Any basic paraphase "tricks" (for smoothness...)

[EFK]

Just some hypotheticals, really. I dig into a lot of Supros, and their 50s paraphase setups were very different than the later 60s "standard" split cathode versions. I have one now that is set up with a shared 2.2K cathode, unbypassed, and 270K anode resistors. 270K/4.7 K split to feed the paraphase grid, so the grid R to the power tubes are 270K. As I see it, this is set up for maximum gain. So, the amp essentially breaks up as soon as you touch the volume. I don;t have a problem with that, as I personally think most Supros are "crank it to 10" amps anyway, but the problem I have here - what I think the problem is, anyway - is that paraphase itself is beginning to overdrive by about 2 on the volume, and the distortion seems to go from gritty to full-bore solid state square wave almost instantly. A very splatty, blatty, hard clipping kind of sound. Once its up to 10 everything sounds great and raunchy, but there is a really ugly span in there that I'd like to smooth out. So, how does one "smooth" a paraphase? I temporarily replaced the cathode R with a 5K pot, and it seems to improve when I get it up to about 3.4-3.8K. It gives me a hair more headroom anyway. Should I try decreasing the plate resistors? Maybe back them down to something like 180K? My primary concern is the drive to the power tubes when it's cranked - I'd like to retain the "10" sound of the amp, but smooth and slow down the transition between 1 and 9! Just wondering if anyone has any paraphase-specific tricks or ideas.

[kg]

one thing you can do is reduce the grid leak resistors on the output stage. lowering their value will load the paraphase driver more, which will limit it's output voltage swing.

the splatty crap distortion sounds like output stage grid blocking. when the driver attempts to push the output stage grids above 0Vgk they will conduct current (intercept electrons, making the grid more negative). this charge is momentarily stored in the coupling caps until it has been drained off through the grid leak resistance. obviously the higher the time constant (coupling C * leak R) the worse this problem will be. reducing the cap and leak resistance will also reduce (but not eliminate) the blocking distortion.

the only way to really eliminate it is to reduce the DCR of the g1 node to a very low value, either by transformer coupling or by direct coupling a follower of some sort.

[pdf64]

Seperate those cathodes!
I experimented with paraphase a few years ago and experienced the same thing. My conclusion was that the horrid bullets of distortion are the result of a positive feedback mechanism - can't quite remember my hypothesis at the moment, not enough gain for free running oscillation, but at a certain point on the curve (with sufficient input signal amplitude) for a quick blip of oscillation.
Whatever, fully bypass the cathodes, or preferably seperate them, and everything is fine. Pete.

[EFK]

I did initially think it was blocking distortion of some kind, although it doesn't quite sound like it. I have the output coupling caps at .01 and currently 10K grid R, although I experimented with up to 47K. None of that made any difference. Oddly enough, Pete's use of the term "bullets of distortion" (above) is PERFECT in describing what I am hearing. As the volume is raised, if I cut back the guitar volume or play quite softly, all is fine, but if I dig in there are these clacky, splatty "shots" that really do almost sound like the crack of a rifle shot, very intermittent. I should mention the amp is two channels, one is less drive and it's not nearly as bad as the other. I truly think what is happening here is the first stage of the inverter (I know technically only the second triode is the inverter, but I'm considering the whole 12ax7 as a single package since they're tied together and both feeding the output) is being SLAMMED with a too-hot signal, and doing something nasty. Probably the first part, being slammed and making too much gain on top of that, then it is feeding this distortion to the inverting stage where it is also getting amplified. You know when you are playing a reverb amp w/ the reverb cranked up, and someone bumps or moves the amp and there is this "CLACK clack clack clack...." spacey nasty sound? That's it.

So, if I have a shared 2.2K, I can try split at around 4 to 4.7K each? On the other hand, I am unclear as to how a bypass cap can help here. I have though about it, but I didn't want to juice up the signal even more - but you suggest a big one to fully bypass, i.e. 25uF plus? Relative to this, Merlin mentioned on his site, "The two triodes could also share a common cathode resistor. No bypass capacitor would be required since the two valves amplify out of phase, and this will help to compensate for any imbalances" and I've been trying to understand the implication of that. So, I'm a little confused as it seems that a bypass cap here is somehow treated or functioning differently than in a typical preamp stage.

[Steve Conner]

Well, theory is one thing. In theory, the shared unbypassed cathode resistor is a feedback path that helps to balance the circuit, and bypassing it negates the advantage.

(The simple theory already assumes that the circuit is balanced, so to a first approximation the capacitor does nothing at all.)

But in practice, Pete gave you a report from the front line saying that he tried it and it helped, so might as well give it a go. Sometimes experimental evidence is worth more than theory. Circuits with more than one feedback path are hard to analyze even for the gurus like Merlin. I certainly have no idea why it's playing up.

The only paraphase I've tried was a floating one with two separate cathode resistors. One interesting thing about the paraphase is that the feeds to the two power tubes have different frequency responses. Because one of them goes through an extra stage, it has one extra HF pole from Miller effect.

So when used in a Class-AB power amp with a global NFB loop around it, it could show instability on only one half-cycle at high powers. I've seen this myself, but with a circuit where the cathodes were already separate. If anything, joining them should make it better, not worse.

[pdf64]

A shared cathode resistor here actually bypasses itself, because the grid signals are inverted compared to each other, they cause the signal at shared cathode to cancel out - check it with a scope.
So bypassing it won't 'heat' the gain up. 25uF should do it.
But watch the unbypassed cathode, and I'm sure that I can remember seeing those bullets appearing as a blip, visible on the scope.
Hopefully Merlin can offer some insight.
But the issue will almost certainly be cured by splitting the cathodes (yes double the shared value)> It's your decision whether to bypass them, either fully or with a resistor in series, but bypassed shouldn't increase the gain - it should be quick to confirm this .
If not fully bypassed, the output impedance will increase. Pete.

[Steve Conner]

Pete, did you have a global NFB loop in the circuit where you saw the "bullets"?

[octal]

I've had good results with paraphase- built a Gibson GA-40T Les Paul (conventional paraphase, shared BYPASSED cathode resistor) and Tweed 5D4 Fender Super output stage (this is what I'd term a semi-floating paraphase- there's an extra resistor to ground attenuating some of the 'antiphase' signal.) This is also unusual in that the 'phase' triode section has a bypassed cathode R and the 'antiphase' does not. I added 1.5K grid stoppers to the output tubes in both of these builds.

I think paraphase is unusally sensitive to layout... the gain on these PIs is insane.

Can you get a 'scope and and a signal generator on this Supro? Are you sure the problem is not in a proceeding stage?

Another option to reduce gain is to change the PI tube. A 5751 would offer some gain reduction with otherwise similar characteristics. You would need to increase the 4.7K resistor in the 270K/4.7K voltage divider to restore the PI balance.

Nathan

[pdf64]

There wasn't a global NFB loop on the amp. The sonic bullets were self contained in the paraphase, I could disconnect the feeds to the output tube grids and they still appeared; that led me to the conclusion that there was positive feedback path within the paraphase circuit somehow. Or maybe the power supply decoupling cap for that stage was bad (just joking I hope).
I'm sure that the 5D4 http://www.webphix.com/schematic%20h..._5d4_schem.pdf is a floating paraphase, the 50k grid leak resistor is at the virtual earth point, so it's probably not doing much.
Because that stage has almost 100% NFB, the cathode can be unbypassed, all the other NFB will keep the output impedance low.
The Vox AC50 has a floating paraphase with a shared bypassed cathode resistor - they sound great!
I just checked Merlin's 'floating paraphase' page and it offers some insight;
'The cathodes do not have to be connected together, although traditionally they are, as this saved on the cost of an extra bias resistor. It has been suggested that leaving a shared bias resistor unbypassed will improve the balance of this circuit, however, it actually introduces a little positive feedback. Therefore, this old-fashioned arrangement should not be used for hifi, but is ok for guitar.'
I have to disagree with him on that last point though!
Pete.

[octal]

I'm sure that the 5D4 http://www.webphix.com/schematic%20h..._5d4_schem.pdf is a floating paraphase, the 50k grid leak resistor is at the virtual earth point, so it's probably not doing much.

After thinking about it a bit, you're right. The 5d4 IS a floating paraphase with a "Leo" trick to save a resistor. (Combining the virtual earth resistors with the grid leak resistors, instead of keeping them separate.) Sorry for any confusion I might have caused.

I do have to respectfully disagree with you on another point, although perhaps I've misunderstood you. In a conventional (non floating) paraphase with separate cathode resistors the gain will be quite a bit higher if the 'first' cathode resistor is bypassed. If the 'second' cathode resistor is bypassed, the grid voltage divider will need to be adjusted to maintain balance. Without a shared cathode resistor, there's no NFB.

Nathan

[pdf64]

No disagreement, sorry if I wasn't clear, by 'that stage' I meant the inverting stage with all the feedback, with the 50k grid leak in the case of the 5D4. Pete.

[EFK]

Here's an update if anyone is interested. Keep in mind I am crippled by not having a scope, so all I can use to try to make sense of what is going on are (1) my ears, (2) an LCR meter that I use to inject 120hz or 1Khz signals and (3) a DVM. So: the best VAC balance on the output tube grids that I can achieve is 59.3 VAC and 52.1 VAC. Not bad, I think. This is w/ 2.2 to 2.7K shared cathode R for the paraphase (doesn;t really change anything between those values) and raising the 4.7K R to ground off the paraphase grid (See description at beginning of thread) to 6.8K. When I raise the shared cathode R, the balance goes downhill, and different values of grid R also throw the balance off - even tweaking the two together (I used two pots), balance is best w/ the 2.7K cathode R and 6.8K grid R. The cathode R is *unbypassed* - putting a bypass cap on it throws the balance off by about +3/-3 VAC at lower output levels and throws it WAY off at maxed output level, like 20 VAC off between output tubes.

Also, using my ears, splitting the cathodes made no difference at all. Raising the cathode R as noted above does give me a bit more headroom, but the onset of clipping is still harsh. Also, adding the bypass cap made no real audible difference despite the balance apparently being thrown off.

I feel that what I find objectionable is being able to hear the paraphase going into full-on clipping while the output tubes are still very clean. Probably it would be best if I could hold off the PI from clipping until the output tubes are beginning to overdrive. I'll have to think on that. I tried a 12at7 and a 12ay7 but they cut back too much on the tone and drive at full bore.

ONE OTHER side note: maybe someone can explain this. When I inject the 1khz signal, it's into a load, but I can still hear it in the tubes somewhat as I bring up the volume. As the volume knob is rotated, I hear the signal level coming up but then between about 10 o'clock and 1 o'clock the sound kind of fades out, then comes back as it is rotated up to max. Interestingly, this is exactly the span where the sound is objectionable to me when hooked up to the speaker. Does this mean anything, or is it something that can be explained?

[pdf64]

Sorry that the cathode thing didn't work out for you.
Something puzzles me - what makes you think that the paraphase is clipping before the output tubes? The paraphase has the highest output voltage swing of all PI types, so early clipping is unlikely unless the operating point is way off.
Do you have a link to your schematic?
Pete.

[EFK]

I'll have to draw out a schematic. *Something* is clipping really early - I admit I'm just assuming it's the paraphase, although both channels are making a lot of gain prior to the inverter. I have around 45 VAC on the output tube grids by maybe 9-10 oclock on the volume knob, then it's a much slower rise to max from there.

I wonder if their could be an oscillation issue going on here, something inaudible but manifesting as a harsh distortion sound through a part of the signal rise?

[octal]

You should be able to easily achieve perfect balance if you're varying the ratio of the divider resistors with a pot. I think the best combination will be separate, equal value cathode resistors both bypassed, though shared and fully bypassed should work OK too.
Here's a diagram (cathode resistors are show unbypassed, but you get the idea.) If you can't achieve almost perfect balance, something is either wrong with your amp or your test set up.