Dealing with high inrush current

Thread Starter

Veracohr

Joined Jan 3, 2011
772
I'm trying to make a 200W audio amplifier, but on power-up it burned through a trace because, I think, of the high current charging the high capacitance. Here's the power supply:

Screen Shot 2018-06-17 at 7.30.37 PM.png


I need the high capacitance because of the ripple current capability that comes with it.

This is the first higher current circuit I've designed, and I obviously overlooked some things. In the next image, the upper green trace between the transformer and bridge rectifier is the one that burned. It's 2.5mm wide and about 4.5mm long - which should be enough for the maximum steady-state current, but I didn't account for the capacitor inrush current.

Screen Shot 2018-06-17 at 7.24.39 PM.png


I don't know how high the transient current is, but in simulation it can be 40A peak or more depending on the inductance of the transformer, which I don't know.

My question is: how is this sort of thing dealt with usually? Should I go with point-to-point wiring using large gauge wires up to the capacitors? Or some kind of active inrush limiting?
 

dendad

Joined Feb 20, 2016
4,451
A couple of ways to deal with it, with just changing the trace size, use the largest trace you can fit, with traces top and bottom of the board.
Also, you may have seen on some boards, having the solder resist left off over the trace in total, or on lines along the trace, allow the addition of solder to thicken the conductor to increase the current capability.
That is not taking into account various electronics to limit the current. If the trace is burning out with just the inrush current, your trace is way to small anyway.
 

Thread Starter

Veracohr

Joined Jan 3, 2011
772
o_O Too bad there’s no good facepalm smiley here.

Ok, I’ll have to fix that too, but the problem remains of needing to account for possibly several tens of amps of transient current. I suppose an electronic current limiting solution would be the most elegant, but using fat wiring would be the simplest. I can figure out some way that works, but I’m curious what one would do if designing such a circuit for commercial purposes.
 

Hymie

Joined Mar 30, 2018
1,277

ebp

Joined Feb 8, 2018
2,332
Typically for low voltage things like this it is dealt with by using a bridge rectifier that can handle high peak current and relying on the transformer resistance to limit the peak. The inductor resistance will also help flatten the peak. The bridge in question should be adequate. It will [EDIT] most likey require a heatsink for normal operation. 40 amperes peak current for a couple of cycles is nothing to be concerned about. The 1 mH inductors may very well saturate at that current, but that is not going to be problem (it may take a few cycles for them to have the magnetic remnance in the core "fixed").

You can approximate the transformer equivalent resistance referred to the secondary (i.e. essentially assuming all of the resistance is in the secondary winding and none is in the primary) by looking at the full-load versus no-load output voltage.

The AC input will require a time-delay fuse.
 
Last edited:

Hymie

Joined Mar 30, 2018
1,277
You could place a suitable resistor in series with each inductor L1 & L2 that had relay contacts shorting the resistors. The coil of the relay could be powered by VCC2, energising once the voltage had risen sufficiently, reducing the inrush current.

Alternatively you could place suitably rated NTC devices in series with each inductor. Hi-fi purist would react with horror at such an impedance in the power supply circuit. But they would only need to be of a few ohms when cold to sufficiently limit the peak inrush current.
 

ian field

Joined Oct 27, 2012
6,536
I'm trying to make a 200W audio amplifier, but on power-up it burned through a trace because, I think, of the high current charging the high capacitance. Here's the power supply:

View attachment 154624


I need the high capacitance because of the ripple current capability that comes with it.

This is the first higher current circuit I've designed, and I obviously overlooked some things. In the next image, the upper green trace between the transformer and bridge rectifier is the one that burned. It's 2.5mm wide and about 4.5mm long - which should be enough for the maximum steady-state current, but I didn't account for the capacitor inrush current.

View attachment 154627


I don't know how high the transient current is, but in simulation it can be 40A peak or more depending on the inductance of the transformer, which I don't know.

My question is: how is this sort of thing dealt with usually? Should I go with point-to-point wiring using large gauge wires up to the capacitors? Or some kind of active inrush limiting?
NTC thermistors are a traditional remedy, but transient response will take a hit if you put them after the secondary.

try a regular type in series with the primary and hope for the best - some commercial units have a fixed resistor in series with the primary that's bypassed by a relay after a short delay.
 

dendad

Joined Feb 20, 2016
4,451
The arrangement of the transformer & bridge rectifier seem OK to me to produce a dual rail output – could you please explain exactly what is wrong?
The problem starts with the circuit diagram. The pin numbers on the bridge rectifier are wrong I have done exactly the same thing, and made a PCB that is correct to the circuit but....
Here is the difference. (excuse the "CAD" please)
Bridge.jpg
Also, I note the T1 terminals have a "+" and a "-" label. That also is incorrect. Try something like "AC1" and "AC2".

I would not worry about the surge on turn on. As has been mentioned, the transformer will not supply the 40Amps anyway, so just make the tracks heavy as above and it will be ok.
 
I'll chime in.

Put ZNR's across the banks of caps. That will help, but not necessarily inrush current.

I dealt with it almost the right way. My inrush current limiter doesn't have a timeout, but it was simple.

1. ZNR's (without them, the circuit described below will fail)
2. Arranged an optoisolator resistor and backwards Zener diode to conduct at "about 2/3 V" for each supply. In my case four. 2 positive and 2 negative.
3. Add a relay and metal oxide resistor. The resistor is in series with the AC line. The relay jumps out the resistor.
4. Have a relay that can close the speaker terminals.
5. When all power supplies are >2/3 Vcc turn on speakers and short out resistor.

Other notes:
I had fuses in each power rail If one blows the metal oxide resistor will have to be replaced.
The fuses are good protection anyway.

Time constant is part electrolytic capacitors and the series resistor. Mine was on the order of 50 ohms and 10,000 uF total capacitance on 50V supplies.

The audio actually ramped up exponentially using a simple OP-Amp circuit and the FET optocoupler in series with the audio input. 200 ohms minimum R. triggered at the 2/3 Vcc too.

The system protected the output transistors when they were installed backwards. The only failed part was the resistor.

Troubleshooting, requires bypassing the resistor.

It does require another power supply not surge limited, About 12-15V.

There is no timeout. e.g. If turn on isn't achieved in x-time, then remove AC power and indicate fault.




Amp also had a FET optocoupler in series with the ausio
 

Thread Starter

Veracohr

Joined Jan 3, 2011
772
From what I found about pcb trace fusing current, that trace should have been able to theoretically handle up to 352A for a half cycle (8.3ms), but the rectifier flub screwed it all up. However, that leaves me wondering how the transformer supplied enough current to fuse the trace without getting damaged itself when it's only rated to 13A. I guess it's also a transient capability. When I redesign the board I'll beef up the traces and not worry too much about limiting the current.

The 1 mH inductors may very well saturate at that current, but that is not going to be problem (it may take a few cycles for them to have the magnetic remnance in the core "fixed").
I have to admit to not knowing much about saturation current. I only included those because during simulation it looked like they enabled a somewhat lower RMS current from the transformer under maximum load, and I was trying to find a way to get the lowest cost transformer that would do what I needed. Going back to my simulation now, it doesn't look like it's that much less. I can't even look up the saturation current because I wound them myself. They're just sloppy 12-turn toroids made with 14-ga wire. They measured close enough to 1mH.


IMG_1191.JPG
 

ebp

Joined Feb 8, 2018
2,332
Without knowing the core material characteristics it is hard to predict saturation current. Given how few turns there are for a millihenry, my expectation is that it is a very high permeability material, which generally means it saturates at quite low flux density. The winding looks pretty good for that wire gauge on a small core. I've seen lots of commercially wound small toroids that looked considerably worse for the same sort of wire and core size. I count 13 turns - it is passes through the hole that count.

Transformers can deliver current far above their rating for a short time without damage. Like the PCB trace, it comes down to a question of time. Copper has quite high specific heat (ratio of temperature rise to amount of heat input), so even with a lot of current it takes a bit of time to get the mass of copper hot. The iron core really doesn't work any harder with a short circuit than it does with no load. The limitation in the amount of current that can be delivered is really the resistance of the windings, ignoring the temperature rise limit - and a fuse on the input.

I think your circuit will be just fine with the connection error fixed and no additional bits for surge limiting.

The potential for pinout errors is something that always made me uneasy between the time I sent off plot files and I the boards back. It is kind of like proof-reading your own writing - you look, but see what you "want to see" rather than what is really there. Unless you can get someone else, who will be very careful, to check your work, I don't think there is any way of avoiding errors like that other than staring at your library footprints and datasheets until your eyes bleed. I've heard of far worse errors (e.g. half of a large pin-grid array microprocessor "upside down"). A friend recently had a problem because the pin numbering on an IC incremented clockwise instead of the usual counterclockwise as viewed from the top, and the datasheet showed the pinout as a bottom view - just perverse!
 
A friend of mine (in the embedded systems business as a one-man shop) had problems with a circuit he designed for himself. He kept thinking it was the IC and replaced it from multiple vendors. He thought they were counterfeit and they were obsolete as well. I offered to take a look at it. It turned out that most of the inputs were TTL compatible except two. They were CMOS. Those were usually designed to be strapped and not be programmable. The fix was easy. A FET and a couple of resistors for each input and invert the programming for those bits.
 
Top