'69 SL100 HT fuse blowing

[bordonbert]

Hi guys. I have had this amp since 1970 so I know its history intimately. A fair way back now it had a mains Tx transplant with a Marshall supplied replacement, (Coca Cola and HT don't mix too well). It has also had a recap only a couple of years ago. All looks as it should inside. There are no bulging caps, scorched resistors or dodgy looking wiring.

I normally use the amp with a Fryette Power Station SL100. It has never had a problem with this setup though it should be mentioned as unusual. While in use at home with the Fryette, out of nowhere, it gave a very short crackle and went quiet. I immediately turned it off checked and found the HT fuse blown. I swapped out the valves for another set and tested. The HT fuse blew again, time to look deeper. I pulled all of the power valves and fired up again. It stayed happy. It didn't seem to be a valve at fault so I wondered about the bias section and worried about the output Tx. The Output Tx meters up on both primary and secondary as still intact at least. The supply line voltages check out all down the line.

The bias setup is the usual "27k//220k + diode + 15k + 27k slide pot" arrangement with supply from a dedicated 100V winding on the mains Tx. All of the components in that chain meter out correctly and the voltages seem to be in line with what they should be. I know the slider pots can give trouble and that presumably is of two types. Either the wiper can lose a clean contact with the track or the track ends lose clean contact with their terminals. Both of those cases should increase the resistance in circuit and increase the bias voltage negatively which would reduce current. Surely that can't have caused this problem? Incidentally I'm planning on replacing that pot with a modern cermet, (I'm no cork sniffer, this is a working amp not a museum piece, though I will keep the original in storage). A 25k will drop in there fine and should give me a usable working range considering the tolerance of the original.

I also took a look at the PI output isolating caps. The voltages at their outer end, towards the power valve grids, are exactly equal and pretty much match the bias voltage to a volt or so.

I know my electronics very well but I'm no valve guru. The theory is one thing to an engineer but a tech's knowledge of models and their common problems and solutions is priceless. Can anyone else suggest any ideas on what else could produce the fault?

[martin manning]

Will it idle with the power valves in, without blowing the fuse?

[pdf64]

Did you check that the second set of EL34 weren’t idling super hot? It’s a good idea to cool the bias down beforr powering up with different valves.
Are you using slow acting fuse? I’ve read reports that if using UL rated fuses, the value needs increasing from IEC spec types, so you may need 1.2 or 1.5A SloBlo for it to hold.
To check the coupling caps, lift a leg at the 220k grid leak end, and measure the voltage there. Anything more than a few mV indicates potential for trouble.
Maybe power up via a light bulb limiter, 150-200W bulb, until this is resolved.

[Stevem]

The first thing you want to do before slapping any outout tube in the amp is to Crank the negative bias on pin 5 up to highest you can.

Also for the sake of reducing the load on the V+ winding I would cut in half the amperage rating of that HT fuse.

[bordonbert]

Thanks guys there is some great advice in there which raises some interesting points. Dropping the bias current when putting in a different set of valves is of course the sensible way to do it. Did I? We-e-e-ell, mea culpa, nope! Even though I knew I should I just hit the "Go" button and hoped. That's my stupidity. I can't check whether it is ok at the moment Martin, I pulled the bias control out to update it to a modern Cermet when I saw the state of the original pot. (Oh God! A modern part in a classic amp. I've sucked all of the tone!!! ) Once I get the new pot in place for the bias control I'll run it up again doing it the way I knew I should have all along, start from scratch as high as possible (in bias negative voltage terms) and increase from there.

Yes, I am using a slow blow fuse PDF64. I have to admit that I have wondered about the 1amp value for years, particularly because I really would like to pull a pair of valves and that more or less halves the current draw. Marshall don't actually recommend valve pulling as there are other issues in terms of grid current etc which are affected so maybe I should just leave it as a 100W, the Fryette takes care of the volume level anyway. I can't help thinking that there is a toss up between a lower current slow blow and a higher current quick blow. In a genuine fault condition you want the fuse to go as quickly as possible but over time a quick blow can become weakened by being taken too close to the brink too many times so selection for current level would be critical. I'd be interested in your ideas. Anyone recommend NOT doing it a particular way with experience to back up their decision?

At first thought, lifting the leg of the coupling caps at the 220k end shouldn't show up a leaky cap fault. There is then no path for the potential leakage current to go anywhere so surely you won't actually see anything of note. Caps don't leak voltage, they leak current and the rest of the circuit resolves this into a voltage which you can then detect. I may be seeing this with a total blind spot but do you mean replace the resistor connection with a meter measuring DC current to ground?

[sluckey]

At first thought, lifting the leg of the coupling caps at the 220k end shouldn't show up a leaky cap fault. There is then no path for the potential leakage current to go anywhere so surely you won't actually see anything of note. Caps don't leak voltage, they leak current and the rest of the circuit resolves this into a voltage which you can then detect. I may be seeing this with a total blind spot but do you mean replace the resistor connection with a meter measuring DC current to ground?

You're wrong about that. Lifting one leg of the cap is a very good way to see if the cap is leaky. Measure voltage. Your meter provides an adequate load.

[bordonbert]

Of course, you're absolutely right, the meter itself is a high impedance load which will show any leakage as a voltage. A senior moment. Thanks for the metaphorical face slap Sluckey. Apologies for doubting you PDF64.

[pdf64]

… I really would like to pull a pair of valves and that more or less halves the current draw. Marshall don't actually recommend valve pulling as there are other issues …

I advise against it too. The heater voltage can rise, and the HT supply rises and becomes stiffer. The net result being that the remaining valves get put under a fair bit more stress.

[Roe]

it draws too much current with power tubes. double check bias circuit and the power supply, esp. mains and screens filtering

[bordonbert]

I've now rebuilt the bias circuit with a new 25k Cermet trimpot and a small 2k2 offset resistor to keep it towards the lower edge. The original problem was definitely one valve which went out in a big way. Blagging a set of valves from one of my other amps has stabilised things and means I am back on track again. That was the obvious first choice issue of course. I'm now running with 4xEL34 at 100W again. I'll take the advice on that from you here and from Marshall. The higher power isn't an issue nowadays with the Fryette in the loop reducing only the volume, and the amp warms nicely when it can be cranked that little bit more.

I do see arguments about whether the 8uF shown in the bias of the original schematics and the 10uF found in some later amps is correct and "best". 8-10uF? Is it really an issue? I am going to try increasing the bias caps from my current 10uF to 22uF. Simulating that shows a big improvement in residual noise voltage as you would expect. It reduces residual hum on the bias voltage at the 2x220k node by a factor of 4x. However it should be kept in mind that it also increases the time constant of the bias circuit at startup and that means it takes more time to come up (negatively) to its correct setting. Spice simulation of the bias setup with both suggests it will take around 5s with 10uF and 10s with 22uF for both. Of course it isn't the case that there is necessarily any danger if the bias is not exactly where we want it to be. It can be comfortably higher but dropping for a few seconds without a problem. I don't see a little extra warmup time to sidestep that as being a deal breaker for me.

During the time taken for the bias voltage to stabilise, assuming the valves have reached a point of conduction, the bias current would be high and dropping! This isn't a problem when the standby is in the silent position with no anode voltage applied. If the amp is switched on with the standby in the play position or if you switch the standby to play too quickly and the valves heat up to emission before the bias voltage comes up, the bias current will be high and will gradually come down to the correct value. I don't see that as being a massive problem but a little extra time given to warm up and stabilise before kicking the standby to play would be a good thing.

Has anyone else changed the values of these caps or is there any reason I'm missing why 8/10uF should be the upper limit?

[martin manning]

I think the time constant is the major concern. Use of the standby as you describe will keep the power valves from exceeding dissipation limits, and from a cold start there probably isn't any problem due to the heater warm-up time. The value of lower ripple in the bias supply has a diminishing return, so you might want to experiment by paralleling some caps across the ones you have to see if there is any noticeable difference in 100Hz hum in the output. !0uF is more or less the standard now that 8uF is a less common value.

[bordonbert]

Just paralleling another 10uF to the originals was what I had in mind Martin, just as you suggest. I mentioned 22uF just as a reflex response to use E12 values. It's by far the easiest and quickest way and is completely reversible, always a consideration in a classic amp like this. I know what you mean by the diminishing returns suggestion, there are other significant routes for hum to find its way into the signal path and they will eventually dictate how effective this cap increase is in noise reduction. That said, this is one of the easiest mods to reduce what little there may be for the cost of a couple of cheap passive component, easily accessible and without influencing other aspects of the design overly. The point about giving the bias voltage more time to settle before kicking the standby must be heeded of course. I wonder how useful it would be in making the amp less susceptible to mains borne transients, though once again that is not an overly common problem to have.