Audibility of output coils

[jan.didden]

Re: Re: Re: Re: Re: Here we go again

Jan,

The point is we cursorily disregard linear frequency response distortions well within the passband for they are indeed small, but they are there.

In this example we get 1e-3 dB at 1KHz /90 e-3 deg., 2 e-3 dB at 10 KHz / 0.9 deg. and so on.

But if we look at the error itself and in dB, it turns out to be in the mentioned -60 dB range. I am not arguing whether this is of consequence or not given the fact one may also posit it is swamped by other much larger errors from cabling, drivers etc. But .... Are we ready to dismiss what others say about audibility without further consideration?

I am not.

Rodolfo

Ahhh... Got it. I think

But then, I don't think we can compare -60dB non-linear distortion (THD) to -60dB linear distortion (1/1000dB freq response deviation). I would confidently declare the latter inaudible.

Edit: I see Bob has made a similar comment

Jan Didden

 

[jan.didden]

Excuse me?

This must have been one of very few moments of un-clarity in Doug's mind, if quoted correctly. Yes, if an amplifier boasts a meaningless damping factor of say 1000, it might then become 500.

But "appreciable power loss" into loads less than 4 ohms? Not in any practical application. Such a typical coil might consist of 15 turns of 1,2mm wire, 6mm diameter. The resistance of that would be about 5,5 milli-ohm. How low must the load go for that to cause appreciable loss?

Johan,

Right, I think my parafrasing was a bit off. In his book he actually speaks about 'measurable power loss in 4 ohms' rather than appreciable power loss. His example was a 20 turn 1.5mm coil 1 x 2 inch of 6uH which he says has about 20milliohms resistance.
My bad.

Jan Didden

 

[PMA]

Agreed, Jan.

Even quite low levels of non-linear distortion are audible (in orders of 0.01%, in case they are hi-order harmonics, like crossover distortion). And even quite high levels of LINEAR distortion are inaudible.

I have my test disc with signals generated for listening test purposes, with different harmonic content (for non-linear distortion audibility test).

 

[jan.didden]

It does, base 7.

Anyway, my questions earlier to John Curl were heading toward something that Mike said- take an amp that doesn't have an output coil, connect it to a highly capacitive load (like my ESLs), then add a 2uH coil and see if there's any audible difference. I have my own hypothesis about what the results will be...

Stuart,

I think you would want to keep the first session pretty short, lest the factory-installed smoke in the amp starts to escape..

Jan Didden

 

[Edmond Stuart]

Re: Re: Re: Re: Re: Re: Re: Re: Re: Re: Here we go again

I wonder what the output inductance is of a vacuum tube amplifier due to the leakage inductance of the output transformer. I'm guessing it might easily be more than a couple of microhenries.

Bob

Hi Bob,

A very relevant question! For the same reason I once asked a Dutch manufacturer of transformers (Rik Stoet) for some figures. They were around 25uH.

Cheers,

 

[PMA]

high capacitive load

Well - this would highly depend on amplifier circuit design and parameters. In case that bandwith is not etremely high, and slew rate is not extremely high, and circuit is designed in a particular way, you really do not need output coil. I can load one of my amps with 3uF in parallel with resistive load and use 1kHz square as an input signal without problem. BW is about 70kHz(-3dB), step response is exponential-like. Exponential-like into resistive load. With capacitor, there are several quickly damped waves on transient responses. The amp is tested with capacitive load in range from 100pF to 3uF. No output coil used.

 

[Edmond Stuart]

Oh, forget about all this nonsense. If you want the ultimate speaker cables, go to your local auto parts supplier and pick up a couple of sets of those really cheap, made in China, jumper leads.
Cheers, Glen

Okay, but isn't it possible that the cable could sound worse after burn-in?

Blasphemer! Burn!

BUT how about this:
http://www.vandenhul.com/cable/thrd.htm

Cheers,

 

[Edmond Stuart]

Re: Re: definition of 'rock solid'

A "Rock Solid" measurement method for the given application.

Cheers,
Glen

Hi Glen,

You missed my point. How do you KNOW that your method is appropriate for the given application? If you do know, you make use of foreknowledge. And where do you get this foreknowledge from? Most probably from an even more rock solid measurement method.

Cheers, Edmond

 

[GK]

Re: Re: Re: definition of 'rock solid'

Hi Glen,

You missed my point. How do you KNOW that your method is appropriate for the given application? If you do know, you make use of foreknowledge. And where do you get this foreknowledge from? Most probably from an even more rock solid measurement method.

Cheers, Edmond

Edmond, what on earth are you going on about? I can't work out if you are being serious or just stirring. I gave you the method I use for measuring the open loop gain and phase of my amplifiers which use differential input stages. How do you think I know that it is an appropiate test method for this application? You agreed that it is so, in a previous post, didn't you?


Cheers,
Glen

 

[GK]

Hi Glen,

sorry to say - I might well be the nincompoop that you refer to.

I dislike the sound of most resistors in the signal path.

I recently changed an electrolytic for a film cap and for me it was MUCH better

then when I DC coupled it was another improvement.

leads sound different even when the basic LCR model has them fine up into the mega hertz region

I have absolutely no technical explanation how it is I can hear this.

I just accept that I do not fully understand any of this stuff

On the subject of a coil on the o/p - I will try it this w/e and report it here ( for what is it worth ) I know in advance whatever I say some will believe me and others will not - It's a funny old world

cheers

mike

Hello Mike.

I think that I can predict your answer. You're going to hear it.

Cheers,
Glen

 

[forr]

Hi,

Quoted from
http://www.diyaudio.com/forums/showthread.php?postid=1217019#post1217019
PMA
---I can see no special "transmission line behaviour" for audio amplifiers and speaker cables. If, then only for RFI (not the signal through amp). This can be effectively damped by placing a resistor of some 50 - 300 ohms directly at speaker binding posts.---
Bob Cordell
---a speaker cable will exhibit some of the characteristics of a mis-terminated transmission line at high frequencies in the MHz range.---

This is the conclusion of Cyril Bateman's more recent work :
quoted from
http://www.diyaudio.com/forums/showthread.php?postid=1202265#post1202265
Forr
---one may think that everything able to diminish RF returns via the speaker cables is wellcome.---

This make me to open this new thread about RFI induced distorsion
http://www.diyaudio.com/forums/showthread.php?s=&threadid=102083
There has been about ten persons willing to read Paul Miller's article on the subject. Some parts of it may be taken with circumspection but RFI in audio amps which should better scrutinized, particularly in this thread :
http://www.diyaudio.com/forums/showthread.php?postid=1203822#post1203822

 

[Edmond Stuart]

Re: Re: Re: Re: definition of 'rock solid'

Edmond, what on earth are you going on about? I can't work out if you are being serious or just stirring. I gave you the method I use for measuring the open loop gain and phase of my amplifiers which use differential input stages. How do you think I know that it is an appropriate test method for this application? You agreed that it is so, in a previous post, didn't you?

Cheers,
Glen

Hi Glen,

I'm damn serious and definitely not stirring. But apparently, you are still missing my point and apparently I wasn't clear enough, sorry.
So, I give it one more try.

My point is that, when examining the NFB LOOP gain (not always the same as gain without feedback), one actually should inject a test signal somewhere into LOOP (i.e. Middlebrook's method).

Your point is that, in case of a differential (read symmetrical) I/P sage. this is equivalent to overall gain with the FB disabled. Right?

And, indeed, I do agree with that, but I didn't on face value. I first verified it by simulation in my way, how else could I agree with you?
Just by (false) reasoning that I/P stage is symmetrical? NO!, because the I/P stage is loaded by an asymmetrical circuit (current mirror and VAS). So the whole thing isn't perfectly symmetric. However, that appears not to introduce a significant error. But how do we know that and how sure we can be about that? Where did we get this knowledge from? Just by comparing the two methods. In other words, you'll need the other method, maybe only once, but still it is an essential part of the learning/simulating/measuring process.

Perhaps, you'll find this academical blah blah, but I like to get things exact and right, as there is already to much non-academical blah blah on this forum, though happily not from you.


Cheers, Edmond.

 

[Bob Cordell]

G'day Bob

I think one can say with a great deal of certainty that a coil, switched in and out of the circuit in an A-B test, will induce an audible effect, provided that the coil is large enough.
For one thing, there will be a reduction of amplitude at high frequencies. If the coil is made big enough, the loss of treble could easilly be percieved.

However, when one considers just how big an inductor would need to be to induce such an audible effect, a uH or two looks quite irrelevant - especially so when contrasted with other factors such as room accoustics and speaker positioning which can provide a drastically larger contribution.

Also, one cannot assume that just because an amplifier may not have an output coil, that it will not have an electrically equivalent output inductance.

Distortions caused by coil non-linearity would would show up on standard linearity measurements. Since we can easilly make amplifers with output coils that measure exceedingly well (Halco as an commercial example) in this department, it is obvious that this is almost a practical non-issue, provided that care is taken with the design, just as it has to be in many other aspects.

Cheers,
Glen

I agree completely with what you have said:

That a big enough coil will likely produce an audible difference (which may not necessarily be the same as an audible degradation), but that the coil might have to be larger than what we have been talking about to create much of a difference based on frequency response.

That amplifiers without coils also have some amount of effective output inductance.

That coil nonlinearity would most likely show up on an HF distortion test.

Cheers,
Bob

 

[GK]

Re: Re: Re: Re: Re: definition of 'rock solid'

Hi Glen,

I'm damn serious and definitely not stirring. But apparently, you are still missing my point and apparently I wasn't clear enough, sorry.
So, I give it one more try.

My point is that, when examining the NFB LOOP gain (not always the same as gain without feedback), one actually should inject a test signal somewhere into LOOP (i.e. Middlebrook's method).

Your point is that, in case of a differential (read symmetrical) I/P sage. this is equivalent to overall gain with the FB disabled. Right?

And, indeed, I do agree with that, but I didn't on face value. I first verified it by simulation in my way, how else could I agree with you?
Just by (false) reasoning that I/P stage is symmetrical? NO!, because the I/P stage is loaded by an asymmetrical circuit (current mirror and VAS). So the whole thing isn't perfectly symmetric. However, that appears not to introduce a significant error. But how do we know that and how sure we can be about that? Where did we get this knowledge from? Just by comparing the two methods. In other words, you'll need the other method, maybe only once, but still it is an essential part of the learning/simulating/measuring process.

Okaaaaay Edmond. I hate to be picky, but it is just obvious to me that the phase and gain imbalance between the non-inverting and the inverting input are negligibly small for the open-loop test I was describing.
The current mirror does provide and asymmetrically load to the LTP, but the push-pull current drive provided to the base of the VAS is symmetrical by virtue of the design's operation and the phase-shift incurred by the current mirror is minor because it’s bandwidth is high – much higher than the output stage, for example.

Cheers,
Glen

 

[GK]

That a big enough coil will likely produce an audible difference (which may not necessarily be the same as an audible degradation),

Hi Bob

This is something I touched on previously, but it did not catch on. Audibility does not necessarily imply sonic degredation. In fact, if your system appears to lack a little bass, it might even sound better with a small treble cut!

Personally, I’d prefer to tweak my equaliser instead of worrying about any output coil, but that’s only because I’m a philistine who likes elaborate tone controls.

Cheers,
Glen

 

[phase_accurate]

Did you guys ever notice that very few of these old-fashioned amps with output caps (not that I am a fan of those) rarely have output inductors ?

Maybe this is because those caps are inductive enough by themselves but maybe it is because those inductors are seldeom needed at all.

Regards

Charles

 

[Hartono]

hi Charles,

Also because those cap have enough ESR.

Hartono

 

[darkfenriz]

How about LC series resonance?
C is chosen to cut infrasonic, L to cut ultrasonic but they can resonate somewhere in between.
Still we have series C in passive HP section of crossover, but of lower value.

Adam

Edit:
However series impedance should still be insignificant in terms of damping, responce etc., some just fear any extra audio resonance...

 

[Hartono]

"How about LC series resonance?"

well it might change the frequency response but should be tiny

 

[Edmond Stuart]

Re: Re: Re: Re: Re: Re: definition of 'rock solid'

Okaaaaay Edmond. I hate to be picky, but it is just obvious to me that the phase and gain imbalance between the non-inverting and the inverting input are negligibly small for the open-loop test I was describing.
The current mirror does provide and asymmetrically load to the LTP, but the push-pull current drive provided to the base of the VAS is symmetrical by virtue of the design's operation and the phase-shift incurred by the current mirror is minor because it’s bandwidth is high – much higher than the output stage, for example.

Cheers,
Glen

Hi Glen,

You ARE right and I don't disagree with your final conclusions. But I don't feel comfortable with your approach, leading to those conclusions. Okay, things I like to verify, are totally obvious for you and maybe for the rest of the world, but, being a very careful man, I never take these things for granted.
If you are damn sure you can neglect the phase of the current mirror or whatever detail, then nothing is wrong with your method.

But why take the risk of overlooking something and why NOT being on the safe (inverting) side? It cost you nothing, in fact, it is even simpler, because you don't have to correct the gain for the attenuation of the feedback resistors and it is appropriate to all topologies.

Oh, I almost forget, what if the FB network includes some form of lead compensation, that tiny capacitor, you know? How do you measure or spice the NFB loop in this case?


Cheers, Edmond.