Recommended BJT/MOSFET:s

Loial

Member
2004-11-21 10:06 pm
Gävle
Hi
I'm in the beginning of constructing a very lightweight, slim amplifier. The amp itself is a Tripath-circuit, already up & running.
But now I'm starting the construction of the PSU, an SMPS PSU capable of approx. 2*[email protected]
I've built a couple 12V DC/DC converters, with no major problems at all. Now this amp is running on the mains ( 230V )
And my question is - what topology do you think is suitable? and what transistor/s would you recommend for primary switching?
 

mzzj

Member
2005-07-02 8:17 pm
65N 25E
Loial said:
I forgot - Is a PFC necessary, or is it considered overkill for 200W?

Definetely waste of time in DIY. Leave PFC to be your power companys problem. Nowadays PFC is required in europe for power supplies over 70W but for DIY projects building a PFC for anything else than fun doesnt make sense. You just end up with lower effiency when running with PFC.
 
I have designed and built a 2Kw PFC for fun :)

Actually PFC provides some advantages since it outputs a high voltage and it's almost regulated, thus a simple unregulated full bridge with low-cost TO-220 IGBTs rated at 10A and a single transformer (with no additional inductors) is enough to step-down the output of the PFC and provide as much isolated and coupled outputs as you want. Note that a single ampere of primary-side current at 400V already produces 400W output.

When a PFC is not used, you either have to mess with regulation or accept huge fluctuations in output voltages. I agree that PFC increases cost, size and complexity but I like it anyway.


Concerning switching devices, as most people will probably recommend MOSFET, I'm going to recommend SKP10N60 IGBTs from Infineon and MJE13009 bipolars from On-Semi. In the past I had bad experiences with high-voltage MOSFETs so I decided to no longer use them.

Concerning topologies, the obvious recommendation for your application is half-bridge, and full-bridge when you want to double power output without using more expensive switches.
 

Loial

Member
2004-11-21 10:06 pm
Gävle
How come the use of a PFC makes the regulation better?
If the output of the PFC is almost regulated, doesn't the SMPS need to have a regulation of some kind? thinking of the difference in current consumption in idle vs. full load.
What makes the PFC 400V different from the 325V rectified mains voltage?

I'm not really into what the PFC does, more than making the PSU produce less EMI?
A PFC doesn't seem too complicated though, maybe ill try building one.
 

mzzj

Member
2005-07-02 8:17 pm
65N 25E
Loial said:
How come the use of a PFC makes the regulation better?
If the output of the PFC is almost regulated, doesn't the SMPS need to have a regulation of some kind? thinking of the difference in current consumption in idle vs. full load.
What makes the PFC 400V different from the 325V rectified mains voltage?

I'm not really into what the PFC does, more than making the PSU produce less EMI?

PFC 400V is already regulated vs. 325v rectified mains is something like +-10% to +-30% depending wether you live in scandinavia or nepal :D
There is still some ripple voltage left on PFC output caps, maybe 5-10% typically, but this ripple voltage can be reached as well in direct mains rectification. Universal-input pre-regulator is another benefit of PFC, but i doubt that you need to desing your power supply as 80-250Vac input for DIY.

EMI can easily be even worse with PFC than without. All PFC does it that it makes your power supply look like resistive load instead of drawing huge peaks of current for 5ms and no current at all for rest of half-cycle. your electricity company will be more happy for easier load and bigger consuption :D

Why PFC is nonsense for DIY is that there is maybe one handfull of SMPS hobbyist in Sweden and still 5 million older-made computer and other smps power supplies in service without any kind of PFC. Lots of load with bad power factor causes unneccessary heating of transmission lines, requires bigger fuses and so on.
 
With 220AC and 940uF of storage I remember measuring 320DC with no load and 280V with 30Vp-p ripple witrh 1Kw load. Neglecting SMPS loses you can scale down these numbers, so for example you would have 60V rails with no load and 52.5V rails witn 5.6Vp-p ripple under 1Kw load.

Having 940uF with 320V is like having more than 13.000uF per rail with +-60V (use E=.5*C*V^2 to compare energy storage).

(Obvious disclaimer: High voltages are dangerous, particularly when they are not isolated from earth, so use them at your own risk).
 

Loial

Member
2004-11-21 10:06 pm
Gävle
That sounds promosing, 10-15% voltage difference shouldn't be a problem. The amp-circuit itself can handle +/- 60Vdc.

with half-bridge topology, what is the voltage stress on the primary switchers? I've got an excellent page on my favorites at home - cant remember the name now, though...

One last question ( hehe, i doubt THAT... )
- Would a UCC3810 circuit do the job as a timer/oscillator?
i know it is a little bit of an overkill that to - but they grow on trees here were i work :-D
 

Loial

Member
2004-11-21 10:06 pm
Gävle
Ok, now I've been sketching, calculating and burnt a few braincells at the same time.
This is actually the first SMPS I'm doing all the way, the other 12V PSU:s have been made from ready-made schematics with a few modifications only, and performance haven't been an issue.

SO, I've discovered that the questions are many. Any SMPS pro that can guide me through the dimensioning and construction of this?

Here's what I've planned so far:
Output capable of driving 2*[email protected] ( lowered the specs a little bit... )
Input voltage 230Vac
The topology I've chosen is half-bridge
Maybe PFC ( I'll look at that later )

So, here are the questions:

1. In half-bridge, there are two caps in series with the
transformer, how do i dimension these?

2. As i asked before - is there a possibility to skip these caps by
using the CC3810 current control?

3. As primary switch, my only choice is MOSFETs. What specs
should i consider? input voltage will be 280-400V ( 420V maybe
with the PFC later on ) Are 500V mosfets sufficient?

4. How big current should these be capable of handling?
2*100 watts, maybe 75% efficency (???) in the PSU, 85% in
the amp ( class T ). Should be around 310 watts? then how big
current peaks should the primary sw. be capable to handle?

5. I'm thinking about the switching frequency - is 60-70 kHz a
good choice? i want to make the transformer as small as
possible, but on the other hand higher frequency means a lot
more switching losses... is there any "magic" frequency where
both transformer size and switching losses are kept as low as
possible, and still affordable for DIY?

6. Running at 60kHz, my transformer core choice is ETD44(F44)
100N1 / 2*25N2 secondary. Is this core large enough?

That's it for now... :)
 

mzzj

Member
2005-07-02 8:17 pm
65N 25E
Loial said:
Ok, now I've been sketching, calculating and burnt a few braincells at the same time.
This is actually the first SMPS I'm doing all the way, the other 12V PSU:s have been made from ready-made schematics with a few modifications only, and performance haven't been an issue.

SO, I've discovered that the questions are many. Any SMPS pro that can guide me through the dimensioning and construction of this?

Here's what I've planned so far:
Output capable of driving 2*[email protected] ( lowered the specs a little bit... )
Input voltage 230Vac
The topology I've chosen is half-bridge
Maybe PFC ( I'll look at that later )

So, here are the questions:

1. In half-bridge, there are two caps in series with the
transformer, how do i dimension these?

2. As i asked before - is there a possibility to skip these caps by
using the CC3810 current control?

3. As primary switch, my only choice is MOSFETs. What specs
should i consider? input voltage will be 280-400V ( 420V maybe
with the PFC later on ) Are 500V mosfets sufficient?

4. How big current should these be capable of handling?
2*100 watts, maybe 75% efficency (???) in the PSU, 85% in
the amp ( class T ). Should be around 310 watts? then how big
current peaks should the primary sw. be capable to handle?

5. I'm thinking about the switching frequency - is 60-70 kHz a
good choice? i want to make the transformer as small as
possible, but on the other hand higher frequency means a lot
more switching losses... is there any "magic" frequency where
both transformer size and switching losses are kept as low as
possible, and still affordable for DIY?

6. Running at 60kHz, my transformer core choice is ETD44(F44)
100N1 / 2*25N2 secondary. Is this core large enough?

That's it for now... :)
I am not any sort of pro but i stick my spoon to here anyways..

1. By allowed input ripple voltage and rms current rating.
2. I dont see any possibility in half-bridge to skip these caps.
3. Low gate charge(as usually in switchers), enough low rds-on
Actually its compromise in mosfet size, switching losses increase when you choose bigger mosfet and on-state losses degrease.
for 420V operation i might choose 560 or 600V mosfets to be on safe side.
4. Dont choose them according what specsheets say about current, they are "speculative ratings" Instead of that calculate rds-on losses and switching losses and check if result is something what you can cool and tolerate. And remember that your rds-on is function of temperature. Multiply by 1.7 to get rds-on at 100C. This is bit of iterative process, repeat until satisfactory. rds-on losses are maximum at maximum load and high line when smps is running at low duty cycle and peak currents are high at swicthes. Switching losses are tricky to predict, but at 60khz they should be around same as your rds-on losses quite easily. IR has some nice appnotes about mosfets and TI power seminars are worth of solid gold.

5. 60-70 khz sounds fine, easier to source ferrites that are low-loss in this frequency than some of more modern ferrites meant for high-freq operation (ferroxcube has a lot of them to choose)
With increasing frequency transformer desing gets also more tricky and with higher power levels even more so. Magic frequency depends on power output, it might be easy to desing 200W smps to run at 400khz but extremely difficult at 2kW.
"Old style" to dimension power supply is to divide mosfets losses equally between switching losses and rds-on losses and divide transformer losses 50-50% between copper and core losses.
Transformer loss should be small partition of total losses, as it is usually desinged for 20-40C temperature rise above ambient.

6. Calculate flux swing in your core and look at core manufacturer datashit how big core losses you end up. Less than 100mW/cm3 is a good starting point. After that calculate winding losses, again for minimum duty cycle and maximum load. If you can do with something like 3Amps/mm2 you should be on safe side. Dont oversize your wires, you just end up with more proximity and eddy losses. Keep wire thickess below 2x skin depth, split primary to 2 layers and make your secondaries between primary layers. INSULATE properly between primary and secondary, 4kV insulation is required also in sweden. 2 layers of OHP film will do this.
 

Loial

Member
2004-11-21 10:06 pm
Gävle
Ok, now I'm on my way...
The ferrite core will be an ETD34, with neosids material F44.
The only thing that worries me, is as before, the primary switchers. I bought a bunch of BUK445's, really cheap ( 25 pcs for 12$ ), but I'm not quite sure they will do the job... I'm gonna use them later anyway, for other projects... what do you think?
perhaps two in parallell? or a fullbridge topology instead ( sharing current load )?
((BUK445-500
Rdson = 1.5 Ohms
I = 2,9 A
U = 500 V
Ptot = 30W
Gate charge = 30 nC ))

EDIT
I've got my hands on some IRFB9N60A:s
They are a lot more powerful, maybe It's a waste of resources :)
And I've only got 6 of them...
Rdson = 0,75 Ohm
I = 9,2 A
U = 600V
Ptot = 170W
Gate Charge = 49nC
 

Pierre

Banned
2004-11-26 6:32 am
Paris
You can also try these:

IRFB11n50 (11A, 500V, 0.52ohm Rds(on) ).
2sk2141 (6A, 600V, 1.1 ohm Rds(on) )

Please follow the thread I started "Assymmetric outputs in half-bridge", as I am basically in the same boat as you.

I am using an ETD44, pretty similar (a bit bigger) to your core. I am trying to get more than 500W out of it, but by the moment I am having problems with mosfets exploding, perhaps due to gate drive issues.

What gate-drive scheme are you thinking of?
 

Loial

Member
2004-11-21 10:06 pm
Gävle
I've looked in to your thread to, interesting.

I'm planning to drive the mosfets from two small gatedrive transformers I've found here at work, really neat little devices capable of 50W:s :)
I'm thinking using the TL494 ( got some... )
or maybe the UCC3810/UC2801

The thing with these two transistors i mentioned, is that I've got them, purchased them for a low price, figured that if i cant use them here, they'll fit other experiments...