Blown 1 OHM current-sense resistors

[Gaz]

Thanks, frus, that's exactly what I needed to know, and I appreciate you humoring my elementary question. I didn't know exactly how to Google it.

Yes, the amp does have a 1A Slo-blo B+ fuse in on leg of the bridge rectifier on the AC side. Since it didn't blow before the resistor, and a few have explained to me why it possibly didn't, I thought a fast-acting fuses there may be more reliable. I thought I might try it like the JCM900 schem g-one posted with the indicator lights, and individual fuses. This way I could know which tube was the culprit as well. Thanks again.

[Jazz P Bass]

Man, I had to look that one up!
Sorry, Gaz.
snarky - definition of snarky by the Free Online Dictionary, Thesaurus and Encyclopedia.

[Leo_Gnardo]

I thought I might try it like the JCM900 schem g-one posted with the indicator lights, and individual fuses. This way I could know which tube was the culprit as well. Thanks again.

In series, like old fashioned christmas tree lights. Every electron that passes through one, passes through the one before it, and the one after it. And if one goes open-circuit, no glow bro.

What worries me, is those 1 ohm resistors will open up before the fuse will. I suspect that's what's getting on your nerves Gaz. You did your best to select what you thought was a good sensing resistor, and in theory it was. In practice, not so much. No problem putting in a JCM900-style indicator. Just put in tougher sensing resistors and all will be OK.

Snarky - isn't that the clip on guitar tuner that's so popular now?

[R.G.]

You will never be able to protect every component in a tube amp against every type of failure.
In that case, you have to accept certain failure modes taking out components.

Well, yes, you always have the chance of any part having a sudden, random failure. That's what failure rate data is about.

What you CAN do is ensure that single failures do not start a chain of destruction that makes things much worse to repair. Things which are just going south because they're too hot or some such can be protected against eventual damage.

As an example:tubes deciding to short and bias supplied dying can't be prevented. They can only be made less frequent by designing for reliability. What can be prevented is losing your filter caps, rectifier, power transformer or output transformer by stopping current flow through the tubes when more than X time*current product has passed through the tube in a given time window. The parts needed to do this can have failure rates much lower than the tubes and other parts they are watching.

[Gaz]

Man, I had to look that one up!
Sorry, Gaz.
snarky - definition of snarky by the Free Online Dictionary, Thesaurus and Encyclopedia.

No worries, Jazz, I was clearly being a bit reactive.

[R.G.]

My green question still remains: If I were to, hypothetically, add fuses between the current-sense resistors and ground, should I anticipate any interplay between them when there is a surge condition? Do the resistor and fuse share the current somehow? What I mean, is will the 1R resistor limit the current before it 'gets' to the fuse?

MMMMMMaybe. If the current in the loop is limited by things with higher resistances than the fuse and resistor, no, the resistor won't change things. If the current is limited primarily by the fuse resistance during faults, yes, the 1R is a similar (and larger) resistance to the fuse, so it will change the fault current, perhaps slowing the fuse blowing. The answer depends on the rest of the stuff in the current loop, not just the fuse and resistor.

Yes, the amp does have a 1A Slo-blo B+ fuse in on leg of the bridge rectifier on the AC side. Since it didn't blow before the resistor, and a few have explained to me why it possibly didn't, I thought a fast-acting fuses there may be more reliable. I thought I might try it like the JCM900 schem g-one posted with the indicator lights, and individual fuses. This way I could know which tube was the culprit as well.

I went through this kind of scenario a few years ago. I used the 1R current sensor for sensing both bias current and the total amount of current in each tube, all the time. There was some filtering to keep sudden spikes from tripping things, and then threshold detectors to say "hey! this tube's gonna be toast~!" The detectors set latches which turned off the B+ in a few microseconds and turned on LEDs to indicate which tube was having problems. I did the original in CMOS and analog stuff, but today I'd use a $2.00 microcontroller, which could do a much better job of figuring the most recent power/energy eaten by the tube.

By the way, I hate replacing fuses. I try to think of ways to stop overcurrents that don't have the sloppiness in both current and time that they have.

In your situation, the 1A slo-blows in the AC side of the rectifiers have a far different current they're working on. They're processing the really high peak current pulses put out by the rectifiers into the filters. This is hardly related at all to the signal (or fault!) current going through one tube after the filter cap. The filter cap holds enough charge to damage things even if the AC-side fuses are open. If you're going to protect with only fuses, you have to actually fuse the current that you're trying to stop, not something upstream.

There is always an issue lurking in putting in protection circuits. You have to be sure that you're not making it worse and less reliable by putting the protection in. It can get complicated.

[g1]

Interesting standoff between theoretical and practical here . A few more considerations:
First we add resistors to sense bias. Then we find they are acting as fuses when tubes go bad. So we add diodes to protect the resistors. Now we add fuses because the resistors aren't fusing anymore . Seems a little funny but I agree it's much easier to change a fuse mounted somewhere accessible rather than open the amp to change resistors. And I also like the lights to show which fuse blew.
But the Marshall in the example is using 500mA Slow-blo for each pair of tubes. This is probably a compromise between protection and "nuisance" blowing when the tubes are running full out. So what are we protecting, OT? PT? and how well can we protect it without blowing fuses at high volume? (I'll leave that math for R.G. )
On the practical side I like to consider the industry "heavyweights". Did Fender ever fuse cathodes? Does Marshall still do it or did they give it up? I'm not up on their current models. I can guarantee warranty claims are a big issue for manufacturers, and if they can reduce claims with protection schemes, they will do so. In my experience Fender xfrmr's are more robust than Marshall's, so maybe that could account for the difference (warranty claims again).
There was another thread where the Crate BV120 had a SIDAC from the heater winding to ground. If the B+ ever got to the heater (plate to heater short?), the SIDAC would clamp the heater to ground, presumably to save the hum balance resistors. Yes, sometimes those resistors get burnt. But often enough to warrant a SIDAC? Now on occasion the SIDAC's short and need replacement. So 6 of one, half dozen of the other. Again I'm thinking, does any other mfgr. do this? Does Crate still do it or was it an experiment?

[Enzo]

And that brings us back around to the primary question: how often do the 1 ohm resistors actually burn up? If we are talking the convenience of something, doesn;t that presume it happens enough to warranty special attention?

[Steve Conner]

This just goes to show how there is no such thing as a simple question in electronics.

When designing a circuit you have to start with a clear idea of what your goal is. I haven't seen that stated anywhere in this thread, so let me offer one: The amp should survive a tube short without sustaining any damage that would require opening the chassis to put right.

Now we have to look at all the possible combinations of tube shorts. You guys have covered the plate to cathode short in excruciating detail. If you use a wire wound cathode resistor and/or a diode across it, then the HT fuse or cathode fuse should blow before the resistor does.

(Aside: the difference between a HT fuse and a cathode fuse is that you can have one cathode fuse per tube. In normal operation, the current is shared evenly amongst all the tubes, but fault current tends to all go through one tube. It follows that in a 4 tube output stage, 4 cathode fuses would be 4 times better at discriminating between faults and normal operation than a single HT fuse. But since the main purpose of the fuse is to protect the transformers, this hardly matters. However, cathode fuses can isolate a blown tube and let the amp continue playing, but that isn't relevant to our design goal either, so we might as well leave them out.)

Another common short is screen grid to control grid. This usually takes out the screen resistor, and it is much harder to protect than the cathode resistors due to its higher resistance. As far as I know no manufacturer even bothers. I've suggested using a MOSFET to current limit the screen supply. This might even prevent screen shorts from happening in the first place, as they are often caused by the screen wires melting from excessive current.

Yet another one I've seen is plate to heater. This will feed B+ down the heater line and take out any hum balance or DC elevation circuit. A grounded centre tap on the heater winding is a pretty solid fix.

Exercises for the reader:

List the other possible tube shorts and try to figure out their consequences.

Map out the fault current path in a screen to control grid short, both with and without a cathode fuse. Could the grid stopper resistors burn up?

Question the underlying assumption. Might it be better to have a cheap sacrificial component inside the chassis that will burn up and force a service call, rather than have a clueless user replace the fuse with a nail and damage the transformers?

[Enzo]

Not on your list, Steve, but not rare at all:

Broken off center post, allowing power tube to be installed in socket facing the wrong way. The filament between pins 2 and 7 is more or less a dead short. So figure out what shorts to what in each wrong position. Off one pin and it shorts pins 3 and 8 - same as plate to cathode. Off two pins and it is a 4 to 1 short. That shorts the screen pin to either a grounded pin 1 or to the grid circuit if pin 1 is used for the grid stopper. Off three pins and now we have pins 5 and 2 shorted together. Assuming a low resistance to ground for the heaters from either center tap or virtual ground, that grounds off the grid circuit. Probably won't damage the bias circuit per se, but if this is a four tube stage, it will surely kill bias to its next door tube. Next up, pin 6 to pin 3. Fairly benign. Shorts the plate circuit to either nothing,or maybe commonly to the screen B+ node. Not good, but probably no fireworks. COntinue to pin 7 back to pin 4. Looks like grounding the screen to me. Then pin 8 back to pin 5 - same as the earlier cathode to control grid short. And finally pin 1 back to pin 6. The two wild card pins. Is pin 1 grounded or a tie point? Is 6 a tie point? In a Twin Reverb, we'd have the B+ screen node shorted to the grid stopper. If you are wired for EL34 with pin 1 grounded, then we have a dead short on that screen node. All depends on what you do with those pins.

Some of those failure modes are not very common in a failed tube, but in a broken post tube, it could be any of them. And believe me, I see this often enough.

[ReadyTeddy]

Another possibility--there was simply a manufacturing defect in the 1-ohm resistor. That's pretty rare, but we're talking about a rare event. All this talk of designing to make this or that component less likely to fail caused me to think of a poem I learned many moons ago, Poem: The Deacon's Masterpiece, or, the Wonderful One-hoss Shay: A Logical Story.

[Steve Conner]

Yes, the ideal modern product is built like the One-Hoss Shay. The whole unit crumbles to dust the day after the warranty expires.

Thanks for the in-depth analysis Enzo. I wonder if the explanation for the blown 1 ohm resistor is that someone shoved the power tube into the socket the wrong way, while the main filter caps were still holding a charge.

[R.G.]

I'm a great fan of the Wonder One-Hoss Shay. One of my favorites.

Actually, that line of thinking leads to some rational approaches to designing for reliability. I worked for a while in the section of our local company that designed for RAS - Reliability, Availability, and Serviceability, which are three different, but interrelated things. (This was, of course, TIC - Three Initial Corporation. )

Reliability is of course designing for it to work without failure. The presumption here is that if one part fails, the unit stops working. Defining "fails" is the issue.
Availability is an issue, as in it works all the time without time down for preventive maintenance and other polishing and oil changing.
Serviceability is how quickly and how cheaply can the unit be diagnosed and brought back on line if a part does fail.

In many cases, reliability and availability work in tandem. Getting things to be more reliable requires finding weak links and either making them stronger to suit the circumstances, or changing the circumstances to make the available parts not be so stressed. This is kind of the basic approach for the One-Hoss Shay.
Availability is what you do when you can't make parts not ever fail. You do maintenance. You can't make car engines work forever without touching them, so you make up long lists of oil and filter changes, tuning procedures, etc. It's easy to confuse availability with serviceability if you think of time to diagnose and repair as down time and preventive maintenance time, but they're not really the same, they just overlap some.

And finally, everything fails sometime. It's better that if you know a part's going to fail, you make it easy to fix. Like with ... sockets! Hey! We can put the most failing parts in sockets! Like er, tubes.

Putting everything in sockets has its own set of issues, not least being that in electronics, sockets and connectors are some of the most likely parts to fail. There's some thinking that was done long ago on this. You put a part in a socket/connector when the part is much more likely to fail than the socket/connector. If this isn't true, you're making the reliability worse to improve serviceability. Sometimes this is a good tradeoff, where the time to fix a non-connectorized part is unacceptable. Sometimes not. NASA has no practical way to do service, so serviceability isn't much use to them.

All of this starts with enumerating the expected life under the expected operating conditions, and the chances of failure, as in mean time to failure or intrinsic failure rate. "Expected operating conditions" includes PM and user abuse. For tube amps, this is the breaking off of the polarizing posts Enzo mentioned. I have a story I won't bore you with, but a trained, experienced service tech once nearly go us sued by using a big pair of pliers to force the polarized, NON-reversable connectors of a power supply in backwards.

But I digress. You have to know the parts, estimate the conditions, and estimate the failure rates under those conditions, and then you can start the process of eliminating weak links.

The real trick is figuring out what's going to fail first, then either changing the part to make it stronger, the conditions to make it easier on the part, or if you can't do those, putting in some kind of provisions to limit collateral damage and make replacements easier.

In the case of 1-ohm sense resistors, you need to figure out whether the 1-ohms are failing, or if they're collateral damage. If collateral, how can the primary failure be prevented from propagating?

I need more coffee.

[Jazz P Bass]

Nice post RG.
If I may add, I have seen where resistors that burned because of a failure, were replaced with higher wattage parts. (Oh it wasn't spec'd correctly)
Sometimes you want a part to fail, to save other parts. (or prevent a fire)

[g1]

And to back up what Steve said about fuses, I have seen JCM900's with tin foil on the OP valve fuses and fried output transformers.