6SL7 Phase Inverter Loading Problem

[GregS]

I've just built a bass amp with a power amp section capable of 160W of headroom. It uses 4 KT88/6550 power tubes and is driven by a long tail pair phase inverter with a 6SL7 tube.

The problem is I'm only getting about 60W rms output before the PI runs out of headroom and starts clipping. I've confirmed this with a dual trace scope with the input to one channel across the load and the other at one of the PI plates.

My assumption is that the signal output from the PI is being loaded down too much at the power tube grids. I have a bias arrangement with a bias pot for each power tube so each tube has a grid leak of 120k. The grids are separated from each other by the coupling caps. However, as I understand, they wind up in parallel with respect to the output of the PI as far as AC is concerned. This would mean each half of the PI is seeing an impedance of only 60K plus whatever else is in the ground path as I understand. I believe this to be the immediate cause of the problem.

I tried decreasing the PI plate resistors from 100K to 50k and values in between to see if that helped with the loading problem but to no avail.

I'm a little reluctant to start increasing the values of the grid leaks as I'm aware they need to be lower for KT88's/6550's. However, I'm not interested in trying to overdrive the power amp or squeeze the most power possible out ot the tubes so perhaps this is a reasonable option. I've seen other 4 x 6550 designs that have one gridleak of 100k for each pair of tubes. However, since I have a separate grid leak for each tube could I not use at least twice the value than for a shared gridleak? If that's not equivalent then I may have to go with only two bias pots (1 for each pair of tubes) to get a higher impedance.

Perhaps the 6SL7 is not up to the task of driving a power section like this? I understand the 12AT7 should be able to but that's an awkward switch because of the octal socket of the 6SL7. Are there any adapters for octal sockets to 9 pin. Would a 6SN7 be able to drive this load better? That would be an easy switch.

Tube cathode followers to provide the drive needed are out of the question as there's no physical room or heater capacity for another tube. I've read about MOSFETS providing the same function but I don't think either of these should be necessary since I'm not trying to drive grids positive or distort the power section or anything like that. Just trying to get max clean output power.

Right now I'm only able to get a 32 V peak signal to the power amp. I figure I need at least a 50 - 60 V peak signal to get to the max headroom of the power amp and get full output power.

Any advice with this dilema I'm facing would be much appreciated. Thanks.

Greg

Edit: should have added there is negative feedback going from the 8 Ohm tap of the OT the the other input of the PI. The series resistor is 150K and shunt is 4.7K

[soundguruman]

The 6SL7 needs more plate voltage, so that it clips after the power tubes clip. You are out of voltage.
You gotta increase the available voltage, not change the plate resistor.
If you are sure that it is the 6SL7, and not the previous stages that clip.
You should be able to get 60V from the 6SL7 before clipping.
6SL7 inverter is used in some very nice hi fi amps, like Levinson, and I don't see why not.

[Enzo]

I'd agree, you are just running out of room for the signal.

Pull the power tubes and see if the PI clips at the same level. I bet it does. And there would be no power tube grids to account for it.


On the other hand, you may be right, so do the experiment and find out.

[Chuck H]

With the power tubes pulled there wouldn't be any grids to consider. But there would still be the considerable load of 60k as compared to a more typical 110k to 220k. That still shouldn't keep you from getting the swing you need though. As noted, do be certain that it's the PI that's clipping and not some prior stage. If so then try a considerable increase in plate voltage on the PI.

[soundguruman]

More voltage on the BB+ is going to be the ticket.
lmao, my friend Kelly paid Bugle Boy $700 to set up the test, and produce the first pair of 6SL7 Bugle Boy tubes.
For his Mark Levinson, of course...and they sound sweeeeeeet!

[GregS]

Thanks for the replies.

It makes sense that the HT supply at the phase inverter is too low. When I was planning what the HT would be I was forgetting that the cathodes are well above ground in a LTP PI. In my case the PI supply is at 300V and the cathodes are at around 67V which only leaves an effective HT of 233V. I was fixated on the low impedance of the power amp because I was worried all along that it would be a problem.

I did scope prior to the PI as well and the clipping isn't occuring there. In fact I also scoped the input grid of the PI and it was not clipping there either which leaves the HT. I'll try a higher HT tonight and see what happens and report back on the result.

Thanks again. Much appreciated,
Greg

[pdf64]

What (roughly) is the output tube bias voltage?
As Enzo mentioned, see if the PI clipping disappears when the power tubes are removed, ie don't get PI clipping mixed up with power tube grid conduction/clipping.
Pete.

[GregS]

The bias voltage is around -62V so a 32V peak signal should be nowhere near being limited there. Thanks for the suggestion though and if increasing the HT doesn't solve the problem I'll try that out.

Greg

[soundguruman]

The bias voltage is around -62V so a 32V peak signal should be nowhere near being limited there. Thanks for the suggestion though and if increasing the HT doesn't solve the problem I'll try that out.

Greg

Put the schematic up on the site and we can tell you.
Your bias voltage may be too much, your grid bias voltage resistors may be choking the swing off.

[loudthud]

Make sure the PI tube is running enough current. You probably need something like 2.5mA per side with no signal. This restricts what B+ and plate load resistors need to be. PI's typically run the plate voltage a little lower than you would in a normal gain stage. This is because clipping on the bottom of the waveform doesn't matter if the power tubes are already cut off.

[soundguruman]

Make sure the PI tube is running enough current. You probably need something like 2.5mA per side with no signal. This restricts what B+ and plate load resistors need to be. PI's typically run the plate voltage a little lower than you would in a normal gain stage. This is because clipping on the bottom of the waveform doesn't matter if the power tubes are already cut off.

I don't think it's the current but rather, I think his voltage splitter for the output tube grids, the resistors may be too small a value, working like a ppi master volume?

[Gingertube]

The 60K "load" impedance is in parallel with the 100K plate loads of the PI, so effective 6SL7 anode laod is less than 40K.

With a KT88 bias voltage of -62V it means you need to swing at least 2 x 62 = 124V peak to peak. That would require 3.1mA peak to peak current swing (through that 40K) in the 6SL7. It just doesn't have that capability.

So YES get the B+ supply to the phase inverter up - a lot..., you said 67V at cathodes, at most negative going peak at anode you still want say 100V minimum across the triode (anode to cathode), add the 62 volts negative going swing to give anode idle volts and then another 62 for then positive going swing and you end up with a required B+ of absolute minimum of 290V. You want to be able to over drive the output tubes so I would go at least another 100V on top of that, say 400V minimum.

You still have a lack of current swing capability. Fix that (and the 6SL7 loading issue) by direct coupling MOSFET Source Follower buffers from the PI anodes. The buffers should idle at a minimum of say 3mA to give you that 3.1mA peak to peak swing capability with some reserve. This is the easiest most convenient fix.

There are other ways of doing this.
2 methods used by the HIFi crowd are:
1) direct couple the output tube grids to mosfet source followers (usually with current source loads) and the put the coupling caps between the mosfet gates and the PI with bias applied at the MOSFET gates. For individual bias adjustments you then need a source follower for each output tube. This requires a negative rail (for the source followers) of say 3 times the required bias voltage (-180V minimum)
OR
2) use a SERIOUS tube for the diff amp splitter instead of a 6SL7 or 12A_7. I've seen guys use a pair of triode strapped EL84 for the diffamp splitter running 10K to 15K anode loads. If you want to stick with octal tubes a pair of triode strapped 6V6 would also work well. You can then drive those low impedances on the KT88 grids with no trouble at all. This is the BFI (Brute Force and Ingnorance) approach but works really well.

Cheers,
Ian

[GregS]

Here are the results of some tests I just did. The reason some of these resistance values are odd is because I clip leaded other resistors in parallel with the original 68K to get the lower values.

1. Changed PS dropping resistor from original 68K to 52K

HT (at PI node) = 343V
Plates = 234V, 235V
Max clean signal at PI output = 74V p-p
Max clean signal at amp output = 70V p-p

2. Changed PS dropping resistor to 40K
HT (at PI node) = 374V
Plates = 255V, 256V
Max clean signal at PI output = 80V p-p
Max clean signal at amp output = 72V p-p

3. Changed PS dropping resistor to 34K
HT (at PI node) = 395V
Plates = 269V, 270V
Max clean signal at PI output = 84V p-p
Max clean signal at amp output = 78V p-p
Cathodes = 79.9V
Above tail resistor = 77.9V
Bias = 2V

4. Changed PS dropping resistor to 22K
HT (at PI node) = 436V
Plates = 297V, 298V
Max clean signal at PI output = 96V p-p
Max clean signal at amp output = 84V p-p
Cathodes = 88.6V
Above tail resistor = 86.6V
Bias = 2V

In the last test I'm still only getting 110W rms output into an 8 Ohm load by my calculation which is still far short of the 160W target output (This is what the transformers are capable of). Also I think I may be pushing it in the last scenario as the maximum plate rating of a 6SL7 is 300V.

Increasing the HT improves things significantly but there's still much farther to go. I notice the bias is over on the HT side at -2V which will restrict signal swing there somewhat so I may get a little more headroom by reducing the bias resistor for a more centred bias. I could also increase the plate resistors from 100K to 120K and get a little more signal swing that way (I think). However I don't think these changes will make enough difference.

It seems the problem of the low impedance at the power tube grids will have to be addressed. I'll draw a schematic of the power amp and bias cicuit and scan it and post it hopefully tomorrow. There are questions I have about the grid leaks that will be more clear with the schematic.

Greg

[soundguruman]

"1) direct couple the output tube grids to mosfet source followers (usually with current source loads) and the put the coupling caps between the mosfet gates and the PI with bias applied"

That's a sin. You are putting mosfets in a tube amp?

[soundguruman]

What is the bias voltage splitter resistors?
What is the DC voltage on the 6SL7 between the plate and the cathode?