Long and Circuitous Bias Question ...

[jbefumo]

There's a bit of a story behind this amp:
It started with an old friend offering me a blonde fender-style cabinet that he picked up at a show in return for repairing the reverb in his '63 Epiphone Pathfinder amp. It was a new aftermarket cabinet, clearly never used, so I jumped at the deal.  It was a 2X12 cabinet, and his reported dimensions suggested that it would fit a Super Reverb chassis, so I figured I'd use that as a platform for one of my project circuits.
Looking for a chassis, I instead found a complete Super Reverb that was too good to pass up.  The cabinet looked like it had been dragged behind a tour bus, and it was advertised as a "black face" when it was clearly silver face, but I figured it would be difficult not to get enough parts out of it to justify the purchase.
At that point, I had never even heard of 70W, Ultralinear Super Reverbs...
What I discovered when I opened it up, however, prompted an immediate change of plans:
I had by that time learned that this was the one model of Super Reverb for which 'blackfacing' really wasn't feasible, so I really wasn't prepared for what I found. It had, in fact, been properly retro-converted, including a complete complement of Mercury Magnetics transformers (power, output, choke, and reverb), a GZ34 rectifier, properly modified bias circuit, and some of the neatest, boutique-approved wiring I'd ever seen. I also learned that this swollen version of the SR was too wide for my cabinet, but I found a  brand new Mojotone Twin Reverb cabinet at an attractive price, so I grabbed that, and the Super Reverb chassis did indeed fit it perfectly.
The conversion process had left holes where the original Master Volume and Normal Channel mid control had been, filled with crappy looking plastic plugs. I decided to restore those controls, mainly for cosmetic considerations.  However, I implemented the new Master Volume as a post phase-inverter cross-line circuit, with an extra two capacitors isolating it from the bias supply, and a push-pull pot set ups so that the default position has it completely out of the circuit, and a 'pull to enable' facility.
Since I rarely ever used the normal channel on a Fender anyway, I didn't feel too bad about restoring the middle control.  After studying some schematics, I decided to modify it to a 6G5-A tone circuit utilizing that oddball 4-terminal treble pot. Sifting through my parts bin I discovered a pot from an old Baldwin organ that measured exactly 6.8k, so that gave me my middle control.
I replaced the 6L6s with a pair of Gold Lion (reissue) KT66, which I'd done on some Fenders in the distant past with good results (after contacting Mercury Magnetics to ensure their transformer was up to the increased filament current), and I also implemented an 'RMS control', that I first encountered in one of Dennis Kager's Sundown amps, and have since used in all of my fixed-bias amps:
 With the control fully clockwise, it has no effect whatsoever on the circuit, but turning it the other way introduces some additional cathode biasing whose effect is more tactile than auditory. Since there were already some extra plugged holes in the back of the chassis, I stuck the pot in one of them.
Finally, I replaced the reverb drive tube with a 12AU7, which I find make the reverb more suitable for my personal tastes.
... which brings me to the speakers, and (finally) the subject of my question:
My cab now contains a pair of vintage Altec Lansing alnicos, with a claimed frequency response of 60-8000 Hz. I've used these, and the 15" version, in many amps in the past, and think they're the best speakers ever made -- they have a crystalline sparkle without the tendency toward harshness that characterize JBLs of the same era. There's just nothing quite like them available any more (other than, perhaps, the ones being reproduced by Great Plains Audio, which I have not heard.)
The end result is, for my tastes, the best sounding clean amp tone I've ever heard, so I'm reluctant to mess with it further,  other than to ensure that it's perfectly adjusted.
My point(s) of uncertainty involve those KT66s, through the 2-ohm-secondary OT, pushing the 4-ohm load presented by those two Altecs.
So, to all the more experienced amp gurus out there, where would you set the measured cathode current to best bias this combination? I'll be pouring over ValveWizard's site to come up with my own value, but would appreciate a sanity check from those more experienced than myself.
Finally, would it be worth swapping that OT with a similar model from Mercury Magnetics featuring multiple secondary taps? What would be the implications (tonal and otherwise) of the properly balanced relationship between OT and speaker load?
Thanks to all for sticking with me through this rather verbose diatribe!

[kagliostro]

I've a lot less knowledge than you have

I can only say that a 4ohm speaker on a 2ohm OT tap will double the reflected impedance seen from the tubes

if there is a mismatch between the best impedance that the tubes expect you'll have a decrease in the output power

What this will mean ? Probably an effect on the tone of the amp (only you can say if this is worse or better than the "standard")

may be also that this situation put some stess on the tubes but I would like to hear the opinion of someone that is more skilled than I

Ciao

Franco

[PRR]

Using a wrong load: changing the bias doesn't fix that.

As these speakers are cherished, I'd encourage getting an OT with the proper ratio.

Or leave it mis-matched and accept that you won't have every Watt the amp is capable of.

[shooter]

I've done just that on 2 or 3 builds and all is happy, as long as the tube side *sees* what they want for Z.
I have been experimenting on my last build by doubling and halving the speaker  Z.  The *loudness* does drop some, but more important, tone does change, at either end my OT "can't deal" with the extremes well, the sound thins out, less bass, more screetchy treble.

[PRR]

Analogy.

You have a car. Engine dimensions set a "happy" RPM for economy and pep. Tire diameter sets RPM per MPH. Say that the gear ratio is "about right" for that engine and tire size.

Now you put on "farm" or "donk" tires *twice* the size. (Jacked springs and cut fenders to fit.)

Car is now way down on pep, and glugs gas. Engine can't get to its happy RPM when cruising.

Will fiddling the idle RPM screw fix this?

[jbefumo]

Thanks to all! At the moment I'm just using it at bedroom volumes, and in the meantime will save my pennies and replace the OT.  I know I could get a less expensive one, but do want to keep that Mercury Magnetics set intact, so will hold out for that.  Should be able to recover some of that by selling the one that's in there now.  Don't want to keep it around for too long or I'll have to build another amp around it ....

[jbefumo]

Was just reading a post at another forum on a related topic, and someone observed that the JTM-45 ran KT66/6L6s into way higher loads, but at lower voltage, which suggested the question: Would swapping out the GZ34 for a 5U4GB rectifier tube ameliorate the situation?  (I'm still leaning toward replacing the transformer, but if I do that I'm definitely going to go for a new baffle board and use my Altec Lansing 15" (418) to effectively turn it into a Vibroverb ....)

[HotBluePlates]

... Would swapping out the GZ34 for a 5U4GB rectifier tube ameliorate the situation?  ...

6 of one, half-dozen of the other...

Using a wrong load: ... Or leave it mis-matched and accept that you won't have every Watt the amp is capable of.

[jbefumo]

Mainly just trying to minimize wear/damage until I can afford to replace the transformer, but I suspect I'm fretting over nothing since I rarely turn it up in the house.

Using a wrong load: ... Or leave it mis-matched and accept that you won't have every Watt the amp is capable of.

[HotBluePlates]

I think a difference in wear or possibility of damage is essentially zero, regardless. It boils down to more power or less.

[jbefumo]

So, sorry if I'm being dense, but I'm looking at the schematics & specs for a JTM-45 side by side with the Super Reverb. In particular, the JTM has 440V going to the center tap of an 8K load, and the Super Reverb has 465V going to the center tap of a 4.2k Load. The Sr has 6L6GCs, and the JTM, KT66s (which I'm running in my SR, having backed the bias off to a value recommended for the latter tubes.) Really not obsessing over the specifics of the application -- it sounds great as it is, but feel like there's some major gaps in my understanding of this part of the circuit.

I think a difference in wear or possibility of damage is essentially zero, regardless. It boils down to more power or less.

[jbefumo]

Coming at this from another angle: Suppose I was starting from a clean sheet of paper, designing an amp to use a pair of 6L6GC tubes, in class-AB, to generate somewhere between 35W and 50W .... it would appear that I could spec an OT with a primary anywhere from 4.2k to 8k, adjusting other components, voltages, etc., accordingly, right?

I think a difference in wear or possibility of damage is essentially zero, regardless. It boils down to more power or less.

[jbefumo]

Ok -- I see -- from the datasheets a pair of KT66s into 8k at 425V are making 30W, whereas 2X6L6GC into 5.6k @ 450V are making 55W, and I'm guessing that the stock super reverb with 4.2k load and 460V is biased so as to produce the claimed 45W or so? 


So the issue is, as others have stated, mainly power and tone, and none of the combinations are necessarily right or wrong, so long as they are not drastically unbalanced, or biased to where the tubes are glowing cherry red, which is an entirely different issue.

[HotBluePlates]

A hypothetical example will be clearer than a real-tube example.

We have a "F4T" pentode tube which will be operated in our amp with 400v on the plate and fixed bias. With 0v on the grid, the F4T can only pull its plate as low as 50v. At the screen voltage we are using, the maximum plate current possible at that 50v on the plate is 125mA.

See the attachment for a graphical depiction. The loadline, because it is drawn for push-pull, extends from a point at plate-to-cathode voltage and zero plate current (400v, 0mA) up to the peak plate current and minimum plate voltage (50v, 125mA). The voltage change times the current change is power output (as with the boring old power equation), but both are peak values so RMS power output is half this. When calculated, we get ~21.9w of power output (see graph for formula).

The slope of the loadline between these points indirectly gives the OT primary impedance; the plate-to-plate load impedance is 4 times the slope of the line. Use Ohm's Law to find Ω: 350v/125mA = 2800Ω  ->  2800Ω * 4 = 11.2kΩ plate-to-plate.


We've drawn a load which exactly maximizes the F4T's capabilities at this supply voltage (and peak plate current governed by the screen voltage), because the loadline goes right to the knee of the 0v gridline. If you drew a rectangle on the graph with vertical sides at 400v and 50v, and horizontal sides at 0mA and 125mA, then the area of the rectangle is maximized.

What happens when we choose a different load, but keep the same supply and screen voltage?

If the load is lower, the loadline is more-vertical; it will intersect the 0v gridline well to the right of the knee in the area where the 0v line has gone mostly-horizontal. Say we choose to use a 7kΩ plate-to-plate load, and doing so the peak current is slightly higher at 130mA but the loadline shows the plate voltage is pulled only as low as 172.5v. Power = 0.5 * 130mA * (400v-172.5v) = ~14.8w. So the lower load gives less power output than our optimized load found earlier. Peak plate current is still governed by the tube, specifically how much current the tube will pass when the grid is driven to 0v at the given screen voltage.

Let's say we decide to go higher, to a 15kΩ load. Now the slope of the loadline is flatter, and so the line intersects the 0v gridline below the knee where the 0v line is bending down towards 0v and 0mA. Let's say it intersects that 0v curve at 40v on the plate and 96mA of plate current. The voltage change from the 400v at idle is 360v, so (400v-40v)/96mA = 3750Ω ->  3750Ω * 4 = 15kΩ.  What's the power output?  Power = 0.5 * 96mA * (400-40v) = 17.28w.

We've ignored distortion, but have still shown that even with zero distortion, clean output power drops if you have a load impedance higher or lower than optimum for the tube's characteristics at the supply voltage used. With real tubes, the distortion components are changed depending on whether you use higher- or lower-than-optimum loads, but suffice to say that distortion increases over that at the optimum load. Overall, both assertions mean clean output power is lower with an other-than-optimal load regardless of whether the load is higher or lower than optimum.

[jbefumo]

GREAT, Thanks -- I'll print out and digest this when I'm more fully awake.


I the meantime, after some hand-waving internal arguments, I'm thinking that I should probably replace the 470R grid resistors for maybe 1K x 5W?


I think at the end of the day, I love the tone, and don't need a lot of power, so my main focus should probably be on ensuring that I don't prematurely kill those KT66s at fifty-bucks a pop ....


thanks again for all the patient hand-holding!

A hypothetical example will be clearer than a real-tube example.

[HotBluePlates]

Coming at this from another angle: Suppose I was starting from a clean sheet of paper, designing an amp to use a pair of 6L6GC tubes, in class-AB, to generate somewhere between 35W and 50W .... it would appear that I could spec an OT with a primary anywhere from 4.2k to 8k ...

You "can't get there from here" in a design without assuming an available supply voltage. In fact, if you're truly working from a blank sheet of paper, you don't specify the tube type at all, but know your desired power output and likely power transformer specs. The power transformer's voltage and current available, and your desired power output, then define the optimum load.

You'd then make a note of the peak current requirement derived from the required power output, supply voltage and load, and decide what type and how many tubes you will need to control that current.

You then look at your proposed output tubes, the idle plate voltage and peak plate current at full power, and figure out if (1/2 peak plate current * idle plate voltage) equals or exceeds the tube's plate dissipation rating. If it does, you only now conclude you cannot operate these tubes in class A, and determine this will be a class AB output stage.

Bias is the last thing figured out. Relatively easy if you will be able to run the tubes class A, but somewhat harder to analyze if you will be running class AB. As you likely saw from my previous post, there was zero mention of bias when looking at load, plate voltage & current, and power output of the stage.

If you stop a bit and think only in terms of Ohm's Law when considering the loadline presented by the OT, the supply voltage and the output stage peak plate current, something fundamental should become obvious:

A class A amp is designed to idle at 1/2 peak plate current; current swings as-high during half the signal as it does downward during the other half of the signal. If you maximize the class A output stage by idling right at max dissipation for the tube at your supply voltage (assume this doesn't exceed any other rating limit), what must you do if you raise the supply voltage to some other value?

• Higher voltage means we need less idle current to stay at the same maximum dissipation, so we idle at lower current.
• Lower idle current means our peak at twice-idle is lower/less-high than it was before.
• Plate voltage swing will be wider/larger than before
• The wider (horizontally, more voltage swing) load line which does not go as-high (vertically, less current swing) means higher load resistance.
• The above mirrors what we developed earlier, and is only another expression of Ohm's Law.
• Therefore: the same tube operated the same way at a higher supply voltage requires a bigger load impedance

All of the above behaves the same if you choose to operate the same tubes in class A at a lower supply voltage: a lower load impedance will become more-optimal.

[HotBluePlates]

I didn't touch Class AB earlier because there are some critical differences with class A which become a pain to describe for others.

Namely, you're raising plate voltage over what is convenient for class A while also lowering the load impedance. This would exceed the plate dissipation rating of the tube if you were still idling at maximum plate dissipation. But we already know bias voltage will be greater to idle less than max dissipation.

But "how much bias" is hard to determine without drawing lines on tube curves for the screen voltage used, and going through a long formula to calculate average plate dissipation over the whole signal cycle. There's not an easy formula to use, which is also why it's somewhat funny in retrospect to not use a similar bias voltage as that indicated on the schematic.

Anyway, the practical way to go is bias the amp however you see fit. Then play it as loud as you're ever going to play it while watching the tube plates. If they're redplating, increase the idle bias voltage until they stop redplating. If there's no redplating, you're likely fine.

Hopefully, all this has shown you why PRR said earlier you can't adjust bias to change/compensate for a different load impedance. Yes, you can re-bias to keep the tubes from overheating, but the issue is a different thing.

I the meantime, after some hand-waving internal arguments, I'm thinking that I should probably replace the 470R grid resistors for maybe 1K x 5W?

Change them if you want. Or leave them as-is. It would take some figuring with plate curves for your amp's screen voltage (which probably aren't in print anyway), or some estimation of your amp's screen voltage just to figure out if the screens are coming anywhere near their dissipation rating.

Or you could insert an ammeter in between the screens & screen resistors to measure average screen current while you play. You'd then calculate screen dissipation at maximum power output. If the screen current times screen voltage (after the drop across the resistors) is approaching the dissipation rating, then it's worthwhile to raise the screen resistors.

Any other recommendation I might make absent this would only be blind guessing.

[HotBluePlates]

Another way to look at power and the output transformer, which I hope will turn on more light bulbs:

There are only so many commercially-available output transformers, and the manufacturer almost also states the rated power handling of the device. So let's look at designing an output stage by starting with the OT instead of any particular tube type.

Let's imagine I have a 10w output transformer with an 8kΩ plate-to-plate primary and an 8Ω secondary. How much supply voltage might I need?

10w at the secondary implies a voltage:
Power = Volts² / Impedance, so
Volts = √(Power * Impedance), so
Volts = √(10w*8Ω) = 8.94v RMS or 12.64v Peak.

A transformer's impedance ratio is the square of the turns (and voltage) ratio:
8kΩ:8Ω = 1000:1 = Impedance ratio, so
√1000 = 31.62:1 = Voltage ratio

The required peak secondary voltage is stepped down from a larger voltage swing on the primary. So working the other direction we need to multiply the secondary voltage of our desired power to know the primary voltage swing which will be required.
8.94v RMS * 31.62 = ~283v RMS
12.64v Peak * 31.62 = ~400v Peak

We need this total voltage across the whole primary; the voltage swing seen by one side of the output stage will be half this amount. Let's stick with the peak numbers for now.

400v Peak (whole primary) / 2 = 200v peak (half primary).

So we'll need to have a B+ higher than 200v by the amount of any voltage below which the tube can't pull its plate downward. This is around that knee in the 0v gridline we saw earlier, and the plate voltage at that knee is our minimum plate voltage. Let's say we're considering a tube whose minimum plate voltage at the knee is 40v. That means we need 200v peak + 40v = 240v minimum as our supply voltage (plus any extra for self-bias, plus any margin we choose to add).

How much current is required? We could divide whole-primary peak voltage by the whole-primary impedance to find the peak current.

400v Peak / 8kΩ = 50mA Peak.
Power (RMS) = 1/2 * (Peak Volts) * (Peak Current)
Power (RMS) = 1/2 * (400v Peak) * (50mA Peak) = 10 watts RMS

So that's on-the-nose, and defines the requirements of the output stage (and even the power supply) when we shoot for a specific power output. This tells us something about what will be needed from any output tube(s) we choose to use, as what will be required of the output tubes must fall within their ratings. Notice how this process was about basic relationships between Power, Voltage, Current and Impedance and had nothing to do with any tube except where we acknowledged tubes are imperfect and can't pull their plates down to 0v (this was just skipping steps to make the process easier).

Theory says an ideal class A stage is 50% efficient, so if we need 10w of output power I should be able to get that with two 10w push-pull output tubes. In class A, the peak plate current is also equal to the idle plate current, so on half the signal the current rises from idle to 2x idle (peak current in one direction), and drops to zero on the other half of the signal (peak current in the other direction).

So can we add peak voltage across half the primary to the calculated/estimated minimum required plate voltage, idle at a current equal to the calculated peak current, and still stay within the 10w tube's plate dissipation rating?
200v Peak + 40v across tube = 240v plate-to-cathode
240v * 50mA = 12w

Hey, look! We can't stay under 100% dissipation with a 10w tube, but the 12w 6V6 appears quite handy!

Or we could start over and look for an OT with a higher primary impedance (like 10kΩ), which would then have a lower peak current for the same required peak voltage  swing dictated by our chosen output power.

We could go into class AB instead with the above example. However confining the discussion to class A operation forces an easy-to-see decision about idle dissipation, and more easily hammers home how the needed plate voltage swing can be balanced against OT primary impedance to reduce the peak current, which is directly related to what idle current we will use.

[jbefumo]

Been a hectic day -- have cut/pasted all of the info into a Word document and will print out and digest over the next few days as time frees up.  Just a cursory glance, however, has convinced me that there's enough here to do for me downstream of the Phase Inverter what Designing Tube Preamps For Guitar and Bass up to that point.  This is just the first excuse I've had to actually need to understand the power section better, so thanks!


Anyhow, falling back on my customary 'reasoning by analogy' while passing the time in a PET scanner, I think I stumbled on something that was right in front of me all along:



Blackface Super Reverb: 2 X 6L6 --> 4.2k Load --> 2-Ohm secondaries @ 460V == 45Watts
Marshall JTM: 2 X KT66 --> 8k Load --> (whatever secondaries were specified) @ 420V == ~30W
 
MY Blackface Super Reverb: 2 X KT66 -- 8.2k Load (4.2k-->2-Ohm-ohm transformer driving 4-Ohm Load) @ 420V (reduced by 40V due to replacing GZ34 Rectifier with 5U6GB) == .... ~30W ...


Have I not inadvertently stumbled onto essentially the same overall topology as a JTM45?


Granted the analogy is imperfect as the output transformers are different, but in terms of what everything is seeing from its own perspective, basically the same.


As others have pointed out, yes, I'm giving up some power, but not, as i feared, in an unbalanced, unsustainable, or unpredictable configuration.


Still eagerly anticipating further expansion of understanding, but no longer fretting that I'm involved in something shameful ;^)

[HotBluePlates]

Have I not inadvertently stumbled onto essentially the same overall topology as a JTM45?

Perhaps yes. 

[jbefumo]

Hiding in plain sight ...


Or is the metaphor of 1000 monkeys on typewriters ....

Have I not inadvertently stumbled onto essentially the same overall topology as a JTM45?

Perhaps yes.