audio SMPS startup

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Hello all.
As some of you know, I was designing a 600-800W offline SMPS for audio amplifiers.
The thing is that I have a prototype up and running very well, but I have a small problem:
When I have re-designed the modulator circuitry in order to make switching times smaller (higher efficiency), it seems that my startup circuit doesn't work properly now.

The problem is that the modulator (SG3525+typical dual mosfet driver + pulse transformer) draws too much current at the start, above 200mA, and if you don't provide them, it latches drawing that current but with no output. If my lab supply is limited to, say, 100mA, I can't get it to start.

My actual start-up circuit is a NPN power transistor with 11V zener in the base. The collector has a power resistor in series to 320V bus. I had 7K before (thus limiting current to 320-11/7k=45mA aprox. Now I have reduced to 700 ohm!!! (450mA max) with no luck.

What can be the cause of such a startup current demand and how can I reduce it? I have tried increasing my slow-start time and is now around 1.5secs, but that didn't help.

Best regards,
Pierre
 
Is pulse transformer suffering from saturation? Is your soft-start soft enough? Is some filter capacitor too big? Show us a schematic, but anyway, you will have to trace the source of that excessive current consumption. For example, the SG3525 should never draw more than 10mA when it's not driving any gate load directly. Are you using a gate drive approaches that uses a single gate transformer both to transfer signal and power to gate buffers?
 
Thanks for the quick response.

Problem solved: the culprit was my mosfet driver: an UCC37324, that was suffering of some kind of latch-up (the datasheet says something about max. latch-up current of 500mA).
As soon as I have changed to a similar driver (TC427; this one switches "only" +/-1.5A) it starts with no problem even limiting at 50-60mA.
Unfortunately, it is somewhat slower, I will reduce the series resistor to see if I can get it any better without having ringing.

Rise time of Vgs from 0 to 11V is now around 150nS, fall is somewhat faster due to the gate schottkys. I think it is quite ok, anyway.
Sw frequency is around 80KHz.

Best regards,
Pierre.
 
Schematic & Board

Pierre,

I, too, would like to see a schematic.

About 8-9 yrs ago, I did a modular off-line SMPS controlled by an SG3525 driving a COTS gate drive transformer to a pair of IRF840s connected in half-bridge configuration. Clock frequency was 80kHz, and switching frequency was 40kHz. Rectification was a pair of double-diodes (MUR1620CT & MUR1620CTR) into an LC filter. Feedback, which monitored the +/- outputs equally, was optocoupled back to the (+) input of the 3525's error amp.

The main problem I had was the start-up ckt. The spec sheet for the 3525 (regardless of manufacturer) shows the start-up/stand-by current to be the same as the same as the operating current (~20mA). For example, the datasheet from TI (which bought Unitrode) shows the stand-by current to be ~14-20mA. For a bootstrap start-up ckt to adequately power the PWM chip during start-up, it needs to be capable of supplying at least 50mA, to prevent under-voltage lock-up. This necessitates the use of two 8.2K 5W resistors in series, followed by a TIP50 (or any high-voltage NPN transistor) in a zener-follower circuit.

Newer PWM ICs like the MC33025 (or UC3825) have a start-up current ranging from 500uA to 1.2mA, enabling a bootstrap ckt to be just a couple of hi-value resistors coming off the +320V line.

Of course, the operating current for my supply came from a bootstrap winding off the main power transformer. I think that the excess parts needed to meet the '3525's high start-up/stand-by levels makes this chip a better candidate for lower voltage SMPS applications, like 12VDC SMPSs (car audio).

The MC33025 (UC3825) and other newer hi-speed PWM ICs with lower start-up/stand-by currents are much better suited for off-line applications.

I'm sure EVA or any of the other SMPS gods or goddesses will chime in regarding the start-up condition. Just mu $0.02 worth

Steve
 
Modern SMPS control ICs feature very low startup currents because they don't turn on their internal stuff until the startup supply voltage is reached, and nobody prevents you from using the same technique.

Just charge a 100uF capacitor through a 2mA (or so) current source until it reaches 16V (.8 seconds), then use a comparator with hysteresis (for example, 16V turn-on and 12V turn-off) driving a PNP transistor to apply power abruptly to everything. The comparator circuit should draw very little current by itself (LM393 with high-value resistors should be fine). The current source should also provide power to the comparator circuit (less than 1mA, but it may be turned-off with the same signal used to turn-on the PNP transistor in order to prevent overheating. A discrete latch with two transistors may also work if UVLO is already implemented.

A 100uF capacitor should provide enough energy storage to circumvent the temporal latching of the gate driver until the first few switching pulses start providing charge to it. Smaller values will probably work (just find out how much capacitor voltage drops with oscilloscope).
 
OK, no problem in showing my sch. Let me arrange it a little bit.

N-Channel: my startup circuit looks exactly like you say: 2x3k9/3W resistors in series with a TIP50 transistor. Base goes to a 11V zener biased by 3x47K resistors from 320V bus.
Control circuit is based on a SG3525 + driver exciting a transformer, that drives 2 IRF840 mosfets directly in half-bridge configuration.
Transformer is a ETD44 operating at around 100mT.
Output is rectified with 4x MUR1560 diodes. Inductance is 47uH per rail and capacitance is 6600uF per rail.

Eva: thanks for the info for refining the startup circuitry.

Best regards,
Pierre
 
Start-up Circuit

Pierre:

Here is a link at Amazon.com books regarding a very good book I have used to design my SMPSs:

http://www.amazon.com/gp/product/075067329X/ref=nosim/002-4224792-2842419?n=283155

Marty Brown (EDN Design Series and former Motorola Engineer) is the author. He goes through, in great detail, all of the sections of s comprehensive SMPS, and his section on the start-up section is very good. Real-world calculations, not just theory.

I calculated the component values for my design, based on his design procedure. The start-up ckt I used was rather robust, and the NPN kept cool at all times. I also used a standard 1N4002 diode coming off the pass regulator's output to effectively turn it off after the IC reaches steady-state operating conditions.

However, as I stated before, and this would hold especially if I wanted to mass-produce my design, I preferred to go with simplicity. The simpler the start-up ckt, the better.

Later, I replaced the SG3525 with an MC33025 (UC3825 equivalent) hi-speed PWM from Motorola and later OnSemi.

The '33025 has the comparator with hysteresis integrated that EVA mentions, which keeps everything off until the bypass cap reached full charge, thus giving it that very low (500uA to 1.2mA) start-up current. As I mentioned earlier, my new start-up circuit is now 2 - 100kW 1/4W resistors in series from the +320V line to the +Vcc pin of the PWM IC, bypassed with 100mF. This gives good start-up results.

Steve
 
Thanks for the info. That low startup current chips seem the way to go for extreme simplicity.

Anyway, I am thinking... I have dimensioned the startup circuit to fully support around 60mA with my SG3525+driver, but what is the minimum current so it can start?
Yes, I can hear you, I will experiment ;-) , but just in case some of you have tried before...

Best regards.
 
Pierre:

I felt a bit inspired this morning so I've drawn that:

An externally hosted image should be here but it was not working when we last tested it.


It's untested, so it will probably require some resistor and zener value adjustments. Zener leakage currents may become a bit troublesome causing the transistors to turn on too soon if B-E bleeder resistors are not of a low enough value. This is particularly true for high power zeners, so 0.6W types or smaller should be employed.

The main feature of that circuit is the absence of high voltage transistors, and of resistors subject to high dissipation. It will start to apply power to the output as soon as the voltage at the 100uF capacitor reaches approx. 24V, and it will latch in an on state when the output voltage exceeded two diode drops.

Output voltage is zener regulated to 19V, thus allowing to drive the MOSFETs hard to reduce conduction losses, but without the risk of blowing the gates due to overvoltage (the auxiliary winding has to provide in excess of 20V in order to get 19V output). It also features under-voltage lockout: the circuit will unlatch and remove power to the SMPS control circuit when capacitor voltage falls below 15V or so. Output current is intentionally limited to 500mA.

A 100uF capacitor charged to 24V should provide enough energy for the SMPS to start switching and powering itself through the auxiliary winding. Smaller capacitor values may also work. However, note that placing big decoupling capacitors at the output is undesirable, as they will steal the charge from the main capacitor when the output is turned on.

Even when modern control ICs featuring low startup currents are employed, such a startup circuit may be useful when more things than the control IC have to be powered, like op-amps, comparators and MOSFET drivers (or even a fan)...

I hope this could provide you with some ideas...


Workhorse:

Don't start your production lines too soon...
 
Start-'em Up!

Pierre- Not sure what you're asking. If you're asking what is the '3525's minimum start-up current, then I think it is 14mA. The '33025's minimum start-up current is 500 m A and its max is 1.2mA. If you're asking what is my minimum start-up cruuent that can be powered by the regulator I made, then it is at least 1mA, and at most, probably 40mA.

EVA- Thanks for the schematic. I don't think I have ever seen a start-up ckt so comprehensive. It looks like you've thought of everything here.
 
Thanks, EVA, your contribution is always very appreciated.
Yes, N-Channel, I was only wondering if I could lower the dissipation in the startup resistors & transistor during power up by limiting the current to, say, 15mA.
I think that keeping startup current as low as possible is always good: if so, even if some part of the circuit (for example mosfets or aux supply) fails and the control circuit don't receive other source of voltage but startup supply, the components won't suffer from excessive dissipation.
However, with a more elaborated version such as EVA's, there will be no concern in this aspect.

(About the production lines comment. It seems that everyone is starting to know each other in this forum... je, je)
 
Yes, EVA's contributions to this Forum are always priceless. EVA- please forgive those of us who fail to recognize it enough! ;)

Pierre- To turn off the regulator after the PWM reaches steady-state operation, I calculated the bootsrtap winding to supply about 15-18V, while the regulator supplies never more than 12-13V, which is sufficient for start-up. This reverse-biases the 1N400 diode coming off the regulator, effectively shutting off the Big dropping resistor and NPN portion of the regulator. For the high-value resistor and zener portion, this just stays on, because its power dissipation is negligible (less than 1m W).

Also, for limiting the power dissipation of the power reesistors and the NPN during start-up, I wouldn't really worry about that, since it is in start-up mode for less than 1mS.
 
Yes, N-Channel, that's exactly what I do: use a dedicated secondary of the transformer to produce 12V and a diode in series with the startup to deactivate it when the supply reachs steady state.
As you say, the only dissipation is the zener bias, and the startup is only a few dozens of mS, that is not enough time even for heating the resistors up.

BUT the problem can come from a failure or continued shortcircuit / overcurrent situation: if the protection stops the SG3525, the aux. winding will produce no voltage, and the power supply will try to re-start continuously with no success, that can lead to an overtemperature in the startup circuitry (resistors and NPN).

I thought of a simple way to solve this: insert a capacitor in series with the base of the NPN startup transistor, so it can start only once. If for any reason the supply stops oscillation, it won't start never, until it is unplugged and cap discharged. Simulation worked very well, but I haven't tried this.
But this can be an inconvenient if the overcurrent triggers occasionally, for example, during a heavy bass. That should be not dangerous, but you will have to be there to unplug the supply, so I don't like it very much.

Do you understand what I mean?

Thanks
 
I had pretty bad experiences with IR21xx series for offline applications, so I moved to transformer-based gate drive.
However, they perform very well in lower voltage Class-D amplifiers, from my experience.

Regulation is not mandatory for moderate power levels, however (the main problem can be no-load or very-light loading voltage rise).

Best regards,
Pierre
 
Pierre-

Yes, I understand now what you're trying to say about a lock-up condition. With respect to your "base capacitor" coupling idea, I like the solution's elegant simplicity. What wopuld be a good value for the "coupling" cap! A nice big value, like 500-1000mF? Or something smaller?

Steve
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.