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

[frus]

Is 96Vpp what the standard Marshall PI puts out on average?

I'm SURE it can put out much more. I'll scope my amp later to tell you exactly (setup and voltages are very similar to yours)

Is that output level before clipping or max output?

Before clipping

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?

That's kinda weird, I wouldn't expect it to sound any louder. Are you sure everything's wired OK?

[capehead]

I'm SURE it can put out much more. I'll scope my amp later to tell you exactly (setup and voltages are very similar to yours)




Before clipping



That's kinda weird, I wouldn't expect it to sound any louder. Are you sure everything's wired OK?

That would be awesome to scope it out. I'm sure it would help to better inform more people than just myself. And about that 4x6V6 amp with the Major PI; yeah everything was wired correctly, and in fact I had to bring down the gain a bit to keep it under control. I increased the Rk value of the pre-inverter stage (where the NFB couples) to 2.7k, lowered that stage's Ra to 56k, the split load values I reduced to 47k, the driver's cathode up to 1.5k, the attenuation between the inverter and driver down to 100k each side, and reduced the coupling caps from the splitter from .047 to .022. Like I said before, it was certainly clipping the hell out of those little 6V6's, but it was very, very loud.

[frus]

I measured the unclipped PI output at some 140-150 Vpp

[pdf64]

Could you report the VB+ on the LTP; I think that should affect the max level before clipping and will vary according to the particular model / design?

Also what load was the LTP driving, eg the regular 220k grid leaks, the 6550 model 150k grid leaks?
Thanks - Pete

[capehead]

Could you report the VB+ on the LTP; I think that should affect the max level before clipping and will vary according to the particular model / design?

Also what load was the LTP driving, eg the regular 220k grid leaks, the 6550 model 150k grid leaks?
Thanks - Pete

The voltage on the LTPI is 380v and the grid leaks are 100k. I'm pretty sure I put that on the schematic I posted earlier in the thread.

I measured the unclipped PI output at some 140-150 Vpp

That is very interesting, I never knew how efficient the old standard really was. I suppose there are other factors at play here besides just the inverter's behavior. I decided last night/earlier this morning to throw caution to the wind and threw together a CCS inverter. I haven't been able to test it yet through a speaker, but I did test the voltages which are quite low in comparison to what I am used to; something like 8.5V on the grids and around 10V on the cathode. I used a variation of the schematic I posted.

[frus]

Could you report the VB+ on the LTP; I think that should affect the max level before clipping and will vary according to the particular model / design?

Also what load was the LTP driving, eg the regular 220k grid leaks, the 6550 model 150k grid leaks?
Thanks - Pete

My B+ on the PI was 360V
You're right, my grid leaks were 220k and lower load should decrease max output swing, good catch

[Dave H]

...I decided last night/earlier this morning to throw caution to the wind and threw together a CCS inverter. I haven't been able to test it yet through a speaker, but I did test the voltages which are quite low in comparison to what I am used to; something like 8.5V on the grids and around 10V on the cathode. I used a variation of the schematic I posted.

It doesn't have to be that complicated for a guitar amp. This should do it -

[teemuk]

If you can get a negative power supply for the CCS it will simplify the circuit even further since you no longer need bias the grids to nothing but ground reference.

This is a great asset because you can simplify with things further, like omitting capacitive coupling of the feedback.


Mesa/Boogie made several amps with CCS tail on the LTP. In first iterations they had those constant current diodes, in the second they just provided a very, very well-filtered voltage reference, which allows subbing the CCS with a plain resistor. The regulation of the reference ensures constant-enough current. Mesa's designs are also a good example how "symmetric" grid circuits no longer are a recommendation in such topology. If you don't need to use feedback you can practically even ground the grid of the differential's side that is the common cathode amp. With CCS on tail there simply is no need to mimic the architecture of grid/cathode circuit of the conventional LTP.

[Dave H]

The grids could still be grounded and the cap and grid resistor omitted without having to use a negative supply. I only included the 5.6V zener to give the transistor a few volts to work with. If the input is from a grounded circuit (e.g. vol pot) the cap and grid resistor on the left side could also be omitted. Both circuits have the advantage of having no DC across the presence pot so it won’t be scratchy.

[capehead]

It doesn't have to be that complicated for a guitar amp. This should do it -

[ATTACH=CONFIG]27486[/ATTACH]

This is a much more intuitive method, thank you! That transistor; can it be practically any NPN? Like a 2N3904 or something similar? With the CCS I just built, I used the 450V/1A NPN that the Peavey Mace uses for it's phase inverter since I wasn't completely sure of what kind of voltages I would be seeing. How it as a whole turned out was a good news/bad news situation. The good is that it was noticeably louder and the high end response was much better than before; the bad is that it sounded like crap, there seemed to be some sort of blocking distortion and weird distortion going on. I will be giving your idea a go as soon as I get some time, so definitely thanks again.

If you can get a negative power supply for the CCS it will simplify the circuit even further since you no longer need bias the grids to nothing but ground reference.

This is a great asset because you can simplify with things further, like omitting capacitive coupling of the feedback.


Mesa/Boogie made several amps with CCS tail on the LTP. In first iterations they had those constant current diodes, in the second they just provided a very, very well-filtered voltage reference, which allows subbing the CCS with a plain resistor. The regulation of the reference ensures constant-enough current. Mesa's designs are also a good example how "symmetric" grid circuits no longer are a recommendation in such topology. If you don't need to use feedback you can practically even ground the grid of the differential's side that is the common cathode amp. With CCS on tail there simply is no need to mimic the architecture of grid/cathode circuit of the conventional LTP.

What do you mean with that part about how symmetric grid circuits are no longer recommended? In my research on this whole CCS thing, I've looked alot into how Mesa did it since they are one of the few major guitar amp companies who have used a CCS. I actually have most of the parts and plans to make a Mark style amp in the future using the Coliseum 300 CCS output circuit; are there any concerns about using this arrangement? The concept of using a fixed negative voltage has crossed my mind, but I made my power supply board as a PCB and the prospect of having to rearrange it in this case is somewhat daunting since everything on that board is a tight fit. As for the use of feedback; I am rather stuck with it in this amp since I already have the 3 holes drilled into the chassis for presence controls and with the preamp circuits I am using, a simple treble roll off style presence would not sound very good.

[Dave H]

This is a much more intuitive method, thank you! That transistor; can it be practically any NPN? Like a 2N3904 or something similar? With the CCS I just built, I used the 450V/1A NPN that the Peavey Mace uses for it's phase inverter since I wasn't completely sure of what kind of voltages I would be seeing. How it as a whole turned out was a good news/bad news situation. The good is that it was noticeably louder and the high end response was much better than before; the bad is that it sounded like crap, there seemed to be some sort of blocking distortion and weird distortion going on. I will be giving your idea a go as soon as I get some time, so definitely thanks again.

I knew it would happen. You posted this as I was writing my next post. Check out the second simpler 'grounded grid' circuit. The emitter resistor will probably need adjusting to set the bias for the B+ you have.

[capehead]

I knew it would happen. You posted this as I was writing my next post. Check out the second simpler 'grounded grid' circuit. The emitter resistor will probably need adjusting to set the bias for the B+ you have.

I like this one too, it is very similar to the Marshall 9200 poweramp schem I posted earlier. The Marshall uses an MPSA43, which is a 200V NPN. Does the voltage rating need to be that high with either of these designs? From what I saw with my previous attempt, I don't think I do considering the highest voltage I saw was just over 10 volts on the cathode. It looks like I'll be able to try one of these out tonight, it seems you like the ground referenced one better, would that be my best bet? The first one I'm tempted to do as it shares many of the same features as a regular LTPI. I don't suppose that 5.6V zener value is too critical? I have 5V, 6.2V and 12V zeners on hand, what effect would it have to either raise or lower that voltage?

[loudthud]

I have 5V, 6.2V and 12V zeners on hand, what effect would it have to either raise or lower that voltage?

Just adjust the emitter resistor value to set the current which will also set the plate voltage at the outputs. 6.2V should give the best temperature stability.

[capehead]

Just adjust the emitter resistor value to set the current which will also set the plate voltage at the outputs. 6.2V should give the best temperature stability.

Awesome, thank you for the tip.

[Dave H]

I think I would choose the first circuit with the zener diode. The second circuit has the advantages Teemuk mentioned but its temperature stability isn’t very good. I calculate a 10V drift in plate voltage for a 30C temp rise. The circuit with a 6V zener should be a tenth of that (if the diode has zero temp coeff.) or better still pick a zener with a temperature coefficient of -2mV/deg C to compensate for the transistor Vbe drift.

I’ve just run a simulation on the circuit you posted (with 380V on the PI and 100k grid resistors on the 6550s) and the PI clips at about 50V peak which is very close to your bias voltage of -48V. It is possible that the PI is clipping first. You’d get a lot more voltage swing out of it by changing its 820 ohm bias resistor to 470.