Standby Switches with SS Rectifier???

[R.G.]

I switched from electrolytics to 800V film caps in the power supply. Never have to worry about surges or voltage limits ever again

Nor about electrolytic capacitor self-rotting. Motor run film caps were part of the Immortal Amplifier I envisioned. Good on you!

[Bergheim]

I switched from electrolytics to 800V film caps in the power supply. Never have to worry about surges or voltage limits ever again

Nor about electrolytic capacitor self-rotting. Motor run film caps were part of the Immortal Amplifier I envisioned. Good on you!

I tried a 40uF motor cap for the reservoir but ditched it eventually as it occupied far too much space. About 50x140mm if I remeber correctly

[R.G.]

The cap industry has started making decent, high(-er) value caps that don't have the metal package. A while playing with the Mouser selection database generally turns up something that's lighter and smaller. HVAC and motor run caps can be a bit heavy and huge.

[ViperDoc]

I’m going to skip the standby and add a thermistor and see how that works.

[ViperDoc]

Where is it best to implement thermistors for slow voltage ramp up? See below:

THERMISTORS.pdf

Or likely something else? Thanks.

[pdf64]

I suggest a 4th option, one on the primary side (I do that anyway) and one on the secondary, in series with the CT / HT fuse.

[R.G.]

One way to design electronics is to relay on the inherent properties of parts. Example: the time delay of NTC thermistors to do a surge delay. They do this by the property of changing their resistance as they self-heat. Another is to decide how you want things to act, and then concoct electronics to make that happen. Example: sense something happening with a comparator and have it turn on/off a power device, or have a microcontroller look at circuit conditions and decide what should happen.

Of these, the guitar amp genre nearly always uses the first. This is partly because the historical guitar amp was so simple and accessible, and the decide-what-you-want-and-make-it-so school requires more obscure electronics. But sometimes you can get good results that are hard to do with the inherent-properties approach.

The MOSFET current clamp is an example of the method where you decide what you want done, then design to get specifically that. A power MOSFET, a transistor and some Rs and Cs set the maximum current that can flow through the MOSFET for as long as the MOSFET lives. So you have to do some stuff to make sure that the MOSFET lives, but given that, you can set the maximum current to whatever you want. For normal amplifier operation, you have to set the current clamp to enough for the current pulses that need to flow into the first filter cap from rectification. But that's far less than what happens when you turn on the amp cold, or turn on the rectifiers with a standby switch. So the current clamp approach will all by itself will give a slow turn-on.

As a side note, you can put an opto coupler into the current clamp to turn it off like a switch. In that case, the MOSFET clamp >> IS << the standby switch, and what the panel standby does is turn the opto's LED on and off.

[ViperDoc]

Thanks, R.G., I recall you mentioning something like that earlier. I’d like to learn more about that, I don’t know how I would set a current clamp up.

[R.G.]

It's in the tube amp section of geofex.com. Here's the link:

http://www.geofex.com/Article_Folders/m ... 0Clamp.pdf

This is an update of an earlier version using a bipolar transistor for the pass device. R1 just pulls up the gate to the drain, so until something else happens, the MOSFET conducts hard when the voltage from drain to source is more than the threshold of the MOSFET, about 5V?ish with today's power MOSFETs. That works until the current causes the voltage in the sense resistor R2 to get higher than the turn on voltage of the NPN transistor's base-emitter, about 0.5-0.7V. When that happens, the NPN steals current from the gate through R1. The more current through R2, the harder the transistor pulls down on the gate of the MOSFET, turning it partially off. In practice, the MOSFET lets the voltage across its drain-source increase to keep the current from increasing. The current through it is effectively clamped to the current that makes R2 have a voltage of 0.6V (or so, depending on the NPN).

This isn't free of course. The MOSFET has to eat the current through it times the voltage across it as waste heat. So you have to heat sink the MOSFET to let it get rid of the heat without burning out. A cold start up means it will "eat" quite a few watts for a fraction of a second as it prevents the power caps from charging with huge pulses of current from the rectifiers. This heat will mostly go to warming the thermal mass of the heat sink metal, as it will be over quickly. In normal operation, the heat sink has to get rid of the "normal" heating of a few volts at maybe half an amp. So the heat sink has to be able to dissipate maybe 10W into the air all the time.

If you contrast that to an NTC thermistor, you'll find that the thermistor is keeping its resistance low by staying hot too. NTC inrush limiters only have low resistance when they're hot, so they don't have a big pulse of heat to start with, but have to stay hot in normal operation.

It's a trade off.

It's also possible to do some fancier stuff. A timing circuit could be made that uses a series resistor in the wire between the rectifiers and filter cap, and a MOSFET "shorting" the resistor. the MOSFET would be connected to a timing circuit that held it for for a few seconds at power on or standby off so the resistor would slow down the power transient. A MOSFET failure in this, just like the current clamp, would probably short the MOSFET, leaving the amp functional, but without the advantages of a slow start.

[ViperDoc]

Forgive my breathing through a straw.

Thanks for the current clamp article. Here's a draft of the 50W British-style power transformer wiring with the current clamp wired in:

50W CURRENT CLAMP DRAFT.pdf

[note: I corrected the orientation of the BJT]

How does that look?

The article did not detail what type or value of BJT/bipolar transistor to use. Any thoughts?

For what it's worth, the complexity of that circuit just makes me want either a standby switch or a commitment to replacing filter caps when needed. But if I recall correctly, the advantage to such a current clamp is reduced component heat (with proper heat sinkage), filter cap protection without the standby, and not enduring constant thermistor heat.

[Roe]

Where is it best to implement thermistors for slow voltage ramp up? See below:

THERMISTORS.pdf

Or likely something else? Thanks.

why not use it after the rectifier? doesn't it work with DC?

[R.G.]

Thanks for the current clamp article. Here's a draft of the 50W British-style power transformer wiring with the current clamp wired in:
50W CURRENT CLAMP DRAFT.pdf
[note: I corrected the orientation of the BJT]
How does that look?
The article did not detail what type or value of BJT/bipolar transistor to use. Any thoughts?

Looks right. I think a 2N3904 NPN would be fine.
Hmmm. Let's see here. the worst case for the NPN is if the NPN has to hold the MOSFET off for the full peak of the voltage coming from the rectifiers. The incoming voltage at the drain of the MOSFET would be, say, zero to 500V and back down to zero over a half cycle of the rectified AC.
If the NPN has to hold the MOSFET off at the peak of the AC half-wave the NPN has to hold the gate voltage at zero with 500V across the drain to gate resistor. A baby NPN like a 2N3904 can handily conduct up to maybe 100ma when switched fully on, and can dissipate up to a watt in just air without heat sinking. The current the NPN has to conduct is (for example) 500V divided by the resistance of the drain-gate resistor. If that resistor is 100K, then the current is 500V/100K = 5ma. A 2N3904 can do that just fine.
As long as nothing is burning up, the zener keeps the gate-source voltage below 12V. So the NPN should never see more than 12V from collector to emitter. A 2N3904 can do that fine. If the MOSFET should happen to need 12V gate to source to conduct full current (and the IRFBE20 doesn't need that much), then the NPN can "get out of the way" by turning fully off at 12V across it.
I notice that I left out a series gate resistor. Some MOSFETs in some circuits can oscillate unless you put a small resistor in series with the gate as close to the MOSFET gate as you can get it. Better to add a 100R to 1K in series from the MOSFET gate as physically close to the gate as you can get it. This doesn't affect the circuit reasoning for the drive and NPN above. It would probably be OK without it, but to make sure, put it in.
The IRFBE20 can conduct up to 7.2A pulsed current according to the data sheet. So the smallest current sense resistor should be 0.6V/7.2A = 0.083 ohms. The current into the filter cap in pulses from the rectifiers has to equal the current going out of the caps as average DC. So if the pulse current is 7.2A max, and the average current out of the filter caps is .... um... half an amp average (that is a guess; anybody got a better number for the B+ current at maximum power out???), then the duty cycle has to be at least 0.5A/7.2A = 0.07, or on for 0.6mS out of the 8.6mS of a 60Hz half-cycle of an AC power wave. That's not too short or too peaky, so there's room to limit the current peak down some. That will spread the width of the charging pulse out over more of the AC half cycle, while reducing the peak current. Good!! That's what this circuit is for!

Finally, this mini design review makes me want to go check if the IRFBE20 is still a good choice. It is - but power handling in MOSFETs keeps getting cheaper. I went off and looked for what power MSOFETs were (1) in stock at Mouser (2) over 800V Vds (3) over 2A continuous drain current (3) over 7A pulsed current (4) in a TO-220 package (5) over 20W dissipation and (6) under $2.00 each. I came up with a higher rated device, the Infineon SPP06N80C3 for $1.99. Another possibility is the ST STP4LN80K5 for $1.13. There are options. If you want to do your own MOSFET picking, I'll be happy to look at what you come up with.

The final issue is heat. MOSFETs tend to only die two ways - puncturing the gate insulation with spikes, and sheer overheating. The 12V zener will stop the transients in nearly all cases. Overheating can short the MOSFET. The amp will still work, but will not have the advantage of clamping. To keep the MOSFET in good health, keep adding heat sink fin area until it stays cool enough. This will be more than simply bolting it to the chassis. It will need an actual aluminum finned sink. At worst, you'll need to get rid of about 10-20W. I can help with his when/if you get to here.

For what it's worth, the complexity of that circuit just makes me want either a standby switch or a commitment to replacing filter caps when needed. But if I recall correctly, the advantage to such a current clamp is reduced component heat (with proper heat sinkage), filter cap protection without the standby, and not enduring constant thermistor heat.

Sorry about the complexity. It's one down side to the decide-what-you-want view of designing electronics. Yes, there are advantages over the thermistor (in my opinion at least). But I do understand how daunting this kind of thing can be. The NTC thermistor will work, as will simply making the standby be a mute switch if you decide to go one or both of those ways.

[turbofeedus]

Just want to chime and, and hopefully not derail too hard from the current clamp discussion.
I had a similar issue to OP, SS rectified, B+ surged to 520V at startup, exceeding my 500V rated caps, falls to 490 after warmup.
I implemented Merlin's standby switch shown at the bottom of the article here: http://www.valvewizard.co.uk/standby.html
B+ now sits around 200V with standby engaged (yes, I added the 47K bypass resistor), and climbs to about 490 when the switch is thrown.
Totally necessary? Maybe not, but it makes me feel better.

standby4.jpg