proper voltage at PI input ?

[Alan0354]

I just feel the closed loop gain of the power amp is a little high. You have 100K resistor minimum( 250K pot adjusted to 0) for feedback and 4.7K for setting gain. Then you top on the 4 ohm tap. That is gain of 41. Of cause you don't get 41 as the open loop gain is not that high for a PI power tube stage that attenuated by the output transformer. Why not lower the gain of the power amp, you might get some advantage of noise performance out of it.

Or else, might as well open the loop and run with no feedback. I don't like a high value resistor as feedback resistor as it can create extra poles that cause stability issue and peaking of frequency response. That will be totally unpredictable from looking at the circuit diagram. Fender use 810ohm for feedback. Marshall Plexi use about 25K.

In the back of my mind, I still a little worry Daz expressed about it sounded something more than just over driving the power amp. Too high a resistance value bugs me as I deal a lot with opamp stability and high value resistance together with long wiring to the summing junction is a recipe for bad things.

If you have to keep these values, make sure you keep the connections from the 100k resistor to the PI grid input ( on the negative side) and also the 4.7K very short. Keep the wire between the 250K gain adjusting pot and the 100K short as possible too. High impedance wires make a good antenna.

[daz]

Why not lower the gain of the power amp

How? You mean with more feedback via a smaller resistor or higher ohm tap?

On a side note, the jcm 800 MV has 100k off the 4 ohm tap.

[Alan0354]

How? You mean with more feedback via a smaller resistor or higher ohm tap?

On a side note, the jcm 800 MV has 100k off the 4 ohm tap.

Ha ha, I am still quite new to the game, I have limited experience, just the resistance is a little high.

You can get the same gain by lowering the resistor value and keep the same ratio. For example, you can divide the value of both resistor by 10. Use a 25K pot, 10K feedback resistor and 470 ohm gain resistor. Now the resistance of the feedback is much lower.

In fact, from talking to you, I just did my own measurement on my new design as you know we are both about the same stage of the build. I just want to share my experience with you. I have mine set a close loop gain of 10 and I measured closed loop gain of about 5.3. There goes to show how low the open loop gain the opamp( power amp) is.

I use 1KHz as reference frequency and I found the output ( at the speaker) start to show -1dB ( 10%) compression at 35V peak to peak. The input is about 6V peak to peak. So my power amp input dynamic range is 6V peak to peak or (6/2)X0.707=2.12Vrms. The output of my gain channel is 30V peak to peak at master volume 10!!!! I am way over the limit too!!! I need to put a resistor in series to lower at least to 12V peak to peak as I do want a little power amp distortion.

But as I said, my way is the clinical way, nothing to do with how the amp sounds. For all I know, it might suck doing this way.




To the experts here:

I also want to confirm that using my data, I calculate the output power at -1dB compression:

Tubes are 6V6GT

RMS output voltage=(35V/2)X0.707=12.37V RMS.

Dummy load=R=8ohm

W=(V^2)/R= (12.37^2)/8=19.1W output power.

Am I doing this right? 19W sounds reasonable as I only drive to -1dB.

[Chuck H]

Draw it on paper Chuck. Look....voltage divider before the master: a series R to the input lug, a load from input luig to ground. Thats a VD before the master, right? Now look at what it also does....the load, since it's ground to input lug is a R across the pot, thereby changing the pot value. You still have that series R there like you suggest. If that pot were a 100k instead of 1M and just had a 100k series in front of it, wouldn't the result be the same? Maybe i'm wrong somehow, but when i did what you say I looked at it and had exactly as i did before, a VD before the master. Only difference was the values i used. You mentioned a R from the in to out lugs, but only as a temporary R to determine the pot value i need for the level i want, right?

Ah! I must have read through it too fast (that happens on forums when you have five technical posts to wade through). I thought you had the divider AFTER the MV. It's all good then. Of course the better option is to minimize parts count and series resistance by including the MV as part of the divider circuit as I described. But the end result is almost the same. So your intuition and experiments are on the right track... As is often the case

[teemuk]

The biggest obstacles in trying to understand much of what i read are words/phrases i don't understand like "open loop gain".

Open loop gain = gain of the circuit without the global negative feedback loop enclosed

Closed loop gain = gain of the circuit with the global negative feedback loop enclosed

Why not lower the gain of the power amp --- How?

There are several ways to skin the cat. Ultimately, the closed loop gain will be the most prominent factor in defining power amp's gain since enclosing the feedback loop practically reduces gain from open loop gain to closed loop gain. Here you have to attack the negative feedback and do note that negative feedback circuit is almost always a VOLTAGE DIVIDER, not just a single resistor. Yes, you may have 100K going from output to input, but there will also be a shunt resistor that forms a voltage divider with that 100K resistor. The presence control is most likely wired in parallel to it, which is how presence control also alters the magnitude of feedback. So concentrate on the divider, not just a single part of it.

Open loop gain is determined by many factors. The first part is PI section's gain. This can be altered by varying the cathode circuit (bias or tail resistor) or plate circuit (plate resistors), as usual, but what you might have already figured out is that these modifications are always a compromise between how the stage is biased and will therefore effect how the circuit performs. In particularly headroom and overall symmetry. Next you have your power tubes. For total open loop gain the PI stage gain is multiplied by the gain of this stage. As plate loads of the power tubes are pretty much the transformer the only effective means to alter the gain is to either vary the bias or the screen voltage. Degenerative feedback through cathodes would work too but in a fixed bias circuit it will rob efficiency quite a lot. A cathode bias circuit will also behave very differently to fixed bias so even a "hybrid" of them will once again effect overall performance of the amp.

Negative feedback is basically the difference between open and closed loop gains so there's a careful balance: With high open loop gain there's plenty of negative feedback to linearize the amp: You get very clean output with very flat response. Unfortunately all this will also collapse very suddenly once the amp clips and too much NFB practically ensures very hard clipping characteristics as it tries to keep things together up to the point where it just no longer can't. Too little negative feedback and its existence practically makes no difference at all: performance is close to open loop gain performance, overall distortion will be higher, linearity and damping will "suffer" and the response to complex loads is completely unlinear. As negative feedback has less effects the crossovering to clipping distortion will be a tad smoother. You need to figure out what the goal is since neither of this performance is bad or good, and any goodness or badness is practically defined by the application and individual preferences of the end users.

You have an oscilloscope now? Why so little use for it? We still only have your description of what takes place in the circuit once you overdrive it. If it's just power amp clipping then I don't know why so much focus on "fixing" it. It goes away when you back off the volume dial. Little clipping on the higher volume settings never hurt anyone, it's just a bit more gain in storage for those moments when you plug in a guitar or effect with lower output signal. Is there really difference between dialing to "5" and achieving certain volume level and distortion or dialing to "7" instead. You already have a voltage divider to attenuate the PA input: the master volume pot. Go ahead if you want to fine tune but, IMO, it seems a bit pointless.

If the problem is something else than just clipping then we once again are completely oblivious of it because once again we go by your mere description of what takes place in that amp. You have an oscilloscope. Use it now.

[daz]

Yes, i have a scope but i know next to nothing about what i'm looking at other than the wave i posted which was said to look good. I asked for example wave of clipping and all that, but so far no one has replied to that. So theres not much i can do other than look at the scope and see when the wave starts to change radically enough that it must be clipping hard, which i assume is when it turns square looking. In any case, theres not mush i can do with it till i know what to look for.

From what i saw in the last posts it seems the FB resistor shouldn't be too large, so i reduced it to 33k. I've always felt when using NFB with this amp it's just too dull, and that combined with the fact that it all but disappears when i turn it up makes me gravitate towards no NFB, hence the no load variable NFB pot.

[daz]

Chuck, i followed your directions and found 109k between the jumped lugs and ground when the PA clips with a 220k in series with the input lug. The tone gets very dull tho when i used a 120k across the 1M pot to simulate 100k and with the 220k series. But reversing them, IE: 220k across pot and 120k in series seems to work great. Tone seems intact and the master acts like a volume control instead of a switch that goes between normal tone and total saturation and harshness. This may not be the final fix, but just reporting to ya on your idea.

Also wanted to show you what happens to the wave when i change the PI cathode back to 470R from 2k that i've been using because it sounds better. At the typical 470R it looks like the right wave with that thick bottom part. Sorry about the hand drawn thing but i don't have my camera with me at the moment. Can you tell me what that thick bottomed wave depicts in tonal terms?

[Dave H]

The thick bottom part could be an oscillation which could be why it doesn’t sound as good with 470R. I was going to mention the 2k bias resistor. I think 2k is too high, the PI is biased too near cut off meaning you could be hearing the sound of PI cut off before the power tubes clip (reducing your max clean power output) but you did say it sounded better with 2k. Perhaps that is because it is oscillating with 470R? If it is an oscillation and it can be fixed it may sound better biased with 470R. Wind the time base knob on your scope clockwise and see if you can see an HF oscillation. It won’t sync but you may be able to see it in the thickening. A real picture of it would be good.

[daz]

The thick bottom part could be an oscillation which could be why it doesn’t sound as good with 470R. I was going to mention the 2k bias resistor. I think 2k is too high, the PI is biased too near cut off meaning you could be hearing the sound of PI cut off before the power tubes clip (reducing your max clean power output) but you did say it sounded better with 2k. Perhaps that is because it is oscillating with 470R? If it is an oscillation and it can be fixed it may sound better biased with 470R. Wind the time base knob on your scope clockwise and see if you can see an HF oscillation. It won’t sync but you may be able to see it in the thickening. A real picture of it would be good.

thats what i was thinking, and if i magnify it the thick part looks like a wirewound pot's conductor....or like a coiled guitar cord. I tried using 1.5k instead of 2k and it's not there. Next i will try the more typically used 1.2k and whatever the lowest value i can find that doesn't cause that on the wave i will leave.
By the way, it only happens with a certain amount of NFB or less. I have a variable NFB pot and at the far end it's no load so zero NFB. If i turn it to get more NFB and watch the wave, that thick part doesn't start to show till i'm at 18k on the pot, so that plus the set FB resistor (33k) and it starts at 50k as a FB resistor. Thats off the 4 ohm tap. Any higher resistance than that and that thick part doesn't show.

[teemuk]

Yep, thick parts usually mean that a very high frequency wave "rides atop" the signal you are monitoring with the scope. Which usually is a clear sign of oscillation. If scope's resolution isn't high enough (or set high enough) you don't see the "waves" of the oscillation, it merely shows up as thick, blurred line.

The 470R is pretty much a standard value for that kind of circuit, I might even consider it as sort of a "reference", starting point for further tweaks. If you can't get the amplifier stable with generic component values then you probably need to attack other issues aside the circuit: layout, proper routing of ground and power supply currents, etc.

[guitician]

You could split the PI load resistors(41kx2 & 50kx2) and take the signal from the junction with the two .02uf. This would give less signal drive to the P-P pair. Or use whatever ratio needed to get the desired volume range.

[guitician]

Try paralleling the FB resistor with a small cap and see if that oscillation stops, or take it to ground.

[Dave H]

You could try increasing the value of the (5k6) grid stopper restors on the 6L6s so that it doesn't oscillate with either 470R or 2k PI bias resistors.

[Alan0354]

It is hard to tell whether the thick part is oscillation. I have to look at the scope picture to tell more. Take a picture with your camera. Good thing about low open loop gain is they tend not to oscillate.

Also read the this thread:http://music-electronics-forum.com/t35716/

We talked a lot about cathode follower and why the bottom looks like that. It is because the CF stage turn off. So don't just making conclusion just yet. Take a picture, one picture speaks a 1000 words, not the drawing.

[daz]

Too many suggestions to try guys ! Thanks, but i think whats going on is that marshall and others who use the 470R do so because they want the drive which i don't, and they relieve the oscillation with that 50 or 100pf cap across the PI plates. That gets rid of it, but i don't like the effect it has so i'm just leaving it at 1.2k on the cathode, which i just tried and that works.

On the split load PI....i thought about that, but for now i'm going to leave the mods to the master. they seem to work well, but if i decide not i may try the split load. The reason i haven't so far is that the plate resistors and caps there are well in place the way i installed them and are a pain to remove and change.