Electronics > Beginners
MC34063 high voltage dc-dc boost converter
MagicSmoker:
--- Quote from: dazz on January 05, 2020, 01:44:45 pm ---Yes, I need to take a break from the sim and read more instead. The reason I used such a low voltage zener is simply because it seemed to work. The higher voltage ones weren't clamping at all, perhaps because of the high frequency of the ringing, not sure. It's also obviously true that the zener I picked was dissipating far too much power, some 15W on average. I probably should also swap the mosfet for one with lower Cds, I think that will reduce the ringing a bit, correct? Is the output capacitance in the datasheet the same as Cds?
ETA: One question, please. Is the snubber supposed to improve overall efficiency? Or it simply lowers the dissipation in the mosfet by the same amount the snubber dissipates?
--- End quote ---
1. R2 is artificially boosting the clamping voltage at the instant of turn-off; try setting it to 1m to effectively take it out of the circuit without actually removing it.
2. Cds is a bit of a moving target because it varies inversely - and non-linearly - with the applied drain-source voltage (when off, of course; this capacitor is obviously shorted when the MOSFET is on). A MOSFET with lower Cds with all other parameters (except cost) being the same will store less energy which eventually needs to be snubbed so, yes, that is a good direction to go. It's a bit of a 2nd order problem at the moment, though; in other words, concentrate on getting the converter to work properly first.
Note, also, that Cds (plus other stray capacitances) forms two different resonant networks in the flyback: with the leakage inductance of the primary when the secondary diode is conducting, and with the magnetizing inductance if the diode turns off some time before the switch turns on again (ie - operating in Discontinuous Mode). Basically, you can damp the former but not the latter.
3. Standard dissipative snubbers just move losses from the switch to themselves, they don't reduce losses (they actually increase the total loss somewhat because of pesky thermodynamics). There are a class of so-called "lossless" snubbers that utilize resonant networks to recycle energy from switching and/or ringing back to the supply (or output), but don't even think about messing around with them for now.
dazz:
--- Quote from: MagicSmoker on January 05, 2020, 03:12:28 pm ---1. R2 is artificially boosting the clamping voltage at the instant of turn-off; try setting it to 1m to effectively take it out of the circuit without actually removing it.
2. Cds is a bit of a moving target because it varies inversely - and non-linearly - with the applied drain-source voltage (when off, of course; this capacitor is obviously shorted when the MOSFET is on). A MOSFET with lower Cds with all other parameters (except cost) being the same will store less energy which eventually needs to be snubbed so, yes, that is a good direction to go. It's a bit of a 2nd order problem at the moment, though; in other words, concentrate on getting the converter to work properly first.
Note, also, that Cds (plus other stray capacitances) forms two different resonant networks in the flyback: with the leakage inductance of the primary when the secondary diode is conducting, and with the magnetizing inductance if the diode turns off some time before the switch turns on again (ie - operating in Discontinuous Mode). Basically, you can damp the former but not the latter.
3. Standard dissipative snubbers just move losses from the switch to themselves, they don't reduce losses (they actually increase the total loss somewhat because of pesky thermodynamics). There are a class of so-called "lossless" snubbers that utilize resonant networks to recycle energy from switching and/or ringing back to the supply (or output), but don't even think about messing around with them for now.
--- End quote ---
OK, I have calculated my leakage inductance at 60nH (which seems low, right? I guess I cheated a bit by lowering the inductances of the transformer too much) and the mosfet output capacitance is 380pF according to the datasheet, I guess I can do a lot better than that. So with the clamping voltage at 50V I still get spikes in excess of 60V, a 80V mosfet would be the way to go, correct?
MagicSmoker:
--- Quote from: dazz on January 05, 2020, 03:32:34 pm ---...
OK, I have calculated my leakage inductance at 60nH (which seems low, right? I guess I cheated a bit by lowering the inductances of the transformer too much) and the mosfet output capacitance is 380pF according to the datasheet, I guess I can do a lot better than that. So with the clamping voltage at 50V I still get spikes in excess of 60V, a 80V mosfet would be the way to go, correct?
--- End quote ---
Gah... It took me less time to just edit your .asc file than type another response. My usual disclaimer applies: I did not optimize everything to leave something for you to do, but I got you into the ballpark.
That ratio extender trick isn't needed if you select a reasonable base switching frequency vs. inductance vs. peak current so that the flyback doesn't really need to go much above 50% duty and operates as close to continuous current mode as possible without actually entering it. Hence why the timing capacitor was reduced and the primary inductance increased.
An RC damper was added across the switch to deal with the high frequency ringing - it only costs about 0.3W of loss and stops your flyback from doing double duty as an AM (or FM!) radio transmitter. As mentioned earlier, this damper can be placed across the primary, instead, I just felt it kept the schematic neater putting it across the switch.
The primary clamp is now an RCD type with the (51V) Zener acting as more of a backstop. Loss in the clamp once the converter reaches equilibrium should be around 0.5W or so with peak voltage well controlled for a 100V MOSFET.
Output capacitor increased to a more realistic value and output diode changed to a 1A/600V SiC Schottky, though as long as you keep the flyback in DCM you can use a conventional fast recovery diode (ie - with <100ns reverse recovery time [given as tt in the .model statement, btw]).
Finally, the simulation parameters were tweaked to give more accurate results and the total time reduced to just a little more than is needed for the output voltage to reach equilibrium (at which point the '34063 goes into burst mode). All in all, waveforms look pretty realistic now and you have a reasonable jumping off point to keep tinkering.
dazz:
--- Quote from: MagicSmoker on January 05, 2020, 09:12:04 pm ---
--- Quote from: dazz on January 05, 2020, 03:32:34 pm ---...
OK, I have calculated my leakage inductance at 60nH (which seems low, right? I guess I cheated a bit by lowering the inductances of the transformer too much) and the mosfet output capacitance is 380pF according to the datasheet, I guess I can do a lot better than that. So with the clamping voltage at 50V I still get spikes in excess of 60V, a 80V mosfet would be the way to go, correct?
--- End quote ---
Gah... It took me less time to just edit your .asc file than type another response. My usual disclaimer applies: I did not optimize everything to leave something for you to do, but I got you into the ballpark.
That ratio extender trick isn't needed if you select a reasonable base switching frequency vs. inductance vs. peak current so that the flyback doesn't really need to go much above 50% duty and operates as close to continuous current mode as possible without actually entering it. Hence why the timing capacitor was reduced and the primary inductance increased.
An RC damper was added across the switch to deal with the high frequency ringing - it only costs about 0.3W of loss and stops your flyback from doing double duty as an AM (or FM!) radio transmitter. As mentioned earlier, this damper can be placed across the primary, instead, I just felt it kept the schematic neater putting it across the switch.
The primary clamp is now an RCD type with the (51V) Zener acting as more of a backstop. Loss in the clamp once the converter reaches equilibrium should be around 0.5W or so with peak voltage well controlled for a 100V MOSFET.
Output capacitor increased to a more realistic value and output diode changed to a 1A/600V SiC Schottky, though as long as you keep the flyback in DCM you can use a conventional fast recovery diode (ie - with <100ns reverse recovery time [given as tt in the .model statement, btw]).
Finally, the simulation parameters were tweaked to give more accurate results and the total time reduced to just a little more than is needed for the output voltage to reach equilibrium (at which point the '34063 goes into burst mode). All in all, waveforms look pretty realistic now and you have a reasonable jumping off point to keep tinkering.
--- End quote ---
Thanks, much appreciated. I will keep on searching for info to learn how all this works and do those calculations myself, if I can.
ETA: WTF? 98% efficiency? I know this is just a simulation, but wow!
dazz:
I've put aside the flyback research for the moment while I build and test the original boost converter.
I have most of the components already, but I've come to the realization that I need to learn more about inductors and how they're built before I do this. The thing is I simply searched for 470uH 3A toroidal inductors in ebay, and got these crappy ones that I'm pretty sure won't even handle 1A. What's worse, the ones I got had 0.3mm wires, not the 0.7mm advertised. So yeah, all but useless.
I've spent an afternoon googling about toroidal inductors and it's a bit daunting TBH. Apparently it's not as simple as picking a core size that can fit the necessary number of turns of the necessary wire gauge to obtain the desired inductance and current rating/resistance. Looks like there's a host of different core materials that affect the calculations.
I'm going by this right now: https://cromwell-intl.com/radio/copper-wire/
There's a table specifying the different inductance indexes for every core type, and the formulas to derive the necessary number of turns for a given target inductance. Seems straightforward enough.
I also did a search in Digikey to get an idea of the resistance and core size of inductors in the ballpark of what I need. Most seem to be 1.280" Dia x 0.650" W (32.51mm x 16.51mm). Unfortunately the datasheets I checked don't say anything about core material or wire gauge.
Is there a standardized method to derive the kind of inductor you need? Would you star by picking the core material, then the size, wire gauge, etc? The problem is, as far as I can tell, all those things are interconnected: a thicker gauge requires a larger core. Materials with larger Al index can be smaller, but then the wire needs to be thinner too, but a thinner wire will produce more resistance with the same number of turns... What a mess! |O
I'm sure there's an easy way to approach this. Any pointers appreciated
Navigation
[0] Message Index
[#] Next page
[*] Previous page
Go to full version