2xKT88 in UL Twin?

[pdf64]

There are a few of these 135 watt Twin Reverbs in my area and I think they sound great.

I'm not sure if Mark Knopfler is using that model in the Old Grey Whistle Test session but there's a fair likelihood of it Dire Straits - Sultans of Swing - Old Grey Whistle Test - YouTube (I hope to find a version of that clip that has better sound quality). Whatever, the guitar tone on the broadcast version is great.

Mods I tend to undertake are to increase the screen grid resistors to 1k 2watts MO type, open the power amp global negative feedback loop, remove the 120pF treble peaking cap on the master volume and being export types, set the line voltage selector to 260V (in a 240V region), which tends to reduce the static VB+ below 500V.
See http://www.webphix.com/schematic%20h..._135_schem.pdf

Other good mitigation against overdriving the power amp (and thereby stressing the 6L6 screen grids) may be to fit more efficient speakers, as the original ones seem to be a weak link.
Pete

[jazbo8]

And, of course, you can't mention UL/DL in guitar amps without bringing up Dr Z, who's had some success in making UL guitar amps. Though now I really wonder where he specs his screen taps...

40% on the Route 66, as it should... have you decided on what you going to do the amp?

[potatofarmer]

40% on the Route 66, as it should... have you decided on what you going to do the amp?

Has anyone measured that? I wonder if he's tried other tapping points as pretty much the only boo-teek builder out there looking into UL operation...

I'm still leaning towards the pair of KT88s.

Now I'm looking for information about mounting inverted KT88s vertically, both w.r.t. heat and mechanical stability. Those spring retainers are manufactured for a reason, but I can't find much discussion of this issue. The heat is going to be about the same - both plate dissipation and waste heat from, well, the heaters - but it's going to be more concentrated than if it were spread out over a quad of tubes. Then again, will the increased distance from tube to tube compensate enough for that?

In one of his books, Morgan Jones brings up the idea of attaching heatsinks to envelopes, but I'm suspicious of getting good thermal transfer from glass to metal. I suppose I could just refrain from using the tilt-back legs. I'll probably just try it out and reevaluate after I decide if the tube swap is worth keeping; at that point I'll decide on adding a fan.

[Alan0354]

In one of his books, Morgan Jones brings up the idea of attaching heatsinks to envelopes, but I'm suspicious of getting good thermal transfer from glass to metal. I suppose I could just refrain from using the tilt-back legs. I'll probably just try it out and reevaluate after I decide if the tube swap is worth keeping; at that point I'll decide on adding a fan.

Can you put heat sink compound to get better heat conduction? Put a fan?!! I bought a Bassman 100 used that had a fan added!!!

[potatofarmer]

Can you put heat sink compound to get better heat conduction? Put a fan?!! I bought a Bassman 100 used that had a fan added!!!

Yeah, a fan is a better idea.


A Fender Twin Reverb better "Polished Turd" modification used by Rich Koerner of Time Electronics

This guy put a quad of 6550s in an "UL" Twin and apparently wasn't too concerned with heat, or spring retainers. Or screen voltage/dissipation; looks like he took the screen node off the OT center tap. Is that new filter node grounded to the death cap?! And he moved the power cord precariously close to that one tube. Still, it's a good picture of what a quad of big-bottles would look like, at least. Looks crazily crowded to me.

[bob p]

The Sovtek 6550-WE is a good choice. I have 20 of them in service here ... a dozen of them between my two SVT, and another 8 in my VT-100. Can't complain about them.

I think that adding a fan, and perhaps rotating the tube sockets so that the tubes direct heat toward the outside of the amp, would be good ideas.

There are a couple of things about that page that bother me though ... like the image that shows parts soldered in mid-air, and the "half-power" switch that doesn't shut off the heaters.

On my Super Twin I've modded the heater supply to allow me to shut off any 2 of 3 tube pairs, running the amp at 60W. For load matching I couple this with a switch in the cabinet that changes the wiring between series and parallel. This isn't really for volume reduction, it's more for cutting down the heat that's generated when the amp is used for practice.

[bob p]

Forgot to mention -- I'm seeing B+ around +530 VDC. IMO that's just cruel and unusual punishment for a 6L6.

[potatofarmer]

Jazbo8 posted a link to this site in a thread over at diyaudio (in japanese, but google translate does a fair job) 1 UL?????????

On that page there's a formula for determining the screen voltage at a given plate voltage:

Eg2 = Ep0 + Bsg(Ep - Ep0)

where Bsg (actually Beta sg) is the percentage of the turns at which the screen taps are taken. I verified this formula with the ultralinear anode characteristics plot in the GEC KT88 datasheet at a quiescent plate voltage of 425.

Basically, with 40% taps, when the plate reaches the hypothetical Va=0V, the screen has dropped 170V from its idle voltage.
With 12.5% taps though, it only drops 53V.

So, sure, the screen voltage is still tracking the plate voltage via autoformer action, but not nearly as much as it would, were the screens connected to a proper ultralinear tap.

At an idle voltage of 510V, the situation is just as bad; with 12.5% taps the screen drops ~64V from idle, while at 40% taps they'd drop 204V.

So I'm getting the gut feeling that the screen dissipation in this amp is roughly between a regular pentode-connected Twin with 510V on the plates and one with 460V on the plates, in which case KT88s might not be the best idea. Time to keep reading...

[potatofarmer]

This graph from the Langford-Smith articles is particularly interesting:



Again, the x-axis is labeled in percentage of the total impedance, which puts our 12.5% tap at 1.56%. I eyeballed the general neighborhood, and it looks like even at this tiny percentage we've reduced screen current by ~20% compared to pentode mode. So that's definitely something.

So now I'm starting to think that since the screen voltage drops, and the screen current drops, with a modest screen stopper the plain-jane 6L6GC might be fine in this circuit.

Alright, next task is to try and measure the screen dissipation using the procedure mentioned in that RCA paper.

[potatofarmer]

Alright, next task is to try and measure the screen dissipation using the procedure mentioned in that RCA paper.

Finally took some measurements.

I pulled two tubes because my dummy load is only 100W.

Idle conditions:
Va = 487V
Ia = 29.9 mA (49% idle dissipation; a little cold, sure)

at full clean output: 24.1V into 10.1 ohm resistor for 57.5W

So for the screens we have, at full output:

27.3Vac across 1.991k screen resistor = It = .0137 A
DC ammeter in series with screen resistor = .00892 A

And the equation for screen dissipation in distributed load connection is:
P = EI - ei

where e = (1/2)(Ns/Np)sqrt(RlPo)
and i = sqrt(It^2 - I^2)

Sooooo:

e = (1/2)(.125)sqrt(10.1R * 57.5W) = 1.506 V
i = sqrt(.0137^2 - .00892^2) = .0104 A

P = (487 V)(.00912 A) - (1.506 V)(.0104 A) = 4.43 W


...which is a whole pants load different from (487V)(.00912A) = 4.44 W

But considering the maximum dissipation for a 6L6GC screen is 5W, I'd say with 2k screen stoppers we're in the clear. Unless, of course, I'm doing things horrendously wrong. It would be interesting to repeat this with different values of screen stopper; I kind of doubt that I lucked into such a good value of screen stopper.

In lieu of a thermocouple ammeter, I measured the AC voltage across the screen stopper. This is probably a significant source of error. The result does seem reasonable, when comparing how the screen voltage swings at a 12.5% tap instead of a 40% tap. I was expecting to see a minor reduction in screen dissipation due to power being dissipated in the load, and this is what I found, but I guess I was hoping it would be a little more.

On top of that, my plate voltage is lower than what I've measured previously, and the bias is on the cold side.

I do have a new matched quad of Ruby 6L6GCMSTR tubes, so I'll probably put those in, warm up the bias some, and call it a day.

[jazbo8]

So the above measurements were taken with a pair of 6L6s and the dummy load on the "8 Ohm" tap?

[potatofarmer]

So the above measurements were taken with a pair of 6L6s and the dummy load on the "8 Ohm" tap?

Really old, beat up 6L6s, yeah. The dummy load was on the 4 ohm tap to approximate the Ra-a that a quad would see.

[jazbo8]

Really old, beat up 6L6s, yeah. The dummy load was on the 4 ohm tap to approximate the Ra-a that a quad would see.

So Ra-a = 10 x 12.5^2 = 1.56k, isn't that a tad low for a quad, or is that just what Fender used?

[potatofarmer]

So Ra-a = 10 x 12.5^2 = 1.56k, isn't that a tad low for a quad, or is that just what Fender used?

The OT, or rather the plate taps of the OT, are 2.26k:4 ohm (or 8 ohm). IIRC a little higher than the 2k traditionally used by Fender in BF/SF amps for a quad of 6L6s.

The "0.125" I used came from Ns/Np, or the turns ratio of the screen taps to the plate taps. I guess I did introduce a little rounding error there; I should've used the voltages I measured instead (21.47V / 171.1V = 0.12548...) but the error/uncertainty is dominated by the direct and indirect current measurements anyway.

I just realized I typed "12.5" in post #40 but I did use the correct ".125" in my actual calculations.

But yeah, the screens' contribution to the output is quite low. I imagine it would increase with a smaller Rg2, but then so would the DC current going through the screen.

Argh, I'm going to have to do this again with the stock screen resistors. Hopefully on a day when the plate voltage is more in the 510 ballpark, though if it really wants to stay around 480 I won't complain.

[jazbo8]

Fender Twin UL Load Lines

But your 6L6s do not seem "tired", they are pumping out more than 58W continuous sine wave , based on Ra-a of 2.26k (for a quad), with Ebb=Eg2=500V, and Eg1 = -58V taken from the Fender schematic. The screen grid dissipation WITHOUT screen resistor is 5.5W using the equation from RDH4, which is:

Idg2 = Eg2 x (Ig2 max/4 + Ig2 idle/2) = 500 x (0.04/4 + 0.002/2) ~ 5.5W

Here is the UL characteristic chart (one side, one pair) with the load lines, the composite load line is shown in green and the Class B load line (Ra-a = 4.5k for a pair) is shown in orange, as you can see the Po is mostly Class B given the cold bias used. The model that I used may not match the tubes that you have, so you should allow for +-10% error at least, but it's close enough for rock 'n roll as they say...