Hi Eva,
thanks for posting those waveforms. I got a little bit nostalgic looking at them, since my last bipolar design dates almost 20 years ago. As far as I can remember, your base current waveform looks very good, a proper textbook example.
Maybe you could consider another variant of the transformer coupled bipolar drive. I think it was proposed by Bruce Carsten, one of the old SMPS gurus. In his design, he uses grounded base and emitter is driven by secondary winding of the drive transformer. Primary of the drive transformer is driven by a current source. As far as I can remember, advantage is:
- greater sustaining off voltage (collector base is always higher than collector emitter)
-optimal turn off behaviour (negative base current equals collector current)
-automatic short circuit protection (when collector current exceeds emitter current, BJT automatically turns off)
I must confess I never really tried the idea, but it seemed to me a good one at the time when it was published.
Best regards,
Jaka Racman
thanks for posting those waveforms. I got a little bit nostalgic looking at them, since my last bipolar design dates almost 20 years ago. As far as I can remember, your base current waveform looks very good, a proper textbook example.
Maybe you could consider another variant of the transformer coupled bipolar drive. I think it was proposed by Bruce Carsten, one of the old SMPS gurus. In his design, he uses grounded base and emitter is driven by secondary winding of the drive transformer. Primary of the drive transformer is driven by a current source. As far as I can remember, advantage is:
- greater sustaining off voltage (collector base is always higher than collector emitter)
-optimal turn off behaviour (negative base current equals collector current)
-automatic short circuit protection (when collector current exceeds emitter current, BJT automatically turns off)
I must confess I never really tried the idea, but it seemed to me a good one at the time when it was published.
Best regards,
Jaka Racman
Jaka:
As you can see, 20 years later bipolar transistors still compete with MOSFETs
Conduction losses are better than the ones that you will get with latest technology TO-220 MOSFETs, 1V or less voltage drop for 10A collector current, a $15 TO-247 MOSFET is required to beat a $1 TO-220 bipolar in that field. And turn-off is not slow at all, at Ic=10A without RC snubber I get a nice 50ns rise time with 10ns or so current tail. However, turn-on is slow because Ib rise slope is limited by the base drive transformer I'm considering some changes
I don't like that emitter-switched drive scheme because all the colector current has to pass through the drive transformer, this imposes even more limitations and leakage inductance requirements. Also, getting the BE junction reverse biased during crossover time becomes quite hard.
XyeMoe:
It may seem crazy but I don't have a schematic for that board, and all the schematics I've posted on that thread were actually drawn after the boards were made.
Why? Because I usually build things direcly on a breadboard and I debug them with an actual oscilloscope and a multimeter, waching true waveforms and voltages instead of simulating. When I'm happy with the circuit, I draw a PCB by hand (no automatic schematic checking) while observing the breadboard or from memory if I can remember the circuit completely.
I will post a more detailed picture of the layout of the daughter board, you will have to figure out the rest by yourself, though
As you can see, 20 years later bipolar transistors still compete with MOSFETs
Conduction losses are better than the ones that you will get with latest technology TO-220 MOSFETs, 1V or less voltage drop for 10A collector current, a $15 TO-247 MOSFET is required to beat a $1 TO-220 bipolar in that field. And turn-off is not slow at all, at Ic=10A without RC snubber I get a nice 50ns rise time with 10ns or so current tail. However, turn-on is slow because Ib rise slope is limited by the base drive transformer I'm considering some changesI don't like that emitter-switched drive scheme because all the colector current has to pass through the drive transformer, this imposes even more limitations and leakage inductance requirements. Also, getting the BE junction reverse biased during crossover time becomes quite hard.
XyeMoe:
It may seem crazy but I don't have a schematic for that board, and all the schematics I've posted on that thread were actually drawn after the boards were made.
Why? Because I usually build things direcly on a breadboard and I debug them with an actual oscilloscope and a multimeter, waching true waveforms and voltages instead of simulating. When I'm happy with the circuit, I draw a PCB by hand (no automatic schematic checking) while observing the breadboard or from memory if I can remember the circuit completely.
I will post a more detailed picture of the layout of the daughter board, you will have to figure out the rest by yourself, though
Eva said:
You are amazing!!!
....nerd....
Schematic of the entire SMPS only in your brain !I also prefer reality vs. simulation.... ,
But well...... usually I draw the schematic on some 'butter bread paper' first (except ultra simple things) ... !
Uhps, and there is one thing I really love to simulate, because
the simulation results match really good to reality:
OP-Amp-Filters below 100kHz.
I use simulation mostly for linear and bode stuff, but I can't simulate base drive circuits since the usual transistor models doesn't even account for charge storage phenomena or other bipolar vodoo-like effects
Also, believe it or not, I'm still unemployed and looking for a decent job since there is almost no electronics R&D here.
Also, believe it or not, I'm still unemployed and looking for a decent job since there is almost no electronics R&D here.
Synchronous rectification daugtherboard layout:
Part list:
C9,10 - 22u 35V
D1..D8 - 1N4148
IC1 - CD4049
Q9,11,13,15 - bc327
Q10,12,14,16,17,18,19 - bc546
R1,2,3,4,5,6,7,8 - 10
R9,10,11,12 - 47
R13 - 22
R14,15,16,17 - 100
R18,19,22,23 - 10k
R20,21,24 - 4k7
"+V" is 17 volts, could be 19V to further reduce conduction losses but the CMOS buffer may fail.
"_EN" is enable input, it's active when pulled low (an external current comparator drives it). It's pulled up by default so everything is disabled at power-up.
"A_off" and "B_off" are pulled up to turn-off the corresponding set of MOSFETs inmediately before the primary is switched in one or another direction, then the MOSFETs remain turned off until their source voltages "S" fall below the input threshold of the CMOS buffer (V+/2). These inputs are driven directly from the outputs of the main SG3525A control IC.
Q17 and Q18 are intentionally saturated in order to delay their effect.
An externally hosted image should be here but it was not working when we last tested it.
Part list:
C9,10 - 22u 35V
D1..D8 - 1N4148
IC1 - CD4049
Q9,11,13,15 - bc327
Q10,12,14,16,17,18,19 - bc546
R1,2,3,4,5,6,7,8 - 10
R9,10,11,12 - 47
R13 - 22
R14,15,16,17 - 100
R18,19,22,23 - 10k
R20,21,24 - 4k7
"+V" is 17 volts, could be 19V to further reduce conduction losses but the CMOS buffer may fail.
"_EN" is enable input, it's active when pulled low (an external current comparator drives it). It's pulled up by default so everything is disabled at power-up.
"A_off" and "B_off" are pulled up to turn-off the corresponding set of MOSFETs inmediately before the primary is switched in one or another direction, then the MOSFETs remain turned off until their source voltages "S" fall below the input threshold of the CMOS buffer (V+/2). These inputs are driven directly from the outputs of the main SG3525A control IC.
Q17 and Q18 are intentionally saturated in order to delay their effect.
Eva said:I use simulation mostly for linear and bode stuff, but I can't simulate base drive circuits since the usual transistor models doesn't even account for charge storage phenomena or other bipolar vodoo-like effects
Also, believe it or not, I'm still unemployed and looking for a decent job since there is almost no electronics R&D here.
Evita got mail.
...somehow I don't get your layout in line with the
traditional synchronous rectification circuits...
..in example like shown here:
http://focus.ti.com/lit/an/slua287/slua287.pdf
Your common source connection would match,
but you only parallel two drains...
On the other hand your A/B gate drive could match again....
Do you use a double secondary winding with separate rectification for
each? .... and parallel both behind rectification?
There are two transformers whose primaries are connected in series, and two sets of push-pull secondaries that are obviously forced to conduct the same current, so paralelling is not required.
Also, If you try to lay out the board, you will find out that paralelling the secondaries actually makes things harder, particularly with a single sided PCB (where do you put the bridges??? and how do you get the high symmetry required between A and B banks???), and produces bigger current loops thus increasing EMI, while using independent rectificacion for each transformer makes things much simpler.
Note that the MOSFETs are laid out in an A-A-B-B-A-A-B-B fashion, and each A/B set is not turned on until the voltages at the drains of both transformers are not low enough.
Concerning current doublers, they are just a fashion thing that allows to use simpler off-the-shelf single ended transformers and lower current inductors at the expense of more than 4 times higher current ripple while employing the same amount of ferrite and copper. Increased current ripple produces much higher I^2*R and switch turn-off losses, so this topology is almost useless for high power applications, its only advantage is easier manufacturing.
PD: You've also got mail.
Also, If you try to lay out the board, you will find out that paralelling the secondaries actually makes things harder, particularly with a single sided PCB (where do you put the bridges??? and how do you get the high symmetry required between A and B banks???), and produces bigger current loops thus increasing EMI, while using independent rectificacion for each transformer makes things much simpler.
Note that the MOSFETs are laid out in an A-A-B-B-A-A-B-B fashion, and each A/B set is not turned on until the voltages at the drains of both transformers are not low enough.
Concerning current doublers, they are just a fashion thing that allows to use simpler off-the-shelf single ended transformers and lower current inductors at the expense of more than 4 times higher current ripple while employing the same amount of ferrite and copper. Increased current ripple produces much higher I^2*R and switch turn-off losses, so this topology is almost useless for high power applications, its only advantage is easier manufacturing.
PD: You've also got mail.
What other output transformer options have you considered? I have been thinking that the choice in that regard may be able to allow for a simpler circuit and still produce little or no EMI. For example, I have been playing around with ZVS circuits and am tending to find that some transformers made of different formulations do better than others. So far I wonder if one consideration is to try to choose one without too much high frequency coupling capability.
I have been using only ferrite transformers so far power transfer with one exception. I choose low switching frequencies around the upper hearing limit, too. I wonder if something like tape wound silicon steel is an option. I made one project using a 60 hz one of that material type because it had to have extremely high current output at low voltage--on the order of 300a AC. I drove it with a 1000hz square wave and the core never became warm.
I have been using only ferrite transformers so far power transfer with one exception. I choose low switching frequencies around the upper hearing limit, too. I wonder if something like tape wound silicon steel is an option. I made one project using a 60 hz one of that material type because it had to have extremely high current output at low voltage--on the order of 300a AC. I drove it with a 1000hz square wave and the core never became warm.
Humm...
Farnell is now selling IPP60R099CS at 10 euro/each for 25 pcs. It's getting cheaper and cheaper for a 600V 31A TO-220 MOSFET featuring less than 0.1 ohm Rds-on, but I still don't feel like expending 250 euro... I've just got a lot of 25 RoHS-compliant MJE13009 for 25 euro.
Also, the datasheet reveals some interesting facts about that new kind of MOSFETs, like a stronger than usual increase in Rds-on with temperature and much lower than usual capacitances at high voltages
Farnell is now selling IPP60R099CS at 10 euro/each for 25 pcs. It's getting cheaper and cheaper for a 600V 31A TO-220 MOSFET featuring less than 0.1 ohm Rds-on, but I still don't feel like expending 250 euro... I've just got a lot of 25 RoHS-compliant MJE13009 for 25 euro.
Also, the datasheet reveals some interesting facts about that new kind of MOSFETs, like a stronger than usual increase in Rds-on with temperature
and much lower than usual capacitances at high voltages Now even Eva is drooling after coolmosses...Eva said:Humm...
Farnell is now selling IPP60R099CS at 10 euro/each for 25 pcs. It's getting cheaper and cheaper for a 600V 31A TO-220 MOSFET featuring less than 0.1 ohm Rds-on, but I still don't feel like expending 250 euro... I've just got a lot of 25 RoHS-compliant MJE13009 for 25 euro.
Also, the datasheet reveals some interesting facts about that new kind of MOSFETs, like a stronger than usual increase in Rds-on with temperature
If older 20amp coolmosses are good enough they cost less than 4 euros a piece at digikey. Or SPW47N60C3 to247 device for less than 10 euros.
they also have lover peak current ratings and avalance energy ratings due to their miniature chip.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.