Output Transformer Help

[dscottguitars]

Hi, I want to build an amp with EL84 tubes PP and have it switch from 2 to 4 four tubes. How do I figure out the correct output transformer specs to use? I'm guessing it would be about 15 to 18 watts with two tubes and 30 to 36 watts with four.

Thanks...

[sluckey]

Get an OT that's suitable for four EL84s, such as used in a Vox AC-30. Hoffman has a good one.

[92Volts]

When running 2 tubes, plug an 8 ohm speaker into the 4 ohm tap or a 16 ohm speaker into the 8 ohm tap to double the effective impedance. This is "correct" for 2 tubes instead of 4. Or don't, neither the desired impedance for any design or the actual impedance of a speaker are very precise with guitar amps.

[jjasilli]

Alternatively you can kill signal to one pair of tubes with a SW or a pot.  This will preserve plate-to-plate impedance.

[92Volts]

Alternatively you can kill signal to one pair of tubes with a SW or a pot.  This will preserve plate-to-plate impedance.

This will preserve bias (if bias would be upset by changing the number of tubes, in some cases it's not a concern). It won't preserve impedance because impedance responds to the tube's AC behavior. A tube that's drawing constant DC current might as well not be there.

[dscottguitars]

Thanks

[dscottguitars]

Alternatively you can kill signal to one pair of tubes with a SW or a pot.  This will preserve plate-to-plate impedance.

A pot. That seems like a good idea. I was just planning on using a switch on the cathodes with it being cathode biased.

[jjasilli]

I'm not sure 92volts is correct. 


It's incontrovertible that a signal pot will keep all tubes ON and preserve plate-to plate impedance at idle.  Lifting a pair's cathodes will double plate impedance at idle (which is not horrible as 92volts already pointed out.)


Impedance is frequency dependent, not current dependent.  Under signal conditions the impedance of the plates of the tubes with signal will vary.  The plate impedance of the other tubes will not vary, but remain at idle impedance.


Lets say:
1.  At idle, the plate impedance of ea tube = 1000.  Then at idle, the p-to-p impedance of all 4 tubes in parallel = 250. 
2.  With signal to all 4 tubes, the plate impedance of ea tube rises to 2000. The p-to-p impedance of all 4 tubes in parallel = 500.
3.  With the same signal to just 2 tubes: the plate impedance of the 2 non-signal tubes = 1000 ea.; the plate impedance of the other 2 tubes with signal = 2000 ea.;  The p-to-p impedance of all 4 tubes in parallel = 333.33.  Life is good.

Also, with a signal pot you can fine tune the amount of extra volume.

[dscottguitars]

Thanks JJ.


I may try that on my next build. I already have a front panel being made with a switch.

[92Volts]

I'm not sure 92volts is correct. 


It's incontrovertible that a signal pot will keep all tubes ON and preserve plate-to plate impedance at idle.  Lifting a pair's cathodes will double plate impedance at idle (which is not horrible as 92volts already pointed out.)


Impedance is frequency dependent, not current dependent.  Under signal conditions the impedance of the plates of the tubes with signal will vary.  The plate impedance of the other tubes will not vary, but remain at idle impedance.


Lets say:
1.  At idle, the plate impedance of ea tube = 1000.  Then at idle, the p-to-p impedance of all 4 tubes in parallel = 250. 
2.  With signal to all 4 tubes, the plate impedance of ea tube rises to 2000. The p-to-p impedance of all 4 tubes in parallel = 500.
3.  With the same signal to just 2 tubes: the plate impedance of the 2 non-signal tubes = 1000 ea.; the plate impedance of the other 2 tubes with signal = 2000 ea.;  The p-to-p impedance of all 4 tubes in parallel = 333.33.  Life is good.

Also, with a signal pot you can fine tune the amount of extra volume.

That may be true for plate resistance (behavior of the tube) but not load behavior of the output transformer.

Say the OPT presents 1k ohm impedance, which should be independent of idle current-- and say each tube draws +100ma above idle current at peak signal.

With 2 tubes, voltage swings -200v (200ma total * 1k ohm impedance) at peak signal. When only 1 tube is working, current swing above idle is only +100ma so plate voltage swings only -100v. That leaves more voltage across the tube, the tube could draw more peak current as a result, so dissipation is excessive. In other words the working tubes are seeing lower-impedance load behavior, because the other pair isn't helping drive the load. This is true whether the second pair of tubes is connected and idling, or not there at all.

Again, not a practical problem, but load impedance increase/decrease to account for multiple devices "should" be based on working devices, not idle ones

[dscottguitars]

Ok, now you're both confusing me. Any problem with a 4K ohm transformer with a 4 ohm secondary rated at 40 watts? It's from Weber and similar to the Vox specs but not as expensive. 

BTW: thanks for your input...

[sluckey]

That's perfect.

[PRR]

By the math: two tubes will pull 8K and give 18 Watts; four tubes then pulls 4K and gives 36W.

But impedance change is awkward. And is HALF-power really what you want? It is not a lot less loud.

The usual way (as has been suggested) is to load with 4K. Four tubes gives the 36W. Two tubes work at a disadvantage and give nearer 10 Watts, a more significant drop of output. Neat, sweet, sensible.

If this amp plays at home 6 days a week, goes out to the club Friday night, I would put a switch in the *heaters* of an output pair. Why hot-up the bottles all week, run up the electric bill, take hours off the tubes? When you set-up in the club, flip the second pair ON. If you forget, it will take 11 seconds to get the second pair hot, so remember before the bandleader drops the baton.

[dscottguitars]

By the math: two tubes will pull 8K and give 18 Watts; four tubes then pulls 4K and gives 36W.

But impedance change is awkward. And is HALF-power really what you want? It is not a lot less loud.

The usual way (as has been suggested) is to load with 4K. Four tubes gives the 36W. Two tubes work at a disadvantage and give nearer 10 Watts, a more significant drop of output. Neat, sweet, sensible.

If this amp plays at home 6 days a week, goes out to the club Friday night, I would put a switch in the *heaters* of an output pair. Why hot-up the bottles all week, run up the electric bill, take hours off the tubes? When you set-up in the club, flip the second pair ON. If you forget, it will take 11 seconds to get the second pair hot, so remember before the bandleader drops the baton.

I know it won't be that much less in loudness, I just wanted to have the feature and try it out. I do like your idea of a switch for the heaters. I think what I will do is a quick wire of a pot to see the dynamics of turning it up and then wire in the switch on the heaters.


I'm still confused on the impedance aspect. If two tubes requires and 8K, what is the effect of using a 4K transformer? I've read that if there is a mismatch, that lower impedance transformer is better, but not what it does.

[jjasilli]

There's more than 1 way to skin a cat.  I've been trying to be supportive and offer up one alternative to implement your plan.  But now that the cat's out of the bag, ultimately I agree with PRR, to the effect that the outcome is worth neither the effort nor the complexity added to the amp.  The difference in volume by doubling or halving amp output power in Watts is only 3dB, which is barely noticeable.  Personally I'm a fan of KISS (not meaning the band  ). 

Nevertheless, I respect your choice to engage in effort or complexity at your sole discretion.  If so, I would not turn off the heaters or the cathodes, but do it "my way" with a signal pot.     However, if you're going to be turning a pot or flipping a SW anyway, you'll get the same result using all 4 power tubes and just slightly turning down the vol pot that's already there.

[pdf64]

My experience is that a significant part of why I perceive power amp overdrive to sound better than preamp OD is HT sag, both at the plate and g2 nodes; the dynamic response of the HT seems to be a key factor in determining the touch sensitivity and the way and amp 'breathes' and approaches a vocal expressive quality.
Hence halving the number of operational power tubes tends to make the amp sound worse, unless the HT supplies are appropriately amended alongside.
Better to modify the SPL an amp generates at overdrive by changing the speaker cab sensitivity, eg fewer / less sensitive speakers.

[jjasilli]

I think pdf64's explanation on the gear page is good:  https://www.thegearpage.net/board/index.php?threads/output-transformer-speaker-tap-effects-voltage-drop.1882695/


I also think sluckey pointed out in a recent thread that the OT with different current levels behaves linearly (at least until saturation, and within the guitar range); whereby I continue to maintain that 4 tubes ON, but only 2 with signal, keeps overall impedances neat & tidy.  This is supported by the KOC SE Mod, which causes a PP PA to emulate SE by using a pot to cut signal to one Power Tube -- thus reportedly preserving p-p impedance according to KOC.


 


@pdf64.  Though I too like sag, KOC also maintains & advocates that any PA overdrive tone can be recreated in the preamp, with the PA giving only a clean boost.  I'm not sure I agree with him on that point.  And, other than pure "tone" is the massive air movement causing a physical jolt & the feed back effect on the guitar, of playing a PA loud enough to overdrive.  So I'm guessing you would no longer like the mere tone of PA overdrive if it were attenuated down to bedroom volume level.

[pdf64]

I think pdf64's explanation on the gear page is good:  https://www.thegearpage.net/board/index.php?threads/output-transformer-speaker-tap-effects-voltage-drop.1882695/


I also think sluckey pointed out in a recent thread that the OT with different current levels behaves linearly (at least until saturation, and within the guitar range); whereby I continue to maintain that 4 tubes ON, but only 2 with signal, keeps overall impedances neat & tidy.  This is supported by the KOC SE Mod, which causes a PP PA to emulate SE by using a pot to cut signal to one Power Tube -- thus reportedly preserving p-p impedance according to KOC.


 


@pdf64.  Though I too like sag, KOC also maintains & advocates that any PA overdrive tone can be recreated in the preamp, with the PA giving only a clean boost.  I'm not sure I agree with him on that point.  And, other than pure "tone" is the massive air movement causing a physical jolt & the feed back effect on the guitar, of playing a PA loud enough to overdrive.  So I'm guessing you would no longer like the mere tone of PA overdrive if it were attenuated down to bedroom volume level.

Thanks, that was a good thread with some helpful insights / contributions

Regarding the situation where, in a 4 tube, parallel p-p amp, 1 of the p-p pairs has their input signal removed and they're just left idling, I think they will effectively become current sources https://en.wikipedia.org/wiki/Current_source
As such they will have very high impedance, and so for an ac analysis, they may be ignored, ie they're effectively 'out of circuit'.

In a regular 2 tube p-p amp, I think that if the signal is removed to one of the tubes, likewise, for ac that tube will be effectively 'out of circuit'. That will take its half of the primary out of action and so all else being equal, the load impedance for the remaining tube will drop to 1/4 of the nominal primary impedance.

There's a big benefit in leaving the 'inactive' tube drawing dc idle current though, as it keeps the OT magnetisation balanced and working properly. Without that, the imbalanced dc magnetisation of the OT would likely saturate its core, its primary impedance would collapse, and it would sound awful.

Regarding sound pressure level, nothing beats the visceral impact of playing loud. But I'm a massive fan of my FluxTone speaker, which allows me to overdrive my tube amps (and how they interact with an actual speaker load) at any SPL. I'm able to hear how they overdrive at domestic friendly SPL, and without setting my tinnitus off / consuming what little hearing I've got left.

[jjasilli]

Thanks, that was a good thread with some helpful insights / contributions
 


Hadn't heard of Fluxtone before.  Seems like a terrific "attenuation" system, preserving highs, etc.


I've relied on KOC re the preservation p-p plate impedance with signal cut to power tubes.  Maybe that's just wrong?   

[jjasilli]

Mulling over the p-p impedance issue, I'm thinking the following. Let's consider a circuit of 2 power tubes in PP > OT primary > OT secondary > Speaker Load.  The circuit has impedance.  This means that it presents a different but specific fixed resistance "tailored" to each specific frequency which passes through it.  At idle there is only DC flowing in the circuit.  DC has a frequency of -0-.  Therefore the impedance of the circuit at idle is NOT non-existent as 92volts maintains.  Rather the impedance of the circuit at -0- frequency = the combination of the static, DC, resistances of the components of the circuit.


Now feed AC signal with a frequency above -0- to G1 of just one power tube, the "fed tube". The other tube with no signal applied to G1 is the "unfed tube".  The fed power tube will send the AC signal through the circuit, including through the unfed power tube.  The unfed tube will conduct the signal, even though it isn't contributing any gain.  In conducting the AC signal, the unfed tube will present it's own internal impedance to the AC signal, which will be in parallel with the impedance of the fed power tube.


Hence, p-p impedance is maintained and KOC is right.  Comments are welcome.

[sluckey]

The fed power tube will send the AC signal through the circuit, including through the unfed power tube

Wow! 

[jjasilli]

Well, that's a Comment all right!  I'm counting on you for confirmation, rejection or something.

[sluckey]

 Give me some time. I'm still trying to get over the fact that you actually said that.   

[92Volts]

As you point out, the unfed tube sees signal on its plate, as voltage swing occurs across the whole OPT regardless of where tubes are present or active. I'll address that later, but we agree it doesn't add gain (or power).

And that's the problem, or one way you can think about the problem. Given one OPT impedance, gain/power is directly related to voltage swing across the OPT. We agree power is lower with 1 tube driven, meaning the tube that is working pulls the same amount of current (or whatever current it's can under these different circumstances) and sees smaller voltage swing as a result.

This lowers efficiency/output power and leaves more voltage on the plate while the tube is pulling maximum current. The second effect is a problem because it increases power dissipation, particularly in push-pull where the tube can draw higher maximum current without a symmetrical decrease in minimum current (minimum current is already near zero, can't go lower).

Given 2 ways of looking at this (maximum performance from the remaining tubes, and power dissipation), the system is behaving as if it's a lower impedance than "ideal".

Also consider the effect of an unfed tube's plate resistance. When it sees more voltage it draws more current, when it sees less voltage it draws less current.

This counteracts the efforts of the driving tube which draws less current to produce rising voltage, more current to produce falling voltage. This further reduces voltage swing and output power, making the whole system behave as an even lower impedance-- getting less power than you'd expect from the tube(s) that are still operating, and increasing their dissipation.

With that said, pentodes are pretty close to drawing constant current (high plate impedance) so this effect isn't that harmful.

But it's not helpful. If appears helpful if you assume tube "impedance" (plate resistance) is all that matters, but tubes are active devices. You need to consider what each one is actually doing when driven or not driven to understand how they affect the circuit.

One thing to note is OPTs requiring DC balance is a separate effect from this impedance conversation, and does matter real. If you have a PP OPT and want to operate in single-ended mode, leaving 1 tube idle to draw DC current to oppose the bias current of the active tube IS necessary, or a least one good solution to that problem (otherwise you'd need resistors/choke or a transistor or some other current source)

[jjasilli]

@sluckey: I'm just the messenger reporting what KOC said re his SE Mod.  I'm no electronics expert.  I'm not "getting" this issue & suspect I'm not the only one.  What I recall from KOC is that his signal pot mod > PP tubes emulates SE in that the removal of the opposing signal wave in the OT primary restores 2nd-order harmonics (which are cancelled-out when both both waves are present at equal & opposite voltages); and preserves p-p impedance.  Now, some knowledgeable people on this Forum are taking issue with the latter claim.  Maybe someone should inform KOC.


@92volts:  thanks that last discussion is helpful.  But I still don't get it; maybe I'm a lost cause.  Lets just talk about 2 tubes in PP, 'cause it's simpler and makes the point.  Previously in Reply 4 you stated that the unfed tube draws constant current & may as well "not be there".  I think the unfed tube does draw varying current due to signal swings in the overall circuit, under G1 signal conditions to the fed tube -- even if my explanation leaves something to be desired. 


When you state in your last reply that "the whole system behave as an even lower impedance", you now seem to concur with the conclusion (if not my explanation); i.e., that the unfed tube does present a plate impedance in parallel with the unfed tube.  Hence it is not the same as if it weren't there.

A question I still have is: under signal conditions does the impedance of the unfed tube remain constant (as-if there were no signal in the circuit) or does it's impedance vary with signal frequency?

(I agree that the current draw issue is an interesting related topic, but it I suspect it's not directly relevant to the specific issue of p-p impedance.)

[sluckey]

This is a tar baby and I ain't slappin' it! Y'all have at it. 

[shooter]

yup, pools way to deep for me, BUT from the sidelines, you're letting G1 "float/no input"?, or are you clamping it at cutoff?

[jjasilli]

Well, if the pool's too deep for you guys, then there's no hope for me!

[pdf64]

...Also consider the effect of an unfed tube's plate resistance. When it sees more voltage it draws more current, when it sees less voltage it draws less current...

I don't think it does that.
Rather, as a current source, plate current will remain pretty much constant whatever the plate voltage.
Being a very high impedance, in regard of ac circuit analysis, effectively it's not there.

See how plate current is set by g2 voltage, and is basically unaffected by the plate voltage https://imgur.com/a/0rpny

Sorry, I can't find g2 curves for EL84, or even EL34, but I don't see why their overall trends should be much different to those of the 6L6.

I acknowledge that in a real amp, when the active tubes are cranked, the g2 node will tend to sag a bit and so the g2 voltage of the 'current source / inactive' tube/s will drop a little, which will reduce plate current slightly.
But still, for all practical purposes, over the audio band, their impedance will be very high.

[jjasilli]

"An[other] important, though more abstract, measure of tube performance is its plate resistance. This is the measurement of plate voltage change over plate current change for a constant value of grid voltage. In other words, this is an expression of how much the tube acts like a resistor for any given amount of grid voltage. . .  plate resistance may be determined by dividing the amplification factor by the transconductance. . ." https://www.allaboutcircuits.com/textbook/semiconductors/chpt-13/tube-parameters/  "Transconductance (for transfer conductance), also infrequently called mutual conductance, is the electrical characteristic relating the current through the output of a device to the voltage across the input of a device."  https://en.wikipedia.org/wiki/Transconductance


So, what is the plate resistance of an "unfed" power tube when it's plate voltage (and maybe screen voltage too) is modulated by the surrounding circuitry and not by G1?

[pdf64]

For a dc analysis, I think that the tube's plate resistance is just the Va-k/Ik, so for an EL84 in an AC30, 310V/50mA = 6k2.
For an ac analysis, its impedance is very high, so we can just ignore it.
The plate voltage modulation is irrelevant, and the g2 modulation is so minimal, and at such a low frequency which anyway is unrelated to signal frequency, we can probably ignore it.

[shooter]

You folks are giving me nightmares
I was in an upgrade class for Philips, they just came out with a quadrature T/R coil, I'm getting the jist of how it works and what to expect when it don't, but overall the class was LOST.  the instructor says, no problem, one of the design engineers for the coil is here.  He talked for an hour, and for that hour I felt like one of the guys working on the Manhattan project!!  after class I headed to the mountains to get sane 

[92Volts]

I don't think it does that.
Rather, as a current source, plate current will remain pretty much constant whatever the plate voltage.
Being a very high impedance, in regard of ac circuit analysis, effectively it's not there.

It does have that effect-- the definition of plate resistance is change in current in response to plate voltage, with other factors being constant.

You're correct that pentodes don't exhibit this effect much. You'll find the plate resistance listed on their datasheets, but it's very high, such as 38k ohms for an EL84. 1v change in plate voltage causes 0.026ma change in plate current.

So, what is the plate resistance of an "unfed" power tube when it's plate voltage (and maybe screen voltage too) is modulated by the surrounding circuitry and not by G1?

Whatever is specified on the datasheet. The plate resistance is the tube's response to a "signal" at the plate, with the grid held constant.

Plate resistance doesn't depend on frequency like a capacitor or inductor. But keep in mind, plate resistance describes deviations from the steady state and in this sense, it either doesn't work for DC or has some sort of "offset".

Say plate resistance is 1k ohms. If I raise plate voltage 1v I see +1ma current, if I lower plate voltage 1v I see -1ma current. This is true whether I make that change 1000 times a second or over the course of 10 minutes.

So it works at "any" frequency (maybe not RF but it works at "normal" frequencies and down to DC). But it's not perfect at DC... 1k plate resistance doesn't mean plate current is always 300ma at 300v. Steady state/bias is influenced by grid voltage. You can think of this as a current "offset" on top of the plate resistance, but it's easier to understand plate resistance as a dynamic behavior than try to reconcile it with the other factors that go into bias/steady state behavior.

It does change with circuit configuration, shown on datasheets as different plate resistances for various conditions. This will have a non-linear effect based on signal amplitude moreso than frequency, but it would also be a pain in the ass to model.

Screens could be a pain to account for. If they see the same signal as the plate, it will lower plate resistance (increase the effect of plate voltage). You can find plate resistance figures for triode-mode operation on datasheets. Ultralinear would be in between. Depending on your screen hookups and whether you're dealing with an opposing tube in push-pull, it could be the opposite signal and would counteract/increase plate resistance.

[shooter]

maybe not RF but it works at "normal" frequencies

RF IS normal, audio takes like days and has wavelengths measured in FEET
 

[jjasilli]

FWIW I emailed the question to London Power to see if KOC wishes to comment.

[jjasilli]

. . . and KOC of London Power replies twice!! re its SE MOd http://www.londonpower.com/single-ended-tone :


1. Our SEM kit does not change the interface between the tubes and OT; rather, the performance of the output stage is altered to achieve single-end tone.Have fun
 Kevin


2.  The OP question about what impedance the OT should be for 2-tubes switchable to 4-tubes should have been countered with a question: which way does he want to get maximum power?
 It is safest to go with max power related to 2-tubes and just switch in the others for tone. This is the way the TUT3 Plexi and 800 amps are designed, and the Standard and the Swede in TUT5.
 
 Alternately, select OT for 4-tube max power and use the 2-tube mode for a slightly different tone at reduced power.
 
 Have fun
 Kevin

[pdf64]

Was KOC aware this was in regard of an EL84 amp?
I seem to remember he's not keen on pushing pentodes out of their happy place (with good reason).

[jjasilli]

I sent him this thread and it appears that he read it.

[PRR]

> But it's not perfect at DC... 1k plate resistance doesn't mean plate current is always 300ma at 300v.

A pentode has *two* plate resistances. The plate curve is two-slope, kinked.

Above say 70V it is like 25K or so. Below that it approaches 1K or less.

The "70V" is really some (un-specified) fraction of Vg2. In a True Tetrode that fraction is about 1. With an effective G3 the fraction may be 3:1 or 5:1, typically putting the knee somewhere below 100V.

In your example, as Vp is raised from zero to 70V, current rises to 70mA along the low-R part of the plate curve. As Vp is raised above 70V along the high-R part of the plate curve, current rises very little.

In all normal audio swinging, we work in the high-R part of the curve. An un-driven pentode hanging on the OT pulls some DC current but very little (not zero) AC current.

If you actually wanted BIG power, and were forced to bring an idle pentode along for the ride, you could calculate all this. But as I understand our question, this is an attempt to "lame" a 2-pentode amp to a lower power, without re-rigging and without unbalanced DC current. It will work. How "big" it works is perhaps a question for the ear, not numbers.

[dscottguitars]

Wow! Been away for a few days and had no idea that this discussion got so deep. Way too deep for my general knowledge. Hope you all are having fun. When I get it finished I’ll post pictures and results.

[sluckey]

So what did you decide to do?

[dscottguitars]

I'm going to use a switch on the cathodes. I got the 4K ohm to 4 ohm transformer. There is a lot of technical advice from everyone that seems to have pros and cons about which way to do this, or whether to do this. I liked the potentiometer idea, but looking at the circuit I don't think I want to have the leads that long to go to the pot and back to the tubes. Besides, I already had a panel made and a pot wouldn't work.

The chassis is a 17x7x2 going into a 4x10 cabinet.

[sluckey]

I think that's a good plan. Set it up for optimum operation with all 4 tubes. Don't be concerned with any impedance mismatch when switching to half power. Any mismatch just results in less power, which is what you were after to begin with.

I notice you are using a 220Ω resistor for a pair of EL84s. That's gonna bias them a lot colder than I normally see. It's more common to see a 130Ω to 150Ω for a pair.

[dscottguitars]

I used that on one I just built with two tubes, same schematic. I had 350v on the plates. I think I had around 11v on the cathode.

[DummyLoad]

I used that on one I just built with two tubes, same schematic. I had 350v on the plates. I think I had around 11v on the cathode.

that calculates to 9.3W per EL84. or, 66% of 14W max.

datasheet state 130R common K R with 300V B+ - so @ 350V B+ a 220R common K is probably a good choice.

--pete

[dscottguitars]

I thought the plate dissipation for EL84 is 12W. I used Weber Bias calculator and got 8W, or 75% and 24mA. On the chart that shows in the 'hot' column.

http://www.tedweber.com/webervst/tubes1/calcbias.htm

[sluckey]

Cathode biased output tubes usually idle at 90 to 100%. 75% is not in the hot zone for cathode bias. Your chart must be for fixed bias amps.

[dscottguitars]

Cathode biased output tubes usually idle at 90 to 100%. 75% is not in the hot zone for cathode bias. Your chart must be for fixed bias amps.

I didn't know that. I will make adjustments when I get there. Thanks...