More Ghetto UL SE - 2xKT88/6U8/6GH8

[foghornleghorn]

Since my single Ghetto UL KT66/5879 amp turned out so well, I have been wandering even further off the range of late......hopefully this one will turn out around 40W instead of 12, with footswitchable overdrive.

I have bought the iron and the chassis, see pic. Also please take a look at my draft schematic, which took me most of today to do as I am not yet up to speed on .sch file manipulation.

This is a work in process and all comments are welcomed.

[sluckey]

How does the negative bias get to the grids of the KT88s?

[foghornleghorn]

Oooops, that .05 cap shouldn't be there! Duh! I took it out, see revised schem.

Also corrected the .01 bootsrap feedback cap from V1B cathode to junction of V1A plate resistors, which should be .1uF.

Also (Jan 2) revised the bootstrap anti-pop relay contact details to show two DC isolating .2 uF caps either side of the relay contacts with a 4M7 to ground between them. And tried to make the crossed but not intersecting wiring clearer.

And revised PS to 325-0-325 HT winding & revised dropping resistor 10K to 6K8.

[jjasilli]

Some comments:  The way the 6U8 circuit is drawn is confusing.  Where wires cross, but don't connect, it helps to use a semi circle (half an arc). 

The FSW circuit looks incomplete.  It seems that one leg should connect at the point between the pentode's split plate resistors; the other leg at the plate. 

Is the first triode supposed to be a cathode follower? It seems to lack a plate resistor and to have no output.

I don't understand the bias circuit which seems to interact both with signal to the 2nd triode and its cathode resistance.

[tubenit]

Is the first triode supposed to be a cathode follower? It seems to lack a plate resistor and to have no output

That is correct. It has a switchable feature of "typical" DC coupled CF ..... OR a bootstrapped version that has a significant increase in gain and in a "crunch" tone.

With respect, Tubenit

[foghornleghorn]

Apologies for the confusion Gents. I haven't figured out how to make it clear that one wire is crossing another but not intersecting it. Still on the .sch learning curve.

First off, the .01 cap on the bootstrap feedback from V1B cathode to junction of V1A plate resistors should be .1, not .01. This is the 6U8.

The bootstrapping feedback is switched to one side of a 2PDT relay. The other side of the relay switches the two MVs accordingly between Clean and OD. So it's like a Dumble Clean/OD but without banging two triodes together.

The tonestacks are a 5F6A Bassman in the clean line and a highly modified Fender TS in the OD, aimed at a fairly middy tone. I may take out the 470K/.01 in each line depending on how things go, but I left them in for now.

V2, the 6GH8A, is a pentode driver feeding a DC CF fixed bias to the KT88 grids and the CF triode grid. The triode has a separate PS node in case it starts drawing KT88 grid current from its cathode when pushed.

The KT88s are in parallel SE but fed from a large PP OT, in UL mode. Note the KT88 screens are fed from the OT primary center tap. B+ in one side, plate feed out the other side of the primary. This is known as Ghetto UL.

There is no global NFB as the UL internal NFB in the KT88s is enough to prevent the output from going nuts when cranked.

That's the theory, anyway.

[jazbo8]

The 50% LFB is plenty to whip the KT88 into shape without the GNFB The proof is in the pudding, check out the soundclips...

[foghornleghorn]

Thanx jazbo8, that worked out OK on my single KT66 build, so I'm figuring it should here too.

I think that the relay switch on the bootstrap PFB line might pop/click when operated, so I discussed that with a pal who builds relay boards and relay power supply boards for Dumble builders. I use his boards, and he came up with this, see below. Just imagine the .02 caps are actually .20 caps.

[jjasilli]

I haven't figured out how to make it clear that one wire is crossing another but not intersecting it.

Some alternatives.  From Jazbo8's post it appears that the FSW is correctly drawn.  It might help to put "NC" (No Connection) on the unused part of the SW, so it doesn't look like a mistake.  

Where leads cross, but don't connect, use a semi-circle across on one lead to show there's no connection.

Where leads cross, but do connect, you can stagger one of the leads so it doesn't go straight across the other lead.  (This is preferable to sing a dot which might get noticed.  Also printers, copiers and the old mimeograph machines might add stray dots anywhere!)

[foghornleghorn]

1. The OT is a 100W Marshall JMP, push pull design of course. Since the screen is fed off the primary's centertap, it's 50% UL, give or take.

2. Made by Magnetic Components in Chicago. Model 40-18026, 1K7 primary. I figure it should be good for a 40W SE guitar amp.

My previous Ghetto UL (GUL) build used a single KT66 and a 15W Hammond 125E OT. I figure it's about 12W and works fine. This 2xKT88 runs at 100V higher plate & screen than that one, and about the same cathode current per tube, so the jury is out on the Marshall OT until I build it and see what it can do.

If it saturates too easily, I have a huge Edcor 100W PP audio OT, or I can get an Edcor real airgapped SE UL OT--yes they do make 'em, 40% tap.

I do have that GEC datasheet but I don't know how what I'm doing to this poor OT translates from that graph. If you can translate that, I'm all ears.


Thanx jj, AFAIK the convention that plain crossed wires do not intersect is still the norm. I just need to go thru and figure out how to put those solid dots in where they do connect.

[jazbo8]

Man, I got this mixed up with your previous ghetto build... I will try and come up with some load lines for this PSE UL build   should be interesting. But I have to say, as the output power (and the idle current) goes up, the un-gapped PP OPT may be pretty stressed, unlike the single KT88 case, where you got away with it, of course, you don't really know for sure until you try...

[foghornleghorn]

That would be great if you could do that, jaz!

The idle current in this OT I'm estimating at about 170mA total, which is twice the KT66 build. Since it would run at about 85% of max. dissipation that will probably drop a bit when the amp is cranked. The KT66 OT weighs about 1 1/2 lbs and is rated 15W. The Marshall OT weighs 7 to 8 lbs and is rated 100W.

The big imponderable is the ~450V DC in the primary, compared with ~350V in the 125E primary.

[jazbo8]

Man, that's a large OPT! Given the size, is it possible that it is gapped? Anyhow, below is the UL plate characteristic of a single KT88 with 4k load line - for 2 KT88 in parallel, you would double the plate current. The Pdiss is set to 35W (represented by the purple dashed line) - perhaps a bit hot?!

[foghornleghorn]

Thanx jaz, that looks OK. I have the plate figured at about 450V rather than 410, which gives Pd (p+g2)= ~38W at ~85 mA, should be OK as 38/46=83%. 46W being max allowed Pd (p+g2) from the datasheet.

Or is that 410V B+ into the OT primary? In which case that should be about 465V according to the PSUD2 sim of my PS. I was figuring about -50V bias at the KT88 grids rather than -45V.

Also the OT primary is 1K7 rather than 2K. Not sure how much difference that might make.

[PRR]

> I don't know how what I'm doing to this poor OT translates from that graph

A pentode can be triode-wired. Gives different performance. Pentode more power and distortion, low damping. Triode, less power and distortion with some damping.

With a tapped transformer, you can get in-between performance.

One extreme, 0%, is Pure Pentode. Power Output about 52 Watts at 5% THD with little damping.

The other extreme, 100%, is Pure Triode. Power Output about 30 Watts at 1.5% THD with good damping.

What is in between? As you get into the 20%-40% range, power output goes UP (but only 6%, not enuff to hear) and THD goes down (from 5% to 2%, a sorta-significant change). 50% tap is so close to the 40% tap that it hardly matters. If you WIND a transformer, you want to aim at 20%-40%; but if you have a 50% tap in-hand, it's not worth crying about.

See attached image. I've hi-lighted the "5K" curves because they show the trends.

Power out:
0% = 52W
20% = 55W
40% = 55W
50% = 52W
75% = 42W
100% = 30W

Distortion
0% = 5%
20% = 2%
40% = 2%
50% = 2%
75% = 2%
100% = 1.5%

Damping
0% = 0.25
20% =0.6
40% = 1.2
50% = 1.2
75% = 1.4
100% = 1.6

Damping: an un-damped speaker is like a hi-strung drum skin. It has a main resonance and a lot of higher resonances. If you connect the speaker to an amplifier with good damping, it's like laying your hand on the drum skin. Hi-Fi guys want the speaker "flavor" highly damped, so they can hear what's on the record. Damping Factor 10 to 1,000. Guitar players use their speaker "flavor" like a violinist uses the spruce sound-board: you don't wrap a violin in rubber and you generally don't want heavy damping on a guitar speaker. Damping Factor 0.1 to about 2.

[foghornleghorn]

Thanx PRR. Especially about damping. I never really understood why audio guys obsess about damping/Zout so much, now I get it, and the other side of that coin too.

Where I was really going with that query was how does what happens with a 5K primary relate to what happens with a 1K7 primary? And in SE as opposed to PP. And what effect, if any, the DC in the primary has. I guess I should have been more specific. Sorry about that.

[HotBluePlates]

I have the plate figured at about 450V rather than 410...

Are you planning on using that Heyboer PT with 360-0-360?

360 * 1.414 = 509v, and diode-drop is only gonna bring that down to 508vdc.

Yeah, you might cheat a sim program with smallish filter caps to give a lower B+, but your schematic says this is a 200mA PT which means it might have even more voltage rise with only the 170mA draw (plus small tubes and some screen current).

If it was my amp, I'd look for 410vdc / 1.414 = 290vac. Hammond has the 272FX and 272HX at 300-0-300, 172mA and 230mA respectively. Alternatively, they have the 270HX at 275-0-275 @ 230mA. That gives 275v * 1.414 = ~388vdc and probably a bit more due to under-loading of the secondary.

If it really did only give you 388vdc, that might play well with the slightly lower (than 2kΩ) OT primary Z, allowing slightly higher idle and peak current.

Anyway, point is I can't ever remember someone saying, "my PT gave me LESS B+ voltage than I wanted." They all seem to show up asking for ways to waste off excess supply voltage, and you don't have cathode bias to help do that in your amp.

Man, that's a large OPT! Given the size, is it possible that it is gapped?

Probably not gapped.

I used a Hammond OT in my Standel build, and it is 8lbs for a 60w transformer. But core size is about how much power at how much bandwidth (mainly, how low do you have to go). That Hammond is rated for the full 60w down to 30Hz, +/-1dB.

And I'm using that in a circuit pushing less than 30w, because the guy who reverse-engineered an original Standel said that was the closest to the core size of the original OT.

[foghornleghorn]

Thanx, I've had one other person question the voltage drop shown by PSUD2, and that was related to my single KT66 SE Ghetto UL build. All I can tell you is the PSUD2 sim I did on that amp came out only 5 to 10V low compared to actual idle voltages, on a 365V C2 sim.

Having said that, it was a cathode biased amp, not fixed bias like this circuit. I modeled both with a fixed current load rather than a resistive load, since they are idling about 85% or so of max cathode draw. I will take a look at resistive load sims.

I went to the 720VCT PT because the lower voltages would draw too much current, and in fixed bias the 200mA supply works, although there's little headroom. KT88s draw much more screen current than the KT66 I'm told.

[HotBluePlates]

Where I was really going with that query was ... what happens ... in SE as opposed to PP. ...

In SE, your only option for output stage class of operation is Class A. Because if your "single end" ever turns off, you hard-clip the output to the speaker.

Where I was really going with that query was how does what happens with a 5K primary relate to what happens with a 1K7 primary? ...

In SE, you can get an operating output stage with willy-nilly selection of B+ and load impedance. But you picked an expensive output tube type, and decided to use a pair of them... If you want to get your money's worth, then you shoot for something approaching maximum clean output power.

If the tube never turns off (never hits, or just touches, 0 current), then for zero distortion it will swing from idle current to 2x idle, and down to zero current, then return to the idle current. Also for zero distortion, the plate voltage change that happen when current moves from idle to 2x idle will be exactly the same as the plate voltage change when current swings in the opposite direction.

For maximum output power, you will also idle the tubes at 100% dissipation, because anything less will simply shift the loadline downward and give less output power.

To make all this work, you have to balance the OT primary Z, B+ voltage, PT current capability and tube characteristics. You've narrowed the playing field for us by selecting an OT (and its primary impedance) before any other consideration.

You have a 1.7kΩ OT to be used with a pair of tubes; that will look like a 3.4kΩ load to a single tube if we plot loadlines on a single tube's plate curves (which is what Jazbo did). Real-deal KT88's have an absolute maximum plate dissipation of 42w; Jazbo used the design maximum of 35w, which also allows wiggle room if you use 6550's or tubes labeled "KT88" but not living up to all its specs.

35w / 410v = ~85mA idle per tube.

If the tube swings upward to double-idle, the change in plate current is the extra 85mA (peak current of 170mA per tube). Ohm's Law time: what voltage swing is required for an increase of 85mA in a 3.4kΩ load impedance? 85mA * 3.4kΩ = 289v peak. Can the tube & B+ support this? 410vdc - 289v = 121v left across the tube, so plenty even in pentode mode.

You'd plot the loadline to help ensure the swing towards 0 current is a similar length on paper as the swing towards double-current. If they are exactly the same, then you have zero distortion. With real tubes, one will be a bit shorter than the other, and how much different they are tells the amount of one type of distortion.

What if you wanted to use a 5kΩ load, as you alluded? Well, you now know that for class A, SE, you need to swing from idle to double-idle current and your selected B+ (and tube dissipation limit) have pegged that idle current at about 85mA. How much voltage swing is needed to push an extra 85mA through 10kΩ? Why 10k? Because I'm guessing you're talking about a 5kΩ primary impedance for the OT, but you have 2 tubes in parallel so plotting loadlines for a single tube requires treating it as a 10kΩ load. 85mA * 10kΩ = 850v peak. Your B+ is only 410v in Jazbo's loadline plot, and you have to leave 40-80v across the tube in addition to plate voltage swing for most output pentodes/beam power tubes (because they can't pull their plate to 0v).

So you'd need over 900v B+... The result is you'd have to raise B+ voltage to get the needed plate voltage swing + operating plate voltage. You'll also need to drop your idle current to hit dissipation limit with the new B+, and re-figure. But that will be easy because you've exceeded the plate voltage rating of the tube. Obviously, such a high load impedance simply won't allow these tubes to pass enough current.

Starting to see the interplay now?

[HotBluePlates]

Thanx, I've had one other person question the voltage drop shown by PSUD2, and that was related to my single KT66 SE Ghetto UL build. All I can tell you is the PSUD2 sim I did on that amp came out only 5 to 10V low compared to actual idle voltages, on a 365V C2 sim.

Different sims, different transformers, different circumstances.

I went to the 720VCT PT because the lower voltages would draw too much current, ... 200mA supply works...

The transformer manufacturer designed that part to deliver 360v when 200mA is being sucked from the winding.

If you draw less than 200mA, the output voltage will rise. Whether it's a little or a lot depends on the regulation of the transformer, and we almost never have that data from the transformer manufacturer.

I'm not anti-sim, but simulations will lie to you, and you have to think ahead about how they will mislead you.

So here's a different way of thinking about it:
You can get any tube, any PT and any OT and get a sound to your speaker. Any preamp circuit will likewise get a sound to the speaker, assuming you solder everything in the right place.

It doesn't really matter if your B+ is 600vdc... you have fixed-bias, and you can dial it down for less current. But all the extra source of slop mean your final amp will under-perform its potential. Kinda like me using a lawnmower engine in my Chevy Tahoe; with the right gearing, it will go but it won't go as well as using the engine in it now.

Now if you're trying to use only parts on-hand, then I get it. Picking parallel SE KT88 doesn't seem like "gettin by with what I got" but let's say you're chained to the PT you have now. The predicted 508v B+ means idle should be trimmed back to ~69mA. The 1.7kΩ OT (which looks like 3.4kΩ per tube) will be under-utilized because you'll only be swinging up to 69mA * 2 = 138mA with a plate voltage swing of 69mA * 3.4kΩ = 234v peak when you have another 220v or so you could be swinging given the supply voltage. If plotted on a loadline, you would see this outlines a rectangle which is smaller than the other operating condition, which means less power output.

I went to the 720VCT PT because the lower voltages would draw too much current...

Too much current how?

In Jazbo's 410v B+ condition, each tube idles at 85mA, for 170mA total. At this point in time, B+ current draw equals idle current, so 170mA (plus few mA's for preamp tubes and screens).

At full tilt, each tube swings up to a peak of 170mA, or 340mA total peak. But they'll also swing down towards 0mA. If distortion is zero, then the average current drawn from the B+ is equal to the average current of the output tubes at the extremes: (340mA + 0mA) / 2 = 170mA.

Distortion won't be zero, but you'll note all data sheet class A conditions have relatively little change of average plate current from zero signal to maximum output. Pencil in 180-185mA average at maximum signal as a guess, add in the small tubes and screens, and you're at about 200mA.

How about that... there must be a reason why transformer manufacturers picked those currents and voltages.  It's just that yours probably assumed cathode-bias operation, in which case you'd drop 45 or 50 volts across that cathode resistor, so the tube would only feel ~463-458vdc.

[foghornleghorn]

OK thanx! I knew that higher Zpri limits power output, other things being equal, thanx for explaining the mechanics of it.

Here's how I got to the 1K7 primary. Based on my PSUD2 C2 B+ of ~450V, I calculated an appropriate Zpri from V squared/Pd (p+g2). 450V squared/46W (from datasheet) = 4400. Divide by 2 for two tubes = 2200. UL "Zpri discount off pure pentode operation" (SWAG) x 0.85 = 1870.
 
Nearest suitable size OT in stock = 100W Marshall 1K7. Not overly scientific I'll admit, but the same method seemed to work OK with the KT66 build, where I ended up with 3K for a single tube.

Edit: Above written before your last post. I understand that PTs are rated for voltage at full rated load, and that less than rated load will drive up the PS node voltages.

I also did the PSUD2 sim again using resistive load of 2K8 (504V/0.180A). VC1 and VC2 were within 5V of the constant current sim of 475/465V. I used regulation of 7% in the sim.

The PT is a BF Super and so is designed for a fixed bias circuit. If I lower the cathode draw the voltage will go up. What I did forget is the Super uses a 5AR4 rectifier so there is a ~20V drop that won't happen in my circuit.

[HotBluePlates]

Here's how I got to the 1K7 primary. Based on my PSUD2 C2 B+ of ~450V, I calculated an appropriate Zpri from V squared/Pd (p+g2). 450V squared/46W (from datasheet) = 4400. Divide by 2 for two tubes = 2200. UL "Zpri discount off pure pentode operation" (SWAG) x 0.85 = 1870.

Now if you stop & think about it, your approach was to assume a level of power and supply voltage, and calculate primary load resistance. This is a good plan.

What if instead of starting with maximum dissipation numbers, you figured an assumed power output. At the top of this thread, you put forth a guesstimate of 40w output, or about 50% efficiency. SE, you'll often only get around 25% or maybe a bit more.

I never did the math on what the expected power output is for the condition we've already been working with:
We landed at 85mA idle per tube, and so a peak swing above that of another 85mA per tube. Total, that's 170mA peak swing, and we're using a 1.7kΩ OT primary. Power is Current² * Impedance, so 170mA² * 1700 = 49.13w.

But 170mA was our peak current swing, so we just calculated peak power output. RMS power output is peak power/2, or 49.13w / 2 = 24.6w RMS. Or, you could convert peak current to RMS current by multiplying it by the square-root of 2, then running through the calculation for power. You'll land at the same number.

I bring that up because rather than use the plate dissipation rating, you might start with an assumed power output (say, 40% efficiency of plate dissipation: 28w), figure a likely available supply voltage based on available power transformers, subtract the minimum voltage that the tube can swing its plate down to (based on a peek at the plate curves), and then plug everything into your formula.

It's also normal if you want to get the best performance from your output stage to have to run through multiple iterations of this process as you start figuring out some supply voltage isn't workable, or current draw is wrong or loading doesn't make sense, etc. Or you steal from operating conditions given in the data sheets  (though this doesn't apply in your case cause GEC never figured someone would "waste" such big tubes on SE operation).

[PRR]

> 5K primary relate to what happens with a 1K7 primary?

Well, but 1.7K with two tubes at HIGH current.

With one tube at modest current, 5K 6k and 7k all favor 20%-40% tap, 50% just a hair less Watts. Two hot tubes will probably be fine at 1.7K and 50% tap.

Also: you clearly are not trying to build a Stadium Buster. You could use 100W iron and 2*KT88 to make 100 Watts, but instead you gone SE at maybe 25W. Who cares if it comes out 28W or 23W?

> And in SE as opposed to PP.

That's an issue. We don't know how much 2nd/3rd distortion happens which will or will-not cancel. Or how much a guitarist cares. I do know that SE UL is well recognized in funky hi-fi circles, so it probably won't stink.

> And what effect, if any, the DC in the primary has.

A big issue. Without proper gapping, you need like 4X the heap of iron, maybe more. You do have 4X the iron (100W vs 25W). It may be fine. If the heavy/costly parts are in-hand, I don't see why not try.
___________________________

> never figured someone would "waste" such big tubes on SE operation

6550 is essentially the same tube.
http://www.mif.pg.gda.pl/homepages/frank/sheets/135/6/6550A.pdf
Bottom of page 2 shows SE mode. 400V and 3K gives 20 Watts at high THD. 3.4K (1.7K/2 tubes) is very near, just suggests like 6% more B+ and 6% less Ib.

And again: if SE is just heavy and hot for what you get, move a plate lead and add in a cathodyne, turn to top of page 3, you get 60-77 Watts with esentially the same iron and glass. (KT88 may need slightly different G1 bias to hit these idle currents, but any new-made tube is liable to need different G1 bias than shown on Golden-Age data sheets.)

[HotBluePlates]

> never figured someone would "waste" such big tubes on SE operation

6550 is essentially the same tube. ...
Bottom of page 2 shows SE mode. 400V and 3K gives 20 Watts at high THD. 3.4K (1.7K/2 tubes) is very near, just suggests like 6% more B+ and 6% less Ib.

They did something else tricky that I just clued-in on for the first time:
Plate voltage is 400v, so idle current is close to the 85mA we were looking at. But screen voltage is very much lower, so G1 bias gets reduced to make the bring idle current back up. The net result is the tube gets driven with a small ~16v peak signal instead of a big 45-50v peak signal, lightening the load on the preamp/driver.

Now I see why Langford-Smith mentioned in an AWA publication that you get more gain from pentodes when screen voltage is low. Yes, you can't get as-big peak currents as high G2 voltage, but we don't need big peaks in SE.

Too bad UL operation completely rules out taking advantage of this feature...

[jazbo8]

Here is the revised load line using your design targets:



Please note that the Ik (left blue dot) greatly exceeds the maximum rating of the tube (175mA) at maximum output, so use with caution! I also want to re-iterate what HPB said, it may be hard to get ~40W output with 2 KT88s in SEUL and the OPT is under-utilized (assuming it can handle the idle current).

[foghornleghorn]

Thank you all for your input. I didn't stay up real late last night, last time I did that was for the new century!

In general my aim with this amp is to get more power than the single KT66 build I did, but with the same pentodey/Ghetto UL SE topography that worked so well in the single KT66 build. I did add a couple of wrinkles (TV tubes with one pentode bootstrapped, and footswitchable clean/overdrive).

The output power is subordinate to the tone, but I would like more girth than the 10/12W single KT66. Like the clear difference between a 50W PP and a 100W PP amp. My regular speakers I use with these amps is two 2x12s in parallel closed back cabs, each cab runs a pair of '77 Celestion blackbacks with original 55Hz cones. Very nice tone, a bit like the G12-65, and a decent soundstage even with a low powered amp.

This difference in available voltage between what the sim software PSUD2 is telling me and what I am hearing here puzzled me, so I took another look at it without using PSUD2.
Let's assume for the sake of argument that with a 180mA current draw I will get 500V directly off the rectifier diodes. The choke with 65 ohms DCR will drop  65 x 0.18 = 12V. The OT primary DCR is 74 ohms, to the plate it drops a further 74 x 0.18 = 13V. So this drops the 500V at the diodes to 475V. So if this holds up I'm 25V high of where I thought I'd be at the plate.

As it happens I have a different 200mA choke, a Triad C-14X, that is the same size as the 65 ohm choke and fits the chassis perfectly. It is 6H not 10H and its DCR is higher at 154 ohms. So if I plug in the 154-65 = 89 ohms additional DCR I get a further drop of 89x 0.18 = 16V. That would get me down to 459V plate. Other than increasing the ripple noise a bit, does this analysis work? And if so, why does PSUD2 show a significantly lower RMS plate voltage?

Alternatively I can stay with the 10H choke and reduce the cathode current draw by adjusting the bias. So I'd have higher plate/screen voltages at lower current. What would that do to the audio power figure? Would it increase? I'm not up to speed on on tube operating point efficiency calculations.

[jjasilli]

Now I see why Langford-Smith mentioned in an AWA publication that you get more gain from pentodes when screen voltage is low. Yes, you can't get as-big peak currents as high G2 voltage, but we don't need big peaks in SE.

Too bad UL operation completely rules out taking advantage of this feature...

There is another, similar way, which Stromberg Carlson used on its 2X KT-88 amps.  There's another type of winding, the name of which escapes me.  Instead of a UL tap off the primary winding, the screen supply has a separate winding which is overwound on the primary winding.  Induction eliminates the need for a choke for the screen, but the screen supply is separately filtered.  This enabled Stromberg Carlson to have a PP 100W power amp with 600V plates, 300V screens.  The trick is to get that wide difference in plate - screen voltages.

Foghorn may need a custom PT.  Maybe from Edcor?  Either a 30% - 50% UL tap,  or an overwound screen supply like Stromberg Carlson's.  With 400V plate, 225V, screen a single 6550 (or KT88) puts out 20W.  That might be enough. Or 2 should do 40W less circuit inefficiencies.  Interesting design concept, but for practical purposes the gapped SE PT would be huge and very heavy.

[PRR]

Cap-input power supplies are VERY sensitive to transformer resistance.

For an unrealistic 1 ohm resistance, PSUD gives me 490V. With a somewhat high 90 ohm resistance I get 412V.

If the PT is in hand, you might measure the DCR of primary and secondary and put them in PSUD.

Expect discrepancy.

I used a 1.7K resistive load. With a 1.7K OT and a good pentode, you *will* want to adjust current and voltage to something in that area. Meaning the power stage "acts like" a 1.7K resistor when optimized.

490V/1.7K= 288mA, 141 Watts total Pdiss (unhappy)
412V/1.7K= 242mA, 100 Watts total Pdiss (unhappy)
373V/1.7K= 219mA, 82 Watts total Pdiss (at the limit)

But hmmm. Jazbo8 is computing for 450V supply. And turning-down idle current to avoid meltdown. If you plot the peak Plate swing from this idle point, it is 425V down and 285V up. That's 1.5:1 different. But in audio we want both sides to swing the "same". Say within 20%, which runs around 10% 2nd harmonic.

Consider the bias in my attached scribble-over. Ebb= 380V, Rl= 3.4k, Ib= 110mA (42W Pdiss) gives 360V down, 350V up. Suspiciously symmetrical for an SE tube. The flip-side is that the drive is 40V up 60V down. We gonna have 1.5:1 asymmetery one way or the other. I suspect the "optimum" is split-the-difference. 415V supply, 100mA Ib, -45V G1 bias.

Another point: You may want to try 40V 45V 50V bias. But if you raise the current at the same B+, the tube will cook. It is perhaps better to start with a "saggy" supply, so as Ib increases the B+ drops. That is wasteful. But it does mean a "too-high" raw B+ is "good" because you can insert resistance for sag.

We have suggestions of (two tubes):
450V 186mA
373V 224mA

We could keep the same Pdiss in both cases if the power supply was 2,025 Ohms from 827V.(!) The 2K resistor would dissipate roughly as much as the tubes, wasteful! The 800+V is awkward for filter caps (and chokes!).

However it does make sense to start fairly high, like 475V, and put about 300 Ohms (20+W!) in series (perhaps in series-with, or in-place-of, the choke). Then small changes of idle current will sag B+ in a direction to limit Pdiss.

________________________________________________

> Ik (left blue dot) greatly exceeds the maximum rating of the tube (175mA)

The 175mA number is read with a DC ammeter. Other sheets use the phrase "averaged over a complete audio cycle". A few dozen milliSeconds of >175mA is balanced with time under 175mA (and going near zero). For audio you compare 175mA to the idle current. Your plan is way-safe.

[foghornleghorn]

Thanx again folks.
jj, I thank you but I don't want to do a custom PT at this point.

PRR, I had previously put the PT pri and sec DCR into my PSUD2 sim, but I went back and measured them again. Corrected for my DMM residual probe resistance the pri is 1.5 ohms and the sec is 74 ohms. I put these figures in and assumed 7% regulation on 360-0-360. I get VC1 = 471V, VC2 (after the choke) = 459V. VC2 is the B+ into the OT primary. Using a 180mA constant current load.

When I switch to a 1.7K resistive load I get VC1 = 452V and VC2 = 435V. I R1 = 255mA, too high. 127mA/tube. So I see where adding a series resistance to knock down the plate voltage a bit would help.

The OT primary DCR is 33 ohms. (In my earlier non-PSUD analysis I mistakenly used the PT sec value of 74 ohms). So VC2, now at 435V, goes in one side of the OT primary and out the other side, during which it drops a bit more before it hits the KT88 plates.

I'm having trouble relating this to Jazbo8's loadline, but I imagine that is because the PSUD2 numbers are somewhat bogus for whatever reason? If so, how can I figure out what my PS is going to give me in actual plate voltage and what output power is reasonable to expect?
I'm puzzled as I went to higher voltage and fixed bias in the first place so I could dial the current down, and keep it within the PT's capability to supply but without frying the tubes. Should I consider going back to cathode bias??

[jjasilli]

The OT is a 100W Marshall JMP, push pull design of course. Since the screen is fed off the primary's centertap, it's 50% UL, give or take.
Sorry I missed this before, when I advocated a 50% UL tap.  Maybe use a bucking tranny, even if you have to mount it on a supplemental chassis.  A 12VAC bucking tranny should drop about 50 VDC in this B+ supply.  If you need more drop, then a series resistor as posted above should get you to about 400VDC plate volts.  With about half that on the screens from the UL tap, I think you make Bingo.

[foghornleghorn]

Not sure I follow you on the screen voltage there. In my previous SE UL build, I had B+ into one side of the OT primary about 365V, 358V on the screen from the OT CT, and 350V on the other side of the primary out to the plate. At idle the screen is higher than the plate.

[jjasilli]

I was thinking that a 50% UL tap would send half the plate voltage to the screen.

[foghornleghorn]

I just got confirmation that the PT is 325-0-325 @ 200mA, not 360-0-360. That makes life easier I think.

http://www.tubesandmore.com/products/P-TF22798

[jazbo8]

I just got confirmation that the PT is 325-0-325 @ 200mA, not 360-0-360. That makes life easier I think.

Better double check, did you plug it in and measure the actual un-loaded voltage? Here is the spec-sheet from Weber/Heyboar:

Heyboar PT

As for the output power, you can estimate it by Ikmax^2*RL/8 x no. of tubes, so in the above example, 0.21^2*3400/8*2 = 37.5W (assuming you are comfortable with the power dissipation and exceeding the Ikmax at times).

[foghornleghorn]

I will check the unloaded voltage, but that Weber PT is not the one I have as the wiring is different. The one I have is the exact same as the AES one I posted. Maybe the Weber one is rated unloaded.

[jazbo8]

I see, I would double check just to be safe. While you are at it, I think it's also a good idea to test the OPT to see if it can handle the rather high idle DC current say 200mA.

[foghornleghorn]

The OT is a 100W Marshall, so it should be able to handle maybe 300 mA when cranked. If it can handle four 6550s it should be able to handle two KT88s at the same Zpri.

I know it handles less in idle with 4 tubes in PP than in my circuit with two tubes, and a 100W Marshall probably puts out maybe only 80W, but those 4 tubes surely have to draw 75mA each or more when cranked, assuming 450V plate/screen, ie a model 2203.

[jazbo8]

The OT is a 100W Marshall, so it should be able to handle maybe 300 mA when cranked. If it can handle four 6550s it should be able to handle two KT88s at the same Zpri.

I know it handles less in idle with 4 tubes in PP than in my circuit with two tubes, and a 100W Marshall probably puts out maybe only 80W, but those 4 tubes surely have to draw 75mA each or more when cranked, assuming 450V plate/screen, ie a model 2203.

The test is for DC current on the un-gapped OPT, just to make sure that it does not overheat with all that idle current flowing through it...

[foghornleghorn]

Sorry I guess I didn't see "DC". So...0.20 ^x 33 ohms OT DCR = 1.3W. Am I missing something here?

[HotBluePlates]

Am I missing something here?

The difference between a SE OT and a push-pull OT: unbalanced d.c. and an air-gap.

FWIW, I don't think anything will burn up with the idle current Jazbo is asking about. Rather, the issue is whether you'll saturate the core with that much unbalanced d.c. And probably no one will know that until you build the thing and see if bass or powe output is degraded.

[PRR]

> ]I was thinking that a 50% UL tap would send half the plate voltage to the screen.

Half the AC/Audio voltage, yes.

The DC Screen voltage will be same-as Plate DC voltage. (Couple volts higher even, due to half the winding resistance.)

That's how UltraLinear works. It's OK.
____________________________________________

HBP> But screen voltage is very much lower, so G1 bias gets reduced to make the bring idle current back up. ....Too bad UL operation completely rules out taking advantage of this feature...

As said, there's tricky ways.

But this is a guitar amp. I'm thinking we LIKE high G2 voltages, so we don't *have* to drive G1 all the way to zero to get full clean output, and are further away from grid-blocking when we exceed full clean output. The drawback is more gain and level from the driver, but that's often a non-issue.
____________________________________________

> Should I consider going back to cathode bias??

Since it lookes like you *may* have "too much" B+, and need to lose some, and because Self-bias is VERY much safer than fix-bias, I would think you should. At least don't weld pins 8 hard to the chassis, leave a way to wire-in cathode resistors.

I would have said it sooner but this aint no Champ and total power consumption IS an issue. The extra 5%-10% heat from cathode bias may matter. But since you can't get a "right-on" set of iron with "perfect" voltages currents and impedance, you gonna have some shimming to do, resistors lose voltage, and can also be good bias.

[jazbo8]

Am I missing something here?

The difference between a SE OT and a push-pull OT: unbalanced d.c. and an air-gap.

FWIW, I don't think anything will burn up with the idle current Jazbo is asking about. Rather, the issue is whether you'll saturate the core with that much unbalanced d.c. And probably no one will know that until you build the thing and see if bass or powe output is degraded.

I don't think it will burn up either but it may get warm, I guess we'll have to wait and see then...

[foghornleghorn]

Thanx again folks. PRR, did you notice that the PT I have now appears to be 325-0-325, not the 360-0-360 that I had assumed it was (because the BF AB 763 Super Reverb schem shows 360-0-360). So instead of 504V DC unloaded that figure is now 325 x 1.4 - 455V. ~50V less. And PSUD2 is now showing 393V on VC2 (B+). Which FWIW is close to your 380V on jazbo8's loadline sheet, and even closer to your 410V compromise. Unless I'm somehow misreading this lower PT voltage it looks like I am well within the ballpark of this loadline.

I'd like to stay with fixed bias as that allows me to use the DC CF driver coupling to the KT88 grids so potential blocking distortion is not an issue, and bass is better.

[HotBluePlates]

I'd like to stay with fixed bias as that allows me to use the DC CF driver coupling to the KT88 grids so potential blocking distortion is not an issue, and bass is better.

Maybe high B+ and cathode bias is a non-issue now...

However, pretend your -175v bias supply was only -125v. Same voltage drop across the driver cathode load puts the cathode (and KT88 grid) at "0v" even though it's not directly connected to ground anywhere.

So you could cathode bias and still use your d.c.-coupled driver. Or you could split the difference and have a smaller cathode resistor (for pat of the total bias voltage) and a negative grid voltage somewhere between 0v and -50v. In other words, a combination of fixed and cathode bias. Primary reason to use it in your case would be you want to waste a bit of B+ voltage. You could always add this to your amp after-the-fact, and simply dial the bias adjust for less-negative voltage at the driver cathode.

[foghornleghorn]

If the PT spec is correct (370V @ 200mA) I'm a little bit light on voltage, rather than way high like I thought I was. However this is a Heyboer built piece and they build in some extra in my experience, same as Hammond does. So it looks to me that I'll be pretty close now.

I'm not quite sure what the bias voltages are going to come out as. The numbers in my schem are a duplicate of my Steel String Singer, which is a 100W PP 4x6L6GC and runs about 440V on the plates/screens. This build now looks like about -40V rather than -50V.

I think I will just have to put it together and see where it settles!

[foghornleghorn]

I'm wondering if I should put a triode stage in front of the 6U8 first stage, in order to ensure the 6U8 pentode's grid can be overdriven. I have plenty of gain into the tonestacks already, but the 6U8 pentode itself is only being driven by the guitar's pickups at present and I'm not sure that is enough oomph to get some worthwhile preamp distortion.

Power amp distortion is not a problem as there is lots of voltage available into the 6G8's pentode.

My SE KT66 had a second stage triode after the 5879 so that preamp had lots of overdrive when I wanted it.

[jjasilli]

I've been toying with the same design issue.  Gets complicated.  There are 2 pentode stages to possibly overdrive.  The addition of a double triode tube then makes it possible to overdrive triodes, pentodes and/or pentodes in triode mode.  Where to put the tonestack can become another issue. 

[foghornleghorn]

Yeah, I'm still trying to figure out a revised topology for it. And six pots for two tonestacks is getting a bit complex. Maybe disconnecting the triode in clean mode, reconnecting in OD mode. Then maybe put the 6U8 bootstrap switch on manual selection.

Have to put my thinking cap on!

[foghornleghorn]

Here's the next step with a 5751 in front of the 6U8 and other changes too. See revised schem attached.

I am wondering if I will have to defeat the TS ground connections with a relay switch to cut one out when the other is operating; otherwise are they not blending with each other?