AO-35 Amp help needed

[skyboltone]

Greetings Lads:
As PRR says, squint the enclosed scheme. I have abandoned the parallel input section in favor of a Leo two stage pre with single tone knob. A brown deluxe scheme 6G3 I believe.
[IMG706]http://i9.tinypic.com/6gkip92.jpg[/img]

Using Leo's values created way too much gain for the tiny space I'm using, (the original AO-35 chassis). I went more or less by the book, adjusting my Rk values to achieve proper cathode and plate values with the available supply. The T7 in V2 is another change in an effort to control gain. Interestingly, if you do the calcs you find that getting rid of the 220K grid resistor to V2B is the most effective step to bring down the gain in stage one. Anyway, I had a pretty nice sounding amp except that I couldn't get any real cleans, so I was tube swapping, cathode resistor swapping, grid load resistor swapping and just generally making things hard on myself for a point to point build (never again on a prototype).

After on of my innumerable swaps I suddenly have a fierce oscillation. As you turn the gain knob up you get a loud "thump" at about 1/3 rotation and the squealing pig at about 2/3 rotation.

I think I've destroyed a component here. I really want to make it clear that lead dress is not the culprit. It worked fine before I made my last move.

When the oscillation starts you see -1.39V show up at the junction of the stage one coupling cap and the volume pot end. I suspect it may be that cap but I don't know for sure and rather than shotgun it I thought I'd try to noodle it. By the way, a 12AT6 in V1 almost entirely eliminates the problem with the exception of the last bit of the turn of the Gain knob. Do the gain calcs and you're only seeing about 14 or so gain factor thru both stages of V1 with the T7 because the plate resistance is so much lower it drops plate and cathode voltages.

Any thoughts. PRR? Voodoo? Darkblue? anybody?

Thanks in advance
Dan H

[dartanion]

Well, you have a whole lot more gain in the preamp now, so you may need to dump some. Since the tone stack that you are using is low loss, there isn't that much gain loss between stages. You might want to try a voltage divider pre or post V1b. Easiest way based on your schematic would be to add a resistor ground right after the 1K going to V1b grid.

Also, now that you have more gain, you may need to back off the tail resistor in the PI as you probably have too much signal swing.

With such a small chassis though, it could just be lead dress. You are using shielded wire right?

Other options would be a more lossy tone stack, increase the value of the grid stoppers on the EL84s, or add some NFB.

[skyboltone]

[quote="dartanion"]Well, you have a whole lot more gain in the preamp now, so you may need to dump some. Since the tone stack that you are using is low loss, there isn't that much gain loss between stages. You might want to try a voltage divider pre or post V1b. Easiest way based on your schematic would be to add a resistor ground right after the 1K going to V1b grid. Thanks for the help Dart. That's a good idea about the voltage divider. I can clip in a pot and dial something in suitable.

Also, now that you have more gain, you may need to back off the tail resistor in the PI as you probably have too much signal swing. I do notice that seems to be the practice in the fender PIs. 1.2k/47k,820/22k, 470k/6800k.
With such a small chassis though, it could just be lead dress. You are using shielded wire right? Well, as I said, it was working with the identical lead dress then something went gunny sack. I've chopstuck it to death. It's point to point, there are no leads other than the component leads. No shielded.

Other options would be a more lossy tone stack, increase the value of the grid stoppers on the EL84s, or add some NFB. Yup, especially the NFB. I think it may help with getting some cleans out of it. The thing was sounding really good, I should have left well enough alone. Thanks![/quote]

[PRR]

> I was tube swapping, cathode resistor swapping, grid load resistor swapping and just generally making things hard on myself .... After on of my innumerable swaps I suddenly have a fierce oscillation.

But you won't tell us what changes you made, or what was the "on" change you made just before it went nutz.

Bah. Blindfold darts is a losing game.

The plan as shown should work. The 6BQ5 have more gain than 6V6, but the 100K/15K divider in V1B plate drops gain.

Layout is ALWAYS an issue. Even if it worked a moment ago, especially if you've been generally swapping. Bring it up to squeal. Better use a dummy load and a speaker with a 100 ohm series resistor across the load so you don't go deaf. Get in there with a dry chopstick and poke. Any poke which changes the squeal, better or worse, is a clue.

BTW: it appears to have two squeals. The "1/3 thump" and negative grid implies an oscillation, but you hear nothing so it may be hypersonic. Keep goosing it, and at 2/3 it is so freaked-out it finds a second mode which this time is something you hear.

The V1A plate cap is evidently good. If it were bad it would be no signal and a positive voltage on the V1B grid end; the negative voltage means that V1A is making a big signal, trying to drive V1B grid positive, and then the cap should charge-down to some negative voltage.

Since VOL has a strong effect (what about tone???? what about MV????), and the voltage into V1B seems too large to be stray coupling, it appears that the feedback is getting into V1A grid. The simplest path for positive feedback is two stages later, V1B plate system is too close to V1A grid system.

But OT plate leads are always potent sources of positive feedback.

If it is too loud yet not clean, lose the cathode bypass caps first. We only do them to maximize gain. Leaving them out sacrifices gain but improves cleanliness. I'd think you DO want a cathode cap on V1A, I think that's an important detail on a guitar amp. But while you have troubles, leave it out. I sure would lose the cathode cap on V1B before I did that tricky thing in V1B plate, but you may need both.

(Suggestion: Marshall tended to use here a 12AX7 with 100K plate and 10K cathode resistor. The stage is starved. Gain is low. The starvation hardly matters because it only has to drive the ~2Meg of V2A grid. Linearity is high to a point, far more than V2A needs. It is "wrong" but it works fine here.)

12AU7 makes a fine clean low-gain phase splitter driver, and is pop-in. (If you live by tube books, you could re-bias for 12AU7; in fact it'll work plenty good with the values you have for 12AX7/12AT7.)

> 12AT6 in V1 almost entirely eliminates

Tee Six??? Won't fit the hole?

> Do the gain calcs and you're only seeing about 14 or so gain factor thru both stages of V1 with the T7[/i]

With 12AT7, parts as shown: first stage is about 40, second stage is in the neighborhood of 60*(15K/15K+100K+20K)= 6, gain of V1A+V1B is 40*6= 240.

6BQ5 will need about 12V peak drive for full output, V2 has gain near 20 from one grid to each plate, so V2A needs about 0.5Vrms drive for full roar. A gain of 240 in front, and negligible tonestack loss, gives input sensitivity near 2mV, This is "awful hot", nothing like Tweed design, and will be very critical about layout. I'd think about losing V2B altogether. Or if very-clean is a goal, use 12AU7 in V1 with unbypassed cathodes.

Long-range: the 6BQ5/EL84 is so sensitive that this type driver and MV probably won't do what you want. The driver can make nearly 100V swings, the BQ only needs 12V swings. You'll have to lose a LOT of gain in the MV to get limiting in the driver before the power stage. Also with MV full up an enthusiastic player can slam the 6BQ5 grids far beyond grid-blocking. The loudest transients actually go silent, then the decay fades back in. The most direct fix is to drop the "C 322V" point down below 100V. In fact if you can wire a transistor, another way to clip in driver before power stage is to build an adjustable 20V-300V 3mA feed for point C.

Non-feedback 6BQ5/EL84 push-pull stages "should be dirty", but are widely used in guitar work and are not notoriously grubby. I doubt you want a lot of NFB unless you'd really rather have a flavorless transistor amp. You may like a touch of NFB to control speaker damping. Using the popular 5K resistor in the PI tail, try 120K from speaker to 5K.

> if you do the calcs you find that getting rid of the 220K grid resistor to V2B is the most effective step to bring down the gain in stage one.

??? What 220K? The leftover mix resistor from the 2-channel original? The 2-in mixer has a loss of 2, of course. If you only got one input, it's a loose end.

The drawing shows a novel way to wire heaters. Obviously your soldering hand knows better than your drawing hand since it does squeal.

[skyboltone]

> I was tube swapping, cathode resistor swapping, grid load resistor swapping and just generally making things hard on myself .... After on of my innumerable swaps I suddenly have a fierce oscillation.

But you won't tell us what changes you made, or what was the "on" change you made just before it went nutz. The word is "one", bad proof read, sorry
Bah. Blindfold darts is a losing game. I agree, but this dope made more than one change at a time, chew me out for that, I need it.

The plan as shown should work. The 6BQ5 have more gain than 6V6, but the 100K/15K divider in V1B plate drops gain.

Layout is ALWAYS an issue. Even if it worked a moment ago, especially if you've been generally swapping. Bring it up to squeal. Better use a dummy load and a speaker with a 100 ohm series resistor across the load series/across? If you could pencil whip that I would appreciate it, I'm using a rug over the speaker, same effect. Get in there with a dry chopstick and poke. Any poke which changes the squeal, better or worse, is a clue. This I will do some more of today. Yesterday I could hardly touch a thing in there without some effect BUT, I think I remember moving those pesky OT primary wires about the time this beast appeared.!!!
BTW: it appears to have two squeals. The "1/3 thump" and negative grid implies an oscillation, but you hear nothing so it may be hypersonic. Keep goosing it, and at 2/3 it is so freaked-out it finds a second mode which this time is something you hear. Yup, that's why I included the detail.

The V1A plate cap is evidently good. If it were bad it would be no signal and a positive voltage on the V1B grid end; the negative voltage means that V1A is making a big signal, trying to drive V1B grid positive, and then the cap should charge-down to some negative voltage. Thanks for this. It's this sort of thinking that separates you sir, from idiots like me. I been "playing" rather than working with this stuff for years, aligning radios, repairing not modifying ham transmitters and receivers, moving broadcast transmitters from AM broadcast to the ham bands etc. None of that prepares one for this simple analysis. I just couldn't figure how ANY DC could make it past that cap. Now I know. Keep it coming for all of us.

Since VOL has a strong effect (what about tone???? what about MV????), I'll play and see. and the voltage into V1B seems too large to be stray coupling, it appears that the feedback is getting into V1A grid. The simplest path for positive feedback is two stages later, V1B plate system is too close to V1A grid system.

But OT plate leads are always potent sources of positive feedback. No changes in lead routing in V1A/V1B area but the OT plate leads were monkied with. I'll keep you posted. I wonder too if I've toasted V1s socket somehow, lots of heat many times. First things first. V1B out is getting into V1As in. I will start with the OT wireing and this.

If it is too loud yet not clean, lose the cathode bypass caps first. I was reading in the tube manual the other day about this. Seems an added benifit of degenerative feedback as well. KF used it in his third stage. I was planning to explore this in stage two.
We only do them to maximize gain. Leaving them out sacrifices gain but improves cleanliness. I'd think you DO want a cathode cap on V1A, I think that's an important detail on a guitar amp. But while you have troubles, leave it out. I sure would lose the cathode cap on V1B before I did that tricky thing in V1B plate, but you may need both.

(Suggestion: Marshall tended to use here a 12AX7 with 100K plate and 10K cathode resistor. The stage is starved. Gain is low. The starvation hardly matters because it only has to drive the ~2Meg of V2A grid. Linearity is high to a point, far more than V2A needs. It is "wrong" but it works fine here.)

12AU7 makes a fine clean low-gain phase splitter driver, and is pop-in. (If you live by tube books, you could re-bias for 12AU7; in fact it'll work plenty good with the values you have for 12AX7/12AT7.) I've played with this before the squeal. Alas, it cleans it up a bit too much methinks
> 12AT6 in V1 almost entirely eliminates

Tee Six??? Won't fit the hole? Typo, forgive me

> Do the gain calcs and you're only seeing about 14 or so gain factor thru both stages of V1 with the T7[/i]

With 12AT7, parts as shown: first stage is about 40, second stage is in the neighborhood of 60*(15K/15K+100K+20K)= 6, gain of V1A+V1B is 40*6= 240. I'm obviously not doing this right. In fact the AX-84 site gain explanation doesn't work without dividing the bottom line by a million anyway. You seem to be using amp factor instead of transconductance? This subject needs it's own thread.

6BQ5 will need about 12V peak drive for full output, V2 has gain near 20 from one grid to each plate, so V2A needs about 0.5Vrms drive for full roar. A gain of 240 in front, and negligible tonestack loss, gives input sensitivity near 2mV, This is "awful hot", nothing like Tweed design, and will be very critical about layout. I'd think about losing V2B altogether. Or if very-clean is a goal, use 12AU7 in V1 with unbypassed cathodes.

Not the goal, I was very close except for a persistant buzziness with MV up and Gain down. Otherwise the thing sang like a canary.

Long-range: the 6BQ5/EL84 is so sensitive that this type driver and MV probably won't do what you want.
Like I said, it was close. Cathodyne PI I think you said in an earlier post?The driver can make nearly 100V swings, the BQ only needs 12V swings. You'll have to lose a LOT of gain in the MV to get limiting in the driver before the power stage. Also with MV full up an enthusiastic player can slam the 6BQ5 grids far beyond grid-blocking. The loudest transients actually go silent, then the decay fades back in. The most direct fix is to drop the "C 322V" point down below 100V. In fact if you can wire a transistor, another way to clip in driver before power stage is to build an adjustable 20V-300V 3mA feed for point C.

Non-feedback 6BQ5/EL84 push-pull stages "should be dirty", but are widely used in guitar work and are not notoriously grubby. I doubt you want a lot of NFB unless you'd really rather have a flavorless transistor amp. You may like a touch of NFB to control speaker damping. Using the popular 5K resistor in the PI tail, try 120K from speaker to 5K.


The drawing shows a novel way to wire heaters. Obviously your soldering hand knows better than your drawing hand since it does squeal. OK, I'll fix it, actually I'm gonna rip it out, the center tap ground worked fine.

[skyboltone]

Nice work PRR. The adjustment of the OT primaries was the main culprit but other things were contributing. The whole scheme was just too hot. Anyway, I've removed the big .1 cap from the bottom side of the PI and put back in the .01 that Fisher and Sampson used. No matter what I did with it the amp complained. There was a missed solder joint on the ring of the input jack to the 1 meg input load resistor. I moved V1Bs plate resistor outside of the circle of the socket; it liked that. Finally I've implemented both suggestions regarding the 12AU7 in V2 and the degenerative gain stage in V1B.

It's now a very well behaved little guy. Still gets loud but more controllable altogether. I'll subject you all to clips in the coming week.

Question: Although much diminished, there is still that persistent buzziness to single notes played with MV up and Gain down. The tone control takes some of the edge off but alters the tone more than I would desire when you back off the dime. I'm thinking can I trade the 500pf treble cap for something else or maybe put in snubbers (270pf?, 350pf?) between pins 1 and 3 and/or 6 and 8 of V1? Increase the size of 100pf shunt across the PI output?

Thanks again for your help, I'm sending you (PRR) a PM question?
[IMG700]http://i14.tinypic.com/4r9o1ew.jpg[/img]

[PRR]

> repairing not modifying ham transmitters and receivers, moving broadcast transmitters from AM broadcast to the ham bands

Then you cudda known. A low-price Class C radio power stage is often biased with no DC voltage, but when the stage before it starts slamming the grid, the grid current drives the grid negative. Tube AM radios start the local oscillator zero-bias, but as it comes up to normal oscillation the grid current in a grid resistor (almost always 33K) pulls the grid negative. The amount of these negative voltages is a clue to how hard the stage is driven (or self-driven).

You also see it in a tube power amp if you meter a grid and then beat the heck out of it. When you ask a 20W amp to make 30W worth of signal, the power stage grids go negative of their resting voltage. If you have an MV late in the amp and over-drive the stage before it, you will see that grid dip negative on pluck-attacks. (Easier to watch on a needle-meter than a DVM.)

>> use a dummy load and a speaker with a 100 ohm series resistor across the load series

> across? If you could pencil whip that

[PRR]

> gain explanation doesn't work without dividing the bottom line by a million anyway.

And transconductance is in microMohos? Where "micro" is 1/1,000,000 ?

Those funny greek letters have to be accounted. (And the US custom of using uMho is annoying.... zeros rolling off the paper in all directions. Many Euro sheets switched to mA/V which is essentially milliMho, and is nicely complementary to resistors in K-ohms.)

> You seem to be using amp factor instead of transconductance?

Mu is plate resistance times transconductance.

12AX7, sample values are 62,500 ohms and 1,600uMho.

1,600uMho is 0.001,600 Mho.

62,500 * 0.001,600 is 100.

Page 4 of GE 1953 shows that both Rp and Gm vary a lot with current. At 0.5mA, Rp=100K and Gm=1,000uMho. 2mA would be near 50K and 2,000uMho. Both of these are essentially 100. Or to save you work, the upper curve shows that Mu is 95-101 for currents from 0.25mA to nearly 2mA, which is probably the full range where we would consider 12AX7. It won't pass more than a couple mA at sane voltage and dissipation, anything less than 0.25mA is probably too weak to be useful.

12AU7 shows more slant of Mu with current. But at any point, Mu is Rp*Gm for that current.

So you only need to know two out of three of Rp, Gm, and Mu to work any problem.

And since Rp and Gm vary significantly with current, while Mu varies very little over the "useful current" range of a tube, it is often simpler to use Mu and a wild guess at either other parameter.

And if the audio load being driven is "much higher" than Rp, the voltage gain will approach 50%-70% of Mu. 12AX7 has a high Rp and can rarely exceed 50% of Mu=100. 12AU7 has a low Rp and with typical audio loads can often do 70% of its Mu=20.