Strange motorboating issue

[DF96]

I found an interesting page here: basic-tube-(3) which does show the disruption caused by GNFB in a tube amp:

That page said about the graph:

The above graph shows the sine wave response a very ordinary tube amp
with average quality OPT without NFB and with NFB, and with no attempt to
tailor the open loop gain or phase shift to prevent the peaks in sine wave
response just outside the audio band.

Which frankly begs the question: Why not just drive the transformer correctly in the first place?

You may choose to define "correctly" as being 'lower impedance than the transformer was designed for'. Others do not.

Stability issues are the obvious manifestation of the pretty dire things happening to the signal that will also audibly happen within the realms of 'stability'. By including an OPT inside the loop you disrupt correct loop operation and guarantee sonic problems. The fight for stability with capacitors, Zobels, careful eyes on the poles etc is symptomatic of a fundamentally flawed approach of putting the OPT somewhere it has no business being - in a feedback loop.

Disagree. By including the OPT within the loop you ensure low impedance at the output, and some degree of correction of transformer problems. It is a way of improving a good transformer. It can't cope with a poor transformer, which is why circuits with poor transformers have to use no feedback or feedback before the OPT.

 

[Globulator]

You may choose to define "correctly" as being 'lower impedance than the transformer was designed for'. Others do not.

Perhaps that's your definition - it's not mine

Driving a transformer correctly is presenting the primary with the correct voltage waveform. The impedance will of course vary at all points during the waveform: the point is to get the voltage waveform right at the primary. Do you get it yet?

Disagree. By including the OPT within the loop you ensure low impedance at the output, and some degree of correction of transformer problems.

You are writing this stuff without thinking it through.
'News Flash' - to get a low impedance at the secondary you HAVE to be driving the primary with a low impedance. How else did you think it worked? You insist it is wrong to drive the primary with low impedance with my method so why is it such a great idea with yours?

At least with mine the correct drive is delivered to the primary the first time. By including the OPT inside the loop your initial drive starts of with a sloppy high impedance attempt until the error is detected at the secondary. That error then gets fed back (along the same error-prone audio amplifier) and moves the primary voltage up or down as required to get the right(ish) answer at the secondary.

Why are you claiming that a near voltage + correction is better than the correct voltage?? Is it because 'it's always been done like that'?

The only benefit to including the OPT in the feedback loop is laziness - the secondary is a nice low voltage low impedance source. In all other respects including it is a terrible idea and fertile source of problems.

 

[Globulator]

BTW Thorsten Loesch explains it here:

http://www.diyaudio.com/forums/tubes-valves/40911-can-partial-feedback-effective-global-fb-decent-damping-factor-5.html#post525580

"4) If the output transformer is inside the feedback loop it limits the amount of NFB applicable severely and distorts much more (especially at low levels) than neccesary. Drive it from a low impedance and the problems are much reduced."

And here for the OP (Thorsten again):
http://www.diyaudio.com/forums/tubes-valves/46996-shunt-feedback-effect-input-impedance.html

which includes an elegant implementation of a push-pull amp without the daftness of having the OPT in the loop.

 

[DF96]

Driving a transformer correctly is presenting the primary with the correct voltage waveform.

Not always. Impedance is an issue too.

The impedance will of course vary at all points during the waveform: the point is to get the voltage waveform right at the primary. Do you get it yet?

?

Why are you claiming that a near voltage + correction is better than the correct voltage?? Is it because 'it's always been done like that'?

As it is the secondary voltage which actually drives the loudspeaker that is the place which the feedback should take its voltage from. The reason it has always been done like that is that it is actually the correct way to do it, provided that the OPT is good enough.

Anyway, I am getting bored with this discussion. You do it your way. The rest of us will do it our way.

 

[Globulator]

Not always. Impedance is an issue too.

?

As your OPT, cable and speakers are apparently free of capacitance, inductance and mass then yes - it is a puzzle that impedance could ever change.

For the rest of us however who listen to 'music' - a collection of varying tones with variable derivatives and levels - on speakers often with mass and crossovers - it's our sad burden that driving a mechanical magnetic transducer with music will always involve a non-constant impedance .

The reason it has always been done like that is that it is actually the correct way to do it

I have given a number of reasons why your statement is false but you are too proud to learn a better way. You have come up with no reason why driving a primary with a poor initial signal mostly corrected later is better than driving it correctly in the first place. None.

Oh and if the OPT is good enough you have even less reason to use secondary feedback because it will be more tightly coupled to the primary.

The mere fact that Thorsten Loesch also pointed out these obvious facts (years ago!) should alert you to the fact you are missing something big here.

You do it your way. The rest of us will do it our way.

Your dogmatic view does not represent the DIY community, also see the links above and read them

Incidentally I was advising the OP how to improve his amplifier design - not you, it is you who chose to trot out your dogma here: not me. If we leave dogma behind we have the simple facts, not just supported but written by a world famous amplifier designer:

"4) If the output transformer is inside the feedback loop it limits the amount of NFB applicable severely and distorts much more (especially at low levels) than necessary. Drive it from a low impedance and the problems are much reduced."

Have you got it yet?

 

[costis_n]

Some discipline is required when doing this test. Omit the input coupling cap # replace it with a grid stopper value to the first stage and have a sine audio oscillator that goes down to a few Hz. Set the amp output level to be consistent say 1Khz .. Make sure that very little amplifier output power 0,5W is at the output, as the test at the LF end deliberately goes below the design cutoff frequency of the output transformer. What is under test is the closed loop phase angle with the output transformer in circuit, into a dummy load.
As with Williamson circuits, the response may actually rise around 2-5Hz range, ignore this. For the exercise it's worth plotting the loop gain too, but that can come later.
Once this technique is mastered, there should be no misunderstanding about closed loop amp design !!

richy

The 180 degrees point appears to be in 3-4 Hz. So I think that 12nF to give a -3db at 28 Hz is good?

I should also note that HF stability is good, with even freq response and -3db at 74 KHz.

Oh and I have swapped different ECC83's and EL84's , stability does not change.

 

[costis_n]

??

Curiously, the 180 phase shift is now happening at ~10 Hz with the 12 nf caps. Are electrons trolling me ?

 

[Globulator]

Maybe you'll need to get used to motor-boating?
I used to have one - they can be great fun

You already know the cure

 

[costis_n]

Maybe you'll need to get used to motor-boating?
I used to have one - they can be great fun

You already know the cure

Maybe you should actually read what I am posting -- there is no motorboating.

 

[Globulator]

Ah - you've cured it - hurrah!!

I only read that HF stability was good and that you had 180 phase shift at 3-4Hz then 10Hz - but if all that means it's cured then you have a result - congrats!!

I'll be off to the optician now then...

 

[costis_n]

I would actually appreciate comments/input on the topology in general. I am thinking the optimal position of the anode compensation instead of the first stage is between the anodes of the P.I.

 

[DF96]

If by 'anode compensation' you mean the CR network which adds an HF shelf to stabilise the feedback loop then the earlier in the amp the better. This is because this network adds HF load to the stage it appears in, so increases distortion (at least for triodes). An early stage will have much smaller signals so less distortion.

 

[costis_n]

The turd has been polished

I actually like what I hear.