Bottom Line: Tube Dissipation and its effects on tone

[Lydian]

Hello again comrades,

Last night I played a show, wound up talking to a guy who played before us. He had some cool tone and turns out he's been making amps for about 6 years. I took a look under the hood and saw some JJ 6l6gc's. From there we jumped into class AB and fixed bias talk until we ended up on power tube dissipation.

I commented that I have some 6l6gc GE clear tops running at about 12 and 14 watts dissipation, far below the 70% threshold commonly referred to. I was telling him I like to run my tubes low, a High Bias?, because it creates an enjoyable distortion and less headroom. This is an AA165 scheme Pro Reverb. This is where he tells me I have it backwards. He continued that the better distortion and less headroom is at higher dissipation levels, 18-20 watts for example or closer to 70% and that where I have my bias is producing more clean tones.

Correct anything that is incorrect with either my or his thoughts on this please. My own ears have shown me that the lower dissipation creates more gain/ or a richer harmonic content interpreted as distortion. The guitar amp handbook by Dave Hunter backs this up as well. However, all over the internet people are confused on what a high bias is, and surmise that high dissipation creates a shorter life span (speculatively true) and higher gain/less headroom(i'm saying false).

The higher the bias, negative volts to the plate, the colder the tube and equivocally lower dissipation. Low bias equals hotter tubes, and hotter tubes create cleaner and potentially firmer tones.

I'm sure I may have misused certain amp terms feel free to slap me in the face. As long as we get some facts straight I'm all ears.

Cheers amigos!

[Justin Thomas]

The few times I've tried twiddling my bias controls so radically, I find that running my tubes at high dissipation results in more hum, a fatter sound, less clean at a particular volume setting. Great for blues. Running them at lower dissipation was less hum, cleaner/thinner - sparkly? But also allowed me to crank the amp on 10 and not kill the tubes. Also seems more pedal-friendly...

When you crank either on 10, they both distort nicely, depending on what you're after - blues, or metal? Personally, I wonder if the eq of the tone stack doesn't have more to do with it. I play Fenders, so they all play fat farty blues on 10 no matter how hot I bias them.

I don't follow an internet "rule." I bias where they sound good for what I do, without destroying a good tube in t h e process. I hate to tell folks, but tubes won't be getting any cheaper in the future. Usually ends up somewhere in 50-60% range; Also, I don't want to run so hot that if something in the tube or circuit "self-tweaks" a little, then one tube is pushed over the edge and self-immolates... a slightly cool amp is a whole lot better than a dead amp.

Justin

[Lydian]

As a note, I'm almost always at 6.5 -10 on the volume knob, commonly attenuated. Thx Justin

[R.G.]

How much theory about this do you want to know?

[Lydian]

R.G.,

As deep a theory as I can understand. There is no way for you to know my exact whereabout in this field, but if you know of any good articles. I would love to read them.

Thanks

[Leo_Gnardo]

Higher dissipation = more gain*. Go in the other direction far enough and you'll have flea fart tone. For me the "sweet spot" is enough bias current to wipe out the crossover notch, then add maybe 10% more. Depending on the tube & amp this can be a surprisingly low percentage of recommended max plate dissipation, say 50% or so. Other tubes seem to need a chunk more maybe 85%. I use a scope & load resistors to set bias, then check again by ear, rarely need to readjust.

High bias, low bias ? ? ? Are we talking voltage, or current? One moves opposite to the other so there's often a misunderstanding right from the start. Make sure which parameter is being described as high or low. Some folks who don't have the knowledge, or equipment, or experience set bias according to the spec or schematic. "All Fenders with 6L6's set bias at -47 volts," like that. Techs who know better measure bias current, and who cares what the bias voltage is, as long as it's not miles away from what's expected.

Amps set with low bias current on purpose tend to sound dry & distort at all levels; the crossover notch does this for you. OK, you can make output tubes last a bit longer this way but it's not much of a tradeoff unless you're never going to use your amp for clean tones, or don't care that your "clean" sounds like it has a shovelful of gravel in it.

Bound to be a good subject for a thread, now here's the punchline indicated by the asterisk above:

* "Sounded GREAT, just before it blew up!" attributed to Joe Walsh, playing thru an amp that likely had an output tube meltdown.

[eschertron]

What Leo said. I'm a relative newcomer to tube amps; from my understanding "higher bias" refers to higher bias current and resultant higher dissipation. So if you like like biasing your power tubes with a relatively low negative bias voltage, you are actually biasing high from a bias current and dissipation standpoint. That's the way I think of it.

... and yes, most guitarists really like that warm creamy sound they get right before the tubes turn from red to brown

[The Dude]

The only addition I have as it relates to the original post is that bias has little effect on headroom/output. We're adjusting IDLE current not some magic pot that makes the amp put out more power.

[Enzo]

You have been happy with cooler bias. The other guy is happy with hotter bias. There is little to be gained by rationalizing which way must be "better". All that matters is that YOU like the results. This steak is too salty. No, this steak needs more salt. There is no right and wrong as long as the amp works and doesn;t damage itself.

We can do thought experiments and theorize all day, and maybe we still will. But ultimately, you have the amp. Turn the bias up (or down) to the 70% level, and see for yourself if it does anything positive. If you like it better, then stick with the new setting. it will have shorter tube life, but that is the price you pay for hotter tubes. it isn;t like the tube life will go from two years to ten minutes. if we lop a month off the tube life for better tone, who cares? if there is no tone advantage to your ears, then why burn up the tubes?

it also matters greatly how you use the amp. Picking power tubes for earlier breakup is fine if you plan to push the amp into power tube distortion. But if you mainly play at polite levels, then power tube breakup just doesn't happen and so is irrelevant.

[R.G.]

Some theory. This will sound long and winding, but it's mostly necessary to get to the point. Sorry - skip the parts that you already know backwards and forwards.

A single power tube has a fixed range of voltages and currents it can control. There is some maximum voltage over which it will arc over, and some maximum current over which it is damaged. There is also a power limit. Heat generated inside the tube is the product of the instantaneous voltage and current. A tube can only get rid of so much heat on average; it can exceed that amount of heat for a short time, but if the average over time exceeds how much it can get rid of, it overheats inside and dies.

Edit: As PDF64 correctly notes:

Hi RG, might it be better to put 'For max audio power output' at the beginning of the above section of your excellent post?
I know most guitar amps seem to be aiming for that anyway, but it may be beneficial to point out that it's not a given thing.
Plus there are some amps that run things a little more conservatively, eg AC4.

Bear this in mind as you read this. I'm focused on the maximum power available, as most amps seem to be chasing.

Given that a single tube has limits, it can only produce a limited amount of AC signal power from its plate. Exactly how much it can produce depends on how it's biased and what the signal voltage causes it to dissipate; those are different.

Bias, in the electronic sense, is an offset in one direction or the other. Things can be biased on, or biased off. The tubes in guitar amps are all normally on, and will conduct the maximum possible current if not biased in the off direction by a grid voltage. This is the source of much nonsense and confusion. A tube with zero voltage from grid to cathode will conduct a current limited primarily by the external power supply and resistances, not the innards of the tube. This very high current must be cut back by an external negative grid-to-cathode voltage. The word "bias" has been sloppily used for both the current and the grid-cathode voltage to get to that current.

Every type of tube (e.g. 6L6 or EL34) has a different bias voltage to get to a certain current; they have different sensitivities to the grid off-bias voltage. Also, every single tube within a tube type (e.g. different 6L6s compared to each other) has a slightly different sensitivity to grid bias. One 6L6 may need -35V to get to some plate current, the next may need -37V; this is a rich source of tube suppliers charging you more money for "matched" tubes. But in general, any 6L6 will be closer to another 6L6 than any 6L6 to any EL34, and so on.

Tubes can't drive low impedance speakers directly, so they need output transformers. This forces the designer to make some choices of tube bias voltage/current and operating point. It's easy for the RF guys. They can bias the tube completely off (high voltage on the plate, no current flowing) and pulse it suddenly fully on (high current, low voltage on the plate) and back off at the desired frequency, which does not vary much. That lets the tubes have the lowest possible dissipation, being either fully off, or maximally on, not spending much time in the middle where the product of voltage times current is a maximum. It makes for a really nasty, distorted signal, but the RF guys can use LC filters to clean that up into the desired signal.

Audio guys can't do that. We have to have our sines pure, and that means not turning the tube fully on or fully off. Guitarists are a sub-class of that; we like a little bit of distortion, but (trust me on this) not fully on or fully off, at least not for much of the signal. And tube amp designers (not build-a-kit tinkerers) will bias a single tube so it is (mostly) not fully on or fully off.

Hold on, we're getting there.

A single tube then has an output transformer, and since it has to not turn fully on or fully off for very long, must be biased right in the middle of its current range, half it's maximum current. That lets it swing half its max current upwards (to full max current) and half its maximum current downwards (to fully off, zero current). That also means it has to swing from (nearly) zero voltage across it at max current peaks to fully max voltage (on full current cutoff). Since the transformer in its plate circuit will not support any substantial DC voltage across it, only AC voltages, then the plate voltage on the tube at idle must be (about...) half the tube's maximum voltage. The transformer's stored energy will provide the other half of the maximum plate voltage as the tube turns current off and the transformer, by inductor action, causes the extra voltage swing.

So for a given single tube, the tube's limits of voltage, current, and power and the use of an output transformer force it to be biased at half it's maximum plate voltage before arcing (or less!) and half its maximum plate current (or less!). If it's not, you get either gross distortion or less undistorted power. You also have to make sure that the idle power - the product of idle voltage and idle current - is less than the tube maximum power, otherwise the tube will overheat and die.

And as I've noted, guitarists are OK with a little distortion, much less than the RF guys, but they're picky about what kinds of distortion and how much. Design for lowest distortion and then tinker is what has been successful for guitar amp design.

What I've described is a single ended Class A bias tube setup. This is limited in power output because of the tube constraints, which is why I went through all that "a tube has a maximum..." this and that. If you want more power than a single tube can provide, a few watts of audio, then you have to use more tubes. You can go parallel or series; or parallel AND series with more tubes.

This is all the prep work to get to biasing classes and power dissipation issues. Sorry for the baby steps, I had to set a level starting place.

[R.G.]

With that as background, we can talk about audio output power, biasing, biasing classes, and power dissipation.

A single tube must be biased to be on all the time or suffer gross distortion. You can set it to be anywhere inside its maximum power envelope of voltage, current, and power limits, but there is a single point of voltage, current, power (and loading through the OT) that produces maximum AC power out.

RF uses can use Class C biasing; the tube is biased off, and banged on and off by a massive-overdrive signal. This produces pulses, not linear signal, on the plate, and the output networks filter out the hash. Class C is defined by maximum grid off-bias and zero plate current as an average condition. We can't use this one in guitar amps.

A single tube biased at roughly half its maximum plate current and voltage will not have gross distortion like Class C, and will be limited to a few watts for most of our tubes. It will dissipate its maximum dissipation all the time, and will, in the internet dissipation parlance, be biased at 100%. It will be HOT, enough to blister fingers, and will eat about 75% of the DC power supply, or much more, just to generate those few watts. You have an electric space heater that produces a little audio as a byproduct. This is Class A bias.

For more audio output, you can parallel more tubes, changing the OT design to get the extra audio potential out. This makes a bigger electrical space heater, and a few watts more audio. You can do a little better by setting up two tubes in Class A push-pull, where one tube is driven on while the other is driven off. This doubles the (abysmal) efficiency of Class A single ended. Still, it's nothing to write home about.

Long ago, someone figured out that if you could have one tube spit out just the positive half-cycles of signal and another tube spit out the negative half cycles, using a center tapped output transformer, you could get more audio out with less dissipation. Each tube is biased exactly at zero plate current, so its plate circuit dissipates zero in the absence of signal. The heaters and screen still heat it up, but this is a minor fraction of the tube's full dissipation. This is Class B, push pull. It's a variant of Class A push pull where the tubes are just barely turned off at idle.

Class B gives you not twice the theoretical power of a pair of Class A tubes, but four or more times the audio power out of the same tubes, depending on how well or poorly the Class A and Class B circuits are set up. It has the side effect of using the DC power supply much more efficiently, as less of the DC is wasted as heat and more goes out the output jack as audio. It's how you get tube amps with more power than about 10W, except for really strange, room-heater designs. Using 1kW to get 10W of audio is not all that much fun.

Class B is efficient, but it introduces crossover distortion. It is not easy - impossible, in fact - to get one tube to turn off **exactly** as its partner is turning on. There is some inevitable glitch there, and it turns out that the human ear does not like crossover distortion in general.

And now we get to Class AB. Here's how this went: Class A single ended - good sound, but low power and lots of heat. Class A push-pull - good sound, a little more power for the same heat. Class B - lots more power, but crossover distortion.

If you think about it, the operating class dictated how big the power supply had to be. Class A single ended needed a power supply voltage at most half the peak plate voltage and half the peak plate current, plus the peak screen current, and the rest of the amp, which is trivial compared to the plate dissipation. Class A push-pull demanded the same voltage and the full plate current, as there was always one tube's equivalent on. In both cases, the power supply voltage had to be kept to some lower voltage (that "at most" thing) to keep the tubes alive.

In Class A push-pull, the physics works out that the power supply voltage has to be lower than the tubes could withstand to keep them from melting. Significantly lower. The tubes are biased at 100% of dissipation for this lower power supply voltage. Tube dissipation limits Class A and Class A push-pull.

In Class B, each tube dissipates only during its conduction time, and each tube conducts for only half the time. So its internal heating is zero half the time, and the same as a class A amp the other half time. The math involved means that the cool-off times make this more efficient than half a Class A power dissipation. In Class A, the tubes dissipate 100% all the time - they don't heat up - at least any more than they would otherwise! - with big signals, and they don't cool off with zero signal. Not so with Class B. In Class B, the tubes cool off with zero signal, down to some minimum dissipation. They heat up with larger signals, until some signal level makes them heat the maximum that the DC power supply voltage and matched audio load let them heat.

To an amp designer that understands amps and gets to design the power stage and the DC power supply at the same time, this means that they pick the power tubes (and hence the power tube dissipation limits), calculate the peak voltage, current, and power swings the tube can reliably stand, and then design the power supply to match that. That's how Class B gets to four or more times the same tube's Class A limits - it gets a custom and much higher DC power supply voltage because it's designed so that even at max audio swings, it just touches max tube power dissipation. Class B lets you use a bigger DC voltage plate supply for the same tube heating.

To the first Class B experimenter, this had to come out as "YAY! More power!" followed pretty soon by "... but that sounds gravely and ugly. What the?" when they noticed the crossover distortion. They cleverly realized that all they had to do was to back off on the off-bias a little, making the tubes not just toss the audio signal over the zero-voltage fence (so to speak ) but to overlap a little. This worked so well that they kept increasing the tube zero signal idle current by backing down ( i.e. less negative) on the grid bias voltage little by little until the crossover distortion became unnoticeable. AHah! Success!

Then the test rig overheated and smoked. They had carefully ratcheted up the power supply voltage as much as possible to let them get the maximum Class B power , and then increased the overlap conduction of the tubes back into the high-dissipation Class A region little by little, raising the power generated in the tubes more, with the now much-higher DC power supply until the smoke came out.

"No problem," they said. "we'll just back that DC power supply voltage back down until the tubes survive with the amount of Class AB bias" (as they quickly named the biasing scheme) " and have a hugely better amp. Much more power, less than pure Class B, and nearly as low distortion as Class A." And they did, and the world as we now know it happened.

And all was well with the world until today, when people started tinkering their own amps' biasing. What people don't know, because they haven't lived through the design-from-zero analysis of a Class AB power stage, is that Class AB biasing is a modestly delicate dance, interacting the maximum power out, the crossover distortion received, AND the power supply voltage to get to a certain point.

And now we're at the point of this whole mess of typing. Biasing to X% of maximum tube dissipation doesn't mean anything, at least without knowing how the AC loading, DC power supply voltage, signal levels, and tube type/dissipation interact. It amounts to a rule of thumb that bets that if you set things to 70% of the max tube dissipation, the original output stage design (which contains things you don't know) and the original power supply design (ditto) and how the original biasing setup was designed (ditto) and your "typical" signals will not scorch your tubes.

As you recognize, this is at least a guessing game, where you're playing "You Bet Your Tubes".

Sure, it's probably fine as a rough rule of thumb, but it doesn't mean much of anything other than that, a starting point maybe. Certainly not the electronic correctness principle I've seen it quoted as. 70% of dissipation may be OK, may not. And it depends on the power supply - and the cabinet ventilation! - as to whether it's safe for the tubes, or turns out an acceptable degree of crossover distortion.

[Jazz P Bass]

I had a Jazz player bring me a Fender HR Deluxe. (say what?).
Only used on the Clean channel.

He wanted the amp biased 'hotter'.

'Why?'.
'I read that it will make it sound better'.
'Better than what?'
'Just better'.

So I did.

He brought it in the next day.
"Please put it back to where it was."

[pdf64]

I suspect that the key aspect being adjusted by the bias control is the conduction angle.
The plate dissipation at idle required to achieve a reasonable conduction angle may be immaterial.
ie with a class A amp, if under certain operating conditions, 360 degree conduction occurs with say 80% plate dissipation, there's no theoretical benefit running the plate/s hotter; but might a 'deeper' class A, idling at 100%, sound any different? If so, why?

The problem with conduction angle is that it seem hard to assess (anyone got a method?); and as most designs push their power tubes to/beyond their limits, setting operation to a suitable idle plate dissipation may be the best solution in the real world.

[R.G.]

Depending on how an amp's output stage and power supply interact with its audio loading, you get different kinds of distortion.

Single ended output stages produce my favorite kind of output stage distortion. It's big, soft, and tasty. My favorite output stage for tonal considerations is a single-ended 6BQ5/EL84 or a 6AQ5 running about 250-300V on the plates. Very nice indeed. It's unlikely that you're hearing real output stage distortion on any other kind of amp.

Pushpull amps vary. P-P Class A can distort if you can drive it hard enough. It's a pure high-current squashing of peaks, and sounds good for symmetrical distortion. There aren't many amps like this.

The Vox AC30 is not a Class A amp as it's normally set up. . It's deep class AB, nearly to A. At some point nearly to output peaks, one tube cuts off. There is a little "crossover" as one tube cuts off entirely, but not nearly as bad as real Class B. I suspect, but have never seen any real clues about this: I think they wanted to make a 30W Class A amp and couldn't. I think they tinkered with bias and got up to 30W by moving it a little out of Class A and increasing the B+ voltage. Just my suspicion. It's the kind of game a tube amp designer that understands the design could make.

The AC30 is cathode biased. Cathode bias does not mean "Class A". Nor does fixed bias necessarly mean Class B or Class AB.

Class A amps and nearly-Class-A Class AB amps don't have noticeable crossover distortion. The sensitive spot for human hearing crossover is down around the zero crossing.

At a fixed power supply voltage, a pure Class A amp will only have limiting distortion - a big enough signal on the grids will drive it to maximum available current when the grid just hits the cathode voltage, and the grid suddenly starts sucking current. This generally prevents the PI/driver from driving it more positive, so the tube stops increasing plate current, and the plate output voltage change limits here. For a properly biased Class A amp, this gives equal + and - peak limiting distortion. And that's all the undistorted audio power you can get out of it at that power supply voltage. If and only if the plate AC load is set up so that this maximum current and peak plate voltage are just so, you get the maximum audio power out; the load is matched.

If you increase the negative grid bias, turning the idle plate current down, the tubes cool off a bit with no signal. So you can increase the B+ voltage a bit, and get more power out of the same tubes without them melting.

Increase the negative bias a bit more, you can then raise the B+ voltage some, and get more power out without melting the tubes.

There is a continuum of bias points and plate supply voltages from Class A and always melting down, down to nearly fully-off biased tubes and only hitting peak dissipation when driven to full power. The thing is, full power at Class B can be much larger for the same tubes because you can use much higher plate voltage, a different loading ratio, and the tubes actually run cooler at any power output level below full power.

At each notch you decrease idle plate current by increasing negative grid bias, the points in the tube conduction where it goes from both tubes conducting to one tube conducting on opposing half cycles moves toward the zero crossing. The little glitches from one tube turning off get closer to 0V, more into where the ear is sensitive, and more noticeable. And down near pure Class B, they are clearly visible as "crossover notches", little glitches in the output signal that don't get covered up by feedback very well.

If you are playing with an already designed amp, you probably don't have the luxury of changing plate supply voltage to optimize output power as you tinker bias. You're almost certainly working on an amp designed for operation significantly into Class AB, well away from crossover distortion and also reasonably away from tube meltdown, while getting what the design engineer's boss thought was "enough power".

As you increase idle plate current by decreasing how negative the grid bias voltages are, you move the tube deeper towards Class A. This moves the crossover notches away from 0V, and makes it sound subjectively cleaner. Your PI can put out some amount of signal to the outputs, and since rebiasing didn't change how big this signal is, the tube can now reach high-current limiting and positive grid-cathode voltage relatively sooner on the signal peaks fed to it from the PI. So it "clips sooner", relative to the volume knob setting. It also gets much hotter, as you cannot easily back down the plate voltage so that you back down the idle dissipation.

If you go the other way, the amp gets cooler - there is actually less heat generated inside the box! - and sounds "colder" as you're further away from the plate current and grid voltage limiting at peaks. It probably sounds "colder" or more "sterile" in a bad way because the crossover glitches get more prominent as they move toward the zero crossing. And you can't easily take advantage of the extra power output you would have if you increased the plate voltage to take advantage of the bigger range of dissipation on the tubes.

As you get closer to Class B, the crossover glitches get near your ear's sensitive zone, making the sound "grainy" or harsh. Down really close to Class B, the crossovers are visible on a scope, sound scratchy or gravely, and sometimes mimic a rubbing speaker they're so bad.

So that's the big picture of bias versus tone and distortion. Thinking in terms of only bias at a fixed power supply voltage and tube dissipation means you're really limited in what you can do. You can't take much advantage of the "bigger headroom" that colder biasing could theoretically give you as you accept more crossover distortion, and your tubes may not be able to stand the higher dissipation as you let therm run hotter.

They probably DO sound good just before they burn out but it's really hard to get them to stay there.

[Jazz P Bass]

If the bias concept is to simply get out of Class B & into Class AB (thus minimizing crossover distortion) where did the 70% thingy come from?

And if you want to get picky about it, wouldnt a scope be required to actually see the crossover distortion?

I would imagine that you could accomplish the fact by listening as the bias is slowly raised.

While there very well may be sonic differences when the bias is raised too high, I don't get why you would want to do that.
(from a tech stand point)

Then again, tinkering with tube amps is kind of like the what the hotrod dudes did with there cars back in the '50s.