Higher power switching psu, any design advice?

[Immo_G]

I am planning on eventually building a car amp, was wondering if anyone had any tips for larger (1000+ watts) amp power supplies.

Firstly , I presume i'll have to use 2 small toriods rather than the standard 1. Is there a trick to getting the voltages matched so the ground is evenly between them.

Secondly, what sort of constant load is required for stability, like i could use a 1k ohm on a 35 rail and only pull 1 watt, or does it require more.

It appears that the secondary side ground can be totally floating compared to the primary. Is it best to totally separate them, to have a medium resistor linking, or to actually connect them directly?

Any advice would be welcomed, thanks.

[dunderchief]

I can't say I can help you at all, but I would really like to see what you come up with. I've asked a couple of times if there are designs out there and have gotten little responce. Keep us posted.

[sss]

check out rod elliots sote
www.sound.au.com
u can find a 350W smps , use couple of those , one 4 each channel , i think that will be enough

[dkemppai]

Originally posted by Immo_G
I am planning on eventually building a car amp, was wondering if anyone had any tips for larger (1000+ watts) amp power supplies.

Any advice would be welcomed, thanks.

Do you have an oscillscope? A good one? If not, my advice is to get one. Get a good one. You can't fix problems you can't see.
(Tektronix TDS1000, TDS2000 are good. TDS3000 series, are my favorites)

Stablility depends on your feedback network. It's different for every design. The bias current in my amp is enough load on power supply at idle.


Use good fets. Uset 55 to 75 volt rated devices, it'll give you a little room for leakage inductance on your transformer. Watch the transient voltage on your good fets! (Thus the good scope)

Wind both of the secondary windings at the same time (Cut two peices of wire, and wind them exactly in parallel). that will help to ensure that you get the same rail voltages. And, you only really need one transformer. (I get 2Kw from one xfmr)

The biggest thing, is go slow. Test everything at small loads, and work your way up. Even build some small supplies... ...when you burn things up, it'll be cheaper (Ask me how I know).
There are lots of possible gotcha's...

Oh, yeah... ...start collecting parts as soon as you can! Unless you have bottomless pockets, cash can dissappear fast!

Good Luck!
-Dan

[Claude Abraham]

At a power level of 1000+ watts, I'd recommend a full bridge converter. This requires a PWM controller, four MOSFETs (n-channel recommended), two high-side n-channel MOSFET drivers, a power inductor, and a single-primary, center-tapped secondary transformer, two ultrafast or Schottky rectifiers, and input and output filter capacitors.. The PWM controller should preferably be current-mode control. There are a lot of high-powered SMPS schematics floating around that use no filter inductor. Definitely avoid those. Everyone who attempts to build them has chronic trouble. Check out the Texas Instruments and National Semiconductor sites for application notes. I hope this helps.

[Immo_G]

Yeah i've heard someone mention a full bridge before, hadn't considered using it, but since i'll make my own transformer, i may as well use it.

Re the transformer, parts store here has a outside diam 75, inside 21, height 30mm, prewould one i would rewind, or I could go a powdered iron one thats a fair bit smaller, though i don't know how the various materials affect it. One of the schematics mentioned only to use a ferrite core.

Thanks for the advice about cutting both wires the same length, that seems a good way to match the rails.

[subwo1]

I have been working of SMPS designs which do not require full bridge. They are half bridge design with capacitor coupling between the mosfets and transformer. They are capable of lots of output power and can do positive and negative rails from the same supply. Using half-bridge provides advantages like stopping DC bias from reaching the transformer, and only needing half the drive circuitry and number of output mosfets. The switching power does not have to pass through two sets of mosfets in series, as would be required for full bridge, so the circuit is simpler. Check out the link below this post.

You would use a ferrite core for the power transformer, not powdered iron for efficiency. You use two windings on the same transformer, both are full-wave rectified to each power rail.

[Claude Abraham]

have been working of SMPS designs which do not require full bridge. They are half bridge design with capacitor coupling between the mosfets and transformer. They are capable of lots of output power and can do positive and negative rails from the same supply. Using half-bridge provides advantages like stopping DC bias from reaching the transformer, and only needing half the drive circuitry and number of output mosfets. The switching power does not have to pass through two sets of mosfets in series, as would be required for full bridge, so the circuit is simpler. Check out the link below this post.

Actually, the half bridge does have a problem with dc bias reaching the transformer. Charge imbalance between the two capacitors can produce flux imbalance in the transformer core. The volt-seconds in the positive and negative directions are unequal, and runaway and core saturation will take place. This can happen with either voltage-mode or current-mode control.

The full bridge, and the push-pull topologies which employ current-mode control do not have this problem. A full bridge does use four MOSFETs vs. two for the half bridge, but each FET in a full bridge need only withstand the input supply voltage. In a half bridge, each FET must withstand *twice* the input voltage. The full bridge can use lower voltage MOSFETs, which have lower on resistance and switching losses. Four parts in the full bridge will run cooler than the two parts in the half bridge, and higher efficiency is obtained.

A rough rule of thumb is to use push-pull for power levels around 200 to 500 watts. Above 500 W, the next choice has traditionally been half bridge. Then, full bridge is recommended for the highest levels of power. I personally eschew the half bridge for the full. If the push-pull is not quite enough, I go for the full bridge. Simpler is not always better. Fewer parts generally means more stress on each one. I just want you to know the tradeoffs involved.
As far as the schematics go, I'd avoid any circuit without an inductor. The transformer output should be followed by a power inductor. National Semi, Texas Instruments, and Linear Technology have good app notes on their sites. Best regards.

[subwo1]

Originally posted by Claude Abraham


Actually, the half bridge does have a problem with dc bias reaching the transformer. Charge imbalance between the two capacitors can produce flux imbalance in the transformer core. The volt-seconds in the positive and negative directions are unequal, and runaway and core saturation will take place. This can happen with either voltage-mode or current-mode control

I have not noticed this kind of problem happening maybe because of the capacitor coupling of the transformer. I also place a capacitor between the transformer and the recitfiers for overload protection.

Originally posted by Claude Abraham







The full bridge, and the push-pull topologies which employ current-mode control do not have this problem. A full bridge does use four MOSFETs vs. two for the half bridge, but each FET in a full bridge need only withstand the input supply voltage. In a half bridge, each FET must withstand *twice* the input voltage. The full bridge can use lower voltage MOSFETs, which have lower on resistance and switching losses. Four parts in the full bridge will run cooler than the two parts in the half bridge, and higher efficiency is obtained.

Does having to withstand double the input voltage matters when the voltage is only 12 to begin with? I don't think they run cooler for the same output power, but I guess that the power efficiency may be better due to less current having to pass through the circuit. But in the mosfets themselves, no theoretical efficiency benefit is gotten since the mosfet channel resistance is doubled, which admittedly does not consider that manufacutures of mosfets make the lower voltage ones better.

Originally posted by Claude Abraham





A rough rule of thumb is to use push-pull for power levels around 200 to 500 watts. Above 500 W, the next choice has traditionally been half bridge. Then, full bridge is recommended for the highest levels of power. I personally eschew the half bridge for the full. If the push-pull is not quite enough, I go for the full bridge. Simpler is not always better. Fewer parts generally means more stress on each one. I just want you to know the tradeoffs involved.
As far as the schematics go, I'd avoid any circuit without an inductor. The transformer output should be followed by a power inductor. National Semi, Texas Instruments, and Linear Technology have good app notes on their sites. Best regards.

An output inductor is definitely important for efficiency. Thanks for your thoughts. Best Regards.

[Claude Abraham]

Does having to withstand double the input voltage matters when the voltage is only 12 to begin with?

Yes. Yes. Emphatically yes! Log onto any MOSFET producer's web site. Compare the premium 20 volt part against the premium 40 volt part. Likewise with 30 V vs. 60 V, 100 vs. 200, etc. The lower voltage parts offer less on resistance without increasing the switching losses (rise time, fall time, total gate charge). The full bridge has four FETs, while the push-pull and half bridge have only two, but the four parts in the full bridge are each subjected to lower stress and losses.

However, one mistake I made needs correcting. I shouldn't rely entirely on memory when making comparisons. The FETs in a push-pull are subjected to the same current as in a full bridge, but twice the voltage. I earlier stated that the stress on each FET in a half bridge is twice the voltage, and the same current as a full bridge. Actually, it's twice the current, and same voltage as full bridge. Since there are two FETs in a half bridge, only one conducts at a time, but the conduction loss is four times that of each FET in a full bridge. The loss is equal to current squared times Rdson, or (I^2)*Rdson. Twice the current yields four times the loss, but since only one drop is incurred vs. two in the full bridge, the total loss is half of four, or twice that of a full bridge. So, the half bridge incurs twice the conduction loss as a full bridge. Even worse, the higher power loss is dissipated by only two FETs, vs. four in the full bridge. The temperature rise is four times as great per individual FET. Since the half bridge FETs run a lot hotter, their Rdson will increase even further. It should be very clear to all who read this, why the full bridge is the premium topology for applications which demand the highest power. At 1000+ watts, stated earlier by the person who started this thread, the full bridge would be the best course of action.

As far as the dc bias issue goes, a series coupling capacitor would indeed eliminate the problem. Best regards.