First smps

[brady]

How are you probing the high side gate? Improper probing of a floating high side switch can cause all sorts of issues. A high-voltage differential probe is the proper way to do it, but they are expensive and most people don't have access to one.
Also, the IR2110, when used with a boot-strapped high side supply, doesn't work very well at very low duty cycles. The capacitors that store the charge to drive the high side gate get charged up only while the low side switch is turned on. If the low side switch has a very low on-time then the bootstrap caps may not fully charge and can cause the high side to switch erratically since it does not have enough charge to drive it properly. This might not be your issue, but it's something to check for.

Brad

I'm using a 100x probe. I actually noticed that it did have an affect on the performance. When I removed the probe there was no screeching sound as there was when the problem occurs. The problem probably does come down to the bootstrap supply as you said.

Regards,
Brady

 

[Silvio]

I am using a new PCB design. I made a jumper to shorten the lead connecting the source pin of the high side and that didn't change anything. I also made the gate resistances larger and that didn't change anything. Thing is, it seems to work perfectly at maximum duty cycle. I can also lower the duty cycle some before things go wrong. As others said, maybe I just cannot lower it that much with the bootstrap supply. However, even though I can lower the duty cycle some, it doesn't effect the output voltage. I'm thinking this could be because I don't have a large enough load, is this correct?

Regards,
Brady

It could be that you are not loading enough until the voltage starts to decrease. You did not say what is the current drain on the output is. you can try a 100 watt bulb (110v) on the two outputs hooked in series. At 60v from your output does not give so much current I guess it will be around 0.8 amps or so. This will be enough for the first test. You can also try a car bulb rated 24v 21watts and load one end of your double output at a time. These are cheap to buy.

I guess Brad was right in what he said about a bootstrap driven high side, as later yesterday I was reading Tahmid's Blog and the same things more or less was said having an IR2110 driving a regulated output. I guess it would be needing a separate supply so that you can isolate the high side VB pin and ground referenced to the mid point VS. One other thing that I noticed was that Tahmid when using the IR2110 powers the VDD pin 9 with 5v taken from the 12v regulator with a 7805 chip. I don't really know the purpose but its done like this for some reason Maybe Brad can explain why. It may driving this pin separately make things run smoother without much distortion on the output waveform.

Regards Silvio

 

[blasphemy000]

I am using a new PCB design. I made a jumper to shorten the lead connecting the source pin of the high side and that didn't change anything. I also made the gate resistances larger and that didn't change anything. Thing is, it seems to work perfectly at maximum duty cycle. I can also lower the duty cycle some before things go wrong. As others said, maybe I just cannot lower it that much with the bootstrap supply. However, even though I can lower the duty cycle some, it doesn't effect the output voltage. I'm thinking this could be because I don't have a large enough load, is this correct?

Regards,
Brady

If you have updated your schematic and board design, could you please upload the current version of the circuit you are using?
The lowest duty-cycle that you can use will depend on the size of your bootstrap capacitors, the current that your VCC can supply, how fast the bootstrap diode can begin conducting when the LS switch turns on, and the amount of gate charge that your FETs require to fully turn on. The IR2110 can be kind of tricky to get right, but they do work fairly well when properly setup.
The no load/light load condition is most likely correct. Your schematic shows a large inductor and capacitors on the output rails so a light load would require very little duty-cycle to maintain the output voltage after the caps are initially charged up.

I'm using a 100x probe. I actually noticed that it did have an affect on the performance. When I removed the probe there was no screeching sound as there was when the problem occurs. The problem probably does come down to the bootstrap supply as you said.

Regards,
Brady

Probing the HS Fet with a 100x probe is an acceptable method of doing so, just make sure that you are hooking the ground clip of the scope probe to an EARTH referenced point so you don't create a ground loop back to the mains. Honestly, based on the one schematic that you posted, you should be able to probe the HS Fet without using the probe's ground clip at all. Obviously the waveform will have the positive and negative bus voltage riding with it, but you'll at least be able to see the shape of the waveform.
Also, if your power supply is making screeching sounds with the probe connected to the HS, then either the probe is connected improperly and it's causing the transformer to saturate, or the capacitance of the probe is loading the circuit and causing osculations. Usually screeching sounds coming from a transformer is due to saturation.

Also, what type of diodes are you using for your gate discharge diodes (D3 & D4) and your bootstrap diode (D5)?

It could be that you are not loading enough until the voltage starts to decrease. You did not say what is the current drain on the output is. you can try a 100 watt bulb (110v) on the two outputs hooked in series. At 60v from your output does not give so much current I guess it will be around 0.8 amps or so. This will be enough for the first test. You can also try a car bulb rated 24v 21watts and load one end of your double output at a time. These are cheap to buy.

I guess Brad was right in what he said about a bootstrap driven high side, as later yesterday I was reading Tahmid's Blog and the same things more or less was said having an IR2110 driving a regulated output. I guess it would be needing a separate supply so that you can isolate the high side VB pin and ground referenced to the mid point VS. One other thing that I noticed was that Tahmid when using the IR2110 powers the VDD pin 9 with 5v taken from the 12v regulator with a 7805 chip. I don't really know the purpose but its done like this for some reason Maybe Brad can explain why. It may driving this pin separately make things run smoother without much distortion on the output waveform.

Regards Silvio

VDD (Pin 9) is the logic side supply. This can be supplied anywhere between about 3.3V and VCC. The VDD supply voltage basically just sets the voltage threshold for the LIN/HIN pins (On is about 60% of VDD, Off is about 40% of VDD with about a 20% hysteresis in between). VDD can be connected to the VCC supply with no problems.
The easiest explanation I can give is this:
Say you are using an IR2110 to drive a set of IGBTs that require a 15V gate drive voltage, but you are using a 3.3V microcontroller to generate the PWM signals. The VCC pin can be supplied with 15V and the VB supply can be bootstrapped with the same 15 volts which will drive the gates. On the other side, you can supply the VDD pin with the same 3.3V used to power the microcontroller so that the LIN/HIN pins can be properly triggered by the 3.3V outputs of the micro. The main thing to remember when doing this is that the VSS and COM pins should be at the same potential as everything is internally referenced to COM. The datasheet says that VSS may be +/-5V from COM, but I don't see any reason for that and I'd keep the both at the same potential.

 

[Silvio]

@ Blasphemy000 Thanks for the good explanation

Tahmid was stressing that this pin (9) should not be below 4v as he said he found problems, and said that if it is so this supply will better be boosted up to 5v.

 

[brady]

Hello,

Here is the schematic and pcb layout.

View attachment HalfBridgeSMPS_PCB_Rev2.1.pdf
View attachment HalfBridgeSMPS-2D.pdf

I need to mention a couple of things:

1) The auxilliary supply, startup circuit, and current sense circuit are not being used currently.
2) The control circuit is currently built up on a breadboard and interfaces with the header on the top of the PCB.
3) There is no snubber installed

Also, I apologize if the part numbers do not match between the schematic and PCB (I'm sure there are many).

As far as probing, When I probed the gates of the mosfets, I believe the capacitance of the probes was slowing the switch on time due to the gate resistors and affecting the performance. I have the circuit in series with a 250W bulb, and when the probe was connected, the bulb glowed where it did not otherwise.

For the gate discharge diodes, I am using MBR1100 schottky diodes.

On another note, I put the two outputs in series and connected a bulb to get a load of about 15 watts. the mosfets got pretty hot after several minutes. I measured the input power and found it was 85 watts. So that an efficiency of 17%. Any explanation?

Thanks for all the help guys. I'm having lots of issues as you can see and I probably wouldn't have any idea where to go without your advice.

Regards,
Brady

 

[blasphemy000]

@ Brady, I have been reading on the IR2110 on Tahmid Blog and reading further on through the posts it was said by Tahmid that the IR2110 can be driven from 0-100% duty cycle only if you have an isolated supply driving the high side gate.

That is correct. The easiest way to get super low duty-cycles with the IR2110 is using an isolated supply to feed the capacitor between Vb and Vs. With the isolated supply for the HS, you don't have to worry about the bootstrap capacitor not having enough time to charge up.

Lower your gate to source resistors to 1k. You can try lowering also your gate resistors to 4,7 ohms instead of 10. I have seen circuits with IR2110 driving 1RF460 fets in this configuration.

The IR2110 is a +/- 2A driver. With a 15V drive voltage, any gate resistor less than 7.5R will exceed the peak drive current of the IR2110 and can cause heating. Monitor the temperature of the IR2110 when using really low resistance to drive the gates.

Blasphemy000 Just a small question to you. In your opinion what will be the best way to switch the gates in a regulated supply? Is it with a gating trafo? Or there is another way while still using the IR2110?

It really depends on the topology and what conditions the supply will be operated under. If the supply is operated where the frequency and duty-cycles used are within range of what the IR2110 can handle, then I think it is a fairly solid choice. I've personally never had good luck with getting them to work, but I know plenty of people that have used them successfully. The IC obviously can work well, otherwise they wouldn't still be making it.

Gate drive transformers, GDTs, are a good choice and can be very simple to implement effectively. Again though, the driving needs of the topology will determine how easily and effectively a GDT can be implemented.

Opto-isolated gate drivers are also a good choice. Many of these have slow propagation delays and a limited frequency that they can be used at due to the nature of the internal LED/Detector setup. There are advances in these opto-drivers that I've come across recently that use some sort of RF/Modulation inside of the IC, instead of LEDs, to bridge the isolation barrier and trigger the output of the driver. From what I've seen, these have the potential to be faster than the LED type. I'm actually going to be using TI's UCC21520 as the gate drivers in the new design that I'm working on for my welding machine. They appear to be very solid gate drivers from the info that I've found and they should work great with the needs of the supply that I'm building.

 

[blasphemy000]

As far as probing, When I probed the gates of the mosfets, I believe the capacitance of the probes was slowing the switch on time due to the gate resistors and affecting the performance. I have the circuit in series with a 250W bulb, and when the probe was connected, the bulb glowed where it did not otherwise.

That is certainly a possibility, the MOST important question is though, where did you have the ground clip of the 100x probe attached?

For the gate discharge diodes, I am using MBR1100 schottky diodes.

I don't see any issues with these. Schottky rectifiers usually have more capacitance than a regular switching diode, about 50pF for this one with 15V across it. It could cause some oscillations, but I don't think that is the problem here. These should be fine.

On another note, I put the two outputs in series and connected a bulb to get a load of about 15 watts. the mosfets got pretty hot after several minutes. I measured the input power and found it was 85 watts. So that an efficiency of 17%. Any explanation?

The 250W dim bulb you have in series with your input would have been glowing pretty good with 85W being pulled at the input. Remember that due to the way a full-wave bridge draws current from the mains, most power meters don't read accurately due to the low power factor.

The heating of the MOSFETs with only 15W being drawn at the output definitely indicates a problem though. Honestly, other than not having snubbers across the FETs, I really don't see any obvious issues with the driving stage of your supply. Kind of odd.

 

[Silvio]

Brady, did you see the wave form arriving at the HI in and LO in of the ir2110? quite a long path you got there from the sg3525
Try a twisted pair of wires instead of the links to HI in and LO in. Any pick up of stray coupling might cancel out. Use different colors to avoid a mistake. You can also try some ferrite beads treaded in the links instead.
As for the IR2110 I think is positioned well near the gates of the mosfets.
I also noticed a ground trace ending near the trim pot of the current sense which is connected to nothing????
R9 and D4 in my smps are 10 ohms and 1N4148. My switching frequency is 60Khz.

One last thing I like to mention, You did use good magnet wire and mylar tape between the windings in your trafo as if the insulation was old it might crack and having shorted turns.

As always trying to help you solve the problem

regards Silvio

 

[blasphemy000]

I just got around to looking at the board layout.

Silvio: That trace that just ends looks like it should be connected to the GND trace that is connected across the upper two pins of the trim pot. From the looks of how it's laid out, every other trace is a GND trace. I've seen other boards made like this, and flood filling also, as a way to shield other traces nearby from interfering with each other. It's a handy trick for signals with really fast edges, really high frequencies, and high current traces. Have to watch out at very high frequencies though because there will be some capacitance between the traces, but at these frequencies I don't think that would be an issue.

I do agree with checking the signals at the LIN/HIN pins of the IR2110. After looking at the board layout, Silvio is right about the long traces from the X1 pin header across the board to the 2110. I know the SG3525 does have a little bit of drive power to it, but the inputs to the 2110 are high-impedance. So while the switching signals at the X1 header might be good, the stray inductance created from the long traces could be allowing interference at the other end of the traces where they connect to the 2110. The main place that I could see this kind of interference being created is where these signal traces cross at 90* to the high current trace between the DC blocking cap and the current transformer. The current switching through the large trace on the bottom of the board could easily cause disruptions in the signals going through the jumpers on the top side. Again, just an idea because I haven't been able to spot any obvious design flaw.

Brad

 

[brady]

The main place that I could see this kind of interference being created is where these signal traces cross at 90* to the high current trace between the DC blocking cap and the current transformer. The current switching through the large trace on the bottom of the board could easily cause disruptions in the signals going through the jumpers on the top side. Again, just an idea because I haven't been able to spot any obvious design flaw.

Brad

I believe since they are at 90 degrees, the magnetic field will not create currents in the signal jumpers since the flux lines are parallel to the cross section of the 'loop.'

 

[brady]

That is certainly a possibility, the MOST important question is though, where did you have the ground clip of the 100x probe attached?



I don't see any issues with these. Schottky rectifiers usually have more capacitance than a regular switching diode, about 50pF for this one with 15V across it. It could cause some oscillations, but I don't think that is the problem here. These should be fine.




The 250W dim bulb you have in series with your input would have been glowing pretty good with 85W being pulled at the input. Remember that due to the way a full-wave bridge draws current from the mains, most power meters don't read accurately due to the low power factor.

The heating of the MOSFETs with only 15W being drawn at the output definitely indicates a problem though. Honestly, other than not having snubbers across the FETs, I really don't see any obvious issues with the driving stage of your supply. Kind of odd.

I had the probe attached with the ground lead at the drain of the low side mosfet and the probe clipped to the gate of the high side. By the way, I have the supply isolated so I can place the probe anywhere in the circuit.

I could replace the schottky diodes on the gate drives with the SF28 if necessary.

I tried measuring the input current with a shunt resistor but all I got was periodic oscillations and I couldn't make any sense of it. That could be something to do with the input filter but I am not sure. Otherwise I am using a Fluke 87-V to make current measurements.

I had one thought on the possible causes of dissipation: perhaps the body diodes in the mosfets are taking current at switch off?

 

[brady]

Brady, did you see the wave form arriving at the HI in and LO in of the ir2110? quite a long path you got there from the sg3525
Try a twisted pair of wires instead of the links to HI in and LO in. Any pick up of stray coupling might cancel out. Use different colors to avoid a mistake. You can also try some ferrite beads treaded in the links instead.
As for the IR2110 I think is positioned well near the gates of the mosfets.
I also noticed a ground trace ending near the trim pot of the current sense which is connected to nothing????
R9 and D4 in my smps are 10 ohms and 1N4148. My switching frequency is 60Khz.

One last thing I like to mention, You did use good magnet wire and mylar tape between the windings in your trafo as if the insulation was old it might crack and having shorted turns.

As always trying to help you solve the problem

regards Silvio

I may not have any ferrite beads on hand but I do have an ethernet cable I could use to make twisted pairs. I'll for sure try that. I'll post captures of the signals coming from the SG as soon as I can.

For the transformer I purchased new 24 awg enameled copper so it should be okay. I used polymide film tape rated at 7.5 kV dielectric strength between all layers. I probed all coils before using it and no coils were shorted together to my knowledge.

 

[brady]

Hello,

So I did some messing around and discovered that the timing capacitor on the SG was not connected at all!! Turns out that it was probably the cause of the overheating -). The supply seems to be working pretty well now. It still will not function below near maximum duty cycle. I suppose I should start looking into snubbers? with the current 15W load I am using and at full duty cycle, the primary waveform is a near perfect square wave so that doensn't seem to point to the need for a snubber. When I lower the duty cycle it doesn't seem like the gate drives aren't being driven properly. It seems more like the parasitic effects are causing some sort of oscillation. That seems like where a snubber may help but I am unsure.

Regards,
Brady

 

[Silvio]

Hello Brady, Well I guess you didn't check your soldering and did not check out the waveform on your pin header board. This will teach you a lesson that when you do something you have to check things out separately when you can. For instance you could check the dead time frequency wave form etc before hand. All you had to do is plug it in a bread board and load the SG with a couple of LEDs and check things out from there. That's what I did with my 1000 watter and did not burn a single component.

Coming to snubbers you have to load your supply firstly at full duty cycle at around half or more of the rated power so that you will see if you really need them or not, 15 watts is rather a small load. Mine did not show any spikes until I started loading it at 300 watts or more. After that you can lower the output voltage and see the results and maybe you can trim your snubbers further for better results. Wave shapes tend to change at lower duty cycles. I did an overall check with mine and connected the scope across the primary using X 10 probe and 50v/ div (I had separate isolated scope and separate Isolated power supply ok so be careful) That will give a result of the HI and LO side simultaneously.

Good luck and I am happy that you advanced a bit more and manged to find your problem.

Regards,

Silvio

 

[blasphemy000]

Hello,

So I did some messing around and discovered that the timing capacitor on the SG was not connected at all!! Turns out that it was probably the cause of the overheating -). The supply seems to be working pretty well now. It still will not function below near maximum duty cycle. I suppose I should start looking into snubbers? with the current 15W load I am using and at full duty cycle, the primary waveform is a near perfect square wave so that doensn't seem to point to the need for a snubber. When I lower the duty cycle it doesn't seem like the gate drives aren't being driven properly. It seems more like the parasitic effects are causing some sort of oscillation. That seems like where a snubber may help but I am unsure.

Regards,
Brady

The timing cap is a fairly important piece. Lol. At least it turned out to be something simple in that regard.

Snubbers are almost always needed. Your original spec was for 300W (150W + 150W); Silvio is right in that a 15W load isn't nearly enough to evaluate the need for snubbers. As you put a heavier load on the supply, more current will flow through the transformer and other inductive paths, and ringing and spikes can pop up that wouldn't normally be seen at lower powers. Very happy for you that you've figured out the overheating issue. Sometimes it is the simplest of things.

 

[brady]

Hello,

I've done a load test and want to share some results.

Load: 82.03 W (bulb connected to both outputs in series)
Input Real Power = 91.8 W Input Apparent Power = 192.7 VA
Efficiency = 89.4% Power Factor = 0.48

I also found that it is significantly more stable with a varying duty cycle under this load. The only complaint is some transformer noise at lower duty cycles. I took some screen captures of the primary waveform.

View attachment Screen_Captures.pdf

As you can see from the last capture, the ringing frequency is 5.841 MHz.

What other information do I need to come up with values for a snubber? I would like to just use one snubber on the primary instead of one per MOSFET, as that is what I have the board set up for currently.

 

[Silvio]

Hello Brady, good results you got there at last lol. Stability is better at lower duty cycles is due to regulation I guess.

Calculation of snubber values is rather simple when you learn how.

1) Measure the ringing frequency and put a ceramic or film capacitor across the element to be snubbed that brings 3 times the ringing frequency (Ex. 5Mhz becomes 15Mhz)

your series resistor will be as follows R= 1 / 2pie FC where F is the new ringing frequency in Hertz and C is the capacitor you inserted in Farads ( EX 470pf)

EX R = 1 / 15 X 10^6 X 470 X10^-12 = 1/ 15000000 X 0.000,000,000,47 = 1/0.00705 = 142 ohms (Find nearest value possible)

You can either put a snubber across each fet or you can put only one across the primary. Putting 2 snubbers the resistor wattage will be divided. The snubber resistor will be in the region of 2 or 3 watts. This I learnt from experience.

If you want you can experiment with different values of capacitors and resistors to get the best result.

Hope that helps

Regards

Silvio

 

[brady]

Hello Brady, good results you got there at last lol. Stability is better at lower duty cycles is due to regulation I guess.

Calculation of snubber values is rather simple when you learn how.

1) Measure the ringing frequency and put a ceramic or film capacitor across the element to be snubbed that brings 3 times the ringing frequency (Ex. 5Mhz becomes 15Mhz)

your series resistor will be as follows R= 1 / 2pie FC where F is the new ringing frequency in Hertz and C is the capacitor you inserted in Farads ( EX 470pf)

EX R = 1 / 15 X 10^6 X 470 X10^-12 = 1/ 15000000 X 0.000,000,000,47 = 1/0.00705 = 142 ohms (Find nearest value possible)

You can either put a snubber across each fet or you can put only one across the primary. Putting 2 snubbers the resistor wattage will be divided. The snubber resistor will be in the region of 2 or 3 watts. This I learnt from experience.

If you want you can experiment with different values of capacitors and resistors to get the best result.

Hope that helps

Regards

Silvio

Thank you for the information. But before I do that I should probably address another slight issue. With the 82 watt load attached, the outputs drop to around 32v each which is the voltage I need at a full load. And this is at max duty cycle as well. So I believe I will need to add a turn or two to each output to get the output high enough. And I'm sure this will change the parasitics and thus the required snubber. I'm also still trying to figure out the current limiting configuration. I really need this implemented because not having it has already caused me to lose some parts... Please see the attached schematic for the current limit circuit.

View attachment C_Sense_Comparator.pdf.pdf

 

[Silvio]

Hello Brady, regarding voltage drop depends on two main factors really it is the B that the trafo is running at as the higher it is the more regulation you get.
Do not misunderstand me as it is all a compromise as there is a way to calculate the the B max for certain frequencies and not every frequency can handle the a high B. My observations using Excellent IT software I noticed that say around 20Khz I saw B max rising to 2300 gauss or more but say at 60Khz it was not more than 1800. Personally I don't know how to choose it. I just compare things and get there.

The other issues with voltage drop mainly is the first 10% is very normal as the voltage rises to the peak with no load. When I do tests and calculate voltage drop I take it in account. On heavy loads voltage drop can go up to 15-19%. remember that the voltage will sag into the ripple and drops accordingly. These drops I noted where in the bulk caps, copper losses, switching losses, core losses, output rectifier output inductor loss and leakage inductance. These all add up together. As a rule of thumb I will take it as 15% plus the nominal maximum voltage wanted on your lowest mains input voltage. I will explain better:- if say your max output voltage at max duty cycle at lowest input volt be 32 volts then calculate plus 15% here. I don't exactly know what is your minimum input voltage calculated but do not exaggerate in this as the mains input rarely falls below 6% of the nominal. Be practical here. You know your mains voltage at home and also know its fluctuations so take your assumptions from there.

CURRENT SENSE. I am not so keen on circuits but I try things the way I know how. What I did on mine was that I loaded slowly bit by bit and adjusted according to my findings until I got the cut off current limit where I wanted it give and take an amp or so. This has to be quick as possible as in the event of short circuit which is very likely in a bench power supply has to save the fets from over current as they will blow in the blink of an eye. It is necessary that the trigger will have a very short rise time.

What I saw recently that someone built an smps with an sg3525 and using a current trafo then loading the output the usual way a and triggering the base of a small npn transistor which then pulls to ground the soft start pin. From what I understood here is as current starts to rise near the threshold the transistor starts pulling this pin to ground and lowering the duty cycle thus the current as well. I think shorting this pin totally to ground will switch off the sg altogether. I still have to try it out in my next smps.

Best regards

Silvio

 

[brady]

Hello Brady, regarding voltage drop depends on two main factors really it is the B that the trafo is running at as the higher it is the more regulation you get.
Do not misunderstand me as it is all a compromise as there is a way to calculate the the B max for certain frequencies and not every frequency can handle the a high B. My observations using Excellent IT software I noticed that say around 20Khz I saw B max rising to 2300 gauss or more but say at 60Khz it was not more than 1800. Personally I don't know how to choose it. I just compare things and get there.

The other issues with voltage drop mainly is the first 10% is very normal as the voltage rises to the peak with no load. When I do tests and calculate voltage drop I take it in account. On heavy loads voltage drop can go up to 15-19%. remember that the voltage will sag into the ripple and drops accordingly. These drops I noted where in the bulk caps, copper losses, switching losses, core losses, output rectifier output inductor loss and leakage inductance. These all add up together. As a rule of thumb I will take it as 15% plus the nominal maximum voltage wanted on your lowest mains input voltage. I will explain better:- if say your max output voltage at max duty cycle at lowest input volt be 32 volts then calculate plus 15% here. I don't exactly know what is your minimum input voltage calculated but do not exaggerate in this as the mains input rarely falls below 6% of the nominal. Be practical here. You know your mains voltage at home and also know its fluctuations so take your assumptions from there.

CURRENT SENSE. I am not so keen on circuits but I try things the way I know how. What I did on mine was that I loaded slowly bit by bit and adjusted according to my findings until I got the cut off current limit where I wanted it give and take an amp or so. This has to be quick as possible as in the event of short circuit which is very likely in a bench power supply has to save the fets from over current as they will blow in the blink of an eye. It is necessary that the trigger will have a very short rise time.

What I saw recently that someone built an smps with an sg3525 and using a current trafo then loading the output the usual way a and triggering the base of a small npn transistor which then pulls to ground the soft start pin. From what I understood here is as current starts to rise near the threshold the transistor starts pulling this pin to ground and lowering the duty cycle thus the current as well. I think shorting this pin totally to ground will switch off the sg altogether. I still have to try it out in my next smps.

Best regards

Silvio

Every time I have measured the current sense pin on the header I get a square wave-like signal with a lot of ringing. I get a characteristic ramp only when measuring directly across the resistor before the bridge. I think the diodes I am using may be some of the problem because they are schottky diodes, thus high capacitance. I will change them to ultra fast silicon diodes and try that. I also will add an RC filter/divider like you used in your 1000W supply to reduce high frequency noise. The comparator circuit I showed in the previous post functions by tripping at one point and resetting at a different threshold below the initial trip point. I believe this should perform as a pulse-by-pulse limit and shouldn't over-current the mosfets as long as the transformer doesn't saturate before the delay of the shutdown signal lapses. If it doesn't work in a pulse-by-pulse fashion it will at least run in a discontinuous mode and self recover once the over-current condition is removed.

Regards,
Brady