim

FWIW, a lot of people still think that the math says a bumblebee can't fly. Turns out that the wrong Reynolds number was applied- bumblebees can fly just fine according to the math. But people go with the older story instead of the correct one.

Same for an emitter follower transistor. I don't know where the idea that negative feedback is evil came from, but it's just not true. Negative feedback is a valuable and necessary part of circuit design. At least it's necessary if you want to reduce distortion to inaudible levels!

--Ethan

Its probably more accurate to describe negative feedback as a design tool- one that is more useful in some places than in others. You don't need it to reduce distortion to 'inaudible levels'. You need it to reduce distortion in circuits that have significant distortion. Its a different matter entirely if the circuit is already low in distortion. Then you might be thinking of using it to reduce output impedance rather than anything to do with distortion.

The problem that is poorly addressed is: if you already have a circuit that is very low in distortion, why does adding loop negative feedback cause it to sound brighter while degenerative feedback does not (assuming the circuit is also wide bandwidth)? If you want to know where loop negative feedback got its 'evil' reputation, it is the lack of a defining answer to this question (most answers are something like 'that doesn't happen' which does not fly well). Many designers already know the answer (which is above); others that are more seat of the pants (hobbiests??) just know that if that they add it, it sounds wrong so they avoid it.

 

[Ethan Winer]

Its probably more accurate to describe negative feedback as a design tool- one that is more useful in some places than in others. You don't need it to reduce distortion to 'inaudible levels'.

My philosophy is for reproducing equipment to be as transparent as possible. That gets you closest to the artist's intent.

Then you might be thinking of using it to reduce output impedance rather than anything to do with distortion.

Sure, that's important too.

if you already have a circuit that is very low in distortion, why does adding loop negative feedback cause it to sound brighter while degenerative feedback does not

Who says it does? "Brighter" is frequency response, which is easily measured and confirmed.

--Ethan

 

[Atmasphere]

Who says it does? "Brighter" is frequency response, which is easily measured and confirmed.

--Ethan

Yes- and 'brighter' is also higher-ordered harmonic distortion. The ear interprets trace amounts as brightness (like it does with other distortions) but also as harshness. This is a bit harder to measure, as the ear is so sensitive in this area; in fact this is one area where the ear is more sensitive than our instruments.

 

[Don Hills]

Vinyl can do 50KHz reliably, if the cartridge is up to it, the old CD-4 LPs of the 1970s are an example. We can cut 30KHz easily enough on our mastering lathe but we don't see much above 18KHz or so from any of our customer's source files or analog tapes. ...

In practice, CD4 discs started clapping out after as few as 10 to 20 plays. Somewhere in the chaos I laughingly call my library I have some microphotographs of groove damage after 25 plays. Not pretty.
What velocity can you get at 30 KHz before you let the magic smoke out? Or do you cut at half speed, like the original CD4 discs were?

 

[Atmasphere]

A lot depends on how well a cartridge can track; of course in the old days that was a bit of a challenge!

Its all about not putting too much current through the cutterhead- as long as you keep an eye on that you're fine. We've not been doing any half-speed.

 

[Atmasphere]

^^ You are correct.

The head has a suspension made essentially of springs. If you don't use the feedback, you have two problems: there is a head resonance in the middle of the passband that has to be controlled, and you don't get any separation. Feedback seems to solve both. The feedback module provides about 30db. Attempts to build a zero feedback head by others in the past have so far not yielded effective results.

 

[Ethan Winer]

Yes- and 'brighter' is also higher-ordered harmonic distortion.

Good point, though the distortion in most modern gear is low enough not to hear.

The ear interprets trace amounts as brightness (like it does with other distortions) but also as harshness. This is a bit harder to measure, as the ear is so sensitive in this area; in fact this is one area where the ear is more sensitive than our instruments.

The ear can hear stuff better than test gear can measure? Not by a long shot!

--Ethan

 

[Atmasphere]

^^ with most 'modern gear' the distortion is audible as a brightness and a harshness while looking great on paper. This is what fuels the endless ad nauseum debates of tubes/transistors.

And yes indeed, our ears are very much tuned to listen for the higher ordered harmonics, in particular the odd ordered harmonics, as it uses them as loudness cues to determine how loud a sound actually is. Quite often these harmonics are occurring in the same regions to which our ears are also tuned: that of birdsong frequencies.

In this regard they are more sensitive than the best test gear, again the ongoing tubes/transistors debates bear that out in spades (or else many solid state designers are failing to understand what is 'negligible' and what is not). If the test gear were really getting to the meat of it, IMO the transistors would have snuffed out the tubes ages ago. But instead, there is still a vibrant tube market. If one really wanted to see the tube/transistor thing finally put to bed, the first thing to do would be to admit that there is something to it rather than assuming that it is the result of nutbag audiophiles. If an engineer fails to be pragmatic, he will also fail to be a good engineer.