WARNING! Risk of Death with Mismatching an SVT's Speaker Load!!!

[jazbo8]

As promised, here are my attempt at drawing the composite graphs for the Ampeg SVT 6550A. Sorry the graphs aren't annotated, if you have questions, please let me know. In any case, here is a brief description of what you are looking at:

Note: The following results assume zero power supply sag, zero voltage drop due to screen resistors, ideal OPT with purely resistive load. The actual circuit performance may be quite different.
- Class A and Class B load lines (light blue and dark blue), note the Class B line is largely covered by the Class AB1 load line;
- Class AB1 load lines (turquoise), you can see it extends to the saturation voltage of the other side (Min Ep);
- Point A - quiescent point at 700V, 16mA (nearly Class B bias!)
- Point B - Max Ia, Min Ep1, tube at saturation;
- Point B' - the other tube at cut off;
- Point C - tube at saturation, corresponds to Point B;
- Point C' - the other tube at cut off, corresponds to Point B'.

Ampeg SVT with 5K load (only one pair shown, there are three pairs in the actual amp):


Ampeg SVT with 10k:


From the above graphs, we can see that for the 5k load, Epmax = 1400-73.3 = 1326.7V and for the 10k load, Epmax = 1400-27.8 = 1372.2V, so not that much of a difference really. But the screen grid dissipation is a different story as I shown in my earlier post.

I've been looking for specs on breakdown voltage for a sound 6550. I haven't been able to find a good reference for just how much voltage a good 6550 will stand. Anecdotal experience tells me it's got to be at least a kilovolt.

Also found this at diyaudio. If you follow his formula, then the 6550A ought to be able to handle 660*2.5=1650V before arc-over occurs. Assuming the amp is working fine, then simply doubling the output load should not cause a catastrophic failure, unless the switch is done on the fly without any precaution, then it would not surprise me that something bad could happen. Talk about a large d(i)/d(t)!

Since I have never even work on an Ampeg, my take on the situation may very well be far from reality... so experienced designers and techs please feel free to set me straight.

[nickb]

Earlier you had mentioned that you're not familiar with the SVT's internals. Just so you know, it does have some of the "immortalizing" mods designed into the circuit. Of particular interest to note are the fusible-link screen resistors and screen diodes, and diodes across the OT primary.

example: SVT-2 Pro & SVT-CL:

[ATTACH]25892[/ATTACH]

Slightly off topic, I hope you'll forgive my indulgence.

Maybe someone can explain it to me: The resistor / diode is only going to do something where the plate goes open or the tube stays hard on. In that case the diode goes followed by the resistor. In normal operation the drop is so small as to be negligible so the diode isn't doing much. Also, will the diode blow before the screen is damaged? I doubt it. Then, it's quite possible for the diode to fail short rather than open in which there's no protection at all.

Seems you'd better off with just a fusible device ( fuse, resistor, PTC thermistor) set to a low current. But then I'm maybe missing something, and that wouldn't surprise anyone...

[bob p]

The most effective of the "immortalizing" mods in my experience is the 220 ohms resistor in each cathode of the 6550īs, although is not represented in the diagram.

Yes. Excellent point. The SVT is a very well designed amp.

The current production SVT are even better designed than the originals. The current production amps have current sensing resistors in the cathodes of each output tube, and they also employ a set of comparators that monitor current flow through each output tube. If something really bad happens, like a tube failure that causes increased current flow through a tube, the amplifier recognizes the problem and engages it's protection circuitry, putting the amp into shut-down mode instead of allowing the amp to self-destruct. I think that it's a very well thought-out design.

I didn't post the full schematic, because the full schematic and assembly diagram for these amps is comprised of eight multiple-paged PDF files. But since we're talking about the protection circuitry I think it's relevant to post another file that shows what we're talking about. This example is from an SVT-2 Pro, though the same protection circuitry is used in the other current-production SVT models. (see attachment at bottom)

Real life is always more complicated than theory. Consider this possible scenario.
(1) The tubes are not quite matched (almost certainly true)
(2) Therefore one tube screen melts before the other
(3) Therefore, at some point only one tube is working so now you have large d(i)/d(t).
(4) High d(i)/d(t) causes large voltages and things start breaking down and, as I think RG already said in one way or another, it's gets really hard to predict, but none of it is good.

The comparator circuitry is designed to specifically address these kinds of problems. It shuts the amp down if condition (1) exceeds a pre-defined limit, or if condition (2) were to occur, thereby preventing the occurrence of (3) or (4).

[bob p]

Still working on the composite load lines, but in the meantime, I found one of R.G. old posts that's relevant to the discussion in addition to what he posted earlier. So the tubes do see nearly double the B+ under normal conditions (more than what the SE graphs suggest), and perhaps even more if the OPT windings have significant parasitic inductance. But even with twice the B+ on the plate, it should not kill the tubes or cause damage to the OPT. So unless there are actual field data that suggest otherwise, I am still going with the "screen dissipation kills" thesis.

Thanks for posting that link. I remember that discussion. In fact, I was using the search function to try to find that exact post the other night but I wasn't able to find it. I guess I should have been searching the threads for flyback diodes and/or looking at the MOV threads.

Here's what he said:

Steve is correct, as usual. Can't put the diodes on the B+ side. Those primaries swing both directions relative to B+.

Those flyback diodes work by clamping the down-going side of the transformer winding to ground. The power tubes can only pull down, from B+ to ground. When you try to interrupt this current, the inductive kickback flips this voltage upwards, above B+, to try to force the tube to accept the same current it was carrying. The voltage flys upwards until the current can be satisfied. This also makes the voltage on the non-current-carrying half head downwards from B+, because the two primary halves work in see-saw fashion.

Power tubes in Class AB normally experience something a bit less than twice the B+ voltage. When one tube pulls down the voltage across its half-primary, the other half primary swings up by an equal amount, but above B+. The inactive tube can see nearly twice B+; it's lower by the "saturation" voltage across the active device, and that can be as low as 50V for some power tubes.

On a current-stop transient the transformer half-primary that was being pulled down flips in voltage above B+; so does the inactive half-primary, so the inactive side heads for ground. Now the clamp diodes come it. As the inductive kickback forces one side above two times B+, it also forces the other side below ground. When that happens, the flyback diodes conduct and clamp the down-going half to ground, so there's no more than B+ across that side. By see-saw action, that also limits the up-going side to two times B+, not a lot more than happens in normal action.

What this action cannot protect against is the inductive kickback of the transformer's leakage inductance. Leakage inductance is by definition the "inductance" that is unique to one half-primary, not the other. The leakage inductance component of the flyback voltage cannot be clamped by the other half primary. On the other hand, the energy carried in the leakage is enormously smaller, so even an arc forming will not cause much burning.

The inability to clamp leakage inductance flyback spikes is why I like a chain of high voltage MOVs across the primary. If their voltage is more than twice B+, they never conduct in normal operation, but when they are tripped over, they eat any discharge energy including leakage inductance energy. Diode clamps are nice, but the leakage inductance flyback may degrade them over time and eventually short them. This is why those diode clamps are usually strings of three 1N4007s - you have two chances to not die. One 4007 may die at something just over 1KV, two gives pretty certain normal operation, and three gives you an active spare.

It's interesting to note that the SVT schematic that I posted earlier shows that Ampeg is putting 3 x 1N4007 across the output transformers primary B+ and plate leads. The 1N4007 are rated at 1 kV each, so the string of 3 of them suggests that Ampeg is concerned about cases where the OT might see over 3,000 volts on a peak. Looks like the SVT OT is designed to withstand one hell of a lot of voltage.

[bob p]

Slightly off topic, I hope you'll forgive my indulgence.

Maybe someone can explain it to me: The resistor / diode is only going to do something where the plate goes open or the tube stays hard on. In that case the diode goes followed by the resistor. In normal operation the drop is so small as to be negligible so the diode isn't doing much. Also, will the diode blow before the screen is damaged? I doubt it. Then, it's quite possible for the diode to fail short rather than open in which there's no protection at all.

Seems you'd better off with just a fusible device ( fuse, resistor, PTC thermistor) set to a low current. But then I'm maybe missing something, and that wouldn't surprise anyone...

Interesting question about theory vs. practice.

In real world conditions I think that the resistor goes first, because it's wattage rating is low enough that it acts as a fusible link. It'd be interesting to hear from Enzo, Leo and/or John about this, as they've serviced more SVT in their lifetimes than just about anyone I can think of.

[Gregg]

The 1N4007 are rated at 1 kV each, so the string of 3 of them suggests that Ampeg is concerned about cases where the OT might see over 3,000 volts on a peak.

2-3kV diodes doing the same job are often found in other amps with "normal" voltages so I guess they put 3xIN4007 because they are just cheap and available.

[bob p]

Could the common factor among amps of different designs be something as simple as that 3kV is the voltage range in which the insulation is prone to fail?

[Pedro Vecino]

Interesting question about theory vs. practice.

In real world conditions I think that the resistor goes first, because it's wattage rating is low enough that it acts as a fusible link. It'd be interesting to hear from Enzo, Leo and/or John about this, as they've serviced more SVT in their lifetimes than just about anyone I can think of.

I'm service several units of Classic, VR, SVTII Pro, etc. and have never found one of those components broken or burned. It can happen but I have never seen. When the comparators have not worked, the problem has gone to the cathode resistor.
Although theoretically is a very good design, I am surprised by the number of problems that occur with the tubes.
We know that the 6550 Sovtek or Svetlana not have the mechanical/electrical quality of General Electric, but other designs of similar amps like Marshall BVA400 are much safer in this aspect.
In the SVT is also fairly common the glass damage produced by the tube retainers. Itīs convenient to cover them with heat shrink tubing to prevent metal/glass contact.

[jazbo8]

Pedro,

Where are you in Spain?

Jaz

[Pedro Vecino]

Hello, Jaz. I live in Valencia (fallas, paella, firecrackers...)

[R.G.]

Could the common factor among amps of different designs be something as simple as that 3kV is the voltage range in which the insulation is prone to fail?

Yep.

[jazbo8]

Nice! Haven't been there in years, perhaps a trip is order. Do you know any good shops that deal with vintage guitar, and amps in Madrid? I go there regularly but all I have seen so far is junk - must be looking in the wrong places...