Alme,
I'm interested in how you achieved 1600W with a similliar sized torroidal core. Could you post a schematic or answer these questions.:
What is your switching frequency?
How much dead time do you have?
How many primary / secondary turns do you have?
What gauge wire are your primary and secondary?
What value resistance did you use from your controller IC to gate?
How many Mosfets and what type did you use for the xformer primary?
I'm curious as to why I'm struggling for 850 watts while your able to achieve almost twice the power I am? Could it be that I'm just not pushing it as hard as I can? I notice you allowed yours to run much hotter than mine 120degC vs 67degC. If pout vs pdiss is a somewhat linear relationship I could probably go for 1400-1500 watts. Before I try it I want to know what you guys think. I dont want to blow the MOSFETS from overload if not possible because it sucks replacing them!!
Thanks,
Randy
I'm interested in how you achieved 1600W with a similliar sized torroidal core. Could you post a schematic or answer these questions.:
What is your switching frequency?
How much dead time do you have?
How many primary / secondary turns do you have?
What gauge wire are your primary and secondary?
What value resistance did you use from your controller IC to gate?
How many Mosfets and what type did you use for the xformer primary?
I'm curious as to why I'm struggling for 850 watts while your able to achieve almost twice the power I am? Could it be that I'm just not pushing it as hard as I can? I notice you allowed yours to run much hotter than mine 120degC vs 67degC. If pout vs pdiss is a somewhat linear relationship I could probably go for 1400-1500 watts. Before I try it I want to know what you guys think. I dont want to blow the MOSFETS from overload if not possible because it sucks replacing them!!
Thanks,
Randy
Randy, supposing room temperature is 20degC, my transformer overheat is about 100degC, and yours is 47degC, that means you can try to pump about double power thru it to reach same overheat. Just make sure that you use sufficient number of FETs - plainly saying, their total maximum current rating should be 3-4 times higher than maximum primary current consumption.
My converter is quite conventional, using TL494 and standard diode-PNP follower to drive gates. Switching frequency is 27kHz, deadtime about 1-1,5us, primary 5 turns for car battery, secondary is different depending on application (usually 15-22 turns). Primary copper cross-section is about 12 sq.mm (15xD1,0mm), secondary according to transformer ratio. Gate resistors are 22ohm. MOSFETs are IRFZ48Nx10pcs. (5pcs. each side). There's really nothing special.
My converter is quite conventional, using TL494 and standard diode-PNP follower to drive gates. Switching frequency is 27kHz, deadtime about 1-1,5us, primary 5 turns for car battery, secondary is different depending on application (usually 15-22 turns). Primary copper cross-section is about 12 sq.mm (15xD1,0mm), secondary according to transformer ratio. Gate resistors are 22ohm. MOSFETs are IRFZ48Nx10pcs. (5pcs. each side). There's really nothing special.
Randy,
I had the same problem with burnt mosfets (my last layout gave me a lot of problems), so I implemented the next simple circuit:
Rsense can be your ground track (it has minimal resistance, but enough to develop a voltage drop which can be easily measured with your oscilloscope). You can amplify that voltage (if you need it) with an LM324 ( it has very limited slew rate but the upper current waveform it’s still well amplified and that’s what we need to monitor). That way I have never burnt a mosfet again, even in a long short circuit condition (I shorted the transformer’s secondary by mistake) the transistors survived (they only get little warm).
I hope this help you
Juan Carlos
I had the same problem with burnt mosfets (my last layout gave me a lot of problems), so I implemented the next simple circuit:
An externally hosted image should be here but it was not working when we last tested it.
Rsense can be your ground track (it has minimal resistance, but enough to develop a voltage drop which can be easily measured with your oscilloscope). You can amplify that voltage (if you need it) with an LM324 ( it has very limited slew rate but the upper current waveform it’s still well amplified and that’s what we need to monitor). That way I have never burnt a mosfet again, even in a long short circuit condition (I shorted the transformer’s secondary by mistake) the transistors survived (they only get little warm).
I hope this help you
Juan Carlos
Primary side current limit
Mr Carlos;
Your suggestion is very good. If you replace Q1 with an NCP100 or TLV431 (programmable low referenve voltage shunt regulators) it is possible to build a high gain protection switch circuit with a precision threshold so that temperature dependency of the overload limit is eliminated.
The gain of either shunt regulator device will also make the switch point very fast. Please beware if you use my idea that both these devices require a minimum cathode current. This current is used to bias the internal voltage reference. Also note that the device will be equally fast in restoring current once the value drops below the trip threshold.
Mr Carlos;
Your suggestion is very good. If you replace Q1 with an NCP100 or TLV431 (programmable low referenve voltage shunt regulators) it is possible to build a high gain protection switch circuit with a precision threshold so that temperature dependency of the overload limit is eliminated.
The gain of either shunt regulator device will also make the switch point very fast. Please beware if you use my idea that both these devices require a minimum cathode current. This current is used to bias the internal voltage reference. Also note that the device will be equally fast in restoring current once the value drops below the trip threshold.
Reference Power
Hi Eva;
I don't know why, at this power level I had jumped to the conclusion that this was an "off line" supply", where any low voltage loading causes the regulator chip to heat up fast due to the extreme voltage drop from the rectified mains voltage to the low voltage reference pin.
I went to the front of the thread; you are exactly right borrowing some power is perfectly reasonable
.
The comparator offers more flexibility and few surprises, I support your recommendation for a comparator. As you pointed out there is a bonus, the sense resistor can be made much smaller because there is no need to develop that much emitter or source error voltage. This reduces total wasted power and increases efficiency.
If you ever need this function where there is no available "spare" DC the shunt regulator trick works OK within its own limitations.
Hi Eva;
I don't know why, at this power level I had jumped to the conclusion that this was an "off line" supply", where any low voltage loading causes the regulator chip to heat up fast due to the extreme voltage drop from the rectified mains voltage to the low voltage reference pin.
I went to the front of the thread; you are exactly right borrowing some power is perfectly reasonable
.The comparator offers more flexibility and few surprises, I support your recommendation for a comparator. As you pointed out there is a bonus, the sense resistor can be made much smaller because there is no need to develop that much emitter or source error voltage. This reduces total wasted power and increases efficiency.
If you ever need this function where there is no available "spare" DC the shunt regulator trick works OK within its own limitations.
If the talk switched now to overcurrent protection (OCP), I want to remind some solution which I find the most simple and efficient way to implement OCP in high-current circuit with resistive current sensor.
I attach its picture. Its operation based on p-n junction physical effect.
I've successfully used this thing for over 4 years in my designs.
There must be matched pair of Q1+Q2 transistors on the same die, otherwise it loses sense.
Q3 is any type of same polarity.
Threshold point is defined like
I=(Et/R3)*ln(R2/R1), where Et=25mV at room temperature.
For example, if R3 shunt is 0.001ohm, R2=100kohm and R1=20kohm, then I=40A, shunt voltage drop is 40mV at this point, its power loss is 1,6W maximum.
I've found difficult to sense such low shunt drop of 40mV with LM393 comparator because comparator was more sensitive to voltage spikes which are produced by shunt parasitic inductance during converter switching operation. Transistor sensor proved itself as better choice for it.
There's no difference in which rail to put shunt: if needed into plus rail, then just to change all transistors for PNP type.
Disadvantage of this sensor is temperature dependent threshold. But it's always possible to provide some current margin for wide temperature range because OCP main function is just to help all design survive during overload.
I've also used such sensors in simple Pb-acid battery chargers with constant current/constant voltage charging curve.
I attach its picture. Its operation based on p-n junction physical effect.
I've successfully used this thing for over 4 years in my designs.
There must be matched pair of Q1+Q2 transistors on the same die, otherwise it loses sense.
Q3 is any type of same polarity.
Threshold point is defined like
I=(Et/R3)*ln(R2/R1), where Et=25mV at room temperature.
For example, if R3 shunt is 0.001ohm, R2=100kohm and R1=20kohm, then I=40A, shunt voltage drop is 40mV at this point, its power loss is 1,6W maximum.
I've found difficult to sense such low shunt drop of 40mV with LM393 comparator because comparator was more sensitive to voltage spikes which are produced by shunt parasitic inductance during converter switching operation. Transistor sensor proved itself as better choice for it.
There's no difference in which rail to put shunt: if needed into plus rail, then just to change all transistors for PNP type.
Disadvantage of this sensor is temperature dependent threshold. But it's always possible to provide some current margin for wide temperature range because OCP main function is just to help all design survive during overload.
I've also used such sensors in simple Pb-acid battery chargers with constant current/constant voltage charging curve.
Attachments
Yes, a positive feedback is usually added into comparator circuit to help improve response time or make latching effect.
Looks like it is basically possible to add hysteresis into transistor sensor as well; I didn't try it though because used such sensor for current limiting type OCP, not latching shutdown. I think series diode plus resistor connected from Q3 collector to Q1/Q2 base can do such job. When thresholding, feedback resistor would be connected in parallel to R2, thus decreasing R2/R1 ratio. Certainly there also must be Q3 pull-up to allow this, actually pull-up will take part in R2 paralleling action, but feedback resistor is going to be of much higher value than pull-up (say, 470k vs. 10k).
Looks like it is basically possible to add hysteresis into transistor sensor as well; I didn't try it though because used such sensor for current limiting type OCP, not latching shutdown. I think series diode plus resistor connected from Q3 collector to Q1/Q2 base can do such job. When thresholding, feedback resistor would be connected in parallel to R2, thus decreasing R2/R1 ratio. Certainly there also must be Q3 pull-up to allow this, actually pull-up will take part in R2 paralleling action, but feedback resistor is going to be of much higher value than pull-up (say, 470k vs. 10k).
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.