Thanks for the link to the other thread. Though I would like to think that I am more competent then simply being able to build scematics like that individual, this project is making me believe otherwise, but it will be good for the experience. OK, I have taken some pics for your viewing pelasure!
Unloaded secondary waves (note overshoot I mentioned) Pay no attention to the RMS on these, they are square waves out of phase, and I believe the ocilliscope just averages what shows onscreen. Note the amplitude matches.
unloaded output Huge difference! The arrows for channel 1 and 2 show ground, and though they look to be the same distance, arrow 2 is going off of a 50V per major division scale, while Ch. 1 is only 20V.
loaded secondary waves The waves are still very close to the same, and I am beginning to think that the voltage problem is in the spike, though I've been told that a spike is normal.
Loaded output About a six volt difference between the two...still pretty bad in my opinion. they should be 3-6 volts difference UNloaded, and the same loaded.
and finally, just a pic of the board I have built so far, to help you all empathize!(pic)
My circuit is fundementally the same as the circuit used in your link, probably because it's just your average rectification circuit.[link]
I am beginning to realize my misconceptions now that the project is loaded. (again, where actually experience outweighs bookwork) I remember being surprised that I was getting 24 Volts on my primaries, hearing that this was normal, and adjusting my windings to try and get about 35V, basically 1:1.5... well, loaded, I am getting 1.5 times my 12V, so I'll have to "wind" back to where I originally was(1:3).
About those spikes now...are they normal, or are they gargantuan? I have rewound the darn thing five times, used two different cores, but the reaction only changed when I changed mosfets, and I ended up getting something similar to the loaded waves while unloaded, but I was also pulling five amps unloaded(so I switched my switchers back). Again, my IC ocillator gives a clean square. The only other thing I can think of is that my huge caps on the primary might be giving a giant inrush every time it switches? Don't know why, I thought they would only help.
Ok, here is how I am winding. I have taken 8 strands of 26 gauge in parallel, and wound 6 turns. then I liink another set of wires exactly the same, and wind another 6 in the same direction. This makes my primary, and I cover the entire toroid, so each wire covers half, and they meet in the opposite side form the link. Now I start at the link of the primary (ground), and wind 10 strands of 30ga. in parallel in the opposite direction of the primary it is on top of, again covering half of the toroid. Then go back to the starting point and wind the other half, so basically both ground points start at the same position, and all ends end together. I realize that my pic looks just a little sloppy, I just soldered it in quickly to take the pics, but I assure you , the results have always been the same on the O scope, no matter how I try to fix the transformer. I have also tried 6 and 8 strands of 22 ga. for secondary and primary.
So...what's the word?
Unloaded secondary waves (note overshoot I mentioned) Pay no attention to the RMS on these, they are square waves out of phase, and I believe the ocilliscope just averages what shows onscreen. Note the amplitude matches.
unloaded output Huge difference! The arrows for channel 1 and 2 show ground, and though they look to be the same distance, arrow 2 is going off of a 50V per major division scale, while Ch. 1 is only 20V.
loaded secondary waves The waves are still very close to the same, and I am beginning to think that the voltage problem is in the spike, though I've been told that a spike is normal.
Loaded output About a six volt difference between the two...still pretty bad in my opinion. they should be 3-6 volts difference UNloaded, and the same loaded.
and finally, just a pic of the board I have built so far, to help you all empathize!(pic)
My circuit is fundementally the same as the circuit used in your link, probably because it's just your average rectification circuit.[link]
I am beginning to realize my misconceptions now that the project is loaded. (again, where actually experience outweighs bookwork) I remember being surprised that I was getting 24 Volts on my primaries, hearing that this was normal, and adjusting my windings to try and get about 35V, basically 1:1.5... well, loaded, I am getting 1.5 times my 12V, so I'll have to "wind" back to where I originally was(1:3).
About those spikes now...are they normal, or are they gargantuan? I have rewound the darn thing five times, used two different cores, but the reaction only changed when I changed mosfets, and I ended up getting something similar to the loaded waves while unloaded, but I was also pulling five amps unloaded(so I switched my switchers back). Again, my IC ocillator gives a clean square. The only other thing I can think of is that my huge caps on the primary might be giving a giant inrush every time it switches? Don't know why, I thought they would only help.
Ok, here is how I am winding. I have taken 8 strands of 26 gauge in parallel, and wound 6 turns. then I liink another set of wires exactly the same, and wind another 6 in the same direction. This makes my primary, and I cover the entire toroid, so each wire covers half, and they meet in the opposite side form the link. Now I start at the link of the primary (ground), and wind 10 strands of 30ga. in parallel in the opposite direction of the primary it is on top of, again covering half of the toroid. Then go back to the starting point and wind the other half, so basically both ground points start at the same position, and all ends end together. I realize that my pic looks just a little sloppy, I just soldered it in quickly to take the pics, but I assure you , the results have always been the same on the O scope, no matter how I try to fix the transformer. I have also tried 6 and 8 strands of 22 ga. for secondary and primary.
So...what's the word?
This negative spike could be caused by very slow diode turnoff combined with high leakage inductance and poor coupling in the transformer [it's like there were flowing very high current when the positive rail diodes turn off]
Anyway, your transformer is improperly wound so you'll habe to wind it back, but this time please do it right :
Isuggest 4T + 4T primary, 10T + 10T secondary [your core appears to be too big to require 6 turns at 55Khz]
At 55Khz and for a simple prototype you don't need to fight with litz wires, just use single solid wire for each winding [for light loads there isn't any difference]
All windings have to be thightly coupled and must have the lowest leakage inductance. This means you must spread EACH winding over the entire core
Wind both primeries first, in a bifilar fashion
Then do the same with both secondaries at the same time, and wind everything in the same direction
Winding this way you will reduce leakage inductance by an order of magnitude [ringing and overshoot is almost proportional to leakage inductance]
Insert some dead time in the control circuit [25-50%] to be able to see clearly how the system behaves when both mosfets are turned off [the spikes could be caused by parasitistic turn-on of one of the mosfets when the other turns-on, look for gate waveforms, or simple cross conduction and/or core saturation due to too low dead time]
I use a potentiometer in my prototypes to regulate duty cycle and look for spikes, cross conduction, core saturation, etc... as the ducy cycle increases
Also, when doing the captures try to use 2uS/div and 10V/div so wen can see clearly the turn-on and turn-off transients
Is it normal to see that much noise or shadow in the traces? If it's ripple from the 12V supply, don't you have a hold function in she digital scope?
Are the output diodes all equal, not blown and properly wired?
Anyway, your transformer is improperly wound so you'll habe to wind it back, but this time please do it right :
Isuggest 4T + 4T primary, 10T + 10T secondary [your core appears to be too big to require 6 turns at 55Khz]
At 55Khz and for a simple prototype you don't need to fight with litz wires, just use single solid wire for each winding [for light loads there isn't any difference]
All windings have to be thightly coupled and must have the lowest leakage inductance. This means you must spread EACH winding over the entire core
Wind both primeries first, in a bifilar fashion
Then do the same with both secondaries at the same time, and wind everything in the same direction
Winding this way you will reduce leakage inductance by an order of magnitude [ringing and overshoot is almost proportional to leakage inductance]
Insert some dead time in the control circuit [25-50%] to be able to see clearly how the system behaves when both mosfets are turned off [the spikes could be caused by parasitistic turn-on of one of the mosfets when the other turns-on, look for gate waveforms, or simple cross conduction and/or core saturation due to too low dead time]
I use a potentiometer in my prototypes to regulate duty cycle and look for spikes, cross conduction, core saturation, etc... as the ducy cycle increases
Also, when doing the captures try to use 2uS/div and 10V/div so wen can see clearly the turn-on and turn-off transients
Is it normal to see that much noise or shadow in the traces? If it's ripple from the 12V supply, don't you have a hold function in she digital scope?
Are the output diodes all equal, not blown and properly wired?
I am using a ferrite material, with a permeability of approx. 3000. Was previously using iron powder, but couldnt get good results below 200kHz, and by good I mean similar to what I have now.
I don't think my diodes are the problem (data sheet), because I get the spikes on the secondaries out of circuit as well, and had them before I added the bridge or capacitors.
I will try the duty cycle adjustment, and rewiring the transformer. Here is how my diodes are wired, I believe it is correct. Now, when you say to cover the entire toroid with the windings, are you saying go all the way around with each primary/secondary, or half way, to meet in the middle, and does it matter if you go beyond that and overlap?
I don't think my diodes are the problem (data sheet), because I get the spikes on the secondaries out of circuit as well, and had them before I added the bridge or capacitors.
I will try the duty cycle adjustment, and rewiring the transformer. Here is how my diodes are wired, I believe it is correct. Now, when you say to cover the entire toroid with the windings, are you saying go all the way around with each primary/secondary, or half way, to meet in the middle, and does it matter if you go beyond that and overlap?
These Diodes are a good choice. Do they get hot without load?. If the reply is no then the problem is in the primary side, secondary side is too simple and too pasive altough RC damping networks are missing, but this shouldn't make a great difference anyway
If the waveforms are almost the same with secondaries disconnected, I think you may be experiencing heavy cross conduction at MOSFET turn-on
About winding the transformer, you have to go all the way aroud the core with each winding and each pair of windings has to be perfectly matched to avoid asymmetric voltage drop in them and core saturation at high powers, so I suggest bifilar [bi-strand] winding
10-20% dead time is also very important to give time to the core to 'center itself' between switching cycles and prevent saturation
Remember that control ICs are not perfect and that the core is not AC coupled so if positive pulses are 1% longer than negative pulses then you'll have some DC applied to the core, 12 * .01 = .12V, enough to saturate it
Each winding by itself has to cover the entire toroid [really, all it has to cover is all the other windings and vice-versa, the toroid itself has nothing to do with leakage inductance or HF coupling, but people seems to think so]
When you have half the windings in one side and half the windings in the other, there is very poor HF coupling between sides of the symmetric windings and this causes huge ringing and overshoots due to huge leakage inductances between halves
Could you post the schematic or the PCB of the primary side?
PD: Transformer winding is more tricky than it appears to be at first
If the waveforms are almost the same with secondaries disconnected, I think you may be experiencing heavy cross conduction at MOSFET turn-on
About winding the transformer, you have to go all the way aroud the core with each winding and each pair of windings has to be perfectly matched to avoid asymmetric voltage drop in them and core saturation at high powers, so I suggest bifilar [bi-strand] winding
10-20% dead time is also very important to give time to the core to 'center itself' between switching cycles and prevent saturation
Remember that control ICs are not perfect and that the core is not AC coupled so if positive pulses are 1% longer than negative pulses then you'll have some DC applied to the core, 12 * .01 = .12V, enough to saturate it
Each winding by itself has to cover the entire toroid [really, all it has to cover is all the other windings and vice-versa, the toroid itself has nothing to do with leakage inductance or HF coupling, but people seems to think so]
When you have half the windings in one side and half the windings in the other, there is very poor HF coupling between sides of the symmetric windings and this causes huge ringing and overshoots due to huge leakage inductances between halves
Could you post the schematic or the PCB of the primary side?
PD: Transformer winding is more tricky than it appears to be at first
Dang woman, you're beautiful! It works perfectly now, from what I can see! It was the transformer winding after all. So hey, nobody ever taught me how to wind a transformer, seems intuitive, right? Wrong. Well, I am glad it was something that simple, at least now I don't feel completely inept.
By the way, the diodes still don't get very hot, maybe just a little warm with lightbulb loads. Should I be worried? it's working fine...
I was reading all these posts, and I was thinking, what's this bi-something winding that everyone keeps talking about. So I did a little studying, and hey, that Tesla guy's not too shabby, he should be famous(just kidding).
I am going to post some pics just in case someone else has a similar problem later, and to confirm that I got a good result with you guys.
Thanks guys.
By the way, the diodes still don't get very hot, maybe just a little warm with lightbulb loads. Should I be worried? it's working fine...
I was reading all these posts, and I was thinking, what's this bi-something winding that everyone keeps talking about. So I did a little studying, and hey, that Tesla guy's not too shabby, he should be famous(just kidding).
I am going to post some pics just in case someone else has a similar problem later, and to confirm that I got a good result with you guys.
An externally hosted image should be here but it was not working when we last tested it.
To those of you who don't know how to wind a transformer, you're not alone! Make sure you use your Ohmmeter so you don't connect the same wire to itself in the middle after you wind the two parallel primaries.An externally hosted image should be here but it was not working when we last tested it.
Look good? This is under load, with very little load, the duty cycle shortens to about 15% and the squares become ramps (I assume this is a short square pulse with a trailing discharge.)Thanks guys.
You have reduced duty cycle to about 60%, haven't you?
These ramps you can see are caused by the magnetizing inductance of the core discharging freely during dead time
As you can see, the flux in the core is somewhat asymetric but not enough to enter deep saturation [when flux is symetric, waveforms are symetric during the time you let the core 'free' to 'flyback' in the direction it wishes]
Now, the transformer appears to have so low leakage inductance that overshoot and ringing are too small to be appreciated at first sight. [Try to remove the RC on the primary and look to the waveforms at MOSFET drains]
Now, you can try to increase the duty cycle to 90% and see what happens [high duty cycle is needed to get good efficiency]
Could you post a capture of the voltage waveforms at the *gates* of the MOSFETS with the maximun load you can use? [with 60% and maximum duty cycles]
I suspect you may have a pending problem of parasitistic MOSFET turn on, altough with the good coupling in the new transformer it no longer screws up the circuit
About diodes : They should get equally hot but only with load [for heavy loads they will need a heatsink], getting hot without load means someting is going wrong
The core may also get 5-10ºC hotter than ambient temperature [independently of load] but this is normal
MOSFETS should not get hot witn light loads or no load, only with heavy load
These ramps you can see are caused by the magnetizing inductance of the core discharging freely during dead time
As you can see, the flux in the core is somewhat asymetric but not enough to enter deep saturation [when flux is symetric, waveforms are symetric during the time you let the core 'free' to 'flyback' in the direction it wishes]
Now, the transformer appears to have so low leakage inductance that overshoot and ringing are too small to be appreciated at first sight. [Try to remove the RC on the primary and look to the waveforms at MOSFET drains]
Now, you can try to increase the duty cycle to 90% and see what happens [high duty cycle is needed to get good efficiency]
Could you post a capture of the voltage waveforms at the *gates* of the MOSFETS with the maximun load you can use? [with 60% and maximum duty cycles]
I suspect you may have a pending problem of parasitistic MOSFET turn on, altough with the good coupling in the new transformer it no longer screws up the circuit
About diodes : They should get equally hot but only with load [for heavy loads they will need a heatsink], getting hot without load means someting is going wrong
The core may also get 5-10ºC hotter than ambient temperature [independently of load] but this is normal
MOSFETS should not get hot witn light loads or no load, only with heavy load
I was under the impression that the duty cycle should be just under 50% on each leg, making two square waves that are never high at the same time. I haven't changed anything from the previous pictures I posted duty cycle wise, but I did zoom in on the time axis like you had asked. I will post a waveform from the gates when I get home. It looks the same as the last picture I posted above, just a little more square rather than rounded off on the edge. I tend to think that you're still right, because I calculated my efficiency at about 54%.
With my lightbulb load, mosfets are about 90 degrees F. , diodes are just slightly warm to the touch. not so surprising I guess, where I am only pulling 300mA through the diodes, and 2.8 amps on the mosfets. 11.8V in@ 2.8 amps=33 watts, Two supplies that are 30V@300mA=18Watts, 54% efficiency, right?
All in all , I am still very excited to actually have a split supply with even readings, and no huge spike. Now for fine tuning.
With my lightbulb load, mosfets are about 90 degrees F. , diodes are just slightly warm to the touch. not so surprising I guess, where I am only pulling 300mA through the diodes, and 2.8 amps on the mosfets. 11.8V in@ 2.8 amps=33 watts, Two supplies that are 30V@300mA=18Watts, 54% efficiency, right?
All in all , I am still very excited to actually have a split supply with even readings, and no huge spike. Now for fine tuning.
Unloaded at gate
Loaded at gate
An externally hosted image should be here but it was not working when we last tested it.
Loaded at gate
An externally hosted image should be here but it was not working when we last tested it.
As I've said, transformers need to be excited in both directions in an alternated and symmetric fashion [there must be no net DC aplied at all]
When I said 50%, 60% or 90% I was talking about the duty cycle of each 'half pulse' by itself. Think that the oscillator of your IC is working at about 117Khz and I was speaking about duty cycle relative to this oscillator, not to the 58Khz frequency seen by the transformer [Datasheets of ICs tend to express duty cycle relative to the pair of symmetric of pulses, ie : 45% instead of 90%]
This image made me think that you reduced duty cycle to about 60% because the voltage drop in the waveform, but if duty cycle was more than 90% [45%] all the time, then this must be crude asymetric transformer saturation [what you see is the voltage drop due to big current flowing through the primary]. Load may be helping to balance things a bit but you should not rely on this [there appears to be no saturation under load]
About gate waveforms, they seem to have too fast rise and fall times to be measured at the gate of the devices instead of at the output of the IC [there should be at least a resistor inbetween to limit peak current and rise times]. Were you measuring directly at the gates as I've said?. Were you taking ground at one of the sources of the MOSFETS?.
Sometimes switching MOSFETS are not doing what you want and the only way too see it is to measure direcly at its legs
Gate waveforms with no load look right, but they show only about 200nS dead time and this is very small for your converter, you should increase dead time to about 500nS-1uS for the reasons I've already explained
Gate waveforms with load show a spike in an asymetric fashion and this means something is still going terribly wrong in the primary side. You should look for the origin of this spike [increasing dead time may make it go away if it was due to cross-conduction]
I suspect your waveforms were measured before gate resistors, and since gate waveforms are delayed [due to the effect of the R and te C of the gate] and altered [due to drain-gate capacitance], these spikes may be huge voltage drops caused by huge currents flowing when both MOSFETS are on at the same time [during a fraction of microsecond]
Beware of where you connect the ground of your probes since in SMPS ground may not have the same potential everywhere, specially if someting is going wrong and is causing fast high current transients and huge voltage drops at PCB traces
Also, beware of what is being induced in the ground wire of the probe and the probe itself since SMPS create magnetic fields, that may be huge when something is going wrong and is causing severe ringing or high current transients [most EMI is actually due to bad design]
When in doubt, try to move the probe or the ground point and look for changes in the measurements
Sometimes SMPS are much more forgiving than people thinks, they may work and provide some output even when they are malfunctioning and generating tons of EMI. Your converter may be generating some output without blowing, but please fix it before using it
When I said 50%, 60% or 90% I was talking about the duty cycle of each 'half pulse' by itself. Think that the oscillator of your IC is working at about 117Khz and I was speaking about duty cycle relative to this oscillator, not to the 58Khz frequency seen by the transformer [Datasheets of ICs tend to express duty cycle relative to the pair of symmetric of pulses, ie : 45% instead of 90%]
This image made me think that you reduced duty cycle to about 60% because the voltage drop in the waveform, but if duty cycle was more than 90% [45%] all the time, then this must be crude asymetric transformer saturation [what you see is the voltage drop due to big current flowing through the primary]. Load may be helping to balance things a bit but you should not rely on this [there appears to be no saturation under load]
An externally hosted image should be here but it was not working when we last tested it.
About gate waveforms, they seem to have too fast rise and fall times to be measured at the gate of the devices instead of at the output of the IC [there should be at least a resistor inbetween to limit peak current and rise times]. Were you measuring directly at the gates as I've said?. Were you taking ground at one of the sources of the MOSFETS?.
Sometimes switching MOSFETS are not doing what you want and the only way too see it is to measure direcly at its legs
Gate waveforms with no load look right, but they show only about 200nS dead time and this is very small for your converter, you should increase dead time to about 500nS-1uS for the reasons I've already explained
Gate waveforms with load show a spike in an asymetric fashion and this means something is still going terribly wrong in the primary side. You should look for the origin of this spike [increasing dead time may make it go away if it was due to cross-conduction]
I suspect your waveforms were measured before gate resistors, and since gate waveforms are delayed [due to the effect of the R and te C of the gate] and altered [due to drain-gate capacitance], these spikes may be huge voltage drops caused by huge currents flowing when both MOSFETS are on at the same time [during a fraction of microsecond]
Beware of where you connect the ground of your probes since in SMPS ground may not have the same potential everywhere, specially if someting is going wrong and is causing fast high current transients and huge voltage drops at PCB traces
Also, beware of what is being induced in the ground wire of the probe and the probe itself since SMPS create magnetic fields, that may be huge when something is going wrong and is causing severe ringing or high current transients [most EMI is actually due to bad design]
When in doubt, try to move the probe or the ground point and look for changes in the measurements
Sometimes SMPS are much more forgiving than people thinks, they may work and provide some output even when they are malfunctioning and generating tons of EMI. Your converter may be generating some output without blowing, but please fix it before using it
I see what you're saying now, you were seeing the curve downward on the falling edge and thinking that this was discharge rather than conduction time. Yes, the efficiency I am getting tells me that there is still trouble. I haven't really had problems with EMI disrupting my readings with the scope, the noise only becomes sizeable when I stick the probe through the middle of the transformer.
You're right, AGAIN... I was reading from the resistors connected to the gates rather than the gates themselves. They're rather hard to reach when they're clamped down to my heatsink, and I had thought you were mainly interested in the duty of the pulse rather than what the mosfet was doing exactly.
So you think that if I decrease the cycle from nearly 50% and add 1uS of dead time that it will help? I was also thinking of adding a potentiometer to adjust frequency and see where I get the best operating efficiency for my core(maybe closer to 40kHz? that seems to be more standard from the ones i've seen), of course I would like it to be working as good as it can before I do that.
You're right, AGAIN... I was reading from the resistors connected to the gates rather than the gates themselves. They're rather hard to reach when they're clamped down to my heatsink, and I had thought you were mainly interested in the duty of the pulse rather than what the mosfet was doing exactly.
So you think that if I decrease the cycle from nearly 50% and add 1uS of dead time that it will help? I was also thinking of adding a potentiometer to adjust frequency and see where I get the best operating efficiency for my core(maybe closer to 40kHz? that seems to be more standard from the ones i've seen), of course I would like it to be working as good as it can before I do that.
Hi Eva for me you are rigth,
Beware of where you connect the ground of your probes since in SMPS ground may not have the same potential everywhere, specially if someting is going wrong and is causing fast high current transients and huge voltage drops at PCB traces
Also, beware of what is being induced in the ground wire of the probe and the probe itself since SMPS create magnetic fields, that may be huge when something is going wrong and is causing severe ringing or high current transients [most EMI is actually due to bad design]
When in doubt, try to move the probe or the ground point and look for changes in the measurements
Sometimes SMPS are much more forgiving than people thinks, they may work and provide some output even when they are malfunctioning and generating tons of EMI. Your converter may be generating some output without blowing, but please fix it before using it
You look These Books :
EMC and Printed Circuit Board
Printed Circuit Boards Desig Thecniques for EMC Compliance.
Son de IEEE ( Montrose) Willley Inter Science.
Jesús
Beware of where you connect the ground of your probes since in SMPS ground may not have the same potential everywhere, specially if someting is going wrong and is causing fast high current transients and huge voltage drops at PCB traces
Also, beware of what is being induced in the ground wire of the probe and the probe itself since SMPS create magnetic fields, that may be huge when something is going wrong and is causing severe ringing or high current transients [most EMI is actually due to bad design]
When in doubt, try to move the probe or the ground point and look for changes in the measurements
Sometimes SMPS are much more forgiving than people thinks, they may work and provide some output even when they are malfunctioning and generating tons of EMI. Your converter may be generating some output without blowing, but please fix it before using it
You look These Books :
EMC and Printed Circuit Board
Printed Circuit Boards Desig Thecniques for EMC Compliance.
Son de IEEE ( Montrose) Willley Inter Science.
Jesús
alright, here's the update...I shortened the cycle time a bit, and these pics are measured from the GATE. I am calculating efficiency between 82%(40kHz) and 91%(57kHz) now, I am guessing that I will want to go with 40kHz if I want to pull more power through it. There is a bit of a hump at the end of one of the clock cycles, and it's followed by a slight wave when the system is unloaded... wondering if it's some sort of hysteresis from the winding not being exactly perfect, or just a crappy mosfet. I think this shows quite a bit of improvement, as I have also removed the snubbing circuit I had between the primaries and center tap, and there is still no spike to be seen, whereas before I relied on it. I think the dead time may be just a hair TOO long now, though, looks close to 1.8uS. I can easily trim it back a bit. So what's the word?
Loaded
Unloaded...sorry, it's a little blurry
Loaded
An externally hosted image should be here but it was not working when we last tested it.
Unloaded...sorry, it's a little blurry
An externally hosted image should be here but it was not working when we last tested it.
This drawing represents all you should find in the waveforms of a switching mosfet under load, details are a bit exagerated
In your circuit the upper gate waveform looks like the one from a mosfet that is not conducting while the lower waveform looks like the one from a mosfet that is conducting [and conducting more heavily under no load than under load, curious, isn't it?
Transformer saturation in the direction of the lower mosfet is obvious, try to increase a bit dead time or switching frequency
As you would see, you allways need more dead time than expected since MOSFETs take some time to turn off and gates take some more time to discharge
In your circuit, even having 1.8uS dead time, gate waveforms are 'too close', maybe less than 500ns clearance between them, look at them in detail...
I willl describe later the processes I've marked with numbers in the drawing
In your circuit the upper gate waveform looks like the one from a mosfet that is not conducting while the lower waveform looks like the one from a mosfet that is conducting [and conducting more heavily under no load than under load, curious, isn't it?
Transformer saturation in the direction of the lower mosfet is obvious, try to increase a bit dead time or switching frequency
As you would see, you allways need more dead time than expected since MOSFETs take some time to turn off and gates take some more time to discharge
In your circuit, even having 1.8uS dead time, gate waveforms are 'too close', maybe less than 500ns clearance between them, look at them in detail...
I willl describe later the processes I've marked with numbers in the drawing
Attachments
So at the gate, I am looking for something that looks like the green line you have in your artwork? Is this getting closer?(loaded, 54.8 watts in, 50.15 watts out)
An externally hosted image should be here but it was not working when we last tested it.
Yep, these waveforms are much more like the ones you will find in a converter working properly
Now, you are geting more than 90% efficiency
Could you post a capture of the drain waveforms in the same order and the same conditions?. Or better, two captures, one of each mosfet showing its gate and drain waveforms
You have increased a bit the oscillator frequency and the dead time, haven't you?
You also could speed up MOSFET turn-on and turn-off by reducing the value of gate resistors, but take into account that the control IC has limited current capability so you may need buffering [the simplest buffer could be made with two emitter followers PNP & NPN with Bs and Es tied together]
Speeding up turn-off is prioritary [it's when mosfets disipate more power] so you may divide gate resistors into two parts and put a [fast] diode in paralell with one of them to speed up only turn-off [speeding-up turn on increases EMI]. Having very fast turn-off, you can reduce substantially dead time and increase eficiency further
If you want to experiment with optimized gate drive, I recomend using the 24V or more you find in the primary not being switched to create a regulated 20V supply and use it to power the control IC and to charge the gates of the mosfets to 18V. I use this technique to reduce the Rds-on and reduce the number of required devices for a given power output [I'm using just two IRFZ48V at about 0.60$ each for more than 60A reliable output at 14.4V 30Khz, but converter topology is somewhat different]
Now, you are geting more than 90% efficiency
Could you post a capture of the drain waveforms in the same order and the same conditions?. Or better, two captures, one of each mosfet showing its gate and drain waveforms
You have increased a bit the oscillator frequency and the dead time, haven't you?
You also could speed up MOSFET turn-on and turn-off by reducing the value of gate resistors, but take into account that the control IC has limited current capability so you may need buffering [the simplest buffer could be made with two emitter followers PNP & NPN with Bs and Es tied together]
Speeding up turn-off is prioritary [it's when mosfets disipate more power] so you may divide gate resistors into two parts and put a [fast] diode in paralell with one of them to speed up only turn-off [speeding-up turn on increases EMI]. Having very fast turn-off, you can reduce substantially dead time and increase eficiency further
If you want to experiment with optimized gate drive, I recomend using the 24V or more you find in the primary not being switched to create a regulated 20V supply and use it to power the control IC and to charge the gates of the mosfets to 18V. I use this technique to reduce the Rds-on and reduce the number of required devices for a given power output [I'm using just two IRFZ48V at about 0.60$ each for more than 60A reliable output at 14.4V 30Khz, but converter topology is somewhat different]
Alright, now keep in mind that I removed the snubbing capacitors, I imagine that the overshoot would look better if they were still there. Then again, I'm not 100% sure what the drain wave should look like, but all the other results seem good. I will keep your gate driving ideas in mind for my next project, I already etched this board...let me know what you think about these. This is both mosfets, Ch1 is drain, Ch2 is gate. The last picture is the secondary output.
An externally hosted image should be here but it was not working when we last tested it.
An externally hosted image should be here but it was not working when we last tested it.
An externally hosted image should be here but it was not working when we last tested it.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.