Inrush current limiter for DC

[minek123]

Here is my problem:
Have an SMPS (700W, +/- 75V rails) powering my audio amp.
Both channels (2x300W) are powered by the same SMPS.
Amp boards have in total 4 x 2700uF on them.
https://www.digikey.com/product-det...nics-cde/380LX272M100J452/338-2435-ND/1699522

Apparently this is too much for the SMPS to charge when I turn this
thing on. Usually it will turn the red LED on (failure to start), and
then I need to power it off, wait 30s, and power it on again.
After 2nd power on, it will start with no problems (caps are already (partially) charged I guess).

Any ideas how way to fix it, without messing with SMPS itself?
Add NTCs to the amp rails?
Or some other simple inrush current limiter can be used?

 

[Mooly]

Use smaller reservoir caps. Honestly.

The SMPS running at high frequency needs correspondingly smaller caps to achieve the same amplitude ripple when compared to a conventional 50/60Hz transformer supply.

Adding NTC's to the rails will destroy the regulation as seen at the output end.

 

[minek123]

This is unregulated SMPS.

To remove the boards from heatsinks and replace caps, will require lots of work - removal of all transistors
from heatsinks (16 in total);, it will be pain in the neck. The chassis is kind of crowded
On the other hand - plugging some little board to the rails between spms and amp would be relatively easy..

Use smaller reservoir caps. Honestly.

The SMPS running at high frequency needs correspondingly smaller caps to achieve the same amplitude ripple when compared to a conventional 50/60Hz transformer supply.

Adding NTC's to the rails will destroy the regulation as seen at the output end.

 

[Mooly]

That won't make any difference to what you are trying to achieve. The high loading at start up will be causing the switching transformer to saturate, and that is detected as a fault or overload condition.

I would at least try it and see what value cap you need to work down to for it to start reliably. Depending on the frequency, a 100 or 220uf cap could be performing the same job as a 10,000uf cap in a traditional PSU. Of course ripple current in the cap is another matter at this power level, and that is something you need to look at, maybe splitting the final value into several parallel caps.

 

[minek123]

Thanks Mooly, I guess that's the way to go.

 

[Mooly]

Well I would certainly try it and see where it gets you. Good luck and let us know how it works out

 

[nigelwright7557]

A SMPS will only tolerate a overload current for a short time on power up.
This is to charge its own output capacitors.
It assumes anything longer is a fault condition and rightly shuts down.

As said by others you don't need huge capacitors at the amplifier end.
The switching frequency is probably many 10's kilohertz compared to 50Hz for a linear supply.

 

[wwenze]

On the other hand - plugging some little board to the rails between spms and amp would be relatively easy..

Add a soft start module then.

 

[Pafi]

If the SMPS can't charge 4x2700 uF, then it's likely to be unable to serve the high impulse current needs of the amp either. Reducing caps can be dangerous, ripple current capability must be preserved! Should have linked the SMPS and amps, not the caps. AC inrush current limiting may help also, depending on the SMPS.

 

[Pafi]

Mooly!

Use smaller reservoir caps. Honestly.

The SMPS running at high frequency needs correspondingly smaller caps to achieve the same amplitude ripple when compared to a conventional 50/60Hz transformer supply.

If the only AC current flowing on the caps were at the switching freq, then you were right. But the load current can have any freq, therefore low impedance must be ensured at any freq. Since the PSU is not regulated, and probably not have synchronous rectifier, its ability to maintain low impedance is limited.

 

[Mooly]

There is a big difference in an SMPS starting from zero (a cold start) and seeing a large capacitive loading on its output compared to when its operating normally.

Yes, the load current can be any frequency (well, not any but from a defined maximum downward) and so that includes all the way down to DC. That's a bit like the old argument saying that speaker current passes thorough the reservoir caps... if that were taken at face value then how is a constant DC loading supported ?

Some switching supplies don't like external soft starts and can react badly (even destructively).

 

[AndrewT]

...............
As said by others you don't need huge capacitors at the amplifier end..................

I completely disagree.
One must provide current to meet the transient demands of the speaker/s
That current normally comes from charged capacitors.
Reduce the capacitors and you reduce your peak current capability.

And Pafi disagrees as well

then it's likely to be unable to serve the high impulse current needs of the amp either.

 

[wwenze]

Pafi has mentioned the keyword - unregulated

 

[AndrewT]

post2186
http://www.diyaudio.com/forums/soli...oney-badger-build-thread-110.html#post4720584
states

The problem that is often ignored is that the speaker demands very high peak currents.
Those currents pass through the amplifier and come from the local supply rail decoupling and/or from the PSU.

If the speaker were to demand a transient of 20Apk, would most of that come from the PSU?
Would the PSU shut down? or go into some form of current limiting to protect itself?

and my next post explains why I think an 8.9A smps may not work well for high transient speaker currents.
http://www.diyaudio.com/forums/soli...oney-badger-build-thread-110.html#post4720682
post1 ("SMPS (700W, +/- 75V rails) powering my audio amp. Both channels (2x300W) are powered by the same SMPS.")
shows a 4.7A smps which will be much less capable of meeting high current demand.

 

[AndrewT]

Pafi has mentioned the keyword - unregulated

that does not increase, nor significantly affect, the peak current capability of the SMPS.

 

[Mark Johnson]

For those who enjoy such pastimes, it is a pleasant and rewarding problem in circuit design to create a SmearJob circuit which fits between an SMPS and a large bank of electrolytic capacitors. This circuit senses the voltage on the electrolytic capacitors; when their voltage is less than 75V (during start-up), SmearJob limits the charging current to a value less than 3.5 amperes. Since the SMPS is rated for 4.6 amperes, all is well. Then when the electrolytic capacitors charge all the way up to 75V, the SmearJob ceases its current limiting action and merely becomes a 150 milliohm resistor.

You can calculate the startup time fairly readily; if the total capacitance of the large bank of electrolytic capacitors is "C" farads, and if the final DC voltage is "V" volts, and if the SmearJob circuit's current limit is "I" amperes, then

StartupTime (in seconds) = C * V / I

Numerical example: suppose (C=68,000 microfarads) and (V=75 volts) and (I=3.5 amperes). Then StartupTime = 1.5 seconds. If I were you, I'd add another forty or fifty periods of the AC mains (500 milliseconds) to let the SMPS wake itself up too, and set my speaker relay (muting) delay to be at least 2.0 seconds.

 

[minek123]

If I were you, I'd add another forty or fifty periods of the AC mains (500 milliseconds) to let the SMPS wake itself up too, and set my speaker relay (muting) delay to be at least 2.0 seconds.

Speaker delay is currently set to 3s.
SMPS itself also have some kind of slow start circuit built in - most likely an NTC thermistor...

 

[Pafi]

That's a bit like the old argument saying ...
if that were taken ... then how is ... ?

I'm not going to protect a statement that is different from mine.

 

[Mooly]

Many SMPS can be likened to constant power devices. A traditional PSU doesn't even blink if you draw a transient of 2, 5, even 10 times its rated output. The conventional transformer may saturate under the transient but that's all it is, a transient.

The SMPS thinks huh ho and probably enters a controlled shutdown/limit/duty cycle limit routine to prevent major damage.

The ripple voltage can be thought of as a change in voltage over time. So to get in the right ballpark, if we take the refresh rate of the reservoir cap (say every 20us) and lets say we are drawing 5 amps, then the rate of change of voltage over the cap is given by

V= (I * T)/C

So for a small 470uf reservoir cap we might expect somewhere around 200 millivolts ripple. The ESR rating of the cap is critically important here because the ripple current can be high which of course will worsen the basic calculated figure. The ESR figure becomes dominant. This is why I mentioned earlier that you may need to split the final cap value into several smaller ones.

The transient current output of SMPS is always going to be worse than a traditional PSU, one where you can effectively add as much C as you want. The SMPS is thought of more correctly as a 'constant power output supply' rather than one that can supply huge long lasting current transient pulses.

 

[Pafi]

There is a big difference in an SMPS starting from zero (a cold start) and seeing a large capacitive loading on its output compared to when its operating normally.

Yes, there is difference.

Capacitive load at DC? It equals to 0A. No such thing. The load is an amplifier. Peak current draw must be calculated and served. Capacitor helps doing this.

Some switching supplies don't like external soft starts and can react badly (even destructively).

Those are faulty by design.