How to diagnose frequency response

[TriodeLuvr]

Here is a square wave across my 8ohm dummy resistor load. I swear it wasn't this bad when I first built it, but that was 10 yrs ago now. How should I start trying to fix it?

I would start by applying 20dB feedback. That's what most pentodes need to tame IMD and produce reasonable dampening. Attach a small phase advance cap across the feedback resistor to stabilize the amplifier sufficiently for the remainder of the work. Try a value around 100pF. I would also use the same Zobel values across the 16Ω output that Heathkit applied (100Ω in series with 0.02uF). Next determine the frequency of the ringing and the small-signal frequency response, and make a note of the severity of the ringing. At this point you can experiment with the values of a step network at the input stage to create a reasonable square wave. Bear in mind it doesn't need to be perfect. A small amount of ringing is not only permissable, but might even be preferable for the purpose of sound quality.

Jack

 

[Selmerdave]

Overshoot and Ringing in Tube Amplifiers.

That's a nice clear site...thanks.

 

[gadget73]

The fact that the ringing is so different on the positive vs negative swing is interesting. I wonder if thats an artifact related to amplitude. If I'm reading that scale right its about 16 volts peak to peak, which is maybe a little much. I normally do these under 5v peak to peak, but so long as this isn't pushing close to the limits of clean output power its probably fine.

 

[wparks]

I don't know if I would =start= with 20dB of global negative feedback. Some transformers will not allow an amp to be stable with that much. These transformers probably will, but beware. The ringing I am seeing in your waveform is a classical under-damped transformer oscillation, because the feedback compensation capacitor is a bit on the light side.

It would be good to confirm how much negative feedback the amp has- but here it's likely academic. To do this I would disconnect the transformer side of the feedback resistor and connect it to ground (so you still pull the same current from the pre-amp cathode) and measure open loop gain ( 20LOG(Vout/Vin) using a very low undistorted amplitude (1W is fine), then reconnect it and measure closed loop gain. Feedback will be open loop gain minus closed loop gain. 20dB is the classical Williamson amount, I consider it to be a maximum, but actual amount used depends entirely upon how well behaved your output transformers are. Usually you will also want to consider how much closed loop sensitivity you want the amplifier to have (how much input signal is required to drive to maximum output power) and that may limit the amount of negative feedback (less for higher gain.)

The feedback resistor only sets the gain level, it has nothing to do with the transient response.

Amplitude at 1kHz into 8 ohms versus feedback resistor:




There are two approaches to dialing in the feedback compensation capacitor (the small "phase advance" cap parallel the feedback resistor)- One is to reduce amount of ringing on the square wave, to make it look much more like an ideal square wave, which visually is trading off ringing duration/amplitude for rise/fall edge rate and rounding of the edges. This is a quick and dirty approach, but can give good results.

Transformer ringing response versus feedback compensation capacitor value:



The ultimate way to adjust the compensation capacitor is to observe the measurable bandwidth above 20kHz. This bode plot is from an arbitrary 6AQ5 push pull amp I built. The blue waveform is open loop, once negative feedback is applied the improvement is remarkable at the low end, and remains constant versus any amount of feedback compensation capacitance. The difference is at the high end. The orange waveform shows a dramatic rising response and resonant peak from output transformer resonant oscillation at around 60kHz. This will vary in frequency and amplitude per transformer design. By applying more and more feedback compensation capacitance, one can find the right amount that creates a flat response all the way out to the resonant frequency.

 

[Mr Boochie]

The fact that the ringing is so different on the positive vs negative swing is interesting. I wonder if thats an artifact related to amplitude. If I'm reading that scale right its about 16 volts peak to peak, which is maybe a little much. I normally do these under 5v peak to peak, but so long as this isn't pushing close to the limits of clean output power its probably fine.

Looks like a possible phase-inverter imbalance, crossover distortion, or an oddball application of global feedback to that inverter.

 

[wparks]

The fact that the ringing is so different on the positive vs negative swing is interesting. I wonder if thats an artifact related to amplitude. If I'm reading that scale right its about 16 volts peak to peak, which is maybe a little much. I normally do these under 5v peak to peak, but so long as this isn't pushing close to the limits of clean output power its probably fine.

Good catch @gadget73! You caught me on that one once before. Yes, amplitude is too high. Dial it down and you will see symmetry return.

For feedback tuning and bandwidth measurement (not "power bandwidth" at maximum minus 1dB) generally the smaller the waveform amplitude the better, so you are not taxing the amplifier, but looking at it's pure ideal response. 1W into rated load is usually used, but I have seen lower used on "flea power" amps.

 

[Mr Boochie]

Good catch @gadget73! You caught me on that one once before. Yes, amplitude is too high. Dial it down and you will see symmetry return.

For feedback tuning and bandwidth measurement (not "power bandwidth" at maximum minus 1dB) generally the smaller the waveform amplitude the better, so you are not taxing the amplifier, but looking at it's pure ideal response. 1W into rated load is usually used, but I have seen lower used on "flea power" amps.

I'm a little confused here. Are we talking about @justinis amp or your amp? Or are you working on his amp?

Thanks!

 

[wparks]

@justinis's amp. Turn the amplitude down and you will see. I have seen this in multiple designs.

 

[Mr Boochie]

@justinis's amp. Turn the amplitude down and you will see. I have seen this in multiple designs.

So, the asymmetric square wave of the too-large input signal, could be due to asymmetric clipping?

Thanks!

 

[gadget73]

Unknown honestly, but I'm curious if it looks different at a lower level.

also sort of curious when this clips on a sine wave if its equal or not. That may be a symptom of whatever sounds "off" in this thing.

 

[wparks]

It's a bit of a head scratcher. I've seen it in multiple push-pull amps, (I don't think I have seen it in SE- I'll have to remember to check that at some point). It can happen to either the top or bottom flat. Very well could be due to the difference in output impedance of the phase inverter driving one output versus the other, or maybe non-linearity or maybe different cathode impedance of the voltage amplifier at the one level versus the other. Perhaps an interesting area to do some study.

 

[Mr Boochie]

Just for clarity's sake, I believe we are talking about a DIY 7591 push-pull amp.

https://audiokarma.org/forums/attachments/st-70-clone-b-schematic-pdf.3669154/

 

[gadget73]

SE will do it if the output tube bias point is poor. If its not center on the load line it will cut off or saturate on peaks depending on where its sitting. Ideally it does both at the same time.

Push pull is usually some unequal drive issue. Usually the uneven impedance from a split load isn't a big deal unless its just badly designed. Too wimpy a triode working into too low a grid resistance maybe. I had it on an ST70 with a bad 7199, there was so much heater cathode leakage that it just couldn't make proper swing on that side. This being a long tail pair won't have that issue unless something is very wrong. The CCS will force equal output so long as the plate resistors are equal

 

[Mr Boochie]

SE will do it if the output tube bias point is poor. If its not center on the load line it will cut off or saturate on peaks depending on where its sitting. Ideally it does both at the same time.

For sure, most single-ended amps will have asymmetric clipping. Overdriving an SE output section tends to push the duty cycle away from symmetry more than push-pull outputs do.

I believe this asymmetry favors even-ordered harmonics, which can make the overtone sound "warmer," and SE usually has a slower onset of overdrive.

This could be why many tube heads like a good single-ended design.

 

[wparks]

The slightly larger than ideal ringing in the square wave response should be adding to a brighter character at the high end, not diminishing it.

Just some spit-balling- as nothing else is really standing out . . .

I notice the driver board schematic indicates this amp will support either the 12AU7 or the 12BH7- I have heard many good things about the 12BH7 tonality (if you can find examples that are not too microphonic) and have also heard many 12AU7 tubes sound rather dull.

You said that you "recently" discovered this amp sounded dull compared to other amplifiers. Do you think this is a recent change in this amp's performance, or do you think this difference was always there, but perhaps not noticed until you tried a class D amp?

Was curious about how you are using the volume control on this amp- During these listening sessions where you notice the sound being kind of dead, is the volume on the power amp way down with a full amplitude source, or volume on the power amp all the way up, with the pre-amp turning the volume way down? I'm just wondering about the 50K volume pot with 7.32K stopper resistor on the 12AX7- with the power amp volume turned way down, that will be putting enough series resistance into the grid of the 12AX7 that you could be getting some Miller roll-off, but not if the power amp volume were up most or all the way.

 

[Mr Boochie]

The 12AU7 can swing more current than the 12BH7 (20 vs 10mA), so maybe that's why it's being used?

 

[justinis]

Thank you everyone for this info! I will check the negative feedback and adjust to see if I can damp the ringing.

The slightly larger than ideal ringing in the square wave response should be adding to a brighter character at the high end, not diminishing it.

I notice the driver board schematic indicates this amp will support either the 12AU7 or the 12BH7- I have heard many good things about the 12BH7 tonality (if you can find examples that are not too microphonic) and have also heard many 12AU7 tubes sound rather dull.

Sadly I am using 12BH7s

You said that you "recently" discovered this amp sounded dull compared to other amplifiers. Do you think this is a recent change in this amp's performance, or do you think this difference was always there, but perhaps not noticed until you tried a class D amp?

I cannot say for sure, but I think it has always been there. I've never really been satisfied with how it sounds. I just didn't have something to compare it to.

Was curious about how you are using the volume control on this amp- During these listening sessions where you notice the sound being kind of dead, is the volume on the power amp way down with a full amplitude source, or volume on the power amp all the way up, with the pre-amp turning the volume way down? I'm just wondering about the 50K volume pot with 7.32K stopper resistor on the 12AX7- with the power amp volume turned way down, that will be putting enough series resistance into the grid of the 12AX7 that you could be getting some Miller roll-off, but not if the power amp volume were up most or all the way.

I have been using the volume on the amp to limit volume, with the preamp maxed out. I will definitely try the other way to see if it sounds different. There's a chance the volume pot is 100k not 50k; I'll check that too.

 

[gadget73]

personally I thought the 12BH7 sounded fairly dark when plugged into a socket for a 12AU7. That was a voltage amp position though, not an inverter.

In my amp that can run several tubes in the long tail inverter spot, a 7AU7 (7v 12AU7), 6CG7, and a 6GU7 (6v 12BH7) don't honestly sound different enough to notice. A long tail pair has a lot of local feedback which does a pretty good job of cancelling out differences in that spot.