Interesting problem cascading EF86s

[3choplex]

I am working on converting a Webster Electric PA. It has three separate EF86 inputs that join at a simple mixer into another EF86 followed by a James stack. Up to that point, it matches this schematic: http://2.bp.blogspot.com/-JySFGsgNk0k/Tdhn3QBaUuI/AAAAAAAAAsA/_0Z3XUzdVJw/s1600/TP45schematic.gif The power section is different.

The EF86 initial channels feature 220k plate resistors, 1M5 screen resistors, .1 uF screen bypass caps, and a shared 750 ohm/200uF cathode. All three are followed by a .02 uF coupling cap and a 1M volume, after which they join together into the second EF86 gain stage.

Although I don't have the voltages in front of me, I believe Va on V1 is 90v, and the screen voltage is 70v; V2 and V3 are more like 170v on Va and 120v on the screens. (Obviously some resistors have drifted.)

Each of these channels work fine on their own, if kind of blah. They have a lot of headroom, but get pretty distorted with the volume up most of the way.

I have been experimenting with just cascading one of the inputs into another, but I'm getting a weird problem. What I've done is connected the output of V1's coupling cap to the grid of V2. I also added a 1M grid leak after the cap. If I crank the volume, I get some very low volume guitar signal with no treble. Mostly what I get, though is kind of a gated "blat" if I hit the strings hard. Am I just dumping V2 into cutoff with the gain of V1?

Admittedly, my plan is to reduce the plate resistors, separate the cathodes, etc., to cascade some milder stages, but this was an easy experiment to try. When I tried some jumpers, both V1 and V2 work fine by themselves, just not cascaded. Any thoughts?

[HotBluePlates]

Look again at the schematic you linked.

Each of those input EF86's has a volume control after it. Don't know what the resistance is of the controls, but you note it gets "pretty distorted with the volume up most of the way."

Now look again at the output of the volume controls: each wiper has a series 1MΩ resistor, and these feed the grid of the next EF86. Notice also that the 2nd stage EF86 has a 1MΩ resistor from grid to ground.

By virtue of the 1MΩ resistors to isolate each of the input channels and that 1MΩ resistor from grid to ground, the signal of each input channel is being cut in half at the grid of the 2nd stage. This is likely reducing distortion and "blat" somewhat.

... I have been experimenting with just cascading one of the inputs into another .... Mostly what I get, though is kind of a gated "blat" if I hit the strings hard. Am I just dumping V2 into cutoff with the gain of V1? ...

Probably. It sounds kind of like you're slamming the last EF86 to cutoff, probably from too much signal level.

I am working on converting a Webster Electric PA. ...

In the schematic I see input pentodes, another pentode, a tone circuit, driver stage and output section. Not to be argumentative, but this doesn't seem like it needs "conversion" but maybe just parts-value tweaks to make it more guitar-friendly.

... Admittedly, my plan is to reduce the plate resistors, separate the cathodes, etc., to cascade some milder stages, but this was an easy experiment to try. When I tried some jumpers, both V1 and V2 work fine by themselves, just not cascaded. Any thoughts?

So don't cascade.

BUT... now that you see the voltage divider inherent in the mixing resistors & grid reference for the 2nd EF86, maybe that will suggest a tweak.

Instead of cascading one of the input EF86's into another, maybe just make the 1MΩ isolation resistors smaller, say 470kΩ or 220kΩ. That will pump up the signal going into the 2nd EF86 considerably.

And maybe you don't need 3 identical EF86 input stages. What if you took one out and replaced the socket with a 9-pin socket, put a !@AX7 in that spot, and cascaded one 12AX7 stage into the next, to be mixed just prior to the 2nd EF86? You could tim the value of the isolation resistor up/down to help match levels between the different input channels.

[terminalgs]

first,  I'd make your conversion approach to be more  of a simplification of the circuit, rather than to add a stage up front in hopes of waking it up.   make it a good single channel amp first, then focus on using the extra two 9 pin sockets (or not).

As for your cascaded problem of "very low volume guitar signal with no treble.",  i suspect an impedance mismatch between your 1st two cascaded stages, OR a miller capacitance problem with the EF86's high internal capacitances and whatever inline resistances are left in your modified circuit.

I wouldn't chase that issue too much, because I don't think cascading gain stages out front with help this amp.  Here's what I see when looking at this circuit that, to me,  look like tone-suckers:

(1) too many mixed inputs.  there are 3 EF86 inputs PLUS a phono input all mixed together with 1M resistors.  thats four resistance mixed inputs.   that attenuates each signal by 75% (each channel sees the other 3 channels as paralleled resistance to ground, plus itself.  that's 4X).  To get a good starting point,  I'd disconnect all 1M mix resistors from that union except one.  Also, to make it simple:  pull 2 of the 3 first stage EF86s, and focus on one channel.  At this point, you'll have a 1M pot, feeding the 1M/1M voltage divider leftover by the remaining 1M mix resistor and the 1M grid leak. (you might later rework these resistors and pots,, but for now..  this is an improvement over the 4X mixes).

(2) tone stack: EF86 drives 1/2 tone stack 6GC7 drives the other 1/2.  This might be a problem.  it might not be suitable to handle  the high frequency ranges of a guitar signal.  I'm sure that looks like  a big mess and you aren't excited to tweak that, so leave it as-is (for now). 

NOTE: There is a 10:1 voltage divider between 6CG7 stages set up with that 1M/100K pair.

NOTE: 6GC7's weren't used in the middle of any good amp's signal path other than to drive reverb tanks (anyone?)

(3)  massive amount of neg. feedback   someone double-check my math: If the OT's yellow wire is 16ohm, at 30W with the neg. feedback circuit of the 13K and the 3K3 Resistors, if the closed loop gain of that 6CG7 stage is 20 (???) that's about -14dB of NFB. (too much!)    I'd disconnect for now,, or change that 13K to be 100K or 220K.

If removing the 4 mixed channels,, and disconnecting NFB  doesn't provide results,  I'd suspect that tone stack.  thus, i'd be tempted to continue the mods with this:  for the remaining EF86, disconnect it's volume pot's center lug, and reconnect it to pin 2 on the 6CG7 gain stage that has the 3.3K Rk underneath it, Also, disconnect everything else connected to that 6CG7's pin 2 (disconnect the .047, the diodes, and the 470K resistor).  that gives you:

guitar to EF86 to volume pot, to low-Mu gain stage to PI.   

also, what at the voltages?  are the 6L6's cold?



This amp was designed for what?  auctioneers at a cattle barn?  announce lunch at a high school?  maybe MCee a county fair?  or to host background music (phono) and occasional MCeeing at a resort in the Catskills?   It'll need some tweaking to sound good as a guitar amp (but maybe not too much).

[HotBluePlates]

first,  I'd make your conversion approach to be more  of a simplification of the circuit ...

I agree that this seems the best plan.

... (3)  massive amount of neg. feedback   someone double-check my math: If the OT's yellow wire is 16ohm ...

Is we assume the schematic is correct, negative feedback comes from a dedicated feedback winding. As a result, you can't know how much feedback voltage there is from that winding.

Also, for hi-fi, 14dB feedback is not much. Maybe more than Fender did, but you can always jigger the 13k resistor higher and see what it sounds like.

... (2) tone stack: EF86 drives 1/2 tone stack 6GC7 drives the other 1/2.  This might be a problem.  ...

I agree this might not be ideal, mainly with the extra series resistance between the EF86 plate (which will have an output impedance close to the 68k plate resistor) and the coupling cap going into the Bass control.

But other than that, this is a James tone circuit, but split into two halves. This is not uncommon, and there are examples of it in the Tone Control section of RDH4. You're not too worried about signal loss, because that's the whole point of a tone circuit anyway.

Following the 1st 6CG7 stage & 2nd half of the tone circuit are two more 6CG7 stages to boost the signal back up prior to the 6CG7 split-load inverter. A guitar amp would tend to economize by using fewer, higher-gain tubes and fewer (usually one) stage between a tone circuit and a split-load inverter. So I'm guessing we're alright.

... NOTE: There is a 10:1 voltage divider between 6CG7 stages set up with that 1M/100K pair. ...

But the Treble wiper defeats that divider. I'm guessing the 1MΩ helps make the Treble tone circuit sound "effective".

... NOTE: 6GC7's weren't used in the middle of any good amp's signal path other than to drive reverb tanks (anyone?) ...

On the contrary, it seems like a "too-good" tube.

Yeah, guitar amps didn't use em much, though they pop up in Ampeg and Gibson amps. But there are a number of low-gain or oddball tubes that didn't appear in many amps.

If you re-do the heater wiring, you should be able to pop a 12AU7 in these sockets, and save money (the 6CG7 seems expensive at places I look, for no apparent reason).

Yes, with some parts-value changes you could use 12AX7's in these sockets, but you may not need the extra gain once you get rid of some of the lossy circuits Terminalgs mentioned; and again, 12AU7's are probably cheaper and easier to find in NOS (I know I have more of em than I'll probably ever use).

[darryl]

I have been experimenting with just cascading one of the inputs into another, but I'm getting a weird problem.

If the cascaded stages still use a common cathode resistor and capacitor, there could be feedback occurring, especially if the 200mfd cap has partially dried out.

NOTE: 6GC7's weren't used in the middle of any good amp's signal path other than to drive reverb tanks (anyone?)

The 6CG7 is a 9-pin equivalent of the 6SN7 - both have often been used in valve audio.

( The 1970's era amplifier in my avatar uses two 6CG7's in phase splitter and voltage amplifier duties    )

[terminalgs]

The 6CG7 is a 9-pin equivalent of the 6SN7 - both have often been used in valve audio.

( The 1970's era amplifier in my avatar uses two 6CG7's in phase splitter and voltage amplifier duties    )

Well, I'll be.... !!  yup.  I always thought 6SN7 became 12AU7, but now I see, the 6CG7 is spot-on 6SN7!  and my last build used a 6SN7GT in the PI!!!    I've always liked the 6SN7,  why did I relegate it's descendant to little more than reverb tank drivers?  Thanks for the correction!

[HotBluePlates]

...  I always thought 6SN7 became 12AU7 ...

It did, also.

The 6SN7 is rated for 5w per plate, the 6CG7 (with smaller-than-6SN7, larger-than-12AU7 plates) is rated 3.5w, and the 12AU7 is rated 2.75w. Otherwise, they are broadly-similar tubes.

[terminalgs]

Is we assume the schematic is correct, negative feedback comes from a dedicated feedback winding. As a result, you can't know how much feedback voltage there is from that winding.

Also, for hi-fi, 14dB feedback is not much. Maybe more than Fender did, but you can always jigger the 13k resistor higher and see what it sounds like.

great catch on the NF winding. I saw "yel" at the top of the windings and assumed 16ohm or so... but I guess, being a PA, with all those taps, i suppose it's for a building PA??? ("attention students, recess is cancelled for the week"??)

being a PA amp,  I suspect its more NFB than any of us would design into a guitar amp, but given the unknown element of that winding, yea,, who knows....

... (2) tone stack: EF86 drives 1/2 tone stack 6GC7 drives the other 1/2.  This might be a problem.  ...

... NOTE: There is a 10:1 voltage divider between 6CG7 stages set up with that 1M/100K pair. ...

But the Treble wiper defeats that divider. I'm guessing the 1MΩ helps make the Treble tone circuit sound "effective".

how does the treble wiper defeat it?  To my untrained eyes, it looks like a 1.1M volume pot set on "1" with a one knob tone control that either treble-bleeds via the 150pf (tiny) or treble dumps via the .001.  or, if its a 1M or 500K pot (who knows) does almost nothing at half-throw.   for any frequency whose impedance isn't swayed by the treble control caps, isn't the 1M/100K a big voltage divider? (I'm always a bit confused when a parallel C&R lays itself on top of a voltage divider like this... or maybe that's not what is happening at all...  ?)

[DummyLoad]

..  I always thought 6SN7 became 12AU7 ...

6F8G and 6J5: i believe they are the predecessors of 6SN7 as they share same RC chart. 6SN7 has up rated anode volt capability.

6FQ7/6CG7 are 9pin descendants of 6SN7 and share 6SN7 RC charts - 6SN7 has higher anode volt rating; 450V, 6CG7 is 300V and 6FQ7 is 330V.

12AU7/6C4 are different tubes; yes similar, but still different.

--pete

[HotBluePlates]

...  To my untrained eyes, it looks like a 1.1M volume pot set on "1" with a one knob tone control that either treble-bleeds via the 150pf (tiny) or treble dumps via the .001.  or, if its a 1M or 500K pot (who knows) does almost nothing at half-throw.   ...

Yeah, you have a point... I think I approached that wrong.

The way I see the circuit, if the 1MΩ didn't exist, the 50pF would have nothing to "work against" to be a method to siphon highs around a big resistance.

I wonder if the 100kΩ is an accurate value, given this is a hand-drawn schematic. But then I'm reminded that in order to have the ability to given an apparent treble-boost (by way of the 50pF cap), you need to have a big loss in the system at "no-boost". The restoration back to "no-loss" then sounds like a "big boost".

And 50pF against 1MΩ is just over 3kHz, so it's probably worthwhile to up that to 150 or so, which will bring that frequency down to the 1kHz neighborhood.

[kagliostro]

If you decide to rewire the amp this will be my proposal

guitar > ef86 input > ef86 gain stage > 1/2 6cg7 CF (as TS driver) > ToneStack > 1/2 6cg7 CF (as Send for FXLoop) > FXLoop > 1/2 6cg7 gain (as Return of FXLoop and gain stage) > 1/2 6cg7 as cathodyne PI > Power Tubes

or

guitar > ef86 input > ef86 gain stage > 1/2 6cg7 CF (as TS driver) > ToneStack > 1/2 6cg7 CF (as Send for FXLoop) > FXLoop > ef86 gain (as Return of FXLoop and gain stage) > 6cg7 as LTPI > Power Tubes

Of course the ef86 must have a tamed gain respect what we see in the Matchless or Vox

K

[DummyLoad]

i'd leave it as is for the most part. if it were my hack, i'd:

disconnect 3rd channel. install 2.2K/47uF under Ch1 & Ch2 cathodes eliminating the 750R and 200uF common cathode R&C.

remove input coupling caps & change 2.2M grid leak Rs to 1M w/ 47K-68K grid stopper.

bond the inputs together with two switched phone jacks - plug into one input jack, you bond both channels; plug into the other input jack, you break the bonding and have split channels.

for less gain switch or remove g2 bypass cap of the summing EF86 - you could also do this for one of the 1st stage EF86, or both.

--pete

[3choplex]

also, what at the voltages?  are the 6L6's cold?

The power section is a bit different from the schem--it has two 6L6s. They both have 405v on the plate, 344v on the screens, and about 26v on the cathodes.

As a side note, this has 6SN7s rather than 6CG7s.

Thanks for the input. Lots to digest!

[3choplex]

If the cascaded stages still use a common cathode resistor and capacitor, there could be feedback occurring, especially if the 200mfd cap has partially dried out.

  Can't believe I didn't think about feedback.

[3choplex]

Separating the cathodes got it working. AND... you guys are right, it has WAY too much gain. With the volume way down, it sounds great, turned up it's very harsh. Easy enough to deal with, though. Thanks for the advice, everyone!

[PRR]

> wonder if the 100kO is an accurate value, given this is a hand-drawn schematic

It is tangle-drawn, but correct-enough.

"100K" might perhaps be 111.111K for best-numbers, but all the same to the ear.

The James starts with a 10:1 divider. Then bypasses that up or down to get "boost/cut".

In the popular one-lump James, the 10:1 divider is around the Bass pot. Actually the 0.1 mid-point of the Audio-taper bass pot at low freq and the 100K+10K above/below the bass-pot at high freqs. The treble pot network works against this.

When you separate bass and treb over two stages, the treb pot still needs the broadband 10:1 divider. 1meg:100K is appropriate. (If we stole the 100K+10K values, we'd need another ~~68K in series so the treble pot isn't loaded; 1Meg+100K works fine.)

Each tone-network has loss of 10:1 (maybe 11:1). Each recovery stage has gain near 15. 0.1*15*0.1*15 is gain a bit over 2 from mixer amp out to driver amp in.

The shared cathode network between cascaded bass and treb recovery amps "could be" a source of instability. However the loss between stages is nearly as much as the gain. When you add the loss from second cathode into the common cathode resistor, loop gain is less than unity. (This WILL change if you boldly convert to 12AX7.)

Agree that the 13K(?):3.3K NFB can't be analyzed from the data known. Agree that for "zing" the 13K should be much higher, at least 50K, even open.

Gain structure, output forward, no NFB: Assume 6L6 sit at -40V on G1. Cathodyne is unity gain. Driver has large cathode resistor, will give gain near 10. So 4V peak at driver grid.

Tone stages gain of 2, so 2V peak at mix-amp plate.

Mix-amp probably has gain of 100, so 20mV at mix-amp grid.

Numbers may be 5X to 15X higher with NFB around the output stage.

With the stock circuit it has WAY more gain from mike-in than any guitar-amp wants.

(Didn't I do these numbers before, elsewhere?)

So it HAS plenty of gain with-OUT the preamps. But no volume control.

A volume-pot before mix-amp grid will work. But hiss won't go down when you turn-down.

The EF86 stage will overload at 400mV in. If you put guitar straight-in, and vol-pot after, it's fine for weaker axes but will overload before the volume control on strong inputs.

Keeping the split-James tone network: yank the pentodes. One 12AX7 stage, volume, another 12AX7 stage, go to Bass network, raise 13K NFB resistor, should play fine and a little less boring.

Keeping the pentodes: break the common cathode net under the mike-amp stages. 2.2K and 20uFd each works full-bass, use 0.68uFd for shave-bass, no cap for strong axes that overload the input. Mix-network may have to be hacked so you don't have Too Much overall gain.