When you build this high power PSU, make sure it's working 'gradually' before the big power-up. Apologies if I'm stating the obvious.
Generally, it's advisable to first check chassis isolation from input and output plus input to output(s). Then use an external DC low power supply (14 volt 200mA or whatever) to feed the auxiliary IC supply to check that there's a full duty cycle drive signal to the O/P devices. Check drive phasing is correct if you have a dual channel scope. Then feed in a few volts to the O/P stage supply to check that a little output is achieved without a lot of current flow, it will probably be well under 100mA, it certainly is with 1Kw PSUs I've worked on, more like 20mA. This should rule out shorted transformers, mis-wired O/P stages and the like.
Gradually wind up O/P stage supply volts and check that regulation is reached well before normal running I/P voltage is reached. If you haven't got the facility to wind up I/P volts gradually to a high enough voltage, then an alternative if you have a scope is to just use the auxiliary supply as above but with no power I/P feed to the O/P stage, and back-feed DC up the O/P and wind this up while observing the drive waveform on a scope. The waveform should shut down when the back-fed voltage reaches the desired O/P voltage.
Hope this helps.
Generally, it's advisable to first check chassis isolation from input and output plus input to output(s). Then use an external DC low power supply (14 volt 200mA or whatever) to feed the auxiliary IC supply to check that there's a full duty cycle drive signal to the O/P devices. Check drive phasing is correct if you have a dual channel scope. Then feed in a few volts to the O/P stage supply to check that a little output is achieved without a lot of current flow, it will probably be well under 100mA, it certainly is with 1Kw PSUs I've worked on, more like 20mA. This should rule out shorted transformers, mis-wired O/P stages and the like.
Gradually wind up O/P stage supply volts and check that regulation is reached well before normal running I/P voltage is reached. If you haven't got the facility to wind up I/P volts gradually to a high enough voltage, then an alternative if you have a scope is to just use the auxiliary supply as above but with no power I/P feed to the O/P stage, and back-feed DC up the O/P and wind this up while observing the drive waveform on a scope. The waveform should shut down when the back-fed voltage reaches the desired O/P voltage.
Hope this helps.
joeliu said:
I was pondering this problem lately. I'm working on a small batch of huge power amps, running off +/-138V. The first ones are using standard 60 Hz toroids in the power supplies for which I've had the pleasure of custom-winding the secondaries. Considering switch-mode of the second batch, but at these power levels I'm not too sure of myself. I've built little supplies before, and 500W DC/DC for car amps, but never anything line operated at several kW other than 60 Hz core-coil. The amps are class H, 3 tier. Since they effectively require six power supplies, would putting six 45 volt SMPSs in series to generate all the rail taps be an option worth considering? Six IGBT pairs, six trafos, each one waaaay smaller than doing the whole thing at once and way easier to get perfected. Good idea, bad idea?
QSC uses an unregulated quasi-resonant ZCS power supply which means no inductor that has to take peak output power without saturating. Output peak power is only limited by heat production in transformer, output capacitors, diodes and transistors. ZCS gives lowered switching losses for the IGBTs allowing it to operate at relatively high frequency.
The parts that incrase most in size and cost with high output powers have pretty long time constants so it's a pretty nice topology for high peak power with less average power output.
The parts that incrase most in size and cost with high output powers have pretty long time constants so it's a pretty nice topology for high peak power with less average power output.
I prefer to have a series inductor in the system because then I can use current control to make an overload and shortcircuit proof power supply. A SMPS with proper current limiting is hard to destroy.
Regulation is not that important but it may be cleverly used to accomodate lower impedance loads by reducing supply rails. A PIC can sample output voltage and current and do that.
Regulation is not that important but it may be cleverly used to accomodate lower impedance loads by reducing supply rails. A PIC can sample output voltage and current and do that.
Hmm, true.
They detect peak current and if over some value the power supply shuts down for a short while and restarts. This limit is set higher than the current limit for the amplifier(s) supplied though IIRC so it should only activate if the output stage is already blown...
They don't use PFC but if they did the PFC could be made to reduce voltage if it's original output voltage is a bit higher than commonly used
One benefeit is being able to have most of the energy storage on the primary side, more compact for the same energy and both (or the many) rails get the energy from the same caps. Isn't this worth it?
They detect peak current and if over some value the power supply shuts down for a short while and restarts. This limit is set higher than the current limit for the amplifier(s) supplied though IIRC so it should only activate if the output stage is already blown...
They don't use PFC but if they did the PFC could be made to reduce voltage if it's original output voltage is a bit higher than commonly used
One benefeit is being able to have most of the energy storage on the primary side, more compact for the same energy and both (or the many) rails get the energy from the same caps. Isn't this worth it?
megajocke said:
But those "long time constants" are on the order of seconds, not minutes or hours. Not the same kind of overload tolerance you get with a 20 pound 60 hz toroid. With switchmode, when you get beyond your experience level things go kaboom without even enough time to take the measurements you need to find out why. Which is why I was condsidering separate cores, each one at a more manageable power level, not some 5kW monster that other people around here have been advised against doing. And a supply that's really only good for 800W or so is a step in the wrong direction. If I wanted that for my sub amps I'd buy them in the store instead of building them. I was thinking about that series topology since seeing that 2500W inverter circuit using 8 cells in series to get to line voltage.
If building switchmode supplies for my next batch of amps is just still out of the question at this point, I'll gladly get out my trusty roll off 11 gage magnet wire and wrap some more 20 pound toroids - those WILL work without a fight. And when I can figure out how to push 5kW through a core the size of a matchbox without starting a fire, I'll go back to it.
I've still go time to play with various solutions - the first 4 amps are committed to using standard rectifier supplies because the chassis are already machined for them, and the cores have already been wound. And I want them operational by winter. The second batch is still in the planning stages and I'm open to any suitable power supply option (same power amp module, however).
Transistor and diode package-to sink are something like that but the windings of a relatively large switchmode transformer still has a bit longer time constants than that.
Copper has a volumetric heat capacity of about 3.5 J/(cm^3 * K). An ETD-49 core has about 20cm^3 of winding volume. Let's say half of that is copper. The heat capacity of that will then be 35J/K. If the thermal resistance is 10K/W this will be a time constant of 350 seconds - five minutes. If the thermal resistance is better because of forced cooling the time constant will be shorter of course but it is still pretty long compared to musical transients, measures or even parts of songs.
But of course, it does not compare to a huge 20 lbs toroid when it comes to overload tolerance...
If you want to do the separate converters maybe you could use one converter to supply the +1 and -1 rail, one to do the +1 to +2 and -1 to -2 and so on. This way you will never have the situation when three converters are supplying max power while three are idle. The primary of the three converters will not see more peak power than the six converter version but average power will be higher. (which is more like traditional SMPS applications where maximum peak power and average power are the same)
BTW, what are you using for output devices in the amp? Do you let supplies drop to 0V on opposite polarity output?
Copper has a volumetric heat capacity of about 3.5 J/(cm^3 * K). An ETD-49 core has about 20cm^3 of winding volume. Let's say half of that is copper. The heat capacity of that will then be 35J/K. If the thermal resistance is 10K/W this will be a time constant of 350 seconds - five minutes. If the thermal resistance is better because of forced cooling the time constant will be shorter of course but it is still pretty long compared to musical transients, measures or even parts of songs.
But of course, it does not compare to a huge 20 lbs toroid when it comes to overload tolerance...
If you want to do the separate converters maybe you could use one converter to supply the +1 and -1 rail, one to do the +1 to +2 and -1 to -2 and so on. This way you will never have the situation when three converters are supplying max power while three are idle. The primary of the three converters will not see more peak power than the six converter version but average power will be higher. (which is more like traditional SMPS applications where maximum peak power and average power are the same)
BTW, what are you using for output devices in the amp? Do you let supplies drop to 0V on opposite polarity output?
I see what you're saying - it won't be buying as much improvement in component stress as I'd be expecting. I'll be spending some time on th power supply side over the next couple months to see what I can get away with. What I was really hoping to do is be able to operate these things at 1/8 power, pink noise, at 2 ohms indefinitely - not just for the usual couple minutes or seconds that you get with store-bought amps. The application will draw about 2200VA under those conditions - which is not quite double the trafo rating in the RMX5050. It uses a pair of 600VA. I'm using two 1100VA (50 Hz) units that have been rewound to make full use of the core at 60 Hz so it should take more. When I go to switch mode, I'd like to keep the same capabilities and not have to baby them.
Output stage is 9x MJL21193/4 pairs. I would have liked to use 12, but it just woudn't fit on a 14" PCB (along with rail switches). I thought about a 2.5-step that goes to zero at reactive crossing, but decided on 3 equal rails because it only needed 50 volt soup cans which are soooo much easier to get cheap.
Output stage is 9x MJL21193/4 pairs. I would have liked to use 12, but it just woudn't fit on a 14" PCB (along with rail switches). I thought about a 2.5-step that goes to zero at reactive crossing, but decided on 3 equal rails because it only needed 50 volt soup cans which are soooo much easier to get cheap.
I have a small smps for my new power amplifier.
I use it in Half bridge, ixus IGBT's driven by a ir2110 at 120k.
sg3525 and E65 ferrite on 3c90 /f44. (foil wound)
i can get 8KW out of it.
this powers my new 2kw per channel amplifier and runs at + - 112 volts.
it took months to do but worth it.
and costs much less than a toridal transformer to make at this power.
I found the psu quite easy to do,and a bit of fun.
so if I can do it anyone can.
I use it in Half bridge, ixus IGBT's driven by a ir2110 at 120k.
sg3525 and E65 ferrite on 3c90 /f44. (foil wound)
i can get 8KW out of it.
this powers my new 2kw per channel amplifier and runs at + - 112 volts.
it took months to do but worth it.
and costs much less than a toridal transformer to make at this power.
I found the psu quite easy to do,and a bit of fun.
so if I can do it anyone can.
Neutrik powercon.
when my amplifier drives one channel into 2 ohms (clipping) into my dummy load,my mains amp meter shows 16amps draw.
both channels driven over ranges the ammeter.
what i wanted to do is Put a big smps in an amplifier.
all the smps Pa amplifiers i have tested have no bottom end.
the plan is to make an amp that does sound like it has a big toridal in it.
when my amplifier drives one channel into 2 ohms (clipping) into my dummy load,my mains amp meter shows 16amps draw.
both channels driven over ranges the ammeter.
what i wanted to do is Put a big smps in an amplifier.
all the smps Pa amplifiers i have tested have no bottom end.
the plan is to make an amp that does sound like it has a big toridal in it.
aandy said:
Hi Andy,
I am a bit afraid when I see SMPS in an amp, becouse I rapeired lot of QSC PL series, and when the AMP's transistor(s) failed, the SMPS burned as well becouse of the short on the rails...
Do you think that this is a problem you was able to solve?
Do you have a PFC in front of the SMPS?
thanks,
Tamas
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.