methods of removing oscillation?

[daz]

I'm getting some very high end oscillation after a few tweaks i did to my cascaded preamp in my JCM style build. When i hit the strings hard it happens more and it's like you have a graphic EQ plugged in and you jacked the ultra highs like 10k way up. A staticy horrible top end that when at it's worse causes the entire signal to drop down in volume for a split second. Mostly happens on bright guitar setting like the 2nd position start which on mine is very toppy. But a humbucker doesn't cause it near as much if at all.

But aside from that the amp sounds better than it ever has which is why i'm asking for different methods of stopping oscillation. You see, i already tried the grid stopper method but while it works, it kills the tone that is otherwise so good. I have 100pf snubbers at V1A and the PI. But are there any other ways to deal with this? I have no scope so i can't see where it's happening.

[nickb]

Since you've already established that grid stoppers do the trick, try using a lower value to try to kill the oscillation and retain the tone and /or perhaps you should think about the most recent changes since it was OK before that.

[daz]

Well, what i meant was grid stoppers helped but didn't really cure it till they were much too large. 470k on V1B and V2A. And of course at that point it was just a tone killer. Smaller amounts ruined the tone to albeit to a lesser degree and also didn't fully cure it. I know what change caused it, but it was also what took the tone to the next level. So i need to keep that and cure the parasite.

[nickb]

Well don't tease, what exactly caused it?

Since the grid stopper on V1 makes a difference and assuming V1 is the first stage then it's a reasonable guess that the input wiring is picking up signals. Try moving them - keep close to the chassis or screen them. Do gain/tone controls affect it? That should help figure out how big the loop is. Ground of input stage could be a factor. Schematic and gut shot will help.

Not having a scope is a big handicap.

[daz]

Well don't tease, what exactly caused it?

Since the grid stopper on V1 makes a difference and assuming V1 is the first stage then it's a reasonable guess that the input wiring is picking up signals. Try moving them - keep close to the chassis or screen them. Do gain/tone controls affect it? That should help figure out how big the loop is. Ground of input stage could be a factor. Schematic and gut shot will help.

Not having a scope is a big handicap.

In the 2204 there are two 470k/500pf peaking filters i think they're called. I had the first one, but i just added a second one just like the 2204 and apparently theres just too much HF gain. But it takes a 470k grid stopper to stop it to a degree that is satisfactory, but that also kills the high end. lowering gain doesn't help till it's low enough till it is no longer much of a OD type amp. Input is 68k right on the pin and shielded cable grounded at the jack end and i have move that along with other wires around endlessly. Gain control does help, but you have to turn it down a lot before it helps to any degree worth getting happy about.

[R.G.]

Oscillation requires a few things to happen.
(1) A path for feedback from a relative output to a relative input. In practice, there are capacitive couplings and inductive couplings so that there is always a path for feedback.
(2) A phase shift from input to output that changes the nominally 180 degree phase shift in the amplifier to 360 degrees, and makes feedback positive at that frequency.
(3) Enough gain in the path from input to output to make up for the losses in the feedback paths, however much that is.

Since there are always feedback paths, it is true that if you keep on upping gain, eventually everything will oscillate at some gain.

The ways people attack this are:
(1) make the feedback paths have so much loss that the forward gain cannot sustain oscillation
(2) lower the gain below the point where it will sustain oscillation; sometimes, especially in opamps and solid state power amps, this is done by lowering the gain only at the very high end where the oscillations happen.
(3) tinkering the forward path phase shifts to be bigger or smaller so that the additional phase shift in the feedback path is no longer positive.

What you've done so far is to either lower gain so low that it doesn't oscillate, or to cut frequency response, which is a way of lowering the gain only at high frequencies.

A really, really good thing to do is to try to find what the predominant feedback path is, and to make that have higher losses. At this point, not knowing what the circuit is, we can only guess at some popular feedback paths.

Wiring to/from controls is always good for some capacitive coupling. Ground impedance, or even ground inductance, is another. Attack wiring issues by figuring out where the sensitive inputs are and keeping wires carrying large signals or signals after large gains well away from those inputs. This is what the "move wires around" practice does, mostly. Ground wires carrying currents from later stages back to the power supply that also carry input signals form a feedback path, and this will cause oscillation at some gain. Star grounding or other technique to not have big signals polluting the grounds for little signals will fix this. The currents are usually so low in tube circuits that unless you have really bad wiring practices, magnetic coupling is usually not a problem.

It is very tough to do phase lead/lag compensation to avoid parasitic oscillation like this without a good high frequency oscilloscope.

One thing that might help is to laminate some aluminum foil in plastic insulating with a ground wire attached and see if you can shield something and get the oscillation to change. When you find how to change it, you've found one part of the chain.

One thing that might help is local RF bypassing of the power supply from the B+ side of the plate resistors per tube to the ground end of any cathode parts. This should be a ceramic disk of 0.01 to 0.1, enough volts to withstand B+. This may lower the high end impedance of the power supply and local ground impedance enough to stop it if the bypass filtering is far away or an older capacitor. You can also try local bypassing with a 10uF/500V electro on a per-stage basis.

[daz]

Thanks. Lots of things to think about and i will try tomorrow. Many i have tried too. I think i just have too much HF gain and my lead dress is good. It's short yet every wire is away from the others and crossing at 90 degrees wherever possible. I just feel it's not the lead dress, but that the HF gain is enough that even good lead dress isn't good enough if you see what i'm saying. Anyways thank for that. I will continue screwing with this tomorrow. Too bad because that last tweak really did what i was aiming for and if not for this darn issue i'd have it back in the cabinet by now and playing instead of tweaking.

[exclamationmark]

Crossing wires at 90 degrees shouldn't affect anything since the feedback path is usually going to capacitive coupling. If you really need to, you can try shielded grid wire (ground the shield of the wire at one end only), but it's really more of a bodge fix than actually fixing the layout. You should strive to keep the grid of the preceeding stage away from the plate of the current stage (ie, keep the grid of V1-A away from the plate of V1-B). The reason being is that the signals are in phase and can form a feedback path. You can also twist the wires belonging to a particular triode to form a kind of a poor mans shield, and since it's now a bunch of wires, it should be much easier to keep away from other signal wires. Make sure you stick the grid stopper at the grid otherwise it can cause some unintended effects.

[Chuck H]

You can have huge and horrible amounts of audible HF without oscillation. The oscillation you describe is just touching the audible spectrum. So if you roll out the HF above that, it should fix the problem without changing the tone much at all. What's interesting is that it only happens with certain pickup settings. Most oscillations I've dealt with involve the amp becoming self resonant at some amp settings. If you are sitting very close to the open chassis when this happens it could simply be a lack of shielding between the guitar and the amp innards. If this is the case you might find the problem goes away when you put the amp back in the cabinet (assuming it's shielded). If you've already tried the amp as it is now in the cabinet and th problem still exists...

The input triode is going to be the most sensitive. Already mentioned was placing the grid stops at the tube pin. This can help. Another thing you could try would be a small value cap from the grid pin to the plate pin on the first triode. The cap should be 4.7pf and should be rated for 1000V. I think I mentioned this in another thread of yours. Almost inaudible. It just takes the glassiest and most annoying part of the top end out. But, if I understand the circuit as it's been described to work, as gain increases so does the capacitance. So dynamics that would cause the oscillation become self defeating.

[daz]

Crossing wires at 90 degrees shouldn't affect anything since the feedback path is usually going to capacitive coupling. If you really need to, you can try shielded grid wire (ground the shield of the wire at one end only), but it's really more of a bodge fix than actually fixing the layout. You should strive to keep the grid of the preceeding stage away from the plate of the current stage (ie, keep the grid of V1-A away from the plate of V1-B). The reason being is that the signals are in phase and can form a feedback path. You can also twist the wires belonging to a particular triode to form a kind of a poor mans shield, and since it's now a bunch of wires, it should be much easier to keep away from other signal wires. Make sure you stick the grid stopper at the grid otherwise it can cause some unintended effects.

I don't think any wire is close to any other actually. I took great care to route them so i got what i think is the maximum amount of distance while trying to at the same time to keep wire short. Today i will make a fw changes, but nothing drastic because i can't see much more i can do. Like i said, i really don't think the dress is remotely bad, but that if the dress IS the cause it's only because theres so much HF gain that the dress might have to be impossibly perfect to avoid it. The only dress i felt was stupidly bad was the NFB. That being because in order to put a variable NFB pot on the front panel i had to run a lot of long wires past every part of the circuit it probably shouldn't be near ! But last nite i removed that and changed it so theres no pot and just one short wire straight to the NFB resistor and it made no difference at all. Anyways, I'm about to do what little i can with the dress according to everything i heard hear so far and report back.

[R.G.]

Thanks. Lots of things to think about and i will try tomorrow. Many i have tried too. I think i just have too much HF gain and my lead dress is good. It's short yet every wire is away from the others and crossing at 90 degrees wherever possible. I just feel it's not the lead dress, but that the HF gain is enough that even good lead dress isn't good enough if you see what i'm saying. Anyways thank for that. I will continue screwing with this tomorrow. Too bad because that last tweak really did what i was aiming for and if not for this darn issue i'd have it back in the cabinet by now and playing instead of tweaking.

I suspected you'd already been through the simple stuff. That's why I went for "here's the principles", in hopes that it would give you new place to look.

A lot of this comes from RF wiring practices, where capacitive coupling is a Big Problem, where lead inductance makes leads look like tuned circuits, and where gains run up to over 120db.

Here's more random thoughts in hopes that you can find something useful.
- The comments on grid stoppers is right - they need to be right at the socket pin, as close as you can get the resistor body to the pin.
- Have you ever had problems with picking up radio stations? Maybe the input is pulling in RF and amplifying it. It's worth trying an RF-killer at the input to see. RF pickup can make an already-there semi-oscillation much worse.
- Multiple high frequency rolloffs can cause rather than cure oscillations because the phase shifts from the rolloffs add up. Sometimes you can get better results with only one HF-killer instead of many of them, leaving the rest of the stages with as good a high frequency response as possible. This may not be of great help, as vacuum triode frequency response is already so low that there may not be enough benefit.
- Can you isolate it to one stage somehow??

[daz]

I suspected you'd already been through the simple stuff. That's why I went for "here's the principles", in hopes that it would give you new place to look.

A lot of this comes from RF wiring practices, where capacitive coupling is a Big Problem, where lead inductance makes leads look like tuned circuits, and where gains run up to over 120db.

Here's more random thoughts in hopes that you can find something useful.
- The comments on grid stoppers is right - they need to be right at the socket pin, as close as you can get the resistor body to the pin.
- Have you ever had problems with picking up radio stations? Maybe the input is pulling in RF and amplifying it. It's worth trying an RF-killer at the input to see. RF pickup can make an already-there semi-oscillation much worse.
- Multiple high frequency rolloffs can cause rather than cure oscillations because the phase shifts from the rolloffs add up. Sometimes you can get better results with only one HF-killer instead of many of them, leaving the rest of the stages with as good a high frequency response as possible. This may not be of great help, as vacuum triode frequency response is already so low that there may not be enough benefit.
- Can you isolate it to one stage somehow??

Any G stops i have or have had are/were right at the pin.
No, never picked up a radio station. Input is 68k right on the pin and lead a coax shielded on one side.
I have removed the one of two roll-offs snubbers on V1A leaving the one on the PI but issue still there.

Isolate a stage? Can you elaborate?

[daz]

I hesitate to say i got it because you know how that goes....next day you plug in and it's back. But so far what seems to have done it is changing the snubber from the V1A plate to V1B and using a 220k grid stopper on V2A. neither by itself works but together it seems to have eliminated the issue.

[R.G.]

Great! Good detective work.

The problem with parasitic oscillations is that they usually depend on something that exists that's not even in the schematic, all the various non-ideal things about everything.

[daz]

I understand. There are no absolutes, and even if there were you could never apply them to every amp. There are infinate variables. I just hope this stays as is because the tone stayed almost exactly the same. It's a tad different but not so much as to feel i lost what i was loving about it. But as i said, we shall see....you know how it goes....ears seem to change by the minute.