5000w Smpsu!

[Emmy]

Hi all,

Okay, this might sound crazy, but I need a power supply capable of peaking at 5000W. I'd like it to be able to produce around 3000W continuously. The amplifiers it will be driving are rated at 2000W RMS continuous (combined). DC output voltages required are 63-0-63. My choices, as far as I can see, are:

1 - Huge and/or multiple toroidal transformer(s). This method might be inefficient, big and very, very heavy!

2 - Design/build an SMPSU. Although this might be more complex, it would probably be much lighter, much more efficient, and be a lot smaller. It will also allow me the flexibility of easily adding extra voltages (for other parts of my design)...

Therefore, looking at option 2, I'd like this to run at 100-200KHz, but I have limited experience of SMPSU's! I have plenty of experience in other electronics, so I'm not daunted by it, but I need more info! Finding relevent info on this subject is quite difficult!

So, anyone got any ideas?

Thanks!

Shash

 

[SmarmyDog]

Sheesh Shash...

Only 5KW?

You could keep an eye on eBay for those Vicor 600W DC-DC converters. They can be wired in series if you use protection diodes (see vicorpower.com for app notes). That means you'd only need to design the AC mains rectifier & capacitors. I have a few dozen I plan on using for exatly the same thing.

-Casey Walsh

 

[Sch3mat1c]

Meh, I'd go for some IGBT's at 50 or 100kHz. Full bridge. Find some heavy ferrite pieces, add the right turns ratio and there 'ya have it.

Tim

 

[Emmy]

Hi all,

Thanks for the replies. Although using those 600W modules sounds good, I'd prefer to build my own SMPS in order to reduce the size and cost as much as possible. So, just thinking about it, let me see if I've got this straight (feel free to stop me if I'm not making any sense!):

I've decided to build for around 3000W continuous INPUT, so that should give me the following:

Input Voltage: 240VAC
Input Current: 12.5A

So, first stage is to rectify the voltage to DC. Using a full-wave rectifier, that should give me:

Vdc = 220 * sqrt(2)
Vdc = 340 VDC

Then I need to switch this voltage using (say) the IRG4PC50W IGBT. Using this device, it looks like the optimum switching speed for my current requirements is around 50KHz (square-wave).

So at this point I'd feed that into a transformer with the "right" number of turns, rectify the output with a single rectifier diode, and feed that into a bank of low-ESR capacitors (say 50,000uF built up of 20 x 2,500uF caps).

Sound about right so far?

The only issue I guess would be producing the 50KHz square-wave. Considering how variable the input supply can be, and the fact that I'd like to be able to support a wide-range of supply voltages (90V-270V AC), I guess I should use an SMPS control IC with secondary opto-coupled feedback. So, anyone got some recommendations on what control IC to use?

Thanks!

Shash

 

[Sch3mat1c]

This is just a scaled up version of what's in your computer, matter'a fact I bet you could pull the very same circuit and just beef up the driver, switchers and transformer. You'll want FWCT rectification and choke input; no, I'm not sure how PWM works on that (unless you have an SG3524, or an external flip flop to do the same thing).

AFAIK, anything over 1kW is full bridge (aka H bridge) topology, so count on using four IGBT's.

Tim

 

[gatedriver]

I don't mean to rain on your parade, but if you have never built a lower power switching circuit you will have a s**tload of problems with such a big SMPS because you don't know about all the parasitics involved.
Without a competently designed PCB it's pretty much a no-go. The transformer construction and general physical layout is critical.

the simplest approach for a SMPS would be one of these self oscillating ICs like IR2153, unregulated with a full bridge rectifier and AC coupled transformer. Even this will be non-trivial to get working at this power level.

For one-off DIY, better look for some old transformer! or spend the cash.. still better than spending 100 bucks on SMPS components only to blow it up

 

[Emmy]

Heya,

I understand the complexity of SMPS design, and understand how critical the PCB/transformer layout is. I would use a standard unregulated supply, but the size and mass of such a unit would not be feasible for my application... The only supply capable of supplying enough power with a small enough form factor would be SMPS...

Also, I see this as a learning experience - and I love steep curves! Even if I blow it up the first time... I've been building and designing things for over 15 years (including high-frequency and sensitive circuits), so I should be able to make some pretty accurate "educated guesses"!

Thanks!

Shash

 

[djQUAN]

with that much power and it being your first time with a SMPS, if something goes wrong, it could be your last.

I suggest you go practice with a smaller SMPS (about 100W or so) then go up. I started with very small ones of ~3W and right now, the biggest is only 500W RMS. I don't plan to go higher than that yet.

 

[Emmy]

Hiya,

I'm actually extremely paranoid about safety, so I'd have extreme precautions in place before activating any part of the circuit - and I'll not be standing anywhere near it!

Honestly though, as long as I'm generous on component ratings and I don't make a huge construction mistake (I'm paranoid about that too - so everything will be quadruply-checked!), it should be okay. I've worked with some very high voltages and power systems (including 415V 3-phase circuits), so I'm not too worried about my abilities!

djQUAN: You say you've built supplies upto 500WRMS - do you use a full-bridge or half-bridge transistor formation feeding the high frequency transformer? What specifications do you build your transformers to? What sort of ripple do you get on your output?

Thanks!

Shash

 

[jackinnj]

before you spend anything on this project, spend some money on Pressman's book, or Sanjaya Maniktala's book -- the peak currents in a switchers are much larger than the average drawn --it is just this worm's eye view of the universe that every time you go from 10 to 100 to 1000 watts the complexity increases by a factor greater than 10.

 

[Sch3mat1c]

FYI, I might be of some help. I'm working on a similar project at the moment, although it is made to produce heat, not voltage.

 

[djQUAN]

djQUAN: You say you've built supplies upto 500WRMS - do you use a full-bridge or half-bridge transistor formation feeding the high frequency transformer? What specifications do you build your transformers to? What sort of ripple do you get on your output?

the circuit that I used has this controller but with mosfet outputs. www.djquan.angelcities.com/smps.html


I haven't tried measuring ripple on the PSU output but I don't get any on the amplifier outputs so I guess it's all good. I used only half bridge topology. the ferrite core that I used is said to be able to go up to 1kW or so but used it to 500W to be safe. for the windings, I just use one 1mm dia enamelled wire for every 1oA of current. with this one, I had 5 for each primary.

 

[Easyamp]

3 speakers running off a 300w amplifier will produce the same volume as one speaker with a 1000watts behind it.

The ol "every 3db gain double the power" will get you.

To increase the volume a barely perceptable level of 3db means you must double the power (100watts to 200watts) and if you want to double the percieved volume then you need 10db's of gain or 10 times the power(100watts to 1000watts). and the rest of that 5000w amp's watts would only get you a couple more db's which makes a 5000w psu just a lot of work for not, when you consider that you increase your output by 3db by just adding another speaker and 10db by adding another all running at the same power.

This means that with 6 speakers and a 300watt per channel amp you can get as much volume from them as: (300/3=100) 100 watts each speaker per channel, with 10db gain from having three speakers equals a single speaker driven by a 1000watt amp and and a total of 2 speakers and 2000 watts. As the volume levels increase this speaker power realtionship becomes much more obvious.


Not to mention the added expense of speakers and eqipment made handle all that power.

The cost to volume ratio doesn't support single high powered amps but that hasn't stopped a few.

 

[Sch3mat1c]

3 speakers running off a 300w amplifier will produce the same volume as one speaker with a 1000watts behind it.

IFF ("if and only if") the speakers have the same efficiency, and are EACH ran from a dedicated amplifier. Three hundred watts is three hundred watts; three amplifiers each, however, totals 900W, "close enough" to the 1000W comparison.

If reality worked the way you worded it, everyone would use 100 speakers connected in parallel, powered by a single 1W amp, as opposed to a single speaker ran by a 100W amp. See the discontinuity here?

Tim

 

[hummhoom]

Speakers are more efficient in numbers. If you have an 8 ohm speaker, and you series it with another 8ohm speaker you get a 16 ohm load that consumes half the power, but makes the same sound as just one speaker.

 

[Sch3mat1c]

Do you have the math to support this?

Tim

 

[Prune]

Huh? It would take half the current to produce the same SPL, but twice the voltage. So there's no change in power.

BTW, I'm wondering, why not wire 100 small drivers in a series array and drive with an OTL tube amp? I like tubes, but I'm prejudiced against output transformers...

 

[Mr Teal]

Though I have no formulas to back it up, the general rule of thumb is that a doubling of total power results in a 3 db increase, and a doubling of cone area results in a 3 db increase.

If you have two drivers with 150W each, they should be ~3db louder than one identical driver with 300W. Two drivers with 300W would be 6db louder than one with 300W.

 

[Prune]

Well, of course -- that's by definition -- you are just changing to a logarithmic scale. Say you have power ratio of 2:1, then:
10 log(P2/P1) = 10 log(2/1) ~= 3.01 dB.
For voltage, the multiplier is 20 instead of 10, so doubling voltage on the linear scale increases ~6.02 dB on the logarithmic scale.

 

[Easyamp]

IFF ("if and only if") the speakers have the same efficiency, and are EACH ran from a dedicated amplifier. Three hundred watts is three hundred watts; three amplifiers each, however, totals 900W, "close enough" to the 1000W comparison.

Sch3mat1c read my post again I clearly state that each of the three speakers running off one amplifier see 100watts (3 speakers 300 watts each speaker get's 100w).

This eactly what I said

"3 speakers running off a 300w amplifier will produce the same volume as one speaker with a 1000watts behind it."

That's three speakers running off a 300 watt amplifer not three 300 watt amplifiers??

Every time you add a speaker you increase your gain by 3db when you add 2 speakers you increase your db by 10,
I'm not just making this up, it's fact.

For every 3db of gain you'll roughly need to double your output power and for 10db of gain roughly 10 times the power.

10db of gain is generally percieved to be about twice as loud. so if I have three speakers each with one hundred watts of power it will be equivalent to one speaker being driven with a 1000 watt's.