500W SMPS for audio

[mag]

Hi all,
I have designed and built a 500W peak offline SMPS for audio use that I wanted to share on this forum.
I use it with a Class-D amplifier based on IRS2092 that I designed and everything works fine.

The SMPS is designed to deliver +/-50V at 2A continuous (for thermal reasons)
and a peak of 5.5A par rail (550W peak power).

It provides also a ground referenced auxiliary output at around 18V useful to
power protection circuits.

The SMPS is based on the Transition mode 2-switch flyback topology with L6565 PWM IC and pulse transformer drive.

In attached you will find the schematics and some pictures of it.

Here are the measured characteristics:

Input range: 187...265VAC
Efficiency @ 5.5A out = 92.5%
Efficiency @ 2A out = 93%
Standby power < 3W
Peak output current: 5.7A
Short circuit current: 18A for 300ms (hiccup cycle)
Output ripple < 200mVpp @ 5.5A out.

Magnetics parts:

Transformer: EER42/20 Core with ~1.5mm gap on the center leg
primary: 32turns 4x0.45mm (split in 2 16 turns layers)
secondary: 9 turns 8x0.45mm per section
bias: 3 turns 1x0.45mm
aux: 3 turns 1x0.45mm
primary inductance: 250uH
leakage inductance: 3uH

Pulse transformer: high permeability toroid core diameter 15mm
pri=sec1=sec2=20turns wrap wire, trifilar wound

EMI filter: stolen from a PC power supply... measured 2x15mH

Any questions or comments are appreciated

thank you

ciao

-marco

[Pwrmad]

Nicely done. I've used the "DC Restore" diode in the gate drive secondary in some of my own designs and it does have one drawback. When the pwm is inhibited, by UVLO for example, the primary coupling cap can ring with the drive xfmr enough to turn on the main FETs. Maybe not all the time and maybe the energy in the bulk caps won't blow out the FETs right away, but it's still dangerous. I've seen it happen. What I've done is to minimize the values of the coupling caps and add damping. In 100KHz designs, I'd use a 0.1uF||(0.47uF + 50 Ohms) on the primary and 0.033uF||1K on the secondary.

[eem2am]

YES, Pwrmad is right.......the DC restore prafo gate drive circuit is HIGHLY SUSPECT.........

http://www.edaboard.com/thread231931.html

I am afraid that this DC restore circuit must be decalred dangerous and a threat to the lifetime of the product.

[eem2am]

Also, may i also say....
1...You need a PNP turn off circuit on your upper fet.
2....Why bother with the lower gate drive secondary....just drive the lower fet directly from the L6565 output.
3.....What is your peak and RMS current in fets and output diodes?.....i believe it will be ridiculously high when at 500W load.
....i think you need a continuous mode flyback......if you make it continuous above , say , 200W, it will be discontinuous at lower powers and thats fine for you.
4......have you multiple sandwiched your seconday.....otherwise each half of the output coil will not "see" similar coupling intothe core, and you will not get symetrical rails.
5.....npn/pnp gate drives are a bad idea, because the tolerance on such discretes is poor....and you cannot gaurantee enough hfe at high-ish Ic values.....also, the fT values willl be poorly toleranced..........you should use a proper gate driver IC instead.
6.........your primary fets and secondary diodes are on the same heatsink...so how have you got sufficient isolation between them.
7......your high primary bus capacitance is going to nuisance trip your fuse too often......the fuse manufacturer will tell you this wont happen...but it will.
8......When low bass notes are played, your split rails will loose symetry.....you need to solve this by doing an smps for each split rail... or at the very least, used weighted feedback from SGND and VREC+
9....C23 and C24 will cause you grief, they will resonate like mad with the leakage inductance when the fets are switched off.
10...R30 is the wrong side of C31
11....You forgot to connect SGND between C20 and C18
12....i think the use of Q5 is erroneous.
13...you have big heatsink for fets and diodes...but why then no heetsink for mains rectifier bridge.
14.....you need active power limiting circuit...becauuse your smps will go up in smoke if on 500W continuously...unless you can proove that under no circumstance will it see 500W continuous.

cheers

All the best

[eem2am]

Hi,

Also what speakers are you driving with that output...it must be several speakers in series?

also..is your start up circuit capable of supplying full operating current...if so , then you have insufficient short circuit protection.
...


ALSO....your SMPS will BLOW UP, because you are reflecting more than 300V to the primary, and at low vin, when the input caps are at the trough, there will be less than 300V there, so you will have big problem....I am sorry, it was great of you to present........i also always make big big mistake, and then some other tells me about it and i am sad.....but i must tell you my friend, so your DIY will be in future sucesful and you will be happy in the long term.

I wish you every future success, thanks for your interesting concept.

[Pwrmad]

I didn't mean to suggest that the DC Restore should never be used, only that it needs to be implemented carefully. I've used it, with damping and minimized tank energy, in several successful designs. I do agree that PNP turn-off circuits would be a good idea, especially with these large FETs. Diodes Inc, make some very nice transistors for the gate drive. Tiny, yet fast and lots of current gain for high peak currents. They even have duals for this application. Dedicated gate drive ICs are good too, but not perfect. I would always drive both FETs identically, not one direct and one by xfmr. When both switch together, it minimizes common-mode voltage/current on the xfmr windings. Interleaving xfmr windings is almost always a good thing. My preferred build would be: half primary in one layer; secondary wound bifilar in no more than two layers; half primary in one layer; aux windings can go where convenient. Inrush limiting is always a good thing. C23/24 do need a damping resistor. 50-100 Ohms. Probably 2W.

[eem2am]

thanks Pwrmad, i hear you say you use the circuit with success, but have you done long-time soak testing of it, to see if the lifetime of the smps is affected......(ie by the gate drive problem)

Those diodes.com or zetex transistors are very expensive, -if you are going to use them you might just as well spend the money on a gate drive chip.

With a two transistor forward or flyback it really doent matter if you just drive the bottom fet directly......no current will flow until both are on so who cares if the directly driven one comes on first....it really doesnt matter........and in any case, even of you drive both off a coil......then each coil will likely have different leakage, and come on at different times any way....not only that but the series resistance of the fet gates is likely to have tolerance, so we can forget the fets ever coming on at the same instant........its best to just use a simple single output gate drive transformer.

[Kv3audio]

Well done mag

great work

[eem2am]

Hi,

I notice that you say you achieved 92.5% efficiency with this boundary conduction mode flyback when Vin was 187VAC, and the output power was 550W.

I am sorry, but i believe that this is an error in the measurement.....there's absolutely no way a BCM flyback at 550W from 187VAC can do 90%+ efficiency.

How did you measure your input power?

Also, your claim that efficiency at 200W output was just half a percent less than at 550W output is impossible to believe.

I think you have a serious measurment error here.

[Pwrmad]

eem2am, I don't think I've ever actually seen a failure attributable to this issue. The potential was first shown to me by a colleague working with me on the same power supply about 15 years ago. We worked out the fix I outlined above. That power supply and others (probably 7 different converters) that used that circuit have undergone extensive development, qualification, production and life testing, including HALT testing, without incident linked to this issue. But, I have seen scope captures showing Vgs ringing to several volts and the bulk supply voltage crashing down when pwm is inhibited at turn-off and that's enough for me. The Zetex parts are not that expensive. For commercial product where pennies matter, maybe. For the DIYer not so much. The difference is minimal. You're right, the drive scheme doesn't really matter. I've almost always driven two-switch converters with identical xfmr circuits in the hope that it would allow symmetrical switching of the primary winding and minimize the common-mode current through the pri-sec capacitance. It's a design choice. I've never tested the difference because that would require building two versions of the unit when development schedules barely allow time to build and test one. The one time I did use xfmr drive on only the top switch in an offline 2-switch forward converter. It worked just fine.