Fixed Bias LTP with feedback

[Gaz]

Hi, just wanted to dummy check something I'm working on. It's a Hiwatt-like amp, but I don't have the extra triode to DC couple, or use as a voltage reference (regulator?) like you can see here and here, respectively.

I've breadboarded this and it seems to work, but I just wanna make sure I'm not doing something stupid with how I'm injecting the feedback. This is how I'm doing it. AC coupled with the bias tapped off the B+.



Does that look okay? It seems to work, I just can't find any other examples of fixed bias like this with feedback.

Thanks!

[Chuck H]

About the only thing I could say is that ideally you would have a capacitor in the NFB line to block DC. Because there's a purely resistive path between the speaker and the HV there's a small amount of DC on the speaker (depending on HV source and the speaker load value). It's really tiny and you might not even notice over the usual amp noise.

[pdf64]

Most every other amp with a global nfb loop around its power amp does that though; I guess it’s a matter of magnitude, when it becomes excessive

[Helmholtz]

Does that look okay?

Sorry, no. R26 is shorted which puts full B+ on both grids.

I don't understand why you don't take grid bias from a small additional common cathode resistor (e.g. 470R) like with any Marshall type LTPI.

[Chuck H]

Most every other amp with a global nfb loop around its power amp does that though; I guess it’s a matter of magnitude, when it becomes excessive

That's a good point, and I never even considered that before now when looking at this circuit!?! I think it's because the usual circuits are tried and true while this one was brought into question.

[nickb]

There is an extra connection between the top of R26 and the grid of the right triode that should not be there. The bottom of the 220k could be taken to ground. The 22uf doesn't need to be that big , 100nF would do. I have a concern that hum from the power supply could couple into the right had triode's grid so it might be better to split the 1meg upper divder resistor into say 100K ( to supply) and 910K with 10uF cap from the joint to ground to filter the supply a bit.

[uneumann]

Hi, just wanted to dummy check something I'm working on. It's a Hiwatt-like amp, but I don't have the extra triode to DC couple, or use as a voltage reference (regulator?) like you can see here and here, respectively.

I've breadboarded this and it seems to work, but I just wanna make sure I'm not doing something stupid with how I'm injecting the feedback. This is how I'm doing it. AC coupled with the bias tapped off the B+.

[ATTACH=CONFIG]58050[/ATTACH]

Does that look okay? It seems to work, I just can't find any other examples of fixed bias like this with feedback.

Thanks!

So - without doing a sim or anything, I suggest you simply remove the CF triode in the HiWatt design. Short the grid signal to the cathode line that feeds the PI. Remove the 100K cathode load - you don't need it. Basically, you'll keep the whole DC bias thing working and simply not have it buffered by the CF. Given the values of the resistors involved and the hi-Z of the PI input, I suspect it will work pretty well - although test or sim would be needed to confirm that.

Alternatively, if you don't have a triode to use as the CF - why not use an LND150? Problem solved that way too.

[Gaz]

So - without doing a sim or anything, I suggest you simply remove the CF triode in the HiWatt design. Short the grid signal to the cathode line that feeds the PI. Remove the 100K cathode load - you don't need it. Basically, you'll keep the whole DC bias thing working and simply not have it buffered by the CF. Given the values of the resistors involved and the hi-Z of the PI input, I suspect it will work pretty well - although test or sim would be needed to confirm that.

Alternatively, if you don't have a triode to use as the CF - why not use an LND150? Problem solved that way too.

Thanks for the thoughts, yes, that's really obvious now for me to see the CF is unnecessary for DC coupling! I would actually do that, but I've since moved the master volume after the tonestack recovery triode to get some more gain, and so I now need to AC couple the last stage to the PI anyway. I hadn't considered using a transistor either, but for the same reason, I am going to try fixed bias off the HT.

People claim the fixed bias PI is part of the sound, but maybe I should see for myself... Cathode biasing would be less parts.

[pdf64]

I think the circuit is self (cathode) bias, not fixed bias.
ie if cathode current should rise, cathode voltage will increase too, thereby reducing cathode current.

[Gaz]

There is an extra connection between the top of R26 and the grid of the right triode that should not be there. The bottom of the 220k could be taken to ground. The 22uf doesn't need to be that big , 100nF would do. I have a concern that hum from the power supply could couple into the right had triode's grid so it might be better to split the 1meg upper divder resistor into say 100K ( to supply) and 910K with 10uF cap from the joint to ground to filter the supply a bit.

Thanks, Nick. Yes, the wire shorting R26 is an error in the drawing.

Here's the original scheme I was working from, from Merlin Blencowe's book:



I am just trying to add the Hiwatt 10K/2.2k feedback to this. I thought the way I had it hooked up was filtering the HT, but I guess not!

[nickb]

Thanks, Nick. Yes, the wire shorting R26 is an error in the drawing.

Here's the original scheme I was working from, from Merlin Blencowe's book:

[ATTACH=CONFIG]58061[/ATTACH]

I am just trying to add the Hiwatt 10K/2.2k feedback to this. I thought the way I had it hooked up was filtering the HT, but I guess not!

The HT ripple does get a lot of attenuation from the 1Meg feeding into a quite low impedance of 2k or so. It was just a concern - something to look out for. FWIW a significant difference between your/ hiwatt design and Blencoe's is the big tail resistor of 68k.

[Gaz]

The HT ripple does get a lot of attenuation from the 1Meg feeding into a quite low impedance of 2k or so. It was just a concern - something to look out for.

Yeah, I didn't notice any noise, but obviously would like to set it up ideally if possible. Does this look better?

FWIW a significant difference between your/ hiwatt design and Blencowe's is the big tail resistor of 68k.

Yes, mine is with 12AT7 instead of 12AX7, and I with the value in my scheme I have the bias voltage right at about 2V like the orginal Hiwatt.

[Helmholtz]

Does this look better?

Sorry, no. C18 is AC shorting the right side grid to ground, thus disabling NFB.

[nickb]

^^^^This. Delete C18.

[Gaz]

Sorry, no. C18 is AC shorting the right side grid to ground, thus disabling NFB.

^^^^This. Delete C18.

Ok, thanks, I just thought C18 was needed for filtering the HT like in Merlin's schem, but of course I see now what you mean with the NFB being disabled.

Is grounding the 220K really much different than my original scheme?