I'm trying (and mostly failing) to make this amp on a budget so I was tempted to not use any soft start for the supplies, but after testing the supplies by slowly turning up the line voltage with my Variac I tried a cold start and the inrush pegged the ammeter on the 5 A scale. I guess I shouldn't be surprised with 30 K uF per supply. The 5 A fuse didn't blow, and this amp is going to be on 24/7 but I decided I just couldn't live with the situation.
I looked at lots of really good soft start circuits but what I wanted was one I could make with parts on hand that was "good enough". I have some 20 ohm thermistors left over from another project, but the resistance was too high to leave them in the circuit all the time. So how about using a relay to short them out after the soft start, and viola here's the "Good Enough Soft Start".
My first try at this worked, but the relay pulled in at about 70 V AC so there was still a fair bit of inrush current. so I added about 2 K of resistance in series with the coil. Now it pulls in at about 95 V AC which is acceptable. I also found that putting 2 thermistors in series made for much nicer looking inrush, here's the positive 50 V rail.
The only downside to the circuit is that the relay coil buzzes as the pickup point approaches, but looking at the rail I don't think it's actually making and breaking so I'm hoping that the lifetime of the relay contacts won't be affected. I wish I could look at the inrush current directly, I really need to get hold of a current probe!
I looked at lots of really good soft start circuits but what I wanted was one I could make with parts on hand that was "good enough". I have some 20 ohm thermistors left over from another project, but the resistance was too high to leave them in the circuit all the time. So how about using a relay to short them out after the soft start, and viola here's the "Good Enough Soft Start".
My first try at this worked, but the relay pulled in at about 70 V AC so there was still a fair bit of inrush current. so I added about 2 K of resistance in series with the coil. Now it pulls in at about 95 V AC which is acceptable. I also found that putting 2 thermistors in series made for much nicer looking inrush, here's the positive 50 V rail.
The only downside to the circuit is that the relay coil buzzes as the pickup point approaches, but looking at the rail I don't think it's actually making and breaking so I'm hoping that the lifetime of the relay contacts won't be affected. I wish I could look at the inrush current directly, I really need to get hold of a current probe!
If it is left on 24/7, why do you need a soft start? It only comes into play when you turn it on.
Use a relay with DPDT contacts to introduce some hysteresis (arrange the 2nd set of contacts to connect a second resistor in parallel with your "2K" coil resistor). May reduce chatter / buzz.
You're 100% right, I don't need it. This was purely an emotional/aesthetic decision.
The thermal temp of actual MOSFETs is often hotter than heatsink temp. That’s how heat is driven from MOSFETs to the heatsink. I think max rating on most active devices is 150C (above which there is possible damage). Automotive and military spec are even higher. 191F is 88C and that’s perfectly reasonable for a MOSFET. That would be 33C higher than max external heatsink temperature. That MOSFET will be perfectly happy at that temp for a long time. I don’t think there is anything wrong with your setup. If it bothers you, try the ceramic heatsink pads and CPU cooler thermal compound.
(Btw, the link for the Amazon insulators goes to some news video on YT)
(Btw, the link for the Amazon insulators goes to some news video on YT)
You should rethink how you fastened the MosFets to the aluminum plate.
Looks like a conical head screw without a fender washer to distribute the clamping force evenly or split washer to keep tight. With this setup I doubt the MosFet is flat against the aluminum plate. Panhead or socket head fasteners, fender washer with split washer, or Belleville washer without split washer is a good choice to securely attach output devices to a heatsink.
Also, did you fasten the MosFets with insulation pads to the aluminum plate, drop down the pcb over the legs, screw down the board to the standoffs, then solder the legs? This method ensures the device is flat against the heatsink without stressing the legs.
https://en.m.wikipedia.org/wiki/Belleville_washer
Last edited:
I expect the MOSFET to be hotter than the heatsink of course, but I didn't expect the thermal gradient to be so high and abrupt. And sitting at 88C case temperature - so probably above 90C die temp - with 10 W out just seemed excessive.
My apologies on the cut and past error on the Amazon link, there doesn't seem to be a way to edit it out. This is the link I wanted.
https://www.amazon.com/dp/B07H1C7JP3?ref=ppx_yo2ov_dt_b_fed_asin_title
That aluminum plate is way too thin, 1/8 of an inch is literally nothing, should have at least 3/8.
According to the online calculator I used even if you figure only a 1 square inch area of the plate is relevant to the thermal resistance, 1 inch by 1 inch at 0.125 inch thickness you the thermal resistance is under 0.025 deg C/W.
https://mustcalculate.com/electroni...mat=al&w=1&wu=inch&l=1&lu=inch&t=.125&tu=inch
https://mustcalculate.com/electroni...mat=al&w=1&wu=inch&l=1&lu=inch&t=.125&tu=inch
Good eyes on the screw type! They were the only 12 mm long M3 screws I had on hand. Today I'm going to cut off some flat head 16 mm and use flat and split washer, I don't have any Belleville washers.
And yes I did solder in the FETs exactly as you mentioned, after I remount them with new mica insulators I'll unsolder and resolder to be sure they're flat.
That formula is not the one you should use to find out how good your heatsink/aluminum plate is.
That formula shows how good a piece of metal is to transfer the heat between a heating element/semiconductor to your actual heatsink.
You're correct, and thinking just a bit more it doesn't make sense, it's orders of magnitude too low. Mea culpa.
Doing some more research I found this link on the forum:
If you figure that an inch or two either side of the MOSFET makes any difference at all in the thermal resistance then squinting at this nomograph puts you somewhere around 6 deg C/W which makes more sense. I can live with that. It will be interesting to see how much - if any - difference the mica insulators make.
Soft starts only work at turn on but that’s when the surge of current (can be in the tens of amps for milliseconds) that causes magnetorestriction induced mechanical stress to the windings. It is the mechanical burp sound you hear. This leads to eventual failure of the windings due to stress of being stretched.
Your simple relay and NTC’s is same in concept to many soft starts but your passive relay coil circuit is interesting. My SFP or SFPP circuits designed by Jhofland have the benefit of a solid state relay and a comparator and RC time constant to create the delay time. The SSR has the benefit of not arcing and noiseless. The SFPP adds a soft on/off switch that can be controlled from a low voltage or digital logic signal. It’s one of the circuits that gives a DIY project a true finished look and “like a commercial product switch” user experience during power on and off.
I can't claim I invented the idea, I saw it somewhere decades ago, maybe The Audio Amateur?
The first time I used it was when I built one of Marshall Leachs "Double Barrel Amps" from Audio magazine. I used a huge bank of caps and the amp popped the circuit breaker in the basement every time I turned it on! Luckily I had just seen the circuit, I used an old lamp and a 75 W incandescent bulb for the limiter. So it was sort of a dim bulb tester with a bypass. People who came to hear the amp commented on the bulb more than the amp!
By the way, the biggest disadvantage of the simple relay circuit is that you don't get to pick the pull in voltage. Typically it's in the range of 80 to 90 V to ensure operation at low line so the input current isn't fully settled when the relay pulls in. But by then the supply caps are mostly charged so the inrush is at least tolerable. There's no doubt that there are many soft start circuits that are much more repeatable but the one I used is "good enough".
By the time the current-limited primary voltage reaches 80/115 = 70% of the mains voltage, the transformer magnetization is completely finished, and the DC supply filter capacitors have been charged at least halfway. Which means: the vast bulk of the inrush event is over, when the relay decides to bypass (short out) the inrush current limiter.