RMS output differences in all-tube amps

[agedhorse]

A UL amp will still distort if over driven just that the distortion will be more controllable.

Actually, it won't be any more controllable, but the transition into and out of clipping will be more abrupt. This is one drawback to feedback, worse with global feedback than local feedback, because the system becomes open loop when the feedback can no longer follow the input signal it's trying to correct for.

IMO, this is a major factor in amp design when intentionally overdriving an amplifier.

 

[SherpaKahn]

Note that UL is not global feedback, it's quite local in fact (though it does derive it's source from the transformer primary taps). If the signal source for the feedback came from the secondary, a better argument could be made for "global feedback" but that term is usually reserved for the feedback around more than a single amplifier stage.

... Yeah...I meant to refer to a global feedback loop wrapped around a pentode power amp, which I employed in my build. Also referenced generally to explain dirty tricks used to linearize pentode amplifiers

A UL amp will still distort if over driven just that the distortion will be more controllable.

Actually, it won't be any more controllable, but the transition into and out of clipping will be more abrupt. This is one drawback to feedback, worse with global feedback than local feedback, because the system becomes open loop when the feedback can no longer follow the input signal it's trying to correct for.

IMO, this is a major factor in amp design when intentionally overdriving an amplifier.

*THIS*. That moment the feedback loop collapses, and linearity goes out the window. Heavenly for guitarists with booster pedals, terrible for thumping bass players.

I estimate that (when seeking power tube distortion), most bass players appreciate a wide range of increasing coloration/texture accessible by a combination of volume knob, playing technique, and perhaps a boost pedal.

Speaking for myself, this sounds nice: With volume knobs dimed, increasing signal proportionately decreasing fidelity. Clean full-power with instrument volume reduced. And all of this available in the range of 90%-120% rated power output. Below that, coloration and envelope compression varying with signal intensity.

...Given an amp with lots of gNFB, if its clean output power is measured @100W, but the PA starts clipping at 110W, it's unlikely that a player would have a dynamic range of coloration/texture to play with....They also might not be getting their "tube watts" worth, unless their band is named "Heinous Overload" or something similar...

...In theory, I value global feedback mostly for the improved speaker damping & improved bandwidth (now the room shakes instead of the cabinet)

Tying this back into RMS ratings: I believe I found a graph which illustrates the general shape of the phenomenon: output power vs. distortion. This compares **local** feedback around the power tube/transformer, but the shape shows the trendline, which I think would be similar for a gNFB comparison. This chart is for a local-ish cathode feedback implementation. Regular global feedback loops should collapse even more gracelessly, I expect.

Feedback results in less distortion below 11W, but the increase in distortion is much more abrupt once you're well into dirty territory. In this example, there's still a little dynamic fuzz range @44Hz between 11W-12W without feedback. With feedback, distortion sets in hard. For a determined bassist armed with a boost pedal, that area from 7-12W could yield interesting sounds. Feedback adds that kink followed by rapidly increasing distortion.

TL;DR:
If designed with lots of global feedback to achieve linearity, and an amp's clean power rating is taken shortly before that kink, you're not likely getting your "tube watts" worth

 

[agedhorse]

As a generality, as the amount of global feedback increases, the shape of the distort curve becomes steeper and the angle between linear and non-linear regions becomes closer to 90 degrees.

There are tricks that designers use to alter the shape of the distortion curve even in the face of high global feedback. There is an entire class of circuitry centered around saturation management.

 

[Mr. Foxen]

Anyone played the proper Sunn Model Ts, fairly sure those are ultralinear (not the reissues).

I kinda feel the UL Twins are what gave UL a bad rep. And it is possibly the magic in big old amps that people can't get from new stuff, not really done in modern amps.

 

[BogeyBass]

UL Twins a bad rep...nah
Well Kinda.

Beam Tetrodes already breakup very late in the game. UL amps even later.
If you want super clean UL is great
If you want distortion UL not so much

The Fender UL designs are picture perfect for anyone trying to blast the living crap out of 6L6GC to squeeze more power out of them. 500 volts plate is a good limit for 6L6. Problem is screens die and running the screens on a UL tap will help them survive. Many manufacturers have pushed 6L6 to 540/560 volts and the plates will survive. But the screens will die quickly . Even at 450 volts.

There is 600 volt sweep tubes and very common radio frequency Russian military tubes very similar to the 6L6 but they use different pin outs and a top mounted plate pin. Many Chinese and Russian tubes in the 90s used the same sweep tube plate design and shoved them in smaller bottle 6L6 cases and pinouts. Notorious indestructible 6L6 tubes do 600 volts no problem. But the small bottles don't offer enough cooling for the screens. 8000 to 10,000 hour radio tubes get reduced to 3000 to 5000 hour tubes in smaller bottles

Same thing with 6550 that most haven't learned yet. The screens are the weak point. Data sheet list max screen at 400 to 450 volts depending on what ratings race data sheet you look at.

In reality it's well noted that 6550 survives with only 300 volts max on the screens.

If you want to successfully run 6550s screens at data sheet 400 volts. A UL tap will increase that likelihood.

Or larger bottle KT88s will help the screens survive longer. 6550 could probably do 700 volts no prob but the screens will pop in the smaller bottles.

Likewise blasting the hell out of a KT88
Quad with high voltages can achieve 250 /280 watts no problem. Same thing the heat will kill the screens much quicker

 

[Wasnex]

Same thing with 6550 that most haven't learned yet. The screens are the weak point. Data sheet list max screen at 400 to 450 volts depending on what ratings race data sheet you look at.

In reality it's well noted that 6550 survives with only 300 volts max on the screens.

So a 100W Marshall running 6550s is still stressing the screens even though the tubes are biased really cold.

For context my 2203 runs 460V on the plates and screens run a few volts lower. I set the amp up per Metropoulos recommendation, treating the 6550s as 30W tubes. Using plate voltage and the 30W rating, I set the cathode current to the 70% plate dissipation figure (45.7mA).

I was under the impression this is a very conservative setup for the tube, but a slightly aggressive setup for the OT, since the amp would normally be biased for EL34s at a 25W rating (70%=38mA).

 

[SherpaKahn]

So a 100W Marshall running 6550s is still stressing the screens even though the tubes are biased really cold.

For context my 2203 runs 460V on the plates and screens run a few volts lower. I set the amp up per Metropoulos recommendation, treating the 6550s as 30W tubes. Using plate voltage and the 30W rating, I set the cathode current to the 70% plate dissipation figure (45.7mA).

I was under the impression this is a very conservative setup for the tube, but a slightly aggressive setup for the OT, since the amp would normally be biased for EL34s at a 25W rating (70%=38mA).

Unless the B+ is well-regulated (not common in MI amps), power supply can be expected to "sag" during heavy output. [edit: for push-pull power stages operating in AB1/AB2. More so as you go "deeper" into class-B, and even more so for AB2]
"Sliding screen" operation is a variation of power-supply sag. Unless care is taken to regulate the screen grid supply, it is likely to sag more heavily than the plate supply. The measured idle screen voltage with no signal is only half the story.

The average (dynamic) screen current will determine the average screen voltage under signal-producting conditions. Page 562 of the RDH4 provides an approximation of average screen current as (screen_supply_voltage)*[(1/4)(max_current_g2)+(1/2)(idle_current_g2)]

An amp with idle plate/screen @ 400V/400V may actually reside at 375V/300V under dynamic conditions. Sometimes things just fall apart right when needed.

This concludes the weekly sermon of "When doing the wrong thing is the right thing"

From this fun thread at AX84: Marshall 100W loadline - yikes!

>>Definitely need to measure PS rails loaded at expected W level (remembering that the DC load to apply is much more than the 'output watts'). Then do the class B line from the sagged level - certainly changes the outlook on many amps!

>>Yes, the original problem came up because the poster's amp only sags 30V, from 480V to 450V at full clip, hence the redplating. The loadline doesn't change enough to get him out of the danger zone.

Apparently some original plexis also sag 100V at full clip, from 480V down to 390V. That's not very good regulation.

RA

>> I was told by my tech at the time, Don Butler, that my '66 JTM-100 Super PA head did in excess of 100 watts at the onset of clipping and did 213 watts at full tilt. This amp is one of the 'high-voltage' large mains PT's that was mis-spec'd by Ken Bran on the 110 v tap and actually delivers a higher voltage than was planned. This one also uses EL34's and not KT-66. Having not done the math nor made the measurements myself I cant say it for certain, but I can say it's a damn loud amplifier...

edit: we're talking about push-pull amps
P.S. that last guy's getting his "tube watts'" worth

 

[agedhorse]

The above comments apply to voltages, currents and resulting dissipations with regard to load lines, there still exists the challenge of excessive voltage causing insulation breakdown within the output transformer, the tube sockets and tubes themselves (assuming the caps are properly specified for the real voltages the circuit will see under use).

 

[beans-on-toast]

The subsequent heat of a pushed transformer can also be very hard on the insulation of a vintage unit. There are people that play so hard that the transformers literally smoke. Players like that don’t care what it does to the tubes or the amps, they want the sound that comes with that sort of abuse.

Then there’s the sagging line voltage that some venues have. It starts off low at idle and only dips more when the band is playing. That taxes the amps even more. All bets are off as to the voltages inside the amps at this point.

It would be great to connect a bunch of data loggers (in a non lab environment) and see what’s going on under these conditions. So long as the instruments don’t alter the state of what they are measuring.

 

[byacey]

Ironically, the ultralinear was the closest that tube technology could approach to the linearity of a good solid state power amp. The purpose was to minimize distortion, contrary to what many desire in MI amps.
Audiophiles in the heyday coveted ultralinear power amps.

 

[BogeyBass]

So a 100W Marshall running 6550s is still stressing the screens even though the tubes are biased really cold.

For context my 2203 runs 460V on the plates and screens run a few volts lower. I set the amp up per Metropoulos recommendation, treating the 6550s as 30W tubes. Using plate voltage and the 30W rating, I set the cathode current to the 70% plate dissipation figure (45.7mA).

I was under the impression this is a very conservative setup for the tube, but a slightly aggressive setup for the OT, since the amp would normally be biased for EL34s at a 25W rating (70%=38mA).

That is a conservative application
Your not pushing the tube to max dissipation. And generating more heat. And sounds like fun I've really been wanting to make a very conservative 6550 quad myself

When pushing 6550s for 80/100 watts per pair with 600 volt plates. Dramatically lowering screen voltage will help considerable.

If problems arise with the amp you could increase the screen resistor value. And try to get closer to the 400 volt max rating. I'll assume you won't have many problems. 400 volts screen is good value to use a big bottle KT66 for that Marshall. Likewise be tiny bit easier on the power transformer. 6550 heaters will draw up to 2 amps
Kt66 up to about 1. 4 amps

460 to 540 volts and conservative amounts of negative feedback. Dramatically improve tube life for 6550.
Likewise using simple inline plate resistors because oscillating is another known trait of the 6550. Start pounding them at higher voltages and the ballgame changes

 

[Wasnex]

That is a conservative application
Your not pushing the tube to max dissipation. And generating more heat. And sounds like fun I've really been wanting to make a very conservative 6550 quad myself

When pushing 6550s for 80/100 watts per pair with 600 volt plates. Dramatically lowering screen voltage will help considerable.

If problems arise with the amp you could increase the screen resistor value. And try to get closer to the 400 volt max rating. I'll assume you won't have many problems. 400 volts screen is good value to use a big bottle KT66 for that Marshall. Likewise be tiny bit easier on the power transformer. 6550 heaters will draw up to 2 amps
Kt66 up to about 1. 4 amps

460 to 540 volts and conservative amounts of negative feedback. Dramatically improve tube life for 6550.
Likewise using simple inline plate resistors because oscillating is another known trait of the 6550. Start pounding them at higher voltages and the ballgame changes

Pretty sure the amp came stock with 6550s. It's a 77 JMP 2203. Per the schematic, the screen resistors should be 1K 5W and the NFB should be connected to the 4 ohm transformer tap.

I haven't done any dynamic measurements, but I would expect the plate and screen voltage to drop down a bit when the amp is pushed hard. Changing the screen resistors to 1.5K seems to be a frequently recommended mod. I get the impression this is an iconic design that doesn't really look great on on paper, but there are thousands 100W Marshalls running 6550s just fine and plenty of them are being properly thrashed.

The GE6550s that came with mine were in pretty rough shape. I kept two and trashed two because the tube guts were not mechanically stable. They sounded fine, but bias would shift drastically if I moved the amp. As result, the amp had to be setup really cold and sounded super stiff. I threw in a pair of JJ KT88s that biased up within a few mA of the two remaining GEs and biased the hottest tube to 40mA.

 

[ThisBass]

I haven't done any dynamic measurements, but I would expect the plate and screen voltage to drop down a bit when the amp is pushed hard.

I'd estimate the voltage drop that happens under hard pushed condition as a very drastical drop.
Expressed in numbers already (roughly) 1/3 off if the power amp was pushed hard into saturation only under dynamic aspects of a given (real practice) signal.

 

[ThisBass]

Let me share with you a couple of pics which may help to better understand what is going on under the hood with loadlines under conditions of real practice.

The following pic shows the loadlines for a Fender Super Reverb that powers a resistive load at different numbers of output power resepectively was feeded with different input levels. Type of input signal is a sinusoidal signal at 3kHz.

For smallish output power (low level signals) the loadline still crosses the bias quiescent point (AP) and the loadline looks very close like a straight line.
It can be seen that for highish output levels the loadline is shifted to the left side at some content, thus the loadline does not cross the bias quiescent point (AP) anymore. It can also be seen that the loadline shows a noticeable winding at the lower end. Its obvious that at highish output power the amplifier will show lots of THD caused by push/pull crossover issues.


The next following pic shows the loadlines of the same amplifier for reactive load, which actually is a real speaker.

Rather then single lines as shown in the pic above for resistive load now the "loadlines" for a real speaker load do mutate into ellipses at low level output and into large "breaking" curves at high level output. The anode disspipation power now strongly excceeds limiting dissipation of 30 watt for the 6L6 but, as the transcendending of 30 watt is of partial characteristic (<180 degree) the average of anode dissipation very likely still stays within 6L6 limits even if the amp was pushed very hard.


Finally here you are with a pic which is meant to demonstrate "loadlines" in pratice IF the Fender Super Reverb was pushed in real practice with a Fender Stratocaster guitar.

Its quite hard to still identify "ordinary" loadlines. Instead of (well known) loadlines in real pratice the amplifier acts quite different from any (serious) theoretical considerations.

The small white circle demonstrates the quiescent current bias point. It can be seen that there is really no way to predict any (realistic) plate dissipation number by a distinct (adjusted) quiescent current IF it comes to consider real practice applications, in other words at least IF the amplifier was puhsed by playing a real guitar and also has to power a real speaker any theoretical plate dissipation considerations on paper become meaningless.

 

[agedhorse]

The above shows why design falls into rule of thumb with suitable estimates of derating made based on the understanding of the principles of what's happening rather than specific examples. Reasonable worst case estimates have to be made in practice, but they are based on theoretical knowledge.

 

[beans-on-toast]

All bets are off for when the amp is pushed hard. One shouldn’t design a musical instrument amp where all boundry excursions are kept tame if you want it to sound good. There’s desirable character in those distortions. How far an excursion to predict and find a good design balance is the hard part.

One thing for sure is that you can never predict the extent to which people will abuse any device. They’ll always find a way to surprise you.

 

[beans-on-toast]

That loadline looks like the Hindenburg going down. Enjoyable excursion that ends with lots of compression.

 

[agedhorse]

All bets are off for when the amp is pushed hard. One shouldn’t design a musical instrument amp where all boundry excursions are kept tame if you want it to sound good. There’s desirable character in those distortions. How far an excursion to predict and find a good design balance is the hard part.

One thing for sure is that you can never predict the extent to which people will abuse any device. They’ll always find a way to surprise you.

This is (or sure as hell should be) in fact a VERY important consideration for any high reliability design.

 

[SherpaKahn]

Let me share with you a couple of pics which may help to better understand what is going on under the hood with loadlines under conditions of real practice.

The following pic shows the loadlines for a Fender Super Reverb that powers a resistive load at different numbers of output power resepectively was feeded with different input levels. Type of input signal is a sinusoidal signal at 3kHz.
View attachment 3189904
For smallish output power (low level signals) the loadline still crosses the bias quiescent point (AP) and the loadline looks very close like a straight line.
It can be seen that for highish output levels the loadline is shifted to the left side at some content, thus the loadline does not cross the bias quiescent point (AP) anymore. It can also be seen that the loadline shows a noticeable winding at the lower end. Its obvious that at highish output power the amplifier will show lots of THD caused by push/pull crossover issues.


The next following pic shows the loadlines of the same amplifier for reactive load, which actually is a real speaker.
View attachment 3189905
Rather then single lines as shown in the pic above for resistive load now the "loadlines" for a real speaker load do mutate into ellipses at low level output and into large "breaking" curves at high level output. The anode disspipation power now strongly excceeds limiting dissipation of 30 watt for the 6L6 but, as the transcendending of 30 watt is of partial characteristic (<180 degree) the average of anode dissipation very likely still stays within 6L6 limits even if the amp was pushed very hard.


Finally here you are with a pic which is meant to demonstrate "loadlines" in pratice IF the Fender Super Reverb was pushed in real practice with a Fender Stratocaster guitar.
View attachment 3189906

Its quite hard to still identify "ordinary" loadlines. Instead of (well known) loadlines in real pratice the amplifier acts quite different from any (serious) theoretical considerations.

The small white circle demonstrates the quiescent current bias point. It can be seen that there is really no way to predict any (realistic) plate dissipation number by a distinct (adjusted) quiescent current IF it comes to consider real practice applications, in other words at least IF the amplifier was puhsed by playing a real guitar and also has to power a real speaker any theoretical plate dissipation considerations on paper become meaningless.

Yowza. https://www.americanradiohistory.com/Archive-Experimental Wireless/30s/Wireless-Engineer-1939-06.pdf