Is It Possible to Incorporate NFB onto the Tail of a Constant Current LTPI?

[capehead]

Hi everybody, I've been looking into constant current phase inverters recently. My primary motivation is that a friend of mine has a 120 watt Peavey 5150 that has been giving him alot of trouble and I have offered to trade him a 50 watt 3 channel amp I built. He loves the sound of my amp, but is concerned about the ability of 50 watts being able to keep up with his band. Of course I told him that there is only like a 3dB difference in actual volume, but his point was about headroom. Basically it's about how high the volume can go before the PI and power tubes begin to distort and lose clarity. I obviously can't do anything about the fact that it a 50 watt amp, but the phase splitter I think can be made more efficient so that it remains linear for longer. The PI I wish to replace is the standard Marshal long tail pair and looking into the numerous designs for CCS PI's, I have seen none that introduces the negative feedback in the same fashion, it is usually put onto the cathode of the preceding gain stage. The problem is that preceding gain stage on my amp is the effects return tube and there is a global master volume in between it and the PI, and all other triodes are being used.

So I guess in short; is there a relatively simple way to drop in a high voltage transistor where the 10k tail resistor is now? I am not stuck on the other aspects of the LTPI being the same, but my primary concern is how to incorporate the NFB so that the power amp controls work and respond the same way as a standard arrangement......

[uneumann]

... is there a relatively simple way to drop in a high voltage transistor where the 10k tail resistor is now? I am not stuck on the other aspects of the LTPI being the same, but my primary concern is how to incorporate the NFB so that the power amp controls work and respond the same way as a standard arrangement......

So first off, it would really help to see a schematic of what you have before suggesting changes.
Second - the tail of a PI may not be the best place to inject NFB. The tail is a "common mode" input that affects the gain of both halves of the PI. You may be able to use the other PI input for NFB - depending on your circuit. At that point, using a SS current source may be more feasible, although there are still issues of setting the current and keeping it stable with some kind of reference.

There is also the question of whether PI headroom is really an issue to begin with. Your output stage may be clipping long before the PI does, in which case there is not much point in any of this. Have you checked with a scope to see what clips first?

[capehead]

Ok, here's a quick schem of what I have going on currently. It is using 6550 output tubes with a fairly conservative PI circuit which I designed to be not overly bassy, so I don't think the power tubes are overdriving before the PI. I unfortunately don't have a scope so I can't 100% prove it, but with my experience playing around with PPIMV's, I can hear the distinctive sound of a clipping PI. As for using the second PI input, I've always been weary about doing this since from what I have seen and read, the NFB can't be messed with too much as it will negatively effect the balance(as you see in my schematic, there's alot of messing around with the NFB to get a different PA response for each channel). It is that, and I sort of like the fact that the NFB controls the gain over the whole PI, it is sort of like a compression effect. Correct me please if my theory is all wrong on this btw, lol! But I am pretty sure I remember Kevin O' Connor saying something that the inputs are not symmetrical using this arrangement. One is the true signal and the other is a sample of the output signal, and the circuit does it's best to match both sides. In a CCS arrangement, I see that there is usually a small value resistor between the emitter and ground; perhaps that value could be raised slightly and the NFB injected there?

[Dave H]

I think the feedback was connected to the tail originally to get better balanced outputs from the PI. With a constant current sink in the tail it will be balanced. Replace the tail resistor with a 2mA (?) constant current sink, make the plate resistors equal (100k) and connect the feedback/presence circuit to the ’other’ PI input.

[frus]

I think the feedback was connected to the tail originally to get better balanced outputs from the PI.

No. Feedback lowers the distortion and expands frequency response (at the expense of gain), and improves the damping factor (making low frequencies tighter). It doesn't influence the PI balance at all

[capehead]

I think the feedback was connected to the tail originally to get better balanced outputs from the PI. With a constant current sink in the tail it will be balanced. Replace the tail resistor with a 2mA (?) constant current sink, make the plate resistors equal (100k) and connect the feedback/presence circuit to the ’other’ PI input.

That is basically what I am looking to do which is replace that 10k tail resistor with a sufficiently rated solid state device (I typically see an NPN being used). Perhaps the Marshall power amp schematic I've attached would be a good reference?

No. Feedback lowers the distortion and expands frequency response (at the expense of gain), and improves the damping factor (making low frequencies tighter). It doesn't influence the PI balance at all

So by that logic, it shouldn't matter how much I 'mess' with the NFB before it gets injected back into the circuit; The PI will balance itself? I am just a bit weary about doing this since I haven't seen another circuit referenced that does that. The presence controls I've seen are usually a simple treble roll offs or in the case of that Marshall PA, an adjustable cathode bypass cap on the input stage. Some of the Mesa amps that used a CCS had a slightly more conventional presence circuit, but it was still rather subtle in comparison to the Marshall/Fender standard.

Marshall_9200_2x100w.pdf

[frus]

I think you are overthinking this. There's no way that standard Bassman PI will distort before the 6L6's. I suggest you scope it, but you needn't bother, I've done it enough times. What happens is that power tubes start overdriving first, and if you continue to increase the signal, PI starts to clip asymetrically (cause it's warm biased by 470R), so one power tube's output goes way down (like 1/2 voltage); OT presumably starts to saturate cause of the imbalance and the sound gets all Marshally and funky. But by then power tubes are long in overdrive. OK, you've got 820R i PI cathodes, so it'll behave longer, but that's even longer after power tube overdrive

Now I presume that people who play 5150's want as much preamp distortion as possible, and as clean power stage as possible, so if 50W is too weak for him, bad luck. Installing CCD won't do anything loudess-wise.

Btw, you don't have to worry about the PI (im)balance in guitar amps either. Try this:

Balancing Long-Tailed-Pair Phase Inverter Gains

and crunch your numbers, I doubt that you would hear any imbalance at that minute gain difference (even if the PI tube and power tubes were perfectly balanced/matched, which they are not)

[Dave H]

No. Feedback lowers the distortion and expands frequency response (at the expense of gain), and improves the damping factor (making low frequencies tighter). It doesn't influence the PI balance at all

Yes, I do know that. I didn’t word my previous post very well. I was trying to say that applying feedback to both the tail and other PI input (as is usually done) rather than only to the other PI input results in a better balanced PI. If feedback is applied only to the other PI input (as it is in a solid state amp) balance is poor.

[capehead]

I think you are overthinking this. There's no way that standard Bassman PI will distort before the 6L6's. I suggest you scope it, but you needn't bother, I've done it enough times. What happens is that power tubes start overdriving first, and if you continue to increase the signal, PI starts to clip asymetrically (cause it's warm biased by 470R), so one power tube's output goes way down (like 1/2 voltage); OT presumably starts to saturate cause of the imbalance and the sound gets all Marshally and funky. But by then power tubes are long in overdrive. OK, you've got 820R i PI cathodes, so it'll behave longer, but that's even longer after power tube overdrive

Now I presume that people who play 5150's want as much preamp distortion as possible, and as clean power stage as possible, so if 50W is too weak for him, bad luck. Installing CCD won't do anything loudess-wise.

Btw, you don't have to worry about the PI (im)balance in guitar amps either. Try this:

Balancing Long-Tailed-Pair Phase Inverter Gains

and crunch your numbers, I doubt that you would hear any imbalance at that minute gain difference (even if the PI tube and power tubes were perfectly balanced/matched, which they are not)

I'm not looking for or expecting an increase in actual volume, that I know is impossible; what I figure what will help is a more balanced drive signal going to the output tubes without the sacrifice in inverter output using a larger value tail resistor. I guess I am looking for the best of both which is why my thought process has been intrigued by CCS. You are right that metal players like a loud and clean output section, but I think it is a generalization that all of them want as much preamp distortion as possible. In fact the reason my friend wants my amp is that it has less actual distortion, more clarity but almost the same amount of sustain. The sound of my amp at a medium volume is perfect, its just when it goes past 6-7 on the master that definition gets lost. As for the assertion that the output section is clipping before the PI, I feel it is inaccurate in this case. I am using 6550's biased around 50-60% max dissipation, and one thing I forgot to add in my schematic is that the screen grids both have 50uF filtering applied directly to them after the 1K/5W SG resistors. On top of that, there is 150uF on the OT plates, and another 50uf at the intersection of the choke and the 2 1K SG resistors. So in my humble opinion, my output section is like a rock. I'm sure there is some clipping artifacts evident, but nothing too extreme. I will temporarily install a PPIMV to be able to pick out which is distorting more.

Yes, I do know that. I didn’t word my previous post very well. I was trying to say that applying feedback to both the tail and other PI input (as is usually done) rather than only to the other PI input results in a better balanced PI.

That is about the same as my thinking as well. I don't know the specifics of the hows and whys, but from my experience of building amps (for over 10 years at this point) it seems that the inverter behaves better when the whole thing is ballasted by the NFB. I have worked with other conventional/traditional concepts, almost none have been as good sonically and/or stability wise. If this CCS thing does not pan out, I may try a boosted concerta, like a Marshall Major. I used this arrangement on a 4x 6V6 combo amp and that amp was loud as f***!

[frus]

I am using 6550's biased around 50-60% max dissipation

what's your bias voltage (on the 6550 grids)?

[capehead]

what's your bias voltage (on the 6550 grids)?

It is at -48 volts with a plate voltage of 420v. I used an online bias calculator (I think the Weber one), which at 70% max dissipation was about 60mA, I brought it down to around 40-50mA for the sake of longevity and less distortion.

[frus]

So if you bring a input signal that has the amplitude of 96V peak to peak, the 6550's will clip, regardless of your filter caps. The tube is going to clip if the grid reaches zero volts (unless you have some class AB2 capable driver, which is not the case here).

So the question is, can the PI provide 96Vpp signal? I vote yes.

[capehead]

So if you bring a input signal that has the amplitude of 96V peak to peak, the 6550's will clip, regardless of your filter caps. The tube is going to clip if the grid reaches zero volts (unless you have some class AB2 capable driver, which is not the case here).

So the question is, can the PI provide 96Vpp signal? I vote yes.

Is 96Vpp what the standard Marshall PI puts out on average? If thats the case, I had no idea it was that high. Is that output level before clipping or max output? If this is the case, why was it that when I used a concerta PI with a driver (like the Marshall Major) it sounded louder than I ever thought possible even with 6V6's? I'm sure there was a high degree of output tube clipping doing that, but basically it dwarfed in volume the LTPI design I had in it before.

[jazbo8]

So if you bring a input signal that has the amplitude of 96V peak to peak, the 6550's will clip, regardless of your filter caps. The tube is going to clip if the grid reaches zero volts (unless you have some class AB2 capable driver, which is not the case here).

So the question is, can the PI provide 96Vpp signal? I vote yes.

+1, the power tubes can be over-driven with the current configuration, also biasing them colder, the distortion would be somewhat higher due to larger Class B output, i.e., deeper Class AB.

[capehead]

+1, the power tubes can be over-driven with the current configuration, also biasing them colder, the distortion would be somewhat higher due to larger Class B output, i.e., deeper Class AB.

I certainly do believe the power tubes are clipping to a certain extent, but I really don't think they do before the PI does. I originally had EL-34's in this amp and switched to the 6550's on the assumption that it was the EL-34's clipping out on me. Since I changed them, I have not heard much of an improvement in that regard. I guess I will know for myself when I finish putting a PPIMV in this amp. I am almost certain that when the pre PI volume gets past 7, I will hear it clipping separately from the power tubes. From there I will make adjustments to the FX return circuit to better optimize the frequencies that get passed to the PI, but it only will help so much before I hit a brick wall with the PI performance.