coupling cap voltage ratings, and standby switching - Mesa Studio 22

[mxrshiver]

coupling caps should ideally be rated for the full unloaded B+ the amp can possibly produce, 630V caps covering most amps out there. but i've come across several amps that use 400V caps throughout, when the unloaded B+ is capable of being much higher. for example, the SLO100 (at least the older ones) uses 400V coupling caps, despite the B+ reaching close to 600V unloaded.

i've been doing extensive work on an early Mesa Studio 22 from 1985, including a conversion to the 22+, that's been a real workhorse and had some heavy heat damage and tons of bad solder joints, but is now turning out great. (landed on a great recipe for addressing the bias issues in a way that enhances clarity, happy to share if anyone's curious) part of that work was moving the Standby out of the DC path, into an audio mute shorting the power tube grids together.

this amp uses 400V coupling caps throughout, Sprague 715P's, and on one of the start up cycles, one of the caps shorted, the .1 bass cap in the tonestack. this cap isn't anywhere close to any area i was working on, or that had issues. and it's one of the ones i would think would be least likely to fail, with both the plate resistor and slope resistor between it and the high voltage supply.

this got me wondering if there's any chance i could have possibly worsened the conditions for these caps by removing the DC standby. currently, the B+ floats up to 470V for about 5-8 seconds on startup, then settles to 400V. i've been under the impression that this more prolonged no-load over-voltage when starting up with no Standby, is generally preferable to the more instantaneous over-voltage along with large inrush current, which happens when a DC Standby switch is closed. but i most often see that discussion centering around the switch itself and the filter caps... perhaps for the coupling caps, the duration of over-voltage is more important?

i'm mainly hoping to evaluate how pressing it may be to upgrade all of these to 630V caps, and to better interpret the datasheet values surrounding this concern. looking at the 715P datasheet, it states that all caps rated 800VDC or less can tolerate 250% of rated voltage for up to 5 seconds, which i would think would well cover these conditions, which would only subject them to about 120% of rated voltage for perhaps 10 seconds at the most. but then, why did that .1 cap fail...

perhaps i should be looking at a different part of the datasheet? the dV/dt Maximum Pulse Rise Time does show that the .1 cap would have a lower tolerance than the other lower capacitance values in the amp, at 1,700 Volts per microsecond. maybe the initial voltage pulse was just too quick? i do have a thermistor in the primary to limit inrush current, and it takes over a second for the B+ to climb above 400V, so this doesn't seem likely either...

[acheld]

Sounds like 400 VDC caps would be right at the edge of being ok with your setup; 630VDC rating should take care of it.   

I have used 800V Panasonic film caps, which have worked well and are quite a bit smaller than the orange drop series. But --  genuine** 715P and 716P caps are great if you have the space for them.  I prefer 716P because of the form factor and the copper leads, but they are sometimes harder to source.

Hard to say why your cap failed, but they do sometimes, and maybe prior history and your current mods pushed it over the edge.  Don't overthink the ratings for a polypropylene film cap; this is a guitar amp.  Voltage and operating temp are most important; it's ok to overspeed them.

** I've been handed a number of fake branded caps over the past 5 years.   You really have to be careful about your vendor. 

[mxrshiver]

thanks acheld! certainly no harm and easy enough to over-spec the replacements. the owner's already put a ton into getting this amp reliable again, so i'd just hate to say replacing all 8 remaining coupling caps is also necessary unless i was sure.

do you think start-up conditions for the 715P's were likely better or worse with the DC standby? my thought is that brief large inrush current with voltage excess when the Standby is closed would be rougher on them, but perhaps the several seconds of voltage excess without the Standby is worse, even though the current onset is more gradual... i'd just like to better evaluate similar situations in the future before deciding what to do with the Standby.

[Merlin]

coupling caps should ideally be rated for the full unloaded B+
this amp uses 400V coupling caps throughout
this got me wondering if there's any chance i could have possibly worsened the conditions for these caps by removing the DC standby. currently, the B+ floats up to 470V for about 5-8 seconds on startup, then settles to 400V.

Don't forget about the bias voltage. The max voltage across the caps is equal to the max B+ plus the bias voltage, which might be another 20-50V depending on the power tube type.  I don't think you've made it much worse; those caps are probably running over 400V whether you use a standby or not! 400V caps always fail in Mesas for this reason.

[Willabe]

Don't forget about the bias voltage. The max voltage across the caps is equal to the max B+ plus the bias voltage, which might be another 20-50V depending on the power tube type. 

I'm sure your correct, but I don't understand how. 

The bias dcv is negative, that wouldn't add to the positive B+. 

Do you mean from the bias tap? But what if the amp gets it's bias from the main B+?

I don't get it. I'm missing something here. 

[Merlin]

One end of the cap is connected to the LTP anode, which might spike up to the B+ under certain conditions like switch-on, e.g. 470V say.
The other end of the cap is connected to the power tube grids, which idle at the bias voltage (and might spike even lower under certain conditions like switch-on), e.g. -20V say.
Therefore the worst-case voltage that might appear across the cap would be 470 - (-20) = 470 + 20 = 490V.

[mxrshiver]

thanks so much Merlin! didn't even think about that factor for the power tube coupling caps...

this amp's weird stock bias circuit used to let the bias voltage slowly charge up to a negative voltage close in magnitude to the B+, before the Standby is closed and the phase inverter begins to draw current to bring it back down! i was aware of this when rebuilding it, and was monitoring it when stepping the amp up on a variac after initial work was complete, to make sure it never exceeded the bias filter cap's 63V rating. but even at only 25VAC primary input, the bias voltage wanted to charge up enough to exceed that! i never let it do so, but given that the only path to ground for the bias voltage (besides the phase inverter cathodes) was through almost 3Meg of resistance, i have to imagine it would have charged up to several hundred negative volts given the chance...

on one hand, it always ensured the bias voltage had the power tubes deep into cutoff before the Standby was closed, which probably did a lot to limit inrush current... but on the other hand, it exceeded the 63V rating of the bias filter cap by several times, and exceeded the -100V maximum negative grid voltage for the 6BQ5's, even before the Standby was closed... and, as you pointed out, could have put a truly huge voltage across the power tube coupling caps!

in case anyone's working on one of these in the future - i ended up adding a diode to make the bias circuit full wave bipolar, reducing the first 120K resistor off the rectifier to 100K, eliminating the 33K series resistor, reducing grid leaks from 330K to 220K, reducing the 2M2 from the power tube grids to ground to 1M, and most critically, adding a resistance to ground right on the bias cap, a 12K in series with a 20K pot. now the voltage only reaches -40V on startup before the phase inverter conduction pulls it back down, the power tubes are much happier with the grid to ground resistance actually meeting the datasheet maximum of 300K, and i can dial the bias anywhere from the roughly 14.5W plate dissipation at stock, down to as little as about 8W (i settled on 11W).

i also changed the phase inverter resistor from 15K to 12K, which brought the cathodes back to the 2V they're supposed to be at. i was concerned about fluctuations in phase inverter bias based on different bias pot settings and drive levels, but remarkably, the cathodes only changed about 0.2V from one extreme of the bias pot setting to the other... and under drive, even heavy overdrive with the Master dimed, the voltage across that 12K resistor stayed rock solid at about 19V, barely fluctuating half a volt!

[Willabe]

One end of the cap is connected to the LTP anode, which might spike up to the B+ under certain conditions like switch-on, e.g. 470V say.
The other end of the cap is connected to the power tube grids, which idle at the bias voltage (and might spike even lower under certain conditions like switch-on), e.g. -20V say.
Therefore the worst-case voltage that might appear across the cap would be 470 - (-20) = 470 + 20 = 490V.

Thank you.