HoSo56 Single Ended KT66, 2x5879, Ghetto UL

[foghornleghorn]

Almost finished, just have to wire up the footswitch to the relay. "Ghetto UL" is using a PP OT as a SE OT with the OT CT feeding the output tube screen, ie about 50% UL.

Second 5879 is the TS recovery/KT66 driver. No idea what it will sound like!

Edit.Posted the most recent pic. Finally broke down and paid for a copy of JPEG Wizard 2 so I don't have that "unregistered" watermark spoiling the view.

[Slimtim]

nice looking work

[TIMBO]

VERY INTERSETING 

[kagliostro]

A pentode as V1 and as V4

Yes, Very Interesting

I would like to hear a sample

K

[jazbo8]

Why is V3's screen bypass cap connected to the cathode instead of ground? With a relatively high UL tap, it would be interesting to hear some clips.

[John]

Can't wait to hear some clips. Nice clean work!

[kagliostro]

The job is really clean

Have you a layout to share ?

K

[foghornleghorn]

Thanx for your interest. The screen cap is grounded thru the cathode on the driver 5879 (V3) just in case the internal feedback in the  UL'd KT66 somehow finds its way back into V3, where it could act as positive feedback and cause oscillation. Just being super careful.

I'll need to revise the layout a bit to properly match the as-built. Ill post it ASAP.

[jazbo8]

No kidding, very careful indeed   And 20H on the screen for additional filtering (oops, it seems the choke is removed in the layout)?

[foghornleghorn]

I tried to put an inductor in place but it ended up sitting right over the tonestack slope resistor, which I also tried to change to 100K but I cannot get either changes to take. Don't know how to work .sch properly. If some kind soul could make those changes it would be as built.


I'm also still wondering if I should ground pins 2 & 6 of the 5879s, as per the RCA datasheet. Some say don't bother as it doesn't have a shield like the EF86, but if so why did RCA advise to ground them if they're not connected to anything?

[tubenit]

I'm also still wondering if I should ground pins 2 & 6 of the 5879s, as per the RCA datasheet

I've quite a few 5879 builds and don't ground those.  I tried grounding them on one build and it made zero difference for me.

You've done a very cool build!  THANKS for sharing it! 

Will look forward to sound clips

With respect, Tubenit

[jazbo8]

Did you happen to record the voltages for the KT66? Like to draw some load lines to see how the 50% UL works... Does 4.5k primary impedance sound about right?

[foghornleghorn]

I haven't got that far yet. I need to put in the AC line, and y'day I discovered I have to change the seven pin DIN jack I installed as the DIN plug on the footswitch cable is five- pin. Duh! So I have to go buy a few five pin jacks, and find/buy the right size strain relief piece for the AC line.

I'm starting with 3K primary impedance. The datasheets show 2K2 for 250V plate & 250V screen, but I'm going in with about 360V plate and 320V screen. Since the OT is a Hammond 125E I can easily change up to 4K2 by switching one wire on the secondary.

[jazbo8]

Just wondering if the prolong use of the OPT in SE mode will saturate the core? I don't think the 125E has an air gap.

[foghornleghorn]

It doesn't have an air gap, at least not as big as a proper SE OT would. PP OTs usually have a small gap to accomodate small amounts of DC from minor DC imbalance in the PP tubes, but nowhere near enough for "proper" SE operation.

So yes it could saturate. I guess that depends on how hard I push it. I've not heard what this sounds like, but I'm told it's easy to spot as it's NOT musical!

I'm not sure what effect on the saturation onset point the effective primary impedance might have, if any? Higher impedance lowers output, so if it becomes a problem I could always increase the pri impedance. Or, keeping it simple, just turn down the volume(!).

[kagliostro]

As far as I can know PP transformers can't have a gap nor big nor small because of the mix of the lamination that composes the transformer

In an SE OT all E are on one side and all I on the opposite side

In a PP OT E and I are mixed

K

[foghornleghorn]

Theoretically yes. In practice they do I think have a very small "effective" airgap.

Lundahl actually specifies it, which I have never seen from any other manufacturer. Lundahl specs 25 microns for PP and 150 to 250 for SE in their 1620 series OTs, for example. Caveat: These are c-core designs so it's easier to control airgap spacing I think.

Case in point, I just recently custom ordered a pair of PP OTs from Edcor for use in a pair of monoblock audio amps each running two 2A3 in parallel SE with shunt feed choke on the 2A3 plates (parafeed). Parafeed takes the B+ and feeds it to the plates thru a choke, then the signal off the plate is cap coupled to the OT, so no DC at all goes thru the OT primary. So the ideal Parafeed OT has 0.00 airgap.

Long story short, I got into this with Edcor and they told me that even though they stack their EI PP cores 1x1, like you show in your pic, they could not guarantee 0.00 airgap. So I said OK, do your best. They were kind enough to waive the $40 custom fee too, and their price was exceptionally good, only 20% of the Lundahls (but not of course C-cores).

[jazbo8]

R.G. has a nice write-up on the subject at MEF, so the saturation is frequency dependent - not much a problem for mid-hi frequencies. Only way to find out is to try - play a bunch of low notes and see what happens.

[foghornleghorn]

Thanx, that's very interesting. So if there is no DC on the transformer, only the bass part of the AC signal could be a potential saturation problem. With my design starting to roll off around 100Hz or so, I should expect no problems there.

However I still have the DC on the transformer too, so the possibility remains then. I'll just have to see what happens when I crank it up!

[kagliostro]

I've often read of satisfied people that used PP transformer for SE amp

the secret seems they use BIG PP transformer, rated for much more W than

the SE amp can deliver

K

[foghornleghorn]

That makes sense, I may be in trouble as the OT is rated at 15W. Still it is a Hammond so maybe my 10-12W or so will be OK. We shall see!

[kagliostro]

May be you are lucky

but what I've seen is something like an SE output of 5W into a PP OT rated for 10W or more

---

As you we are waiting your report

K

[jazbo8]

That makes sense, I may be in trouble as the OT is rated at 15W. Still it is a Hammond so maybe my 10-12W or so will be OK. We shall see!

You might be able to get away with it, I calculated the full output power at ~9W (with 360V supply and -25V bias). With the 125E rated at 15W and F3~150Hz, the low frequency from the amp, is probably even less than 9W, so you should have enough cushion to keep the core saturation in check.

[foghornleghorn]

Very cool! How did you do that?? I would like to get that software.

At 8 ohms out on the 125E I have 3K and 4K2 primaries, so what you're showing is mid-way between those two.

I'm puzzled though by the current indicated, +100mA. Using a 300 ohm cathode resistor and assuming 25V bias, I get 83mA.

Thank you for your work!

[jazbo8]

Very cool! How did you do that?? I would like to get that software.

At 8 ohms out on the 125E I have 3K and 4K2 primaries, so what you're showing is mid-way between those two.

I'm puzzled though by the current indicated, +100mA. Using a 300 ohm cathode resistor and assuming 25V bias, I get 83mA.

Thank you for your work!

You are quite welcome, it was a bit of challenge to get this plot to come out right, but it was worth the effort, since there are not many UL curves for the KT66 floating around (only found a few at audiomatica). Anyway, this is just a simulation and does not consider Rk per se, but obviously if Ik=100mA, then Rk=250 Ohm, in any case the cathode current will be different depending on the actual tube, to you can certainly start with 300 Ohms and adjust it accordingly.

The software I use, is pctube available from the great Ayumi's Lab. You will also need R, the statistical math software. They are not the easiest programs to use, so if you need help, please let me know.

[foghornleghorn]

I downloaded both of those, but I am "computationally challenged" to say the least. Looking at this stuff I feel like I just walked in out of the jungle in a loincloth.

[jazbo8]

No worries, it took me a long time to figure out how to use the programs... Anyway, in my excitement to get the plots done earlier, I neglected the Pdiss, here is the new plot with Eg1=-32.5V, Eg2=360V and RL=4.2k (old plot removed to avoid confusion).

[foghornleghorn]

Thanx again! Could I ask you to run another at 3.0K and 320V screen? You show Eg2 as 360V and it will be lower. With a 5K1 screen resistor and the 20H choke behind it adding another 1575 ohms, at idle screen current of about 7mA the Eg2 will be about 320V.

If I rotate the 4K2 line around the 360V plate point to 3K0 it moves closer to the Pd line but doesn't cross it. What would Po be with 3K0?

[jazbo8]

Thanx again! Could I ask you to run another at 3.0K and 320V screen? You show Eg2 as 360V and it will be lower. With a 5K1 screen resistor and the 20H choke behind it adding another 1575 ohms, at idle screen current of about 7mA the Eg2 will be about 320V.

If I rotate the 4K2 line around the 360V plate point to 3K0 it moves closer to the Pd line but doesn't cross it. What would Po be with 3K0?

The output is a bit lower, although I doubt that you need the 5.1k and the choke, perhaps just a small resistor on the screen will do, since the feedback would be greatly altered by these additional components (actually 50% tap is already too much feedback IMO). In any case, here is the new plot for your reference, the bias has also been changed to -25V to compensate for the change in the screen voltage.

[foghornleghorn]

Thanx very much, I had not thought about the feedback level. Since those parts are already installed, I'll try it as is.

If it's too smooth I'll certainly make some changes, especially as we're down to 7W Po now. I guess dropping the screen 40V more than offset the primary impedance reduction.

[foghornleghorn]

Well, it has some power supply buzz, but it works! And it has attitude in spades!  Also sounds way louder than it has any right to be.  Thru a pair of 2x12s each with old Celestion Blackbacks at 4 ohms total, primary Z is only 1K5(!). I'll unplug one cabinet, that will take it up to 3K at 8 ohms.

All below voltages with 122VAC wall.

The PS hum/buzz is odd. If I touch the red probe of my DMM to the C or D PS nodes, (V2, V3) it completely disappears, with or without the black probe on the chassis. E node (V1) does not do that.
Also the common ground line off the -ves of the C, D & E node caps, which is connected to the ground buss that terminates near the input jack, scratches in the speakers when I run the DMM probe along it, yet shows no DC on the meter.

I have 26.1V on the KT66 cathode so it's drawing 26.1/300=87mA. 356V on plate, 357V on screen. Screen voltage is much higher than it should be. Perhaps there is a ground reference problem here that would account for the buzz.

The driver 5879 (V3) has 303V B+ vs 300 planned. I was looking for 140V on  the plate and have 135V. I was looking for about 65V on the screen and have 81V, so need to increase the 470K screen resistor some. Want to get that down to about 45V, one third of the plate. Although it sounds good as is. 2.35V cathode, 2.1mA, OK.


V2, 12DW7 (in place of 12AX7) B+ 228V vs. 230 planned. 123V plate V2A, 131V on V2B cathode, so 3.4mA, a tad high. V2A cathode 1.6V, 1.1mA, OK.

V1, 5879, B+ is a bit low at 208V, will tweek the dropping resistor. 130V plate vs 141 planned. Screen 45V, tad low vs 56V planned. Cathode exactly right at 0.83V, 1.4mA, OK.

Heaters are at almost 7.2V, too high, I sort of expected that from a 4A supply loaded with 1.9A.

Relay supply is 4.99VDC  off the 5V PT winding. Must have rewired the FSW wrong, not switching.

[jazbo8]

Cool, can't wait to hear the sound clips... For the PS buzz, perhaps you got some cold solder joints - since you can make it disappear by touching the nodes with your probe - try chopstick it real hard and see you can make it disappear completely, if so, heat up those joints again. The heater is definitely too high - got to get it down or you risk shortening the life of the tubes.

[HotBluePlates]

... And 20H on the screen for additional filtering ...

What is the purpose of the 20H choke between the OT and the screen? UL typically has a straight wire in that position.

[foghornleghorn]

Like jazbo8 said, additional filtering. What I didn't expect was a big reduction in the screen current, it's only about 1mA quiescent now. Based on the datasheet I had expected nearer 5 mA.

Does UL reduce screen current over pentode mode or was this the effect of the choke?

jb8, I have got the heaters down to 6.4V by inserting a 0.36 ohm 5W sandbox int the line. Will go over the solder joints tonight.

[sluckey]

Does UL reduce screen current over pentode mode or was this the effect of the choke?

This doesn't exactly answer your question but it may be related...

An inductor (choke) opposes a change in current just as a capacitor opposes a change in voltage.

[foghornleghorn]

I was not thinking too clearly. I know that the choke would oppose a change in current but I was thinking of the supply ripple, not the screen DC itself.

Reading the GEC KT66 datasheet "Circuit Supplement" I find that UL does reduce the screen current (page 3), so this big reduction in quiescent screen current is probably a combination of both the choke and UL operation.

[HotBluePlates]

The choke would do almost nothing to d.c. (which is why we use it in series with power supply current).

My concern would be that it would impact a.c. when you were playing, and perhaps change the way the UL was performing. That said, I have no idea what it may or may not do. I'd try it both ways since you already have the choke mounted.

[kagliostro]

The PP OT + choke ressemble to the OT used in some old radio

like on this link http://www.el34world.com/Forum/index.php?topic=15498.0



but to me seems B+ is connected in a different way in your amp

K

[jazbo8]

The choke would do almost nothing to d.c. (which is why we use it in series with power supply current).

My concern would be that it would impact a.c. when you were playing, and perhaps change the way the UL was performing. That said, I have no idea what it may or may not do. I'd try it both ways since you already have the choke mounted.

I brought up the issue with the choke earlier, I didn't think the choke should be there and in fact I think the circuit is no longer "ultra-linear" per Hafler's design.

[foghornleghorn]

That could be, but if the choke has no significant effect on DC, then it must have some effect on the AC signal? Since both plate & screen are connected to the choke and there is AC signal on both.
 
Is this what you are getting at when you say the choke is interfering with/nullifying the UL process? Presumably because of its inductive effect on the signal.

UL definitely reduces screen current compared with pentode mode, that is clear in the GEC datasheet. And I have about 1mA screen instead of about 5mA.

Still trying to connect the dots here. The amp is definitely overdriving bigtime but I can't really tell whether it is preamp, driver, or power tube or any combination thereof. Will have to play with the gain and master pots some more.

[HotBluePlates]

That could be, but if the choke has no significant effect on DC, then it must have some effect on the AC signal?

Yes, which again is why it is used in power supplies.

A choke opposes a change of current (d.c. is "no change" or steady; a.c. is "constantly changing"). In a power supply, it is used to knock down a.c. hum on the B+ line.

UL operation amounts to feedback:
-  in triode mode, there is no screen (or it is tied to the plate); essentially 100% of plate signal is at the screen.
-  in pentode mode, the screen is pegged to a rock-steady voltage; essentially 0% of plate signal gets to the screen.
-  in UL, some middle-amount of plate signal gets to the screen. This amounts to feedback and brings the plate curves more in-line with triode mode (but not quite).

My concern was that the choke would be nullifying the varying screen voltage, thereby nullifying the UL operation. But you're hearing the amp... so try it one time without the choke in the screen path, and see if/how the sound is different.

[foghornleghorn]

I'll try it and see!

Edit: I did some more calcs and plotted a parallel 4K2 loadline thru 85mA at 360V on jazbo8's simulation graph. It's above his 25W Pd curve but just below a 30W curve. I measured 87mA cathode draw so that's about right, and the operating point remains pretty much center biased.

I also plotted 1K5, where I mistakenly ran the amp Sunday afternoon. Whoops! More than half the loadline was above the 30W Pd curve. But it did sound good! Calcs indicate Po almost 15W flat out, not too kind to my Shuguang KT66.

So I'll probably change the Zpri to 4K2 and put in a switch to go from 4 ohm to 8 ohm cabs. But first I'll just pull one cab so I have 3K to 8 ohms, would like to see how that sounds.

One unrelated question. What rated voltage cathode bias cap for the KT66 would you use, with 26.1V on the cathode? I put in a 50V Sprague TE1304. Yet I see that Merlin (the Amp Wizard) recommends "at least 3 times quiescent" especially if you use a standby switch, on a SE build.

[PRR]

> UL definitely reduces screen current compared with pentode mode, that is clear in the GEC datasheet

At No-Signal, it can't.

With big signal, when Pentode plate goes low, screen current rises. But in UL mode, when plate goes low, screen also drops, screen current rises much less.

With a choke in the UL-tap to screen grid link, no screen cap, I think it becomes a poor triode. Main symptom would be half the power-out you got as pentode. THD might drop, or not. And you might need more than 20H to really decouple the screen.

> I was thinking of the supply ripple

But a ripple-reduction choke is usually teamed with a *capacitor*. Yes, a choke alone will reduce ripple into a resistive load, especially low resistance. But not a whole lot. And not cheap. Since 1928, caps are so cheap that it is "silly" to reduce ripple with a $200 choke when a $10 choke and $5 cap do it 10 times better.

Of course such a filter completely eliminates any UL effect.

Neglecting THD: the optimum load for pentode (and near-optimum for UL) is *always* --

1) Pick Pd as high as you dare.

2) Pick any reasonable B+ (not fussy; the book-number is good).

3) Current is then I = Pd/V.

4) Load is Rl = V/I.

> another at 3.0K and 320V screen?

25 Watt tube?

25/320 = 0.078A = 78mA

320V/78mA = 4.1K load.

Problem: 6L6 (KT66 is the same) is a poor SE tube. THD is very high. This cancels in push-pull, but you aren't interested. THD-optimizing 6L6(KT66) in SE does mean shifting the load impedance until the 3rd harmonic cancels part of the 2nd harmonic for the lowest THD-meter reading. The ear is not fooled: we hear the 2nd and 3rd distinctly, not cancelling, especially on dynamic signal. Still and all, the trend is to use lower impedance, so a 3K load is probably a good ballpark.

[jazbo8]

Would it be possible to get a voltage reading without the screen choke? As is, the plate and screen voltages are quite a bit above the initial design figures, thus my plots are no longer really applicable, will re-do them when I get home later if you are interested in seeing where the load lines sit (just for static, DC condition).

[foghornleghorn]

Thank you PRR. OK, I'll remove the choke. I only put it in because I had it on hand and clearly I didn't think through its effect on the UL setup.

The load is coming out around 4K2 however I figure it, which I can wire up on the 125E OT for both 4 ohm and 8 ohm cabinets.

Since harmonic distortion is what I'm looking for, optimizing as you describe looks worthwhile so I will see if I like 3K better than 4K2.

I'm curious, how does the 3rd harmonic cancel part of the 2nd when you change primary impedance?

Jazbo8, I'll measure voltages as soon as I get the PS ground problem sorted out and post 'em.

[foghornleghorn]

I removed the screen choke, leaving the 5K1 gridstopper in place. Still haven't fixed the hum/buzz, I think I'm going to have to pull the board and redo it.

Anyway, with 122VAC wall, the voltages now are B+ (Node B) 368V, KT66 plate 355V, screen 358V, other end of the screen gridstopper 363V. The screen is 3V higher than the plate, but I have to remind myself this is UL so it's OK. I also changed the final PS dropper to 4K7 (was 12K), this got the voltages on V1 right in line with the schem. 25.7V on the KT66 cathode

Much cleaner thru one cabinet at 8 ohms and 3K primary. Couldn't crank it because it's a bit late now, but I think it will still have some serious attitude when I do. A bit less bass, if anything. Maybe the power tube bias cap should go to 15 or 20uF.

[jazbo8]

Best fix the hum/buzz first, since large ripple on the power supply would influence the voltage readings.

[John]

Still haven't fixed the hum/buzz, I think I'm going to have to pull the board and redo it.

Before you do all that, just try moving the wires away from your heater wires. Even though your heater wires are very nicely twisted, a grid or speaker jack wire a little too close will stick pick up noise.

[PRR]

> how does the 3rd harmonic cancel part of the 2nd when you change primary impedance?

The input-output plot for a perfect amplifier is a straight line. Put in 1V, get 2V; put in 2V, get 20v.

2nd harmonic distortion has a slight C-bend.

3rd harm has a slight S-bend.

It is possible for the extra curvature to "cancel" the simple curvature.

I don't think your ear is as easy to fool as a THD meter.

=============

> screen is 3V higher than the plate, but I have to remind myself

This is always OK in Pentodes. (In true tetrodes, you want G2 *much* lower than plate; but nobody runs a Type 24 anymore, and I dunno how it got to be a Big Deal.)

We usually don't want a separate power supply just for G2, so G2 is usually "similar" to plate supply. Tube makers design audio tubes and suggested conditions to favor this ratio. But "similar" never has to be exact. Sometimes we give G2 more filtering, lose some voltage along the way, and that may be fine. Occasionally the DC drop in the plate's OT exceeds any drop in the G2 feed, the plate is "higher than G2", but not so it cares.

[foghornleghorn]

Thank you all for your responses. I chopsticked a couple of signal wires vs the heaters. No difference. I resoldered the preamp caps ground wire. No change. Still humming, still getting scratching along the bussbar.

So I pulled the relay out of the relay board. Presto change-O! Quiet as a church mouse peeing on a cotton boll.