Paraphase to fixed bias?

[EFK]

I think I'm looking at a paraphase PI here, and I'm wondering if it is possible to covert this cathode bias power amp to fixed bias (without noticeably changing the PI)?

[Steve Conner]

This circuit has a tiny cost-cutting measure that makes it hard for you to convert to fixed bias. They've used the same coupling capacitor to drive the PI second stage grid and the power tube grid. In cathode bias that works, because both grids are at 0V, but now you want one grid at 0V and the other one at -21 or whatever.

So, if you want to convert to fixed bias, you'll need to add one more resistor and capacitor.

[EFK]

I guess you would be talking about the .05 caps off of 1 and 6 - or just the second stage one off 6? So if I wanted to do this, what would I need to add and where would I add it? Thanks Steve!

[loudthud]

A problem you'll run into after you do the conversion is that any ripple on the bias supply gets injected into the second side (pin 7) of the phase inverter. Then it gets amplified and sent to only one of the output tubes. Thus the ripple doesn't cancel as it normally does in a fixed bias amp. Weber had this problem on one of their kits. Their solution was a super big filter cap on the bias supply.

[Steve Conner]

Loudthud: Only if you try injecting bias into the bottom of the divider that feeds the right-hand PI grid, which is a silly idea.

The better way is to leave that divider referred to ground as it is, but disconnect the power tube grid from the top of it, and instead feed the power tube grid from the PI plate, using another 0.05uF coupling cap and another 470k resistor to the bias supply.

(I'm talking about the side whose plate is pin 1)

The component count is the same, because you would have needed a capacitor and resistor to keep the bias out of the PI grid anyway. But the ripple doesn't get amplified.

[loudthud]

Good point Steve.

Here's a link to the old thread: http://music-electronics-forum.com/t14986/

[EFK]

Thanks guys. I'd like to avoid having to add another big filter cap in there. I understand the process mentioned above insofar as disconnecting the V5 grid connection but leaving the voltage divider alone. You mentioned another .05 and another 470K off the pin one PI plate feed - so they are going to now tie to the V5 grid, and the 470K is going to tie to the other 470K going to ground, right, except instead of going to ground it will be tied to the new fixed bias supply? In my mind I kind of envision this part along the lines of a typical marshall .1/220K layout feeding the power tubes, symmetrical, but I'm not sure where you would have me take the .05 and then 470K off the PI plate - am I correct in assuming it would be between the .05 coupling cap and the voltage divider, or would it be taken off the plate before it?

[Steve Conner]

am I correct in assuming it would be between the .05 coupling cap and the voltage divider, or would it be taken off the plate before it?

The plate before it.

[redelephant]

Tubes are best off with cathode bias...but u know that and want the power so I'll shut up about that.
I just had a thought that when u create your negative voltage for the bias, why not make the PI into a LTP and bias it with a resistor or CCS to that neg rail? Complicates things a bit, but once the iron is hot...?

[EFK]

I'm not necessarily after a lot more power, maybe just a little and less compression, a little more clean and mostly curiosity. I've got a bunch of these Supros and they're all cathode bias, like to see what this one would sound like fixed. I think if I went to ltpi I'd get a good deal more gain along with it, wouldn't I? Not really after gain here, more articulation and some more clean is always good though.

Do I have this sketch right? Appreciate the help!

BTW, assuming this is correct, is there any reason to not then consider the values of the bias resistors - 470K? Generally the ltpi value for 6L6 is 220K, I think I may have seen 270K on some older amps somewhere. What is the effect of this high 470K value on what is now a fixed bias amp?

[Steve Conner]

Yeah, that's right, as far as I can see.

Reducing those resistors from 470k will reduce the gain of the amp a little, but make the tubes more stable against thermal runaway.

Tube datasheets give a maximum value of the resistor that guarantees no runaway. In fact they give two values, one for fixed bias and one for cathode bias, and the value for fixed bias is always smaller. So you may need to bring the resistors down to 220k.

[redelephant]

I see ur point that u want to try this since u have other cathode biased amps, besides it's always fun to play around

The origianl PI probably has plenty headroom so no reason to change it. It's just I don't like circuits that depend on feedback for proper function, espescially in stages where distortion can mess things up. Then again, that may actually be a good thing, I've never tried a PI like this so can't say. But regarding more gain in the PI:

For maximum headroom at the amp's output, more PI-gain is often better than the opposite. The reason is u want the
output to crunch much before the stages ahead. That means the preamp and PI stages are clean even when the output
reaches it's max. That means the PI must be able to swing much more voltage than is necessary for getting the output tubes to reach total clip. If the PI does not have gain, u risk getting PI distorting around the same amplitude as the output, and the clean headroom will be less.
I know...Any PI no matter how much it can swing will clip at the same time as the output tubes because of the grids of the output tubes suddenly pull current, and the load becomes a very low impedance. If using a buffer between the PI and output tubes the PI will still swing linearly while the output tubes clip.

The thing that sets some guitarists off is with more PI gain, the volume knob must be set lower for the same volume, and the appearent headroom seems less because the output crunches at 3 instead of 5 on the knob. But if u consider the loudness at the speaker output, more PI gain gives cleaner sounds before breakup, thus more headroom.

Btw the datasheets claim max resistance when fixed bias is 100kohms, but that's been ignored before, so probably no biggy unless u go way off.

[octal]

I'm not necessarily after a lot more power, maybe just a little and less compression, a little more clean and mostly curiosity. I've got a bunch of these Supros and they're all cathode bias, like to see what this one would sound like fixed. I think if I went to ltpi I'd get a good deal more gain along with it, wouldn't I? Not really after gain here, more articulation and some more clean is always good though.

Do I have this sketch right? Appreciate the help!
A long tailed pair actually will have less gain than a paraphase.

BTW, assuming this is correct, is there any reason to not then consider the values of the bias resistors - 470K? Generally the ltpi value for 6L6 is 220K, I think I may have seen 270K on some older amps somewhere. What is the effect of this high 470K value on what is now a fixed bias amp?

A long tailed pair will have less gain and less voltage swing than a paraphase due to the inherent local NFB in the tail.

You might try increasing the 270K/12K resistors...especially if you're going to lower the bias pulldown resistors to the output tubes. The 270K+12K creates extra loading on the output to the PI. As long as you keep the ratio the same, higher values will be OK. Maybe 470K and a 20K 1% resistor (or whatever the closest 1% value is.) As Steve said, lower bias pulldown resistors will keep the output tubes more stable. 220K is already twice the datasheet value for fixed bias 6l6, so perhaps 470K would be pushing your luck a bit. Just a thought. You could also do semi-fixed bias and add some lower value cathode resistors on the output tubes to stabilize things a bit.

[EFK]

I had a chance last night to implement this. Little more of a pain in the a** in the Supro as it's two separate chassis segments - preamp chassis up top and the power amp chassis in the bottom of the cab - but it worked out well. I measured the AC coming off the PT before the changeover and it was almost identical to my JTM45, so I basically used a JTM45 fixed bias circuit with just two resistor changes (10K instead of 15K b/t the caps, and 47K off the bias pot instead of 56K). This works perfectly with 6L6GC as I can bias them at 70% and the pot travel is right about in the middle. I also swapped OT for a bassman transformer while I was at it. I did change out the 470K resistors for 220K and everything is running very smoothly. I'm going to put it through a really thorough listening workout today.

I'm a little unclear as to what you are talking about insofar as the two voltage divider resistors, the junction of which is feeding the second input grid. How will the higher values reduce loading? I would have though it would be the other way around. I don't think I have 20K but I know I have 22K, although I think that would change the ratio a bit. Could do two 10K in parallel.

[octal]

One section of the tube has to drive the combination of the "divider string" (270K +12K) and the grid bias pulldown resistor (220K) in parallel-which makes a total load of something slightly greater than 100K- can't be bothered to do the math at the moment. That's a really low impedance load for a 12Ax7 with a 270K plate resistor to drive. The reduction of the existing pull down resistors from 470K to 220K will also throw the balance of the PI off- you may need to tweak the ratio of the divider string to get the balance back in line.

You can check the PI balance by feeding a sine wave generator into the amp and measuring the signal going to the grid of to each output tube. (A 'scope is good for this, but something as simple as DVOM set to read AC volts will work fine too, assuming you keep the input frequency around 400Hz or lower.) You can trim the value of the the 12k resistor in the divider string to bring it into balance. It needn't be perfect, but it should be close.

I'll assume that you were momentarily confused when you wrote about making a 20k resistor by putting two 10Ks in parallel Series will do quite nicely!

Of course, if you like the way it sounds now, perhaps you should just leave well enough alone!

Nathan