Troubleshooting 6550 Jadis Clone

[multi]

Not sure if the correct thread to Ask.
I was asked to look at an amp that keeps blowing the 10 ohm resistors for measuring the current draw of the an amp based on a Jadis amp; has 8 x .1uf separate coupling caps to the output tubes 8 X 6550’s; 8 x 1 meg separate grid resistors on the 6550’s. The amp is based on the GEC kt88 x 10 amp same basic circuit.
1/Is there any problem using 1 meg instead of the max 47k recommended for the grid of the 6550’s in the tube data; I can see why they have used 1 meg because would be very little bass from .1uf into 47k. The original GEC uses .25uf with 47k and I think for HI fi would need at least .47uf.
2/ If you use separate coupling capacitors and separate grid resistors can you still use .47uf with 47k grid resistors or would the value of the caps need to be increased??.
3/ the amp is wired as triodes so take no notice of the Circuit showing Ultra linear; don’t have drawing program so have just used the original GEC circuit./ I had added the resistor values of the Jadis clone; the voltage amp and phase splitter to me look like it will need adjustment to the cathode resistor on the voltage amp to balance the phase splitter.
Original GEC circuit and circuit with part values used in the Jadis clones
Phil

 

[trobbins]

The manufacturer recommended grid leak is 47k per 6550, whether triode or pentode config.

I think you are saying that each 6550 sees a separate 1Meg grid leak back to the bias voltage, and a separate 0.1uF back to its cathode follower driving circuit.

If a 6550 starts to go gassy, or its coupling cap starts to leak, then that 6550 will see a lower bias voltage at its grid, and will conduct more current than the others.

For a new amp and parts, the amp you have would likely have ok bias current levels for all valves in parallel (given you have no individual bias adjustment).

As valves age, or if operated at high temp or max idle power, then its likely that valves can walkaway at some time if their grid is not appropriately clamped to the bias voltage.

The CR coupling network for original has 14Hz corner frequency for 0.25uF and 47k.

1Meg and 0.1uF has a corner freq of about 2Hz. Did you measure 1Meg for all the grid leaks, or just inspect colour codes?

 

[richwalters]

What type are those cathode resistors ? use wirewound. Metalfilm/oxide types are useless in the high peak pulse current condition and will go 'soft' when the stage is hammered. Don't ignore the peak uS pulse current created by the leakage parasitics of the output transformer, which the tubes have to sink on each cycle which damages modern deposited resistors. The 50K grid leaks for 6550 in fixed bias means so. Using higher values on these soft vacuum tubes is asking for runaway trouble, inconsistent running currents, or revert back to KT88's. In saying this, there are alot of various vendor 6550 types around; all having their peculiarities. Unfortunately I have never been impressed using cathode followers as output tube drivers, esp when it's either a 4x4 or whatever parallel p-p configuration. I've alot of 6550's A,B,C versions, the only way of running these mass market tubes is to keep the B+ down, and use them in parallel p-p or even triple p-p as I am currently working on a design.

rich

 

[DF96]

1M grid leak resistors for a power tetrode output is asking for thermal runaway; it is seriously incompetent design. You can just about get away with 470k resistors for a genuine EL34 pentode, but for some reason tetrodes have more grid current. I would use 220k maximum; 100k might be safer, especially if the valves are poorly ventilated. Fixed bias needs smaller resistor values because the bias cannot adjust itself.

 

[petertub]

This amp is designed by a pig that has been smoking wallpaper.
There is no point measuring cathode current on each 6550 as they are biased with a single source. The amp needs a matched octet of 6550 no more no less.
Grid resistor should be at most 50k ( according to 6550 spec)Above that is asking for trouble.
And there is more ...

 

[kstagger]

I remember Jadis being famous in the late 80s for blowing up tubes - the blame was put on the Chinese KT88s of the time. Now that I've seen this - if this is the Jadis design - I'm thinking the blame can be shared.

 

[kevinkr]

I worked on a number of Jadis Amps including the Defy7 back in the late 1990s and concluded after working on a few that even repaired they were a disaster waiting to happen. (The owners of the amps I worked on were not willing to consider modifications to the design that would have made them safer)

 

[kmtang]

Hello Kevin,

Wondering if you can share with us for those potential issues with the Jadis amps?

Johnny

 

[soulmerchant]

Read this:

Fusing Output Valves

Ian

 

[soulmerchant]

I remember Jadis being famous in the late 80s for blowing up tubes - the blame was put on the Chinese KT88s of the time. Now that I've seen this - if this is the Jadis design - I'm thinking the blame can be shared.

Yeah. A previous employer of mine owned a Jadis DA60 which had the wonderful EI KT90 tubes. The sound of the DA60 with these tubes was pretty darned impressive on his electrostats.

However, after a while he had some "problems" with one or the other tube starting to red-plate so asked me to take a look into it.

Once I realized how the fixed bias was set across all paralleled output tubes, I could only tell him the bad news... and at that time it seemed impossible to find matched quads of the same EI KT90's

Ian


btw - at least to me, individual coupling caps makes little sense unless you can bias each tube individually... and that is somewhat tedious to implement. Maybe the builder had thought of it but sort of gave up?

 

[soulmerchant]

I remember Jadis being famous in the late 80s for blowing up tubes - the blame was put on the Chinese KT88s of the time. Now that I've seen this - if this is the Jadis design - I'm thinking the blame can be shared.

I would assign 95% blame to the designer.

With these kind of designs, when there is some moderate imbalance, the tendency will be for that imbalance to accelerate... get worse faster and faster... the weak tube will conduct more and more.....

Then BANG out of nowhere you have a tube re-plating. At least there is a "fuse" that will blow (the 10 ohm resistor). We can thank the designer for that, but it does look pretty darned cheap doesn't it?

So replacing that little 10 ohm resistor with a wire-wound might not be the best idea, considering the possible consequences.

 

[kevinkr]

Hello Kevin,

Wondering if you can share with us for those potential issues with the Jadis amps?

Johnny

They've already been spelled out earlier in the thread. It's enough to say that the output tubes were run under improper conditions and failed frequently as a result. Replacing all of the fried resistors in the Defy 7 was a chore as the PCB had to be pulled out most of the way to get at cathode and screen resistors.

 

[multi]

Jadis clone triode 6550's x 8

Thanks trobbin and petertub
Sorry I should have told has separate bias pot for each 6550.
I did measure each 1 meg grid leaks on the 6550’s
Would it be worth converting to 47K grid leaks? With separate coupling caps and 47k; one friend said I would need 8uf caps? I found with multiple El34’s was important to use separate coupling caps as so many current production EL34’s go gassy and red plate the separate caps stops the lot being destroyed.
Are 12Au7’s capable of driving triode connected 6550’s with 47k grid leaks?
I was given a circuit supposed to be for this amp but has a lot of differences; I have written in all the part values as on the amps I have.
The 100k resistors shown connected to the 1 Meg grid leaks are not used in the amp I have.
Another Jadis KT88 amp circuit had 1 meg grid leaks and driven by 12Ax7’s; why do amp makers use these circuits? they Give a pleasant distortion while the output tubes self-destruct.
Thanks Phil

 

[petertub]

Yes, i think you should replace the grid resistors to 47k, this will prevent thermal runaway.
Upping the caps is not neeeded, in fact too large ones might create sideeffects that
is dangerous during power off.
( repeated quick-reply as it seemed disappear)

 

[soulmerchant]

Hi Phil

As you probably have realized by now, mr. jadis was somewhat eccentric. That probably helped him sell his amplifiers...

Good to know you can bias each tube separately. Then you can consider using wirewounds on the cathodes. You might consider posting a chicken-scratch of your bias circuit.

F=1/(2*pi*r*c)=1/(2*pi*47000*0.00000047)= 7.2 Hz

So 47k and 0.47uF should be fine. Make sure the coupling caps are rated high enough. I would be looking for Metalized polypropylenes rated for 600V DC or higher.

Ian

 

[kevinkr]

I'm not at all sure that cathode follower can comfortably drive ~12K ohm load to 50 - 60Vpk (four 47K in parallel). I would change those 1M resistors to 200K and see if that is sufficient to solve the problem. You may want to look at beefing up the CF, something like the Russian 6N6P running at 10mA or so per section would be up to driving a relatively low resistance quad of grid resistors.

 

[6A3sUMMER]

The 6550 in fixed bias mode is rated for 50k Ohms (per tube). That is 12.5k Ohms for 4 tubes! Not good, hard to drive.

The 6550 in self bias mode is rated for 250k Ohms (per tube). That is 62.5k Ohms for 4 tubes. Much better!

I would suggest modifying for self bias mode. 8 cathode resistors, and 8 cathode bypass capacitors.
Probably not enough room for that, but that is the way it goes.
The power output will drop because the plate to cathode voltage is less.
But if 8 6550s in push pull is not enough power, you need more efficient speakers (or a smaller venue for a music concert).

Self bias allows the tubes to be able to work in a much more balanced mode (good power sharing between the tubes). And the current balance between push and pull primary windings would be much better too, so there would be no/less saturation of the output transformer.
So often, "matched" tubes are not matched at the operating points you will be using them.

You could up the B+ to get the power back, but you would have to re-adjust B+ for the driver tubes.

With all these changes, I bet the distortion would be better due to a better match of tubes, and a better match from side to side of the push pull transformer.

I did not notice the operating voltages and operating currents of the 6550s, but a check
of these conditions is in order, especially since some of todays tubes can not meet the original specifications. The may glow red, short out, go gassy, etc.
It can be hard to get good tubes today.

 

[richwalters]

I'm not at all sure that cathode follower can comfortably drive ~12K ohm load to 50 - 60Vpk (four 47K in parallel). I would change those 1M resistors to 200K and see if that is sufficient to solve the problem. You may want to look at beefing up the CF, something like the Russian 6N6P running at 10mA or so per section would be up to driving a relatively low resistance quad of grid resistors.

Spot on.....the heart of the problem. Also Miller effect in AB, the cathode follower has only the sink resistor value on the 'off' cycle. There is little point trying to get performance from an output stage that has one-over the driver stage.
I am currently designing an amp using a double parallel triple 6550's with a totem pole driver, sim to the transformerless designs. More complex but far lower source impedances for the 6550's.

rich

 

[6A3sUMMER]

The 6550 and 6550A in triode mode are rated for 500V maximum on the Screen and Plate.
Your schematics show 550V.

The control grid resistor must not exceed 50k Ohms for EACH grid (for the 'adjustable' fixed bias that you are using). 1 meg Ohm is not going to prevent control grid current from causing thermal runaway of one or more 6550s. There goes another 10 ohm resistor,
another 6550, or both.
I hate to admit it, but years ago I had tube-rolled a circuit that was an EL34 to use a 6550.
I forgot to change the 270k control grid resistor to a 47k resistor. I heard awful sounding music, and then I saw the 6550 plate glowing red, so I immediately shut the amp off.

Using 4 + 4 50k grid resistors will require a complete re-design of the cathode followers, probably with a more robust tube, standing lots more current.

If you do a re-design to have self bias for each 6550, then each grid can use 250k Ohms
maximum (a single 62k resistor could be used for 4 grids; 2 62k resistors for 8 grids).
That is much easier for a 12AU7 cathode follower to drive.
Except for needing to find room for 8 resistors, and 8 bypass caps, that is by far the easiest thing you can do to make the amp more reliable.

Common practice for triode mode is to use at least a 100 Ohm resistor from each screen to its plate (8 resistors).

The 1 Meg plate resistors of the 12AU7 that drives the grids of the 12AU7 cathode followers is really starving that stage. You need more current there. Of course, that
changes what the plate voltage is, so the grids of the followers will be much higher, and the cathodes of the followers will be much higher. Now you will need a separate filament
transformer secondary that is elevated by DC voltage that is nearer to the cathode DC
voltage.

From what I see on the "original" schematic, I am not sure that it is exactly what was in
the original Jadis amp. For example, the note about the meter series resistor that says
it is dependent on the meter sensitivity is not something a manufacturer would normally do. They would already know what meter they were going to use. For another thing,
the feedback circuit is shown at the input tube, but not at the output transformer secondary.

I short, I do not see the circuit as shown would be reliable without major changes.
Every change you make in one spot can require a change in another spot.

I look at this amp as having potential for great performance, but not without some changes.