question about power scaling

[Lindz]

I have a question for some educated ears and folks that know a bit about power scaling..

I have a 18 watt tmb (1974x style circuit with a few mods) that exhibits what I think one would call crossover distortion - here is a sound clip where you can hear fizzy overtones that trail the notes as they decay

http://snowboardmaterials.com/tmb_fizzz.mp3

As a noob when I built the amp I spent a lot of time and effort killing this problem and found that the Ruby zener mod combined with a diode clamp across the cathode resistor got rid of about 95% of it. Had a long thread here about it.. http://ampgarage.com/forum/viewtopic.ph ... zener+ruby

Well I was happy with things until I recently installed a Hall vvr on my amp that is also now fixed bias.

The amp exibits no fizzy artifacts when it is at full power but as I drop down the voltage with the vvr the fizzies come back. Kind of makes sense as the Ruby mod zeners clamp at a certain voltage and I guess as the amp scales down the values are too high to clamp the lower voltages??

In any event I am looking for a solution to this - I suppose the best solution would be to not have to use the Ruby mod and Bias clamp "band aids" in the first place so the scaling would work without issue

So question 1 is:

What tends to cause the crossover distortion in the first place? Since the diodes reduce it obviously there is voltage going too high around the resistors where the Ruby mod is installed and it also seems to be bias related since the clamp across the cathode resistor really helps tame what the Ruby mod did not, at least in my amp.

Any suggestions on simple changes I could make in the circuit that would kill the fizzies without resorting to the Ruby mod etc?

I had also thought about dropping the b+ a bunch to see what happens to the tone and fizzies at say 330, 320 or even 300 or less volts (its now at about 360 vp, biased at about 9.5 watts at idle). Maybe even drop way down and bias class a?

but I would prefer to see if I can change a few component values yet more or less keep the amp as I have now since I am quite happy with it and I am concerned that radical changes will also change the tone (though that will be a good learning experience anyway). I've tried a few things but am shooting in the dark due to my still limited understanding of the tube circuits (I am reading the O'Connor tut books now)

Question 2 is probably a lot more complicated:

Assuming I cannot kill the fizzies and keep tone that I am happy with without resorting to the Ruby mod etc. can anyone suggest a way to perhaps scale the Ruby mod along with the b+ scaling?

Though this idea would not scale I had thought about using really tight zener values that are right at the idle voltage vs the 2 volts higher that I am using now, thinking that it would clamp tighter and I could at least scale down more before I hear artifacts. Maybe even a bit below? (I suppose that might change the bias?) I assume that once I scale down a fair bit this would no longer help as the voltages drop way below the zener values as it does now but I guess I would at least be able to scale down a bit before the swirlies show up with a vengance

I suppose the magic answer would be to scale the zeners or some other sort of voltage clamp along with the b+ but thats above my pay grade in terms of electronics knowledge, hence my post

any thoughts?

[katopan]

Do you have a login at 18Watt? A little while ago we threw around ideas about a scalable Ruby mod to fix what you are asking about, but no one has built and tried it out yet.

The mechanisms for cross-over distortion remain the same for different amps, but it is unknown as to why it sounds fine in one 18W and terrible in another. They all have cross-over distortion when highly overdriven, but the bad sounding ones have the extra spikes on the edges.

[gary]

After trying different grid resistor values, zobel filter, and higher value el84 cathode resistor bypass cap, the only way I removed the 'vvr dialled down fizzies' was to disconnect the el84 cathode resistor bypass cap altogether. The caveat here is that my amp is permanently run at low vvr settings, so I would imagine that at 'normal' volumes there might be a little gain lacking in the power amp. Perhaps you'd like to try an experiment I haven't got around to yet - making the bypass cap switchable for this very purpose.

[Structo]

OK, I don't know anything about VVR or powerscaling as I haven't tried any of it yet.

But I have had the fizzies and tried various remedies.

A while back Andy La Blanc wrote an article where he experimented with different things and what he came up with was that the screen voltages were too high when compared to the plates.

Now I don't know if this applies to EL84's or not but it's worth a try and no semiconductors are used.

[John_P_WI]

Lindz,

Do you have a master volume or other drive compensation before the power tubes? IF you are NOT scaling the whole amp, as the voltage to the output tubes is lowered through vvr, the tubes can not handle as large of a signal from the PI and will distort very easily.

A poor analogy is this, think of a window that is open, with a sine wave that barely fits through the window, as you close the window the wave peaks can not fit through the opening any more and are clipped off. A drive compensation or master will make the signal smaller to fit through the "opening"....

My second guess would be a grounding issue, wiring issue etc. Are all of the wires twisted that need to be?

Voltage clamps can be fairly easy to implement, and usually consist of a mosfet, a few zeners, resistors and a transistor to help current limit. Also IF you are not scaling the whole amp, as the power amp is dialed down - unloading the power supply - the preamp voltages will increase changing the character and feel of the amp.

I have not used the ruby mod before, I thought it was for cathode biased amps??? I believe that you mentioned that this is fixed bias?

Good luck.

[Colossal]

John, that's actually a cool analogy.

I've followed the 18W Fizz issue over the years. As Katopan said, some people get it, some don't. I would be very interested in a survey of 18W builders to find out what transformers they used as well as tubes, speakers, VVR or no VVR, power amp scaled or whole amp scaled or stock, PPI master volume, etc. Some people seem to build these amps with no issues, others battle the dreaded crossover notch to no end.

[xtian]

Colossal, here's my survey response, FWIW. 18 watt lite IIb. No fizz, hum, nor hiss on this amp. Very pleased with result. Build thread: http://ampgarage.com/forum/viewtopic.php?t=17964

Heyboer PT, Hammond OT (a bit under spec'd). Rolled tubes, but not much change, ended up with Packard Bell labeled EL84s, Sylvania and RCA 12ax7s. No MV, no power scaling.

[Colossal]

Colossal, here's my survey response, FWIW. 18 watt lite IIb. No fizz, hum, nor hiss on this amp. Very pleased with result. Build thread: http://ampgarage.com/forum/viewtopic.php?t=17964

Heyboer PT, Hammond OT (a bit under spec'd). Rolled tubes, but not much change, ended up with Packard Bell labeled EL84s, Sylvania and RCA 12ax7s. No MV, no power scaling.

Very cool man. I did follow the build thread; nicely done. Glad to hear it turned out well and did not fizz. As I said, I would love to see a true survey done which might reveal some common thread amongst those who suffer the dreaded fizz.

[John_P_WI]

I have not played too much with EL84's, as I prefer the bigger bottle tubes but if I do recall the design guides are for the screens to be 30 v or so less than the plates. Tom - Structo - may be on to something here. IIRC, KOC - London power offers a powerscaling kit just for the EL84's addressing this issue.

[Lindz]

John - I am scaling the whole amp so there is no drive compensation per se though I do have a pre-pi mv on the amp (that is usually run wide open or nearly wide open)

I won't say outright that it is not a wiring issue etc. but I literally redid the whole amp 5 or 6 times in an effort to kill the fizzies, and it does not seem to matter how I lay it out (though I may repeat some mistakes of course). In fact I have a 2nd similar combo build using essentially the same design but cathode biased/power scaled that does exactly the same thing without the Ruby mod and a diode clamp across the cathode resistor.

Gary - interesting call, though this amp is fixed bias. I will try it for sure on my other build

xtian - did you stay reasonably close to the lite IIb schem? Your thread indicated 330vdc - I'm thinking I might try changing some values to known non fizzy builds just for kicks

Structo - I'll boot up my amp tonight and check plate vs screen voltage. sounds like it might be a possibility for sure

[xtian]

xtian - did you stay reasonably close to the lite IIb schem? Your thread indicated 330vdc - I'm thinking I might try changing some values to known non fizzy builds just for kicks

Reasonably close. But this was an available-parts build, so I made some subs. Notably the filter caps are 47/47/16 instead of 33/33/16, the power cath resistor is 120 instead of 150, and the V1 plate resistor is 150K instead of 120K.

[Firestorm]

From the clip and your description, the fizz is there as the notes decay, by which time the power amp is certainly back below the "Class B transition" (to use O'Connor's terminology), hence it can't properly be termed cross-over distortion since no tube is driven to cutoff.

Ordinarily I'd say look for parasitic coupling in the preamp (wiring of the tone stack especially) or at the output tube grids: those leads ahould be short, twisted or shielded. If the problem actually is in the PI, you should be able to tame it with largish grip stops on the output tube grids.

But it's interesting that the issue arises with scaling and seems curable by eliminating the cathode bypass cap. When you scale the amp, less current flows everywhere (except the mosfet itself), so logically the screen supply will drop less voltage and the screens will approach (or even exceed) the plate voltages. And with a cathode bypass cap in place, you remove local feedback for AC signals above the break point. So Andy's screen supply voltage divider trick may be just the ticket.

[Structo]

Structo - I'll boot up my amp tonight and check plate vs screen voltage. sounds like it might be a possibility for sure

If I recall you can measure from the plate (Red Probe) to the screen (Black Probe).
If it is positive then your plate voltage is higher than the screen.

As Andy puts it:

Measure the voltage difference between pins 3 and 4 at Idle
its more telling than referencing to ground.

He said the power amp gets fizzy when the screen voltage sit higher.

On my Dumble clone, the plates (6L6GC) and screens are generally the same voltage, so I was going to try this experiment but haven't gotten around to it.
All I have to drop the screen voltage is the choke after the main filters.

[katopan]

Tom - Thanks for posting Andy's article. Interesting read.

[katopan]

Another dumb question just to make sure - you did get the VVR3 for fixed bias amps that tracks the bias along with the B+ reduction? Using the one for cathode biased amps would get very fizzy as you turned it down.