5E3 Voltages and PT Bias

[burgermeister]

I just finished my first amplifier build, a modified 5E3 clone.  It was a fun project, and it actually works, no mushroom cloud or anything!  Also sounds great, at least compared to the Crate G60 I have been using for the last 30 years.

I do have a few questions I cannot seem to find answers for.  I would really appreciate clues!  I am a mechanical engineer, so I apologize for the dumb questions in advance. 

Biasing the power tubes (JJ 6V6S):  Everything suggests to use 1/2 the OT impedance. Why???  The DC resistance is quite different, and for setting the DC operating point, it would *seem* (but obviously isn't) that the OT DC resistance would be the governing factor. The DC resistance is one method of measuring tube bias - if it measures the bias, why does it not set the operating point?

More on PT Bias:  I am using a Hammond PT, 660CT, and 5Y3 rectifier.  B1=376, B2=334, B3=267.  This seems high.  Voltage across the 247R PT Bias resistor is 21.3.  This ends up just above 100%+ dissipation for the 6V6S.  Yet, these would appear to be standard 5E3 clone values!  Is this normal?  Given the likely unavailability of tubes for the foreseeable future, I'd like to cool things down a bit.  Is the best way to address this a larger bias resistor, or a lower voltage PT?  I am leaning towards the latter... it would also cut a bit of power, and the thing is far too loud as it is!

I also added an option to switch to fixed bias.  This does allow me to bias the 6V6 tubes much more reasonably.  But I did notice that all the voltages in the amp increase by 50V when I switch to fixed bias.  Is this normal, or does in constitute a wiring error?  It seems the 50V bias wire from the PT, even as it passes through 70K of resistance and a cap, somehow changes the potential of GND (CT of the 660 winding) to -50V.  It doesn't seem like this setup could support much current, given the resistance between the 50V wire and the 660 CT.  So I am really confused.

And, last, the heater supply runs at 6.7V, rather than 6.3.  Will this affect durability?  Diodes will drop that to 6.0V, or I could get a couple 0.1R resistors to get me closer to 6.3.  What is the preferred method?

I'd be happy with any pointers on where to look for answers, as long as they don't involve differential equations :)

[tubeswell]

I just finished my first amplifier build, a modified 5E3 clone.  It was a fun project, and it actually works, no mushroom cloud or anything!  Also sounds great, at least compared to the Crate G60 I have been using for the last 30 years.

…

Biasing the power tubes (JJ 6V6S):  Everything suggests to use 1/2 the OT impedance. Why???  The DC resistance is quite different, and for setting the DC operating point, it would *seem* (but obviously isn't) that the OT DC resistance would be the governing factor. The DC resistance is one method of measuring tube bias - if it measures the bias, why does it not set the operating point?

More on PT Bias:  I am using a Hammond PT, 660CT, and 5Y3 rectifier.  B1=376, B2=334, B3=267.  This seems high.  Voltage across the 247R PT Bias resistor is 21.3.  This ends up just above 100%+ dissipation for the 6V6S.  Yet, these would appear to be standard 5E3 clone values!  Is this normal? …

I also added an option to switch to fixed bias.  This does allow me to bias the 6V6 tubes much more reasonably.  But I did notice that all the voltages in the amp increase by 50V…

And, last, the heater supply runs at 6.7V, rather than 6.3.  Will this affect durability?  Diodes will drop that to 6.0V, or I could get a couple 0.1R resistors to get me closer to 6.3.  What is the preferred method?

…

If you built it stock and it sounds great, it must be all good 😊


The voltages you report look typical. Cathode bias voltage looks normal. Dissipation at Pmax is normal for this amp. (Cathode bias voltage changes automatically with increasing or decreasing current across the fixed resistance, so the 6V6s will stay in their happy place.)
B+ will change (increase) when you switch to fixed bias. What you report is normal. The decreasing load on the PT results in higher voltage across the board. 6.7VAC on the heaters is within 10% of the target voltage so the tubes will tolerate it. A pair of anti-parallel 6A diodes in series with the heater winding will knock the voltage down by about 0.6V, and the tubes will happily run on that as well.

[HotBluePlates]

... I am a mechanical engineer ... Biasing the power tubes ...  Everything suggests to use 1/2 the OT impedance. ...  The DC resistance is quite different, and for setting the DC operating point, it would *seem* (but obviously isn't) that the OT DC resistance would be the governing factor.

Are you talking about a "Load Line"?  If yes, that's an "AC Impedance" not (just) a DC Resistance.

You understand vibration.  Your guitar audio is some number of Hz, so it's AC.  It's not 0Hz DC.  So we would care about the AC values.

The transformer's AC Impedance is much higher than its DC Resistance.  If it were all DC Resistance, there would be no output power because the tube's output (a current pulled through the transformer's primary) would be wasted as heat.  Dissipation/Heat is Current x Current x DC Resistance.

To actually transmit power from primary to secondary, the speaker load is levered-up to form the Primary (AC) Impedance.  Current through the primary creates a voltage drop in accordance with Ohm's Law:  Volts = (tube) Current x (Primary AC) Impedance.

     - Variational Volts (AC) gets transferred from primary to secondary.
     - Variational Current (AC) gets transferred from primary to secondary.
     - Variational Volts x Variational Current = Power, transferred from primary to secondary.
     
     - Variational Current (AC) x Variational Current (AC) x DC Resistance = Power Wasted as Heat in the Primary.

Since the DC Resistance will result in waste heat, but the Impedance at the primary (due to the speaker load being reflected to the primary by the transaofmrer's Impedance Ratio) results in useful output power, we seek to have DC Resistance be as small a % of total Impedance as practical.

[burgermeister]

Thanks for the replies.  I will probably still try to bias the PTs a tad cooler, hoping to make them last longer.  Maybe I'll start with the bias resistor, it's a lot cheaper than a new OT.

I am not smart enough to truly understand anything that uses differential equations, that includes vibrations!  A man's got to know his limitations....  I do understand the concepts, difference between AC and DC, impedance and resistance, and I do understand the transformer function. 

I just don't understand why the Impedance is used to set the quiescent operating point.  When there is nothing to amplify, the load resistor on the output tube is just the DC resistance of the OT.  The preamp tubes get set up with a load resistor, and the quiescent point is set using that resistance.  It just (wrongly) makes sense to me that the same logic would be applied to the output pentode. And then the AC load line gets added to that point.  But the process for the output tubes never bothers with the DC resistance, nor could the line even be plotted on the data charts because it is so different from the AC impedance.  Thus my confusion.  The books I've got don't really discuss setting up output tubes in much detail...

[PRR]

> why the Impedance is used to set the quiescent operating point.

We don't want a radical change from no-signal to full-roar. Setting the idle to be somewhat like full-roar condition does that.

[tubeswell]

It’s another method of calculating Plate dissipation. The DC resistance of each half of the primary winding is a measurable quantum (with your R-meter). The voltage drop across each winding half can also be measured (when the amp is idling/no-signal with your V-meter). The voltage drop/the DC resistance = the plate idle current. Plate idle current x plate voltage = plate dissipation.


However, ^that^ is only one method of calculating plate current. You can also estimate plate current by calculating tube current through the cathode (using the cathode resistor voltage and cathode resistance), and subtracting a quantum for screen current. (Tube current in a tetrode or pentode is plate current plus screen current). Screen current in an idling 6V6 in a 5E3 is about 2mA.


As for the operating point being at 100% Pmax, that’s just how a 5E3 rolls. The 6V6s will continue to operate  within factory specified limits due to the ‘auto-bias’ action of the cathode resistor. When both 6V6s are conducting signal, the amp will be running in push-pull Class A. When the PI is putting out a sufficiently large enough signal to alternately drive each  other 6V6 into cutoff, the amp will transition into Class B. But this is what’s meant to happen, and the 6V6s won’t be harmed.

[HotBluePlates]

I just don't understand why ... The books I've got don't really discuss ...

Whatcha readin'?  They're not all created equal! 

[mresistor]

A pair of anti-parallel 6A diodes

So you would use a pair of 6A10 diodes, for example?  Why 6 amp rated diodes?   Granted 1N4007s aren't suitable according to Merlin. 

[burgermeister]

I have the Blencowe preamp book (my favorite), the Kuehnel basic tube amps book, and a not terribly useful one from Jurich on tube amps in general.  In all cases currently for sale editions.  I did see Kuehnel used to have a power amp book, but it is out of print and not exactly cost effective to buy used.

I am very open to additional reading suggestions!

[tubeswell]

6A diodes will easily cope with the heater current and they have thick leads for soldering heater wires to. But you could use any diodes that can cope with the v.a

[burgermeister]

I used 3A diodes - works great!  They drop voltage approximately 0.7V

[HotBluePlates]

I have the Blencowe preamp book (my favorite), the Kuehnel basic tube amps book, and a not terribly useful one from Jurich on tube amps in general.  ...

Kuehnel & Merlin are excellent.

I think I see the problem:  you're reading about "bias" of preamp tubes, and in books that use preamp tubes to discuss tube concepts & circuit design.  But for output tubes, "bias" most often is mentioned as part of maintenance & tube replacement.

In the output section, the designer already decided where the tube's bias (Grid 1 Volts and/or Plate Current) should be based on how they want the power section to run.  Your job is merely to infer what that designed-bias was, and check that the newly installed tubes are performing as they should.

    When you looked at "checking bias" it seems you found one of the recommendations to measure DC Resistance of the OT half-primary (with the power off), then turn the power on & measure Volts across the same half-primary.  Then use Ohm's Law to get Volts / Resistance = Plate Current.

     You then have to compare the resulting figure to what you presume the intended bias was.  For a tweed Deluxe, probably not much below 100% plate dissipation.  For a Super Reverb, probably more like 70% plate dissipation, and maybe less.

"WHY" the amp designer chose to set up the power section the way they did is a different challenge.  I think maybe I assumed you were trying to go that route, rather than just know why the biasing procedure you mentioned is the way it is.  Tubeswell explained that very well.

...  B1=376, B2=334, B3=267.  This seems high.  Voltage across the 247R PT Bias resistor is 21.3.  This ends up just above 100%+ dissipation for the 6V6S. ...

There's a 5kΩ resistor between the 1st and 2nd filter caps.  Do you have a 10kΩ or 12kΩ or 18kΩ handy?  Bump that 5kΩ resistor up in value & the 6V6 screen voltage will drop, cooling off plate current.  You could calculate "the best resistor" but it's faster to solder one in & re-check plate dissipation.

[burgermeister]

Yes, and .... I am trying to get to where I can intelligently design an output tube stage with an appropriate bias.  I am still pretty clueless how I can figure that out.

I did make the attempt to vary the screen voltage.  The original build had 4.7 & 22K dropping resistors.  I replaced them with 10K and 18K.  The results were interesting.  B1 went up, B2 as a fraction of B1 went down, B2 as a voltage went down considerably less (since B1 went up), B3 went up.  Dissipation changed by maybe 1W, moving me from 14.5 to 13.5, or thereabouts.

I also have, since then, replaced the 250R bias resistor with a 330R version.  Which, on the surface, one might think would increase bias voltage by 30%, so from 21V to 28V maybe.  But, it ended up from 21V to 23V. Dissipation went down to 12.5.  All in all, not bad unless I need to use a NOS 6V6, which appears to have a 12W dissipation limit.

My conclusion is that the output stage really wants to run where it wants to run based on plate voltage, and any changes I make are partially negated by reactions in other parts of the circuit.  ie, Negative Feedback.  One of those things mechanical stuff doesn't have a lot of, so I lack intuition.

I have ordered a lower voltage OT - I imagine that will cool things down a bit without me biasing the output stage really cold.

All of which brings up a different question.  Is output tube life even affected by dissipation?   Obviously grossly exceeding Pmax will melt things.  But will running at 70% of Pmax, vs. 100%, change the life of the tubes?

[shooter]

My conclusion is that the output stage really wants to run where it wants to run based on plate voltage

Biasing an SE amp is sorta like circular logic, it makes sense for a while, then just seems to fall apart 
I aim for 90% Pdis, no signal
I add signal till it sounds like crap, then back off a smidge
 

[HotBluePlates]

I did make the attempt to vary the screen voltage.  The original build had 4.7 & 22K dropping resistors.  I replaced them with 10K and 18K.  ...  B1 went up, B2 as a fraction of B1 went down, B2 as a voltage went down considerably less (since B1 went up), B3 went up.  Dissipation changed by maybe 1W, moving me from 14.5 to 13.5, or thereabouts.
...
My conclusion is that the output stage really wants to run where it wants to run based on plate voltage, and any changes I make are partially negated by reactions in other parts of the circuit.  ...

I have ordered a lower voltage OT ...

That's unfortunate...  Well, they do say, "If at first you don't succeed, give up & buy a new power transformer."    (just kidding)

I wish you'd spent a few dollars on resistors first.  4.7kΩ --> 10kΩ wasn't enough, but trying 12kΩ or 15kΩ or 18kΩ might have worked.


The 6V6 plate current is the largest % of total current being pulled from the power transformer (PT).  Doing something to reduce 6V6 plate current will reduce loading on the PT.

Since less current is pulled through the impedance of the PT winding & rectifier tube, voltage drop across those is less (Volts = Current x Resistance).  Voltage rises some amount.

Now that voltage overall has risen, you have to re-check it and re-calculate 6V6 plate dissipation in light of the reduced current draw.

... I also have, since then, replaced the 250R bias resistor with a 330R version.  Which, on the surface, one might think would increase bias voltage by 30%, so from 21V to 28V maybe.  But, it ended up from 21V to 23V. Dissipation went down to 12.5. 

Output tubes distort when the peak drive signal is equal to the bias voltage.

We can increase the cathode resistor to reduce plate current, but that also increases the bias voltage which makes the output tube less sensitive.  That may be undesired.

[HotBluePlates]

Now I'll try again the "calculate an answer way" rather than the try & see way.

You appeared to start with B1 (Plate)=376, B2 (Screen)=334, and 21.3v across 247Ω for an idle current of 43.1mA per tube.  (376v - 21.3v) x 43.1mA = 15.3 watts.

     376v - 334v = 42v across the 4.7k dropping resistor:  42v / 4.7kΩ = 8.9mA of screen + preamp current.

Your new condition with 330Ω cathode resistor:  resistor voltage up to 23v --> 23v / 330Ω = 34.9mA per tube.

     You say dissipation is now 12.5w --> 12.5w / 34.9mA = ~358v, plus 23v at the cathode ---> 381v plate.

I'll assume plate voltage will rise again to 390v just to be safe.  We would like a bias voltage no higher than 21v, though 18v of the earlier Deluxe would be nice:  390 - 18v = 372v.  12w / 372v = ~32mA.

     We use the Triode Curves on the bottom of Page 5 of the 6V6GTA data sheet.
     I've eye-balled what looks to be around 3/5 of the way between the -15v and -20v curves to draw in a -18v bias curve.
     I've eye-balled what looks to be ~32mA of plate current, and drew a Red horizontal line.
     I add a vertical Red line where that 32mA line intersects the 18v bias curve:  rough guess it crosses at "265v."

     Although the X-Axis says "Plate Voltage," in Triode Mode the plate & screen are attached.
     So these curves tell us the relationship between Plate Current, Screen Voltage, and Grid-to-Cathode Voltage/Bias.

     So what resistance between "B1" and "B2"?
     390v (B1) - (265v (B2) + 18v (Cathode)) = 107v to be dropped from B1 to B2
     Earlier, we noted current through this resistor was 8.9mA
     107v / 8.9mA = ~12kΩ

     Or we could not-correct for the voltage across the cathode resistor, just in case B1 rises by as much as 18v:
     390v (B1) - 265v (B2) = 125v to be dropped
     125v / 8.9mA = ~14kΩ ----> round up to 15kΩ

Well look at that... That looks suspiciously like when I asked if you had "10kΩ, 12kΩ, 18kΩ handy" in this post.     Except have a 15kΩ, too!

Plate Voltage (B1) will rise some unknown amount, and there will be a non-zero rise of plate current from what is predicted due to the fact the plate will near 400v instead of being at the same exact voltage as the screen.


We could also look to the 5C3 Deluxe, which has voltages printed on the layout portion:

     360v Plate, 308v Screen, 18v Cathode.
     18v / 250Ω = 72mA ---> 36mA per tube

     360v - 18v = 342v Plate-to-Cathode
     308v - 18v = 290v Screen-to-Cathode

     342v x 36mA = 12.3 watts.  Close enough for 1953 Rock & Roll.

     290v Screen-to-Cathode ---> mighty close to the 265v I calculated, given your plate voltage will be higher.

[HotBluePlates]

... I did make the attempt to vary the screen voltage.  ...

So where is the Deluxe at now?

[burgermeister]

Sorry ... I totally missed the replies, the notification ended up in my spam folder.

The 5E3, to the extent one can still call it that, is doing dandy with a lower voltage power transformer.   
B+1=347, B+2=290, B+3=234
6V6 current is 29mA, plate dissipation 10W ignoring the screen current.  Plate voltage is still above the published 6V6GT max, but I've got the 6V6S, so it is well within spec.
The amp is less loud, which is a bonus!  Sound quality is not any different.  Might be noticeable if I had an A-B switch, but given a couple hours to solder in a different transformer, I didn't notice a change.

If I had to do it again, I would probably try a 1K resistor before B+1, basically dropping the voltage for the entire amp and making everything a little more saggy to boot.  Seems that ought to accomplish a similar thing as what I did with the different power transformer?

One thing I *think* I have come to understand is the fundamental difference between biasing a PP power tube and a preamp triode. 
The triode uses a load resistor, and it's 'AC Load' (really just another resistive load coupled by a capacitor) runs in parallel to the load resistor.  So the AC load is always a lower value than the DC load.
The pentode has a negligibly small load resistor for its DC operating point.  The load line would be close to vertical. So the DC operating point is really just set with the grid & screen voltages.  The very same load resistor (aka transformer) has a much higher impedance for AC load, so the AC load is always a much higher value than the DC load.
So two very different animals.  It would have been nice if any of the books I've got would have spelled this out for the dimwitted like myself.
Or do I still have it wrong? 

[PRR]

> The pentode has a negligibly small load resistor

I think you mean "the transformer coupled stage". The tube can be a triode (but that's boring). And a pentode can be resistor-coupled just like your triode preamp example.

And no, I've never explicitly made the generalization you made, and do see its validity.

It might help to come from 1930s radio, where all stages (but one?) were transformer coupled.

[tubeswell]

The pentode has a negligibly small load resistor for its DC operating point.  The load line would be close to vertical.  So the DC operating point is really just set with the grid & screen voltages.  The very same load resistor (aka transformer) has a much higher impedance for AC load, so the AC load is always a much higher value than the DC load.

The load line for a (tube) output stage with a transformer, is an AC load-line (so its never 'vertical'). Its a reactive load (as opposed to a resistive load) - the reactivity being a function of the speaker's impedance reflected back through the OT.  (Transformers only 'work' under signal conditions, because electromagnetic induction only occurs when there is change in current over time). The output tube's operating point needs to be set (mainly) by the grid* voltage because the DC load (in idle condition) is almost non-existent (but the effects of grid voltage and screen voltage on the operating point of the tube are interrelated).

*because the grid is physically closer to the cathode (from whence the electrons are being sourced) than the screen is.

[HotBluePlates]

... The pentode has a negligibly small load resistor for its DC operating point.  The load line would be close to vertical. So the DC operating point is really just set with the grid & screen voltages.  The very same load resistor (aka transformer) has a much higher impedance for AC load, so the AC load is always a much higher value than the DC load. ...

The load line for a (tube) output stage with a transformer, is an AC load-line (so its never 'vertical'). Its a reactive load (as opposed to a resistive load) - the reactivity being a function of the speaker's impedance reflected back through the OT. ...

You're both right.

We want voltage at the tube plate kept off the speaker, so we use a transformer to couple the two.

We could use a resistor & capacitor to couple the power tube, just like a preamp stage.  A side benefit of that transformer is that the DC Resistance between power supply & tube-plate is small.  We waste less voltage across a load resistance, have more voltage at the plate for Power-Output.

The DC Load Line is used in more advanced investigation, when we might consider the effect of power supply voltage sag or similar.  Basically, some few-% of predicted power output would not be realized in a prototype, and the "thorough analysis" shows why that is the case.

In the Old Days or Today, it's a lot faster to consider whether the Operating Point sits on the AC Load Line or not (mainly due to class of operation), do the basic figuring with basic approaches, and know the Power Output answer will over-estimate Reality by some small amount.

[brewdude]

I got lost in all the discussions…
But, we’re the above conditions actually with regard to the 6V6’s running as triodes, as the above posted load lines?
Isn’t the 5E3 a cathode biased push pull pentode power amp?
Or, did I skim right past the details again?

[HotBluePlates]

... we’re the above conditions actually with regard to the 6V6’s running as triodes, as the above posted load lines? ...

The 6V6 were never "run as triodes."

But the problem I had was I wanted to use data sheet curves to predict idle current.  The graphs have an X-axis of Plate Volts and a Y-axis of Plate Current.  But then a note in the corner of the applied screen voltage.  Turns out the graph is only valid at that one screen voltage; change the screen voltage and plate current changes.

The answer for an Idle Condition is to use the Triode Curves.  They assume the plate is connected to the screen, so the X-axis is really a "Screen Volts" axis.  We're not considering any effects with signal, and we're not drawing any AC Load Line, so we're fine.

This doesn't really help with peak plate current & estimating power output with an AC Load.  For that we have to estimate the location of a 0v gridline on the graph that provides such a line for different screen voltages.

[brewdude]

Isn’t there generally another graph which depicts curves indicating the plate voltage at other screen voltages? 

[shooter]

none on the data sheet i looked, but here's the rest.
I like the operational graph for the "big picture" 

[PRR]

Isn’t there generally another graph which depicts curves indicating the plate voltage at other screen voltages? 

Seldom. You figure the 3/2 power of Vg2 to derive the new current. In theory G1 also follows a 3/2 power law, but the maker can put a lot of spin on that. The G2 usually follows 3/2 power pretty exact, for reasonable values of Vg2.

[HotBluePlates]

Isn’t there generally another graph which depicts curves indicating the plate voltage at other screen voltages?

What PRR said.

Also:  I looked super-close at the 6V6GT data sheet.  All curves are for 250v only.

    Top graph on Page 4 doesn't count, because it is used to figure max clean output power, not idle bias.
    All curves of any type show 250v max; I need "what happens at G2 = 400+ volts."

    I have never found a 6V6 data sheet from any manufacturer or any year with curves for G2 > 250v.

    The situation is different/better with other tube types.

[PRR]

Forum has eaten a long message. (May be finger-error.)

Here's the image. Maybe later I will recreate the explanation.

[HotBluePlates]

Thanks for taking the time to put this together!!

I've read you say plate current for different-G2 could be estimated as "3/2 power of the ratio" before, but hadn't spent the time proving it to myself from tube curves or practical experiment.

I know I swiped the use of triode curves to estimate the effect of different-G2 on plate current from you.  Now I've got another tool in the kit!

... Maybe later I will recreate the explanation.

What I think I see:

You figure the 3/2 power of Vg2 to derive the new current. ...

This is a proof of the statement above that "plate current at new G2 voltage is 3/2-power of the plate current at a reference G2 voltage."

    -  Points are plotted at "the knee" of each plate current curve (for a specified G2 voltage) so they're like-for-like.

    -  The "G2 Voltage" is printed in Column C of the Excel spreadsheet.

    -  The "Graphed Plate Current" for each G2 voltage are captured in Column B of the Excel spreadsheet.

    -  Column A calculates the ratio of "Desired G2 Volts" and "Reference G2 Volts," raises that to the 3/2-power, then multiples by the plate current at the reference G2 voltage.  "C5" in the formula changes to be whatever the "new G2 voltage" will be.

    -  Overall, accuracy is pretty good. It gets wonky at very low screen volts, but we either have those figures already or were gonna need to adjust-on-test anyway.


And if we play connect-the-dots, we pretty nearly have a graph of a "3/2 exponential curve."

[PRR]

The curves are for 7027, BTW.