1KW SMPS help needed

gearheadgene

Member
2005-07-05 10:53 pm
CT
greeting to everyone:cool:

I want to build a 1KW SMPS power supply. The plan is to use a PFC preregulator followed by some form of isolation type regulator. I could use some help.

Fortunately, I've found 2 PFC reference designs from On semi that are each 1KW. The first uses the NCP1650 average mode pfc controller (and there is an available eval board), while the second used the MC33262 in critical conduction mode (no eval board available). It looks like either of these would work for me, but am interested in your opinions.

OK, so what about converting the PFC output down? I'm considering a few options:
1) A single supply for the entire 1KW
2) Multiple supplies in parallel
3) Multiple supplies, where each powers a different amplifier or group of amplifiers

Maybe some design details are needed here - the design is for a digital amplifier, probably 5.1 style. I believe the amps will use around 30-35 volts each. There is also some low voltage needs (microprocessor, for example).

Getting back to the supply configurations, I like option 1. My reseach indicates that a half bridge push pull setup would be a good choice. I'm having a tough time findind a good reference for this, however. Also, I am a little concerned about the danger at that power level.

Option 2 concerns me because I am not certain how to put power supplies in parallel without getting into load share problems. I see from other posts that there may be a synchronization problem as well.

Option 3 is neat, in that it is modular and expandable.

Any help or comments?

thanks

gene
 

mzzj

Member
2005-07-02 8:17 pm
65N 25E
gearheadgene said:
greeting to everyone:cool:

I want to build a 1KW SMPS power supply. The plan is to use a PFC preregulator followed by some form of isolation type regulator. I could use some help.

Fortunately, I've found 2 PFC reference designs from On semi that are each 1KW. The first uses the NCP1650 average mode pfc controller (and there is an available eval board), while the second used the MC33262 in critical conduction mode (no eval board available). It looks like either of these would work for me, but am interested in your opinions.

OK, so what about converting the PFC output down? I'm considering a few options:
1) A single supply for the entire 1KW
2) Multiple supplies in parallel
3) Multiple supplies, where each powers a different amplifier or group of amplifiers

Maybe some design details are needed here - the design is for a digital amplifier, probably 5.1 style. I believe the amps will use around 30-35 volts each. There is also some low voltage needs (microprocessor, for example).

Getting back to the supply configurations, I like option 1. My reseach indicates that a half bridge push pull setup would be a good choice. I'm having a tough time findind a good reference for this, however. Also, I am a little concerned about the danger at that power level.

Option 2 concerns me because I am not certain how to put power supplies in parallel without getting into load share problems. I see from other posts that there may be a synchronization problem as well.

Option 3 is neat, in that it is modular and expandable.

Any help or comments?

thanks

gene
how about 2-transistor forward converter? less prone to explode if you have some problem with gate drive. downside is that max dytu cycle is 50% and average currents in coils, trafo, caps and switches are 2 times higher than in half-bridge or full-bridge. 1kW is well within reach with this topology.
 

gearheadgene

Member
2005-07-05 10:53 pm
CT
Without being a smps maven (yet, anyway), I am having trouble telling which topologies are best. On the 2 transistor forward design, I see conflicting info on power. One place I saw claimed it is good up to around 1KW, yet others said it is good up to a few hundred watts.

Can you help out on this front?

gene
 
Have you done some calculations about the magnitude of the currents involved in a 1KW flyback and the energy storage requirements for the coupled inductor? The numbers will quickly discourage you.

Use a half bridge or full bridge topology with a coupled buck inductor instead, almost all multi-output computer PSUs above 150W are done that way.
 

gearheadgene

Member
2005-07-05 10:53 pm
CT
macc : I contacted a transformer manufacturer about building a flyback 1KW transformer. They pretty much confirmed Eva's reply. It would be big and expensive!

Eva: what do you think about the 2-transistor forward design proposed by mzzj? I think that the half-bridge approach is going to win, though.

Has anyone seen anything about Infineon's coolmos? Here's a quote from powerZone:
"Infineon is demonstrating a 1000 W (watt) reference design for a server power supply, with one 99 mΩ CoolMOS CS power transistor. . ."

Looks like this is brand new. So far, I haven't unearthed the reference design. I wonder what topology they are using. Can't wait to see it:smash:
 
gearheadgene said:


Has anyone seen anything about Infineon's coolmos? Here's a quote from powerZone:
"Infineon is demonstrating a 1000 W (watt) reference design for a server power supply, with one 99 m© CoolMOS CS power transistor. . ."

Looks like this is brand new. :


No... it's not new. So called hi spec devices may sound fine until the gate impedance is worked out..i.e device requires higher drive v.s freq.
I've used coolmos in many apps with no real performance advantages other than lower Rds on resistance. The higher drain volts is a plus compared to standard 840 types.
The gate threshold voltage can be lower than standard types.

It all depends on the application.

richj
 

mzzj

Member
2005-07-02 8:17 pm
65N 25E
richwalters said:



No... it's not new. So called hi spec devices may sound fine until the gate impedance is worked out..i.e device requires higher drive v.s freq.
I've used coolmos in many apps with no real performance advantages other than lower Rds on resistance. The higher drain volts is a plus compared to standard 840 types.
The gate threshold voltage can be lower than standard types.

It all depends on the application.

richj


CoolMos is pretty damn fast, gate charge for same Rds-on is lot less because same Rdson is archieved with 5? times smaller chip.
Also output capacitance is substantially smaller.

This can be advantage if you want to run it real fast, or it can be problem if you have specsed your gate drive for similar conventional type mosfet and with coolmos it slews and rings like crazy. (resize you gate resistor in this case)


Btw, Fairchild has some damn pretty IGBT's availlable for SMPS use, especially their new SMPS-II series looks impressive:
FGH60N6S2
Features

100kHz Operation at 390V, 52A
200kHZ Operation at 390V, 31A
600V Switching SOA Capability
Typical Fall Time. . . . . . 77ns at TJ = 125°C
Low Gate Charge. . . . . . 140nC at VGE = 15V

12kW SMPS anyone? :D
 
mzzj said:



12kW SMPS anyone? :D


Yes.....Viva IGBT's....= tram drive.

p.s .....Mosfet chip area is pretty inefficient compared to IGBT's for same current rating. I'm using IGBT's as standard in my switch mode /pfc designs. In well disgned layouts one can scrap the series gate resistor with IGBTS. However I can recommend a mosfet/IGBT driver interface.
The secret is all in the layout.

richj
 

gearheadgene

Member
2005-07-05 10:53 pm
CT
Hi Rich,

I haven't worked out much at all, actually. This is all pretty much new to me, so I am learning as I go.

If you go back in the thread, you'll see that the plan is to use PFC preregulator. I've found at least 2 reference designs from On-semi, plus have seen some from other manufacturers too. This should hopefully be an ok starting point.

Other than that, I'm still poking around trying to get as much info into my tiny brain that I can:clown: Couldn't sleep last night, so was up 'till 4 am reading through the various posts. Haven't finished reading yet, though.
 
Could you tell us about each of the required outputs? ie: output voltage, cross-regulation requirements and allowable voltage error margin, minimum load, typical load and maximum load expected for each output.

Could you also tell us about the input? ie: minimum and maximum input voltages.
 

gearheadgene

Member
2005-07-05 10:53 pm
CT
Eva said:
Could you tell us about each of the required outputs? ie: output voltage, cross-regulation requirements and allowable voltage error margin, minimum load, typical load and maximum load expected for each output.

Could you also tell us about the input? ie: minimum and maximum input voltages.

why soytenlee (my best 3 stooges impression). . . :bigeyes:

Universal input - standard stuff, 85 VAC to 240 VAC
feeds PFC
PFC output (around 400 VDC) feeds isolated DC-DC converter.

Output voltage, probably around 35 VDC
Output current MAX, around 30 A.

I'm thinking that only a single output will ok, but will consider other distributed topologies if needed.

Regarding the load, well that's going to be all over the map. I'd say minimum is pretty low - maybe 50 Watts when idle, and max will be around 650 or 700 W. As an estimate, the typical power consumption will be 200 W.

I've rationalized the 1KW requirment as having enough room for growth without having to redesign - although this may remain as overkill. Started out as needing 650W, then thought to add a little comort zone, then considered some future plans, and then next think I knew - bam, 1KW was just about right.
 
gearheadgene said:
Hi Rich,

I haven't worked out much at all, actually. This is all pretty much new to me, so I am learning as I go.



okay you got a few boundary conditions set up........ ...there are some tricky bits which we shall eventually arrive at and be prepared for compromises.

You mention the pfc rated worst case at 1kW (the inductor size would be based on that figure and run at a lowish frequency)... at 88 V AC the pfc would require nearly 13A from AC supply. This is bad news for EMI input chokes and the input bridge rectifier.

Food for thought....If you assume DC/DC conv 30AX35V =1050 Watts, the overall efficiency or both PfC+ DC-DC 1/2 bridge forward converter would be around 0.85x0.8=0.68.
The loss of efficiency with the pfc at 88V at 13A is to be expected ....You will have to give us some idea of expected physical size of both units
You mentioned <quote> "that all is pretty new to me". Do you have a good oscilloscope and other test gear ?

richj
 

gearheadgene

Member
2005-07-05 10:53 pm
CT
Eva said:
Your requirements seem not too complex, but at first you mentioned several outputs, what are you going to power?

Maybe I'm giving mixed signal here. Let me be a bit more specific - I want to power a 5.1 style amplifier. Each stage is going to be 100 watts into 8 ohm, so that accounts for 600 watts. I plan to run the amps at 35'ish volts, so each channel is going to require
2.9 Amps. I still have to figure out what to do about 4 ohm loads.

Additionally, there's some low voltage circuits that need power (5V, 3.3V, etc.) that may or may not be derived directly from the off-line portion of the supply. Rough estimate of 50 Watts for that stuff.

I am completely open minded on how to handle this supply. That's why I was suggesting other topologies. Heck, I even considered a transformer (torroid) to drop line voltage down to something lower, then run a individual AC-DC supplies on each channel. Then, I could use flyback stuff, simpler (perhaps) and maybe cheaper. Overall, I'm not sure the cost winds up less. For certain, the torroid isn't cheap - and then again, I'd have to handle the switching from 120 or 240.

richwalters said:



okay you got a few boundary conditions set up........ ...there are some tricky bits which we shall eventually arrive at and be prepared for compromises.

You mention the pfc rated worst case at 1kW (the inductor size would be based on that figure and run at a lowish frequency)... at 88 V AC the pfc would require nearly 13A from AC supply. This is bad news for EMI input chokes and the input bridge rectifier.

Food for thought....If you assume DC/DC conv 30AX35V =1050 Watts, the overall efficiency or both PfC+ DC-DC 1/2 bridge forward converter would be around 0.85x0.8=0.68.
The loss of efficiency with the pfc at 88V at 13A is to be expected ....You will have to give us some idea of expected physical size of both units
You mentioned <quote> "that all is pretty new to me". Do you have a good oscilloscope and other test gear ?

richj
That's a bitter pill to swallow - even though you are right! It crossed my mind before, but I have conveniently forgotten about the power loss. With such a lousy loss of power between the PFC and dc-dc, why would anyone bother? Maybe the regulations over in Europe get in the way, or maybe the pfc stuff makes sense for all the other reasons - but the power loss is really ugly.

So where does that leave me now? I could run without PFC, that's another option. If I don't though, and am losing 32% in conversion inefficiency, then the power requirements go up - ugh.

Equipment? Right now, the only stuff I have access to is at work. Moving forward, I'll need to get some here at home.


There is something that I can't put my finger on, and has a big impact on my requirments. In my example of 5.1 system, each channel rated at 100 watts - what are the odds that the system would have to supply 600 watts continuously? Given the nature of audio, it's more like random noise. But what about music? Anyway, I'm rambling.

gene
 
Audio is tricky because peak to average ratio of power consumption may be as high as 10. Also, tight regulation is not strictly required, so you may use the 400V DC internal bus to drive an unregulated DC-DC converter and still get some (average) regulation while making things much simpler.

If you are going to power 6 100Wrms channels then the worst-case peak power consumption will be 1200 W, the worst-case average power consumption will be less than 900 W. The real-world worst-case peak power consumption with music signals at maximum volume may still be near 1000 W, but the average power consumption will hardly reach 200 W, even with everything starting to clip (trust my experience).

With so high peak to average power ratios, having a power supply with poor efficiency at 1000W is not a problem as long as it can repetitively handle 1000W for small amounts of time in a reliable way. Furthermore, you can use a lot of capacitance on the 400V bus in order to isolate the PFC converter from high peak currents, since 400V capacitors store more energy in less space and at less cost in comparison with 35V capacitors.

If you live in an area with 230V mains, you could also discard the universal mains requirement and just cover 176V to 240V in order to make PFC thingy easier.

The low-voltage housekeeping outputs may be derived from the main DC-DC transformer or even from the PFC inductor in order to get power as soon as posible. Then, it's easier to just shut down the main DC-DC converter in case of fault, over-current, over-temperature, etc....