Swirling phasing sound at hi output

[Jana]

double post. server debug thingy.

[Jana]

Very interesting thread. I also read part of the thread referenced on the Metro forum and Randall Aiken seems to confirm something that I discovered about 20 years ago.

Grab a cold one and let me digress for a while.

I was doing a lot of experimenting with amps, had quite an assortment of Marshalls, Fenders, and amps that I built. Most of them sounded "good" when pushed hard, some of them sounded like a Hereford with diarrhea, one amp, an amp that I just threw together from spare parts, sounded absolutely awesome when pushed hard.

Definition: what, to my ears, is sounding awesome? When the amp is pushed hard it has to stay focused. The notes will be distorted, yet still be defined, not turning to an undecipherable mush or have all kinds of weird overtones, undertones, no-tones.

There has to be a reason, I thought, why such a wide range of amps with different tubes, different voltages, different output transformers, etc. could sound good or not. There are those magic marshalls out there, the magic fenders, the amps that have become legend because of the way they sound. Why?

I am not saying that the following answer is the ultimate answer or is the only piece that makes the amps magic. What I am saying, though, is that this is an important piece of it.

A thesis: what most people are assuming is power tube distortion is actually phase inverter distortion (I nearly got tarred and feathered and run out of town on a red plating EL34 for proposing that a few years ago on another forum).

Back to the amps I was experimenting on: I measured voltages, poked and prodded with an oscilloscope, and kept at it until I found that the amp that I threw together from spare parts had this going on:

1. the bias voltage for the power tubes was greater than the peak voltage that the phase inverter was able to deliver before it clipped.

2. the bias voltage for amps that sounded good but not great was close to or even a little lower than the peak voltage that the phase inverter could deliver.

3. the amp that sounded like Hereford excrement had a bias voltage that was much lower than the peak voltage of the phase inverter.

My conclusion:

Never drive the power tube grids to a positive voltage.

My solution:

Drop the voltage to the phase inverter so that it can't deliver as much peak voltage. I do this by tweaking the value of the supply resistor after the choke. Most times it doesn't take much, just changing the 20K to 22K will oftentimes do it (speaking of Marshall circuits here).

Speculations:

Whether by design or sheer luck (think 50 watt plexi here), the 20K resistance (actually 2 - 10K in series) is really close to meeting that requirement. Given component tolerances of the day (10%) it is possible to have 22K there. Then, factor in where the tubes bias and if a pair of tubes that needs more bias are is installed in an amp that happened to have high tolerance resistors, viola, a magic amp. On the flip side if an amp happened to get low tolerance resistors and winds up with about 18K and got a pair of tubes that needed low bias voltage... a Hereford is born.
There is also the variable of the AX7 driver tube and its gain variation.

So, my tweak for every amp I build is to make sure that the phase inverter will not push the power tube grids positive.

Those of you still awake from my rendition of "Goldilocks and the Three Bears," ... flame away.

[Voodoo_Man]

So, my tweak for every amp I build is to make sure that the phase inverter will not push the power tube grids positive.

Those of you still awake from my rendition of "Goldilocks and the Three Bears," ... flame away.

Great moments in history have arrived. I believe you. Usually I don't believe anything a girl says. I have one production line amp that I think sounds outstanding (for a production line amp that is). A 100 Watt B-52. Who threw that shoe? For the small amount of cash I paid, I was blown away. I didn't pass out from astonishment, but it does sound nice. Great preamp distortion. Tried to get them to send me a schematic, but no dice. Think I'll break it open some day and test your theory.

God Bless,
John

[Voodoo_Man]

Sorry about that.

[Alexo]

That is very interesting, Jana. It both jibes with and contradicts some of my experiences:

First, I find that post-pi master volumes can get you all of the ugly distortion you want, and very little of the good stuff.

Second, we all know the Marshall 18 watt, whose well-loved tone is generally regarded as coming from its overdriven power valves. These amps, ime, sound best when they're pushed to the point where the EL84's are breaking up, but the pi remains clean; crank them further and you can hear a fuzzier tone introduced as the pi starts breaking up.

Now for an experience that confirms your findings - the tweed deluxe. The pi is probably the first to break up in this amp, and even when totally cranked, I think the 6V6's are only mildly overdriven. But the 5E3 uses a different pi from most of the amps we're talking about, and by itself, without a little compression from the power stage, it doesn't sound so hot either.

I do think that output tube distortion can get really smeary and blatty and become a big mess o' mush, but it depends on how the output stage is designed - the rc constant of the coupling caps, bias point and load, what the screens are doing, screen resistor values, the sonic character of the preamp, etc., etc., can make or break the big bottle distortion.

I also find that power valves just pushed to the edge of breakup can add a degree of compression and "open-ness" that can do wonders for otherwise so-so preamp or pi distortion.

[billyz]

This is turning into an excellent discussion. Jana, you may be confirming some of Randall Aiken's comments too. He did suggest another solution to the phenomenon, that is, reducing the gain of the PI by increasing the Tail resistor value. It seems there may be more than one cause and even different types of distortion generated. I don't here the Buzzing of Mosquitoes as much, maybe a slight phase shifting. I have heard some clips that sounded rather buzzy and it was not a good thing.

If one were to implement the VVR ,then the Zener diode solution might not be the best as it does depend on the Zener being a few volts over the Bias voltage. So perhaps Jana's approach would work best , of simply reducing the plate voltage to lower it's peak Voltage.

BTW, my 1957 5E3 does not do it even when cranked full up, and it sounds marvelous, all amps should sound so good. I do have a larger Blackface OT in it though and it really kicks the butt out of a stock wimpy tranny, along with the P12Q .

[Alexo]

This is turning into an excellent discussion. Jana, you may be confirming some of Randall Aiken's comments too. He did suggest another solution to the phenomenon, that is, reducing the gain of the PI by increasing the Tail resistor value. It seems there may be more than one cause and even different types of distortion generated. I don't here the Buzzing of Mosquitoes as much, maybe a slight phase shifting. I have heard some clips that sounded rather buzzy and it was not a good thing.

If one were to implement the VVR ,then the Zener diode solution might not be the best as it does depend on the Zener being a few volts over the Bias voltage. So perhaps Jana's approach would work best , of simply reducing the plate voltage to lower it's peak Voltage.

Raising the tail resistor doesn't affect the gain, but it reduces the maximum output voltage from the pi, which is what Dana is talking about. It also stabilizes the pi more, which may be a good thing.

[j-po]

Ok. Tried the zener trick to prevent coupling cap charging and subsequent crossover distortion. Swirling buzz is gone but so is much of the singing feedback of my wreck.

[Structo]

My solution:

Drop the voltage to the phase inverter so that it can't deliver as much peak voltage. I do this by tweaking the value of the supply resistor after the choke. Most times it doesn't take much, just changing the 20K to 22K will oftentimes do it (speaking of Marshall circuits here).

Speculations:

Whether by design or sheer luck (think 50 watt plexi here), the 20K resistance (actually 2 - 10K in series) is really close to meeting that requirement. Given component tolerances of the day (10%) it is possible to have 22K there. Then, factor in where the tubes bias and if a pair of tubes that needs more bias are is installed in an amp that happened to have high tolerance resistors, viola, a magic amp. On the flip side if an amp happened to get low tolerance resistors and winds up with about 18K and got a pair of tubes that needed low bias voltage... a Hereford is born.
There is also the variable of the AX7 driver tube and its gain variation.

So, my tweak for every amp I build is to make sure that the phase inverter will not push the power tube grids positive.

Those of you still awake from my rendition of "Goldilocks and the Three Bears," ... flame away.

Good stuff Jana,
Did you discover a magic PI voltage or what was the ballpark figure on plate voltage for a 12ax7 driver?

I know on the D clones we usually are told to keep the PI up around 285-300v.
I think 300v is max on most 12ax7's isn't it?

I think you are right in that what a lot of people think is power tube distortion is coming from the PI.

[j-po]

Just continuing my previous post. I guess the phasing nonharmonic tones of the crossover distortion were exciting feedback into the strings better than the smooth and sterile clipping of the power tubes with the zeners installed.

[briane]

Geat stuff jana, exactly what my thoughts have been for a long time.

Though I have never investigated in so much detail, I would usually just adjust the power resistors and caps, till I got it right. Sometimes its usually just easier for me to try a fix, and see if it works, though not scientific. Often I have been called a cowboy programmer, and again it is true.

I have also found the swirling can be induced through parasitic oscillations (very small ones) being introduced, and minor lead dress changes can fix.

[M Fowler]

Jana that's not the end of the story is it? Good information thank you.

Mark

[j-po]

I'll take an opposing view on Jana's idea. I feel it's a shame to loose all power tube distortion when the whole swirling problem/nasty clipping is rather in the bias drifting under saturation than the PI itself.

Why not go further and leave out PI distortion too? Just add a stage to compensate for gain. Oh yeah, then you have a modern hi-gain amp. I must admit I like them too but when at amp garage I'm more after that rounder distortion of pentodes.

[jjman]

I had some extreme phasing going on with my homebrew that was being driven very hard by the LTPI. The 6v6s are under 5e3 conditions so the grids were capable of hitting zero volts pretty soon on the dial. I cooled it down by lowering the 220k grid resistors to 82k or so.

Initially I had thought the PI was the source of clipping since the 6v6 grids (PI outputs) would not reach into positive voltage on the scope when driving hard. Then I read about "clamping." With clamping the 6v6 “prevents” the PI out/grid in from moving measurably positive, which displays as a clip at about zero volts.

This also explained why the PI could swing more, before showing a clipped-top voltage on it's outputs, when I used my post PIMV. Preventing the PI outputs from fully reaching the 6v6 grids prevents the "clamping" that looks like the PI is choosing to clip. So with the MV down a little I could see where the PI itself would truly clip, which was much later than where the PI outputs would be “clamped” by the 6v6s with the MV maxed. I’ve wondered how one would determine this point w/o a post PIMV. Maybe by temporarily paralleling some additional grid leaks?

Cooling down the signal to the 6v6s also removed the crossover distortion, during clipping, and the phasing now that I think about it.

[j-po]

Yup, that's it. Grids clamping to zero, but allowing as much negative voltage as the PI can manage results in the bias shifting as the coupling caps hold this shifted value for a while. (Smaller value coupling caps or resistors to bias source/ground limit the ability to hold this incorrect bias as well as limit PI output.) Now as the note decays the bias comes gradually back to spec which results in a decaying amount of crossover distortion = swirling buzz.

The zener trick clamps the PI output at a desired negative level thus retaining symmetric clamping and no bias shift. Stating the obvious for some of you guys, I guess...