5

I'm trying to finally apply what I learned at university in the course of power electronics. However I find that both that course and most of the books / online references are lacking on some points.

In fact (this may be similar to controlling H-bridges for motor drivers, not sure) there are some possibilities concerning how to realize a step-down (buck) converter:

  • Using PMOS as input switch (source at supply voltage) -> gate of the PMOS needs to be HIGH when PMOS is off, while gate of the PMOS needs to be LOW when PMOS is on
  • Using NMOS as input switch (drain at supply voltage) -> gate of the NMOS must be commanded via boostrapping (needs an additionnal Diode and Capacitor)

Additionnally it may be a good idea to use a synchronous buck converter (less losses) using an NMOS in parallel to the output diode. I think I got this part and - anyway - it's simpler to command since it's an NMOS with its source tied to the ground.

Back to the original question: while I agree that it may be (theoritically) possible to easily control the PMOS transistor, I think it's pretty difficult, expecially with high input voltages.

Consider that I take power from the wall outlet: 230V_RMS at 10A max (but for my applications I will go for much less, 1A max). I'm gonna get a pseudo-DC voltage by using a bridge rectifier (Gretz's bridge) with a capacitor at its output (standard practice). This last voltage will be the input of my DC/DC buck converter.

Hence the problem: using a microcontroller to generate a PWN signal to control the output voltage (GPIO: 3.3V output, or 5V at best) it's not gonna be possible to activate the NMOS or deactivate the PMOS.

I think I need the NMOS's gate voltage needs to be around 5-10V above the supply voltage. I'll have to do the bootstrapping for that, yet I didn't really understand it. That's what basically GATE drivers are made for AFAIK.

As for the PMOS a simpler solution may be to use an inverted PWM signal (D = PWM at level LOW, normally it's the reverse) and control an optocoupler which has its collector connected to the supply voltage (same as the PMOS source's voltage). Collectors able to sustain that voltage exist, yet there may be a better solution.

There aren't many high voltage MOSFET drivers available on the market (let alone at low cost) and I would really like to know how to do this. I think step-down/buck converters are quite common nowadays, so I find it hard that no such products exist. This leads me to believe that I'm not looking at the right components (yet). Or the only solution would be to realize the driver in discrete components? Any product reccomandation / reference to satisfy these requirements?

EDIT: as I said to Oliven Lathrop here is what I have in mind to control the PMOS. Basically I use a BJT as a current source and then shunt just enough of the voltage (12-15V) to get the PMOS in conduction mode. Otherwise ideally no current flows in the BJT and the PMOS is blocked. PMOS CONTROL http://img513.imageshack.us/img513/1879/pmoscommand.png.

I have not verified the polarity of the PWM signal (should it be reversed or not) but in principle this may "just" work. NPN transistors supporting > 400V_DC are much more common than PNP/PMOS and their price is small. A small current in the BJT is enough. Therefore R2 has to be quite large (in order to get I_BJT_Collector ~ 1mA) and R1 just large enough (but not too much, otherwise the charge takes too long and I dissipate too much energy). May pose a problem for the discharge though, since the accumulated charges can't be evacuated?

EDIT2: I know on the schematic I represented an NMOS transistor, but there was no PMOS symbol in the schematic program I'm currently using. It's actually a PMOS!

EDIT3: On second though I'm not sure this would work since the current is imposed in the NPN, not through R1. It may just work if the current going into the MOS (I_G > 0) adds up with the collector's current of the NPN (I_C > 0). This way the voltage drop actually increases and conduction is assured. Still doubts on the opposide process though.

user51166
  • 860
  • 3
  • 10
  • 21
  • 1
    Are you absolutely sure that you want a buck converter? If not, a flyback converter may be a better fit for your application (Isolated + low-side switching). – W5VO Aug 04 '12 at 21:27
  • @W5VO: unfortunately 50Hz (even 250W) transformers are way too much expensive (>50$/piece at 50+ pieces). I think if the transformer costs > than twice the cost of the rest of the setup something's off. That's a shame because otherwise it would've been a good idea. – user51166 Aug 04 '12 at 21:35
  • @user: The flyback method that W5VO mentions drives the primary of a transformer at high frequency. Such transformers are smaller and cheaper for the same power level than the "big iron" ones that run at the original power line frequency. – Olin Lathrop Aug 04 '12 at 21:42
  • @OlinLathrop: so the switch is used to generate a presudo-AC high frequency voltage that is then passed through the transformer (if I got this right). Unfortunately I cannot seem to find those on catalogues of the major distributors (e.g. Mouser) under "Power transformers". Maybe they're classified as "Audio transformer" (maybe here: http://ch.mouser.com/Power/Transformers/Audio-Signal-Transformers/_/N-8uav6/). 3$ for 75W is not that bad of a deal. – user51166 Aug 04 '12 at 21:49
  • @user: No transformers intended for flyback power applications would definitely not be called "audio". That's a whole different set of tradeoffs. – Olin Lathrop Aug 04 '12 at 21:56
  • @OlinLathrop: any reference to a POWER HF transformer (or an example out of the catalogs)? – user51166 Aug 04 '12 at 21:59
  • Your proposed circuit works in theory if you carefully calculate the current and make sure the gate is driven to the proper level, not too little or too far. However, you won't be able to drive with significant current else there will be a lot of dissipation. The current will need to be low, so the resistors high and the time constant with the gate capacitance long. You could put a double emitter follower between R1 and the gate to speed things up. You still have to consider dissipation at that voltage. 10mA x 360V = 3.6W, ugh. – Olin Lathrop Aug 05 '12 at 00:23
  • let us continue this discussion in chat – user51166 Aug 05 '12 at 07:27
  • @OlinLathrop: 3.6W doesn't seem very much to me, expecially for high wattages power supply. So there are no alternatives at all? I wonder then how come all AC adapters (universals) for 12V to 0.8A-3A are so dirty cheap. It's true however that the higher wattages, the much more expensive they get. That's not really the case for PC power supplies: 500W at around 30$. They must use some sort of trick too. – user51166 Aug 05 '12 at 07:32
  • @OlinLathrop: furthemore the problem remains with the flyback circuit: how to drive the NMOS higher than the power line voltage or the PMOS at just that level. A possible solution may be to place the NMOS AFTER the primary of the transformer and then connect it to the ground of the power line. – user51166 Aug 05 '12 at 08:25
  • Otherwise the "simple" (and least power efficient way) solution for high-power (and big devices) may be to purchase some ATX PSUs (30-50$ for 500W or so), use the 12V rail(s) for buck/boost purposes and the 3.3V/5V rail(s) for buck purposes (if necessary). Something along the lines of what described here: http://www.wikihow.com/Convert-a-Computer-ATX-Power-Supply-to-a-Lab-Power-Supply. I find it strange however that you cannot buy a transformer of that power for cheap. The whole PSU costs way less than the transformer by itself. – user51166 Aug 05 '12 at 13:04
  • Just a though: I know it's kind of "old tech" (slow, bigger conduction losses, etc) but could thyristors / triac work in this case? There would be no problem in commanding them (standard gate driver, doesn't need very high voltage). However this doesn't solde the isolation/transformer problem. On Mouser I only found flyback transformers of 50W max at around 15$ which is not too bad. Unfortunately they're only rated at 60V_DC max :( – user51166 Aug 05 '12 at 15:57

1 Answers1

6

High side switching is always tricky. There are no easy and simple ways, only various tradeoffs.

PMOS transistors are nice in that they can work within the existing voltage. The gate voltage needs to be pulled below the input voltage by 12-15 V to turn them fully on. The downside is that P channel MOSFETS usually have a little worse characteristics than the equivalent N channel.

N channel may have a better combination of Rdson, voltage tolerance, and cost, but require you to somehow make a voltage higher than the input to drive them. Some high side FET driver chips include a charge pump or other trick for this purpose. Another downside of a N channel high side switch is that the gate must swing a much larger amount, from zero to 12-15 volts above the input. This is because the gate voltage is relative to the source, which is now riding up and down with the voltage being switched. This requires high slew rates to stay out of the partially on region as much as possible, and provides more opportunity for noise pickup elsewhere.

There is no easy solution.

However in your particular case you may not need a high side switch at all. As W5VO mentioned in a comment, a flyback topology only requires a low side switch on the primary. The high side can stay connected to the input voltage.

A center tapped primary with the transformer run in forward mode is another possibility. The center tap goes to the input voltage with a low side switch pulling each end alternately to ground. Again there is no free lunch, which in this case is exhibited by the low side switches now having to withstand twice the input voltage. This is why the center tapped topology is more used for lower input voltages and usually not for worldwide "universal" power, which needs to handle up to 260 V AC or so. That would mean 368 V peaks, and 735 V stress on the low side switches. Transistors with that kind of voltage capability give up other parameters, like gain in bipolars and Rdson in FETs.

There is no free lunch.

Added:

I meant to say this earlier but somehow it slipped thru the cracks. You will most likely need a transformer anyway to get isolation. Unless you really really know what you're doing, you want the resulting supply to be isolated from the power line. The main exception is if the power stays completely inside a sealed box and there is not even a ground connection to the outside world. Otherwise, you run the risk of a user getting connected to the hot side of the AC line should even a few simple things go wrong. There is good reason commercial power supplies are mostly isolated.

Given that you probably want isolation, the problem becomes how to drive a transformer as apposed to how to make a buck switcher directly.

Olin Lathrop
  • 313,258
  • 36
  • 434
  • 925
  • Thank you for your complete answer. So I either need a POWER HF transformer (~100kHz or the likes, maybe even 50kHz will do - hopefully "cheap") or some control circuitry. I see that POWER PMOS are also pretty much twice the price of their respective NMOS. While PMOS may be driven through an opto-coupler (that exist for ~1$/piece), there are not many (if any) MOSFET drivers for that voltage (up to 380V peak). This would require me to do it in discrete components I presume. I'll update my original post with the PMOS solution I have in mind right now. No idea for a discrete NMOS driving though. – user51166 Aug 04 '12 at 22:04
  • Just to let you know: I asked one of my professors and he said the same thing as you did, thus reinforcing the need of a transformer (which he said would've been better if built at home, just because of its price). He said that a transformer - beside galvanic insulation - provides a good dynamic range for the PWM signal, which is good for start/stop of some types of loads. – user51166 Aug 27 '12 at 11:30
  • @user: It sounds like your professor knows what he's talking about. – Olin Lathrop Aug 27 '12 at 12:34
  • Yeah, surely he does. It's not like I don't trust your advices, just wanted to ask him because during the course he gave me it seemed as if transformers were no more. – user51166 Aug 27 '12 at 12:36
  • 1
    @user: Transformers are very much alive and will be for the forseeable future. The vast majority of wall powered power supplies contain a transformer. This is really the only way today to get significant power accross a isolation barrier. Perhaps your prof was referring to "big iron" transformers that handle the wall power directly and scale it down to roughly the desired voltage. Those are pretty much gone. The transformers inside modern power supplies are much smaller and cheaper since they run at 100s or kHz, not the wall power 50 or 60 Hz. – Olin Lathrop Aug 27 '12 at 13:06
  • Yeah, but the strange thing is that he said that high-frequency (~ 100kHz) transformers were bigger because the copper wire had to be larger (because at ~ 100kHz current almost flows only on the surface of the cunductor, whereas at 50/60Hz it flows roughtly across the whole section). He also gave another reason which I cannot remember right now. The "only" reason why ~ 100kHz transformers may be smaller is because the corrisponding energy is way higher (E ~ f) or at least so I read. At higher frequency transformers themselves (cores) are smaller, whereas wire size is larger. Am I correct? – user51166 Aug 27 '12 at 13:14
  • 1
    @user: The wire may need to be a little bigger due to skin effect, but the core can be a lot smaller. Overall, 100 kHz switching power supply transformers are a lot smaller, lighter, and cheaper than those that run at 50 or 60 Hz at the same power level. – Olin Lathrop Aug 27 '12 at 15:04
  • So basically it's worth the trade. My professor said he didn't actually like the flyback topology transformer because of the high voltage "spikes" during switching. However the other topology (forward) seems to be popular "only" for quite low-power applications (< 1kW) and I'm not sure if it doesn't create voltage spikes as well. He said I should use standard 50Hz transformer (DIY) which sounded a bit strange to me too. I think I can just buy a pair of transformer cores (e.g. EPCOS N87, Farnell: 1781876) but I didn't really understand for which frequency these cores were designed. – user51166 Aug 27 '12 at 15:21
  • Anyway in both cases the use of a transformer causes a big problem: suppose I want a 2kW power supply. I step down voltage from 230 V_AC to 48 V_AC which is a better dynamic range for 12 V_DC output (30 V_AC would be even better). This means that at the primary I'll have approximatively 9A and at the secondary approximatively 42A. At the primary side it may be already difficult enough (> 2.0mm [diameter] thick wire), but at the secondary side it's next to impossible (at least with the sizes sold by my local retailer). This leads me to believe I'd have to use either multiple transformers or – user51166 Aug 27 '12 at 15:27
  • multiple secondary windings (secondaries in parallel), which poses yet another problem (even load balancing across transformers). – user51166 Aug 27 '12 at 15:28
  • @user: 2 KW is not a trivial design. You will have to use every trick in the book to get good efficiency, then dealing with the waste heat will still be a significant issue. At that power level you should be thinking multiple phases anyway. Maybe 4 phases of 500 W each 90 deg apart. – Olin Lathrop Aug 27 '12 at 16:19
  • What transformer schema do you propose? Formward converter? Why multiple phases? I can understand that the load has to be balanced on the secondary side (hence my concerns with 40A :S). If I understand you correctly you're proposing of generating 4 PWM signals (hence 4 converters, 4 transformers), generate (e.g.) 12V at each output then converge all that power on a common DC bus, correct? Have to find a good way to create this common bus. Schotty diodesat each converter's output would create "lots" of losses. – user51166 Aug 27 '12 at 16:36