OT Plate Load

[Wilder Amplification]

UL is for one reason: I'm currently using an OT that can do it (Hammond 1650T). At the very least, I feel I should try it. From what I have read from discussion on this forum, screen resistors are a smart thing to use in UL mode, and those would keep the screens from drawing excessive current. If you don't mind my asking, what about your design keeps the power amp from going into clipping? Is it the fixed screen voltage?

Joey, if you have the pictures of that UL amp, I'd be interested in looking at them. I can't imagine there's too much to it, but I've been wrong before.

It's not that it won't clip...any power amp will clip when pushed hard enough. It's the sag/compression effect at high volumes that is common to UL and sliding screen operation that we're trying to avoid here.

The fixed screen voltage has a lot to do with this. When a tube is hit with signal, it draws more plate current as well as screen current. On sliding screen operation, the screen voltage is much closer to the plate voltage and will draw more current than in fixed screen operation as a result, which overtaxes the screens and is the main cause of why tubes don't last very long in most tube guitar amps.

In sliding screen operation, when the screen current increases, the screen resistor causes the screen voltage to drop...or "slide down" (hence the term "sliding screen"), which also drops the gain/transconductance of the tube. This is where sliding screen is very similar to UL in that the rising/falling screen voltage is providing a DC derived negative feedback to the screens, just like UL provides negative feedback to the screens via the OT primary. This results in a compression effect. The effect of this is exaggerated on amps that use low filtering at the screen node since not only is the actual voltage at the screen dropping, but so is the voltage at the node from which it pulls current from, which adds to the screen voltage drop, and hence adds to the sag.

Now, IN THEORY you COULD isntall a screen bypass cap from the screens to ground to hold the screen voltage constant, but the sliding down effect of the screen voltage works to your advantage in that it pulls the grid curves down far enough for them to cross the load line where they should as well as helps to keep screen current from rising any higher than it's already too high value.

With fixed screen voltage, the screen voltage does not drop very much if at all when the tube is hit with signal so as such the gain/transconductance factor remains somewhat constant. With the screen voltage being much lower than the plate voltage, this means less screen current (more current will flow to the plate than the screen due to the much higher voltage at the plate providing for a much easier path), so your screens don't get overtaxed and your tubes will more than likely last a lot longer than they will in sliding screen or UL operation.

By avoiding the sag/compression effect, theoretically this keeps the PA tight right up to the point of clipping.

Reliability is also increased as well since the screens will be running at a much lower voltage. With the screens on their own supply, this allows you to run a much higher plate voltage than you can in UL or sliding screen and also allows you to run into a much higher plate load without screen current increasing.

On top of this, I'm thinking it can handle clipping much better than UL or sliding screen operation since the screen current is much lower than it would be in the other two operation methods. According to a SPICE model simulation of a stock Marshall JCM800 2203 power amp, at max clip bouncing off the rails screen current per tube was up at about 120mA. When changed to a dual rail power amp, at max clip screen current dropped down by more than 1/2 to about 45mA per tube at max clip. I'm sure you've seen the all too familiar "screen glow" that happens inside the plate structure in a tube when it's driven to clip beyond all oblivion, but there should be a lot less of that in this design and I'm sure the tubes will thank me for it.

Supporting this theory is Steve Connor's experience with a dual rail design that he's been running the same tubes in for about 2 years and has yet to have one fail on him.

[Ronsonic]

Yep, it is all about the design of the particular amp in question. I think the difference between "then and now" in respect to impedance matching is that guitar amps are no longer expected to fill an arena on their own and output design reflects this in more conservative and more stable design. Nevermind another 30 years of R&D.

Back in the day a four-tube guitar amp was expected to deliver over 100 clean watts. There were Marshalls and Ampegs that could approach 140 legit watts. Now I get recent designs on the bench where 85 or even less is considered healthy.

Some of that trade off for power is in impedance tolerance and stability. Old Marshalls often become unstable with no load (or for any other excuse they can find, worse than horses) and on the other end of the spectrum the old Fender Champ. Almost commonplace on the repair bench is the Fender Champ that comes into the shop with an 8 Ohm speaker and a burned tube socket. As small as it is, the Champ is another design that was optimized for max output.

My advice would be to go with what the OEM had to say. If the guy who built it says it's tolerant of mismatch, then believe him. If OTOH we've got history of amps melting down, respect that.

[Bruce / Mission Amps]

Sliding Screen?

In sliding screen operation, when the screen current increases, the screen resistor causes the screen voltage to drop...or "slide down" (hence the term "sliding screen"), which also drops the gain/transconductance of the tube. This is where sliding screen is very similar to UL in that the rising/falling screen voltage is providing a DC derived negative feedback to the screens, ...

I've never heard of that "sliding screen" term before. Is that something you made up for this discussion or is it a term used somewhere else in the audio field?
Many many years ago, I used to use a term I made up called... "sticktion" ... because I didn't know what grid conduction or grid blocking was or if it even had a name ... but I still could tell something weird was happening and could see it on my old Heathkit Scope!

[imaradiostar]

"sliding screen" operation sounds like a good description of "too large a screen dropping resistor sounds crappy and sloppy under load."

I don't think it's a form of negative feedback as much as it's a type of compression- screen current goes up, gain and power output go down.

A "dual rail supply" just prevents the screens from drawing excessive current. Good old 6L6's wouldn't make all the power that they make in most guitar amp designs if the screens were at 225 volts while the plates were at 450. By the same token some tubes are at their best this way- KT88's are pretty happy with 700 volts on the plates and 350 on the screens.

I like your description with the JCM800 spice model thing. Now the question remains- have you built an amp with it both ways, comparing the tone and volume output of each? I've never built an amp to switch between the two ideas though I have built designs of each type and found good tones in each.

jamie

[Wilder Amplification]

"sliding screen" operation sounds like a good description of "too large a screen dropping resistor sounds crappy and sloppy under load."

"Sliding screen" came from MerlinB in his article on single ended output stages. I thought the term fit so I kinda copped it from him.

I don't think it's a form of negative feedback as much as it's a type of compression- screen current goes up, gain and power output go down.

As screen current rises/falls at signal frequency, screen voltage does the opposite due to the screen resistor dropping voltage with the increase in screen current. The dropping screen voltage lowers the gain of the tube. Since the screen voltage is theoretically falling/rising at signal frequency and as a result lowering the gain of the tube, this acts like degenerative feedback, much like how fluctuating cathode voltage introduces degenerative feedback on a common cathode preamp stage without a cathode bypass cap.

A "dual rail supply" just prevents the screens from drawing excessive current. Good old 6L6's wouldn't make all the power that they make in most guitar amp designs if the screens were at 225 volts while the plates were at 450. By the same token some tubes are at their best this way- KT88's are pretty happy with 700 volts on the plates and 350 on the screens.

6L6s aren't a dual rail tube. If their plate voltage was rated a bit higher I bet they'd work great. But unfortunately you're limited to 500Va on those. As a result, they will not perform very well with the screens at 1/2 the Va.

It's interesting you bring that up because the only reason why Marshalls even use the single rail supply is because they copied Fender, and at the time they were running 6L6s in the very early days, which work just fine on that supply setup. Their screens are rated the same as the plate so for sliding screen or Ultra Linear they're a perfect candidate. But EL34s with their 800V plate voltage rating and their 425V-450V screen voltage rating, I feel they're much better suited to dual rail operation. Running them in sliding screen your plate voltage is limited by the rating of the screens. Unfortunately Marshall didn't pay much attention to this when they were running some of their amps at 500V+ nor did they pay attention to what plate load they were running at that high a voltage (1.7K Zp-p is WAY too low for EL34s running at 500V unless you have a "not so stiff" supply that can sag enough at max power to allow for that). I think this is a lot of the reason why Marshalls are known for things like taking out OTs, screen resistors, tube sockets, etc etc. I mean I know every amp will do that at some point in its life, but it seems to be almost a given with Marshalls.

I like your description with the JCM800 spice model thing. Now the question remains- have you built an amp with it both ways, comparing the tone and volume output of each? I've never built an amp to switch between the two ideas though I have built designs of each type and found good tones in each.

jamie

All of the amps I've built have been the standard Fender/Marshall 6L6/EL34/6550 output stages set up for sliding screen. But I'm getting ready to build up a dual rail amp and run some tests on it. Hopefully I should be getting the plate tranny in later this week.

It seems a lot of the "mundanes" have been flaming the reliability of current production tubes and have lots of misconceived ideas that they can't handle their rated plate voltage. It's been stated for years "Oh an EL34 is supposed to be rated for 800 volts, but I see them blow all the time at 480 or so". However, what hasn't been taken into account is that the screens on EL34s are your limiting factor there, not the plate. The screens are only rated at about 450V (some are even lower at 425V). The only way to get EL34s to run at their max plate voltage is on a dual rail setup.

Also, as we all know, the screens get taxed bigtime in most guitar amps. One of the benefits of dual rail operation is that your screens won't be getting overtaxed and as such the tubes should last a lot longer than they would in a typical Marshall. This was a big factor in the decision to build up a dual rail model. I figured if I can get 100 watts out of 2 x KT88s and make them last forever, it would save the customer money in retubes in the long run.

[imaradiostar]

It seems a lot of the "mundanes" have been flaming the reliability of current production tubes and have lots of misconceived ideas that they can't handle their rated plate voltage. It's been stated for years "Oh an EL34 is supposed to be rated for 800 volts, but I see them blow all the time at 480 or so". However, what hasn't been taken into account is that the screens on EL34s are your limiting factor there, not the plate. The screens are only rated at about 450V (some are even lower at 425V). The only way to get EL34s to run at their max plate voltage is on a dual rail setup.

Also, as we all know, the screens get taxed bigtime in most guitar amps. One of the benefits of dual rail operation is that your screens won't be getting overtaxed and as such the tubes should last a lot longer than they would in a typical Marshall. This was a big factor in the decision to build up a dual rail model. I figured if I can get 100 watts out of 2 x KT88s and make them last forever, it would save the customer money in retubes in the long run.

Makes total sense and I agree! I'm a full time student and work part time so I can't afford the time or parts to experiment much right now but I've been playing with these concepts and trying similar things.

I had originally planned to use toroids with a bridge rectifier and use the center leg for a "half voltage" tap but I'm a little concerned for the possible offset this could create in the toroid. I've built a few this way and thought they sounded pretty good. For larger amps I'll probably end up with a toroid for the B+ supply and a separate one for the screens and preamp. I still like the idea of a single transformer though.

Another thing I've been playing with that Bruce suggested a while back is running two pairs of diodes from a FWCT transformer and connecting the second set to a choke input filter that runs the screens and preamp. I have a test rig on my bench that'll use either a pair of KT88's or a quad of smaller bottles (probably kt77's or EL34's) with about 525 volts on the plates and the screens at about 350 volts. The choke input supply helps keep the screen voltage relatively constant which I think helps tone.

On my test rig the plate voltage is 525v @ 175ma and the screen supply using a normal old BF Fender choke and a 47uf cap into an 8k resistor made 347 volts- should be about 43ma.

When I increased the "plate load" the plate supply sagged to about 475v @ 600ma output. I loaded the screen supply more heavily at the same time and it only sank to 325 volts into a 4k load- 81 ma! I think it's useful that the screen supply doesn't seem to drop as drastically when loaded. It's also helpful because a transformer that might appear to have too much voltage- say a 800vct transformer making 550ish DC volts suddenly becomes a lot more useful when you can generate 550 and 350 from one PT. I really like that this relationship isn't as drastic as the 50% relationship provided by a bridge/center tap arrangement. Many tubes will play nicely with the approx .9*vsec/1.4*vsec this provides and it'll give a little extra headroom for the preamp and phase inverter.

This could prove useful with 6L6's as well- you'll note both the 6L6 and 7027a datasheets list operation with the screens well below the plates. I think this method should be seriously considered for high power output and longer tube life, especially on amps that already have a choke on board so little modification would be necessary- just move a few wires around and rebias! I'd also wager that this mod and an upward plate impedance mismatch would make an old marshall more reliable while still sounding fabulous and making your pants flap like a good 100 watter should.

jamie

[imaradiostar]

Wow, it seems like a have a way of saying way too much and stopping a thread from moving forward. I'm sorry about that! Hope I didn't distract too much from the original topic.

jamie

[defaced]

Not from where I stand. I just haven't had time to digest, research, comment and ask questions. Like the screen current numbers from his simulation. I'm still trying to figure that out since the data sheet calls out ~15ma as max current for Ig2 (JJ EL34).

I like the choke input idea for the screens. What kind/value of choke did you use for your setup?

I just realized something, the Antek transformers I use have dual secondaries, this would be perfect for a split rail supply. Duh.

[Wilder Amplification]

Not from where I stand. I just haven't had time to digest, research, comment and ask questions. Like the screen current numbers from his simulation. I'm still trying to figure that out since the data sheet calls out ~15ma as max current for Ig2 (JJ EL34).

That's running class A. Average max signal plate/screen current draw = idle plate/screen current draw in class A. Quite a bit different from class AB.

I like the choke input idea for the screens. What kind/value of choke did you use for your setup?

Haven't built mine yet (just got the custom plate tranny in today) but I will be using a 3H choke from MM (MAR100-C).

[imaradiostar]

Good call on the Antek TF's. My last build with one was using the 100 watt 200 volt model. It ran a pair of 6l6's cathode biased with 536v on the plates, 267 on the screens and the cathodes were at about 20 volts with (I think) 500 ohm resistors per cathode. Not a super high power output section but it sounded great. I've been keeping a "lab notebook" so I was able to just look this up- good fun!

I'm using whatever chokes I have laying around. The test mentioned in my previous post was with a choke from a plain old blackface fender twin- I figure they're good for 90ma if the specs are reliable. Even in a distorted 2xkt88 output section long term average screen current + preamp current isn't likely to be above 90ma. Once again- I don't have any completed amps with this type of design but I've played through one on the bench quite a bit and thought it sounded good. Hopefully I'll be able to give specifics soon.

jamie

[tubeswell]

Forgive me for asking this question at the end of this great debate, but where does that leave Zout = Va/(Pa/Va)?

I can see how you get 7k5 from a 12W rated tube with a plate voltage of 300, but I can't see how you get 6k6 from a 2 x 14W rated tubes in PP with a plate voltage of 425. Is there a formula for that, or is it too tricky?

[imaradiostar]

maybe it's part of the practical side of things- if you run a tube with too high a plate voltage you're forced to either drop the screen voltage or reduce the primary z to prevent over-stressing the screens. Perhaps it's about squeezing every last bit of power out of small tubes at the expense of harmonic distortion- I don't really know. I bet someone here does though.

This power supply design (for me anyway) was about trying to use primary impedances closer to the textbook values (much lower screen voltage) rather than the lowish values favored by some guitar amp manufacturers in the 60's. I wanted to see how 6L6's would sound with 500vp and 360vg2 @ 6k6 instead of 450 vp and 445 vg2 at 4k.

I would argue that the difference in sound isn't huge but the tubes are pretty happy with the screens at lower voltages. I'm not sure that I could quantify the differences because I've never tried switching back and forth between the two designs in one amp to compare.

jamie

[Steve Conner]

Tubeswell, where did you get that formula? Maybe it works for triodes in Class A, but more generally, the power output of a tube amplifier has nothing to do with its plate dissipation.

Except insofar as the maximum theoretical efficiency for a Class-B amp, with unclipped sinewave output, negligible tube drop and negligible idle current, is 70%. This figure is derived from the mathematical properties of a sine wave.

So if you want 100W out, then you need to put 100/0.7 = 142W of HT supply in. 100W of this goes to the load and the remaining 42W are dissipated by the plates, so each of the two tubes needs a plate dissipation rating of 21W or more.

In any practical amp, the tube drop and idle current are both significant, and the effect is to lower the efficiency below the theoretical 70%.

I've previously posted the formula I like to use to estimate load impedance:

Impedance is roughly equal to the HT voltage you have available, times 4, then divided by the maximum peak cathode current rating of your tubes.

This also lets you estimate the maximum possible power output: 0.5 times HT voltage divided by peak cathode current.

Note that, if you limit yourself to Class AB1, then the maximum peak current you can get from a tube is set by the screen voltage and screen resistor value. You can read it off from the plate current vs. voltage curve for Vg1 = 0.

So as you lower the screen voltage and/or increase screen resistance, the optimum load impedance gets higher.

[Wilder Amplification]

Tubeswell, where did you get that formula? Maybe it works for triodes in Class A, but more generally, the power output of a tube amplifier has nothing to do with its plate dissipation.

Except insofar as the maximum theoretical efficiency for a Class-B amp, with unclipped sinewave output, negligible tube drop and negligible idle current, is 70%. This figure is derived from the mathematical properties of a sine wave.

So if you want 100W out, then you need to put 100/0.7 = 142W of HT supply in. 100W of this goes to the load and the remaining 42W are dissipated by the plates, so each of the two tubes needs a plate dissipation rating of 21W or more.

In any practical amp, the tube drop and idle current are both significant, and the effect is to lower the efficiency below the theoretical 70%.

I've previously posted the formula I like to use to estimate load impedance:

Impedance is roughly equal to the HT voltage you have available, times 4, then divided by the maximum peak cathode current rating of your tubes.

This also lets you estimate the maximum possible power output: 0.5 times HT voltage divided by peak cathode current.

Note that, if you limit yourself to Class AB1, then the maximum peak current you can get from a tube is set by the screen voltage and screen resistor value. You can read it off from the plate current vs. voltage curve for Vg1 = 0.

So as you lower the screen voltage and/or increase screen resistance, the optimum load impedance gets higher.

Steve, I'm not sure I understand your formula here. Then again I'm also having an issue finding the peak cathode current rating for a KT88. The only rating I can find is Max DC cathode current...which for a single KT88 is 230mA. Is this the spec you go by? If so, do you double this spec when running two valves in Class AB1?

If I double it, the math works for my DRS amp -

(600V B+ x 4) / (0.23 Amps x 2 valves) = 5,217 ohms

Which comes very close to the datasheet spec'ed 5K Zp-p. But I'm not quite sure if I'm carrying out the math correctly so if you could please clarify?

Also, not sure how the 1/2B+ x peak cathode current rating formula works either. I've tried to plug numbers in all sorts of different ways and it just doesn't seem to be working for me.

Now when I calculate the other stuff it all seems to equate...IF I'm doing this correctly so please feel free to correct me if I am -

(2 x B+)^2 / Zp-p = Peak Power

(2 x 600)^2 = 1,440,000

1,440,000 / 5K Zp-p = 288 Watts Peak

288 Watts Peak / 2 = 144 Watts Average into the OT primary load

Or -

B+^2 / Zp-ct = Peak Power

600^2 = 360,000

360,000 / 1250 Ohms Zp-ct = 288 Watts Peak

288 Watts Peak / 2 = 144 Watts Average into the OT primary load

Is any of this correct?

[Steve Conner]

OK, I think I got the power output formula wrong, but I've posted the right one before.

It's basically P=IV or P=V^2/R. But because you're dealing with a sine wave, and V and I are peak values, you have to multiply each one by 0.707 to get the RMS. That's the same as multiplying the product VI by 0.5.

I've never seen peak cathode current ratings for most tubes either. In those old datasheets the DC cathode current rating is understood to be measured by a moving-coil meter or DMM, which gives the average. Again assuming unclipped sine wave output, and a P-P Class-B output with two tubes, the peak cathode current for each tube is (IIRC) about 1.5 times the average cathode current of both tubes combined.

To a certain extent the peak current is under your control as the designer. You choose your desired B+ voltage and output wattage, then calculate the OT impedance and peak current requirement from that, then set the screen voltage to limit the peak current capability of the tubes to that requirement, which is what your dual rail philosophy is all about, not letting any more electrons out to play than you need.