I've been looking at some rectifier characteristics during our snowstorn back here in NJ, and have made a couple of measurements and have a couple of ideas. I like fast rectifiers for amplifier power supplies, and the Vishay HFA25PB60 devices look quite good, combining ultra-fast recovery (23 ns) with a fairly soft recovery. They also appear to have less forward drop at high peak currents than many other FREDs.
But here is what I have been thinking. As a big power MOSFET fan, I know that every one has a fast body diode in it. This body diode is imporatnt for switching supply applications, and usually breaks down at the rated voltage of the MOSFET. They are inherent to the structure of the HEXFET and must be able to handle high peak currents in both directions in switching applications. These applications include Buck converters and class D amplifiers. Fast recovery is very important in these applications.
My thought was, why not use the body diode of an IRFP240 as the rectifier in a bridge. IRFP240's are pretty cheap these days and easy to get. My concern was forward voltage drop under high rectifier spike currents. So I built a single-rectifier half-wave rectifier driven by a 200VA toroid and filtered it with a good 15,000 uF reservoir capacitor. I also put 0.1 ohm in series with the diode so I could look at diode current on the scope. I loaded the capacitor with 16 ohms and the power supply voltage dropped down to about 52V, so continuous DC load current was 3.25 amps. I compared the performance of the IRFP240 diode with that of a diode of a conventional 35A bridge (TB354).
Rectifier performance was indistinguishable, with the IRFP240 body diode showing no more forward drop that the big old bridge diode (both about 1.3V peak at peak current of 15A). This suggested to me that four IRFP240 body diodes would perform just as well as a slow conventional 35A bridge and would have the benefit of less than 80 ns Trr. I was expecting a bigger peak voltage drop from the IRFP240 body diode, which I believe is internally of FRED construction. Has anybody else tried this? Any down-side I have overlooked? Is the cost savings or availability enough to make it worthwhile?
During the same experiments I looked carefully at the voltage drop across the 0.1 ohm series resistor. This is how I measured the peak rectifier current. The 0.1 ohm resistor was built with five 0.47 ohm 2W metal oxide resistors in parallel so it would have low inductance.
Here is where my disappointment came in. I was expecting to see visible reverse recovery current spikes across the 0.1 ohm sense resistor. I was looking to see them much worse on the conventional bridge, which probably has Trr on the order of 350 ns. I saw no eveidence of reverse recovery spikes with either rectifier. Should I have? Something wrong with my setup? Scope was a 100 MHz Tek. OR, is it the case that the dv/dt of the 60 Hz AC driving waveform is too slow to really create reverse recovery effects? If that is so, then that raises questions about the need for speed in power supply rectifiers. What's the deal?
I am not asserting that we should not be using fast diodes in the rectifiers, but I'm wondering why I don't seem to see a measurable difference in diode current behavior.
Cheers,
Bob
But here is what I have been thinking. As a big power MOSFET fan, I know that every one has a fast body diode in it. This body diode is imporatnt for switching supply applications, and usually breaks down at the rated voltage of the MOSFET. They are inherent to the structure of the HEXFET and must be able to handle high peak currents in both directions in switching applications. These applications include Buck converters and class D amplifiers. Fast recovery is very important in these applications.
My thought was, why not use the body diode of an IRFP240 as the rectifier in a bridge. IRFP240's are pretty cheap these days and easy to get. My concern was forward voltage drop under high rectifier spike currents. So I built a single-rectifier half-wave rectifier driven by a 200VA toroid and filtered it with a good 15,000 uF reservoir capacitor. I also put 0.1 ohm in series with the diode so I could look at diode current on the scope. I loaded the capacitor with 16 ohms and the power supply voltage dropped down to about 52V, so continuous DC load current was 3.25 amps. I compared the performance of the IRFP240 diode with that of a diode of a conventional 35A bridge (TB354).
Rectifier performance was indistinguishable, with the IRFP240 body diode showing no more forward drop that the big old bridge diode (both about 1.3V peak at peak current of 15A). This suggested to me that four IRFP240 body diodes would perform just as well as a slow conventional 35A bridge and would have the benefit of less than 80 ns Trr. I was expecting a bigger peak voltage drop from the IRFP240 body diode, which I believe is internally of FRED construction. Has anybody else tried this? Any down-side I have overlooked? Is the cost savings or availability enough to make it worthwhile?
During the same experiments I looked carefully at the voltage drop across the 0.1 ohm series resistor. This is how I measured the peak rectifier current. The 0.1 ohm resistor was built with five 0.47 ohm 2W metal oxide resistors in parallel so it would have low inductance.
Here is where my disappointment came in. I was expecting to see visible reverse recovery current spikes across the 0.1 ohm sense resistor. I was looking to see them much worse on the conventional bridge, which probably has Trr on the order of 350 ns. I saw no eveidence of reverse recovery spikes with either rectifier. Should I have? Something wrong with my setup? Scope was a 100 MHz Tek. OR, is it the case that the dv/dt of the 60 Hz AC driving waveform is too slow to really create reverse recovery effects? If that is so, then that raises questions about the need for speed in power supply rectifiers. What's the deal?
I am not asserting that we should not be using fast diodes in the rectifiers, but I'm wondering why I don't seem to see a measurable difference in diode current behavior.
Cheers,
Bob
Looking on Cathode or Anode? The resistor is damping the effect of parasitics L.
Using a current probe would be better ie Tek hall effect, that way you could look at both anode and cathode. Current transformer w/b 2nd best.
Also a spectrum analzer would be useful for looking at HF hash
HEXFRED isn't that using a HEXFET body diode see www.irf.com/technical-info/appnotes/an-989.pdf
Theory says that using Schottky (a majority carrier device) would negate the need for snubbers, but not to be in the world of SMPS. see APP note here
Using a current probe would be better ie Tek hall effect, that way you could look at both anode and cathode. Current transformer w/b 2nd best.
Also a spectrum analzer would be useful for looking at HF hash
HEXFRED isn't that using a HEXFET body diode see www.irf.com/technical-info/appnotes/an-989.pdf
Theory says that using Schottky (a majority carrier device) would negate the need for snubbers, but not to be in the world of SMPS. see APP note here
Last edited:
We have over 2 feet here -- about 10mi west of EWR.
I measured the capacitance from drain to source on a Fairchild FQA19N20C (like an IRFP240) and it's about 1.3nF.
I measured the capacitance from drain to source on a Fairchild FQA19N20C (like an IRFP240) and it's about 1.3nF.
Thanks for sharing that app note. I did not say that I thought the body diode of a HEXFET was a HEXFRED. I siad that I thought it was a FRED structure, which means that it is a fast recovery epitaxial diode.
In my test setup, the secondary of the toroid goes to the anode of the diode. The cathode of the diode connects to the 0.1 ohm sense resistor, which then connects to the reservoir capacitor. The negative side of the reservoir capacitor returns to the other end of the transformer secondary.
The cathode of the diode is taken as the ground connection to the scope. The rest of the secondary circuit floats. The scope probe then goes to the positive side of the capacitor if I want to measure diode current via the sense resistor. the scope probe instead goes to the anode of the diode if I want to measure diode drop. I use a series resistor and a small diode clamp on the latter connection to limit the reverse voltage that the scope sees so that I can use a more sensitive vertical setting.
Cheers,
Bob
IMO It's not the Trr nor the magnitude of Ir that matters for 60Hz rectifiers it's the snap off or the steepness of dI/dt during turnoff that excites the parasitic or leakage inductance which when undamped that causes EMI/RFI or hash.
So what if Trr and Ir is much bigger on plain old PN rectifiers as long as it has good Vf and it is quiet.
So what if Trr and Ir is much bigger on plain old PN rectifiers as long as it has good Vf and it is quiet.
No, I don't think so. They say that they use a proprietary minority carrier lifetime control doping in the HEXFREDs. Maybe something like the old gold doping that they used to use in TTL logic before they days of Shottky TTL (yes, I'm afraid that I am that old). Actually, the first logic IC's I used were RTL. The first transistors I used were Germanium 2N107 and CK722.
We've gone from microseconds to nanoseconds to picoseconds in my lifetime.
Time flies, and so does technology.
Cheers,
Bob
Hi Bob
First, IRFP240 and the likes are known to have a poor reverse recovery properties compared to something more modern, FDP2572 for instance.
To address the question one needs to rapidly brake the current flow in the diode to see vividly the reverse recovery phenomenon.
In contrast, in the 60Hz diode-capacitor supply the current ceases to flow relatively slow, so the reverse recovery takes more time at very little amount of current, it's RR charge which is constant anyway, not RR time.
Also take into account, that transformer leakage inductance will not allow for any rapid variations in current and the current has no other way to flow but the rectifier diode.
Regards,
Adam
First, IRFP240 and the likes are known to have a poor reverse recovery properties compared to something more modern, FDP2572 for instance.
To address the question one needs to rapidly brake the current flow in the diode to see vividly the reverse recovery phenomenon.
In contrast, in the 60Hz diode-capacitor supply the current ceases to flow relatively slow, so the reverse recovery takes more time at very little amount of current, it's RR charge which is constant anyway, not RR time.
Also take into account, that transformer leakage inductance will not allow for any rapid variations in current and the current has no other way to flow but the rectifier diode.
Regards,
Adam
Soooo, are you suggesting that all this stuff about needing fast recovery rectifiers in the power amp supply is mis-guided? I understand your argument about the speeds involved in 60 Hz supplies, and there seems to be some sense to it. Indeed, I have also looked in the lab for reverse recovery spikes in a test power supply (see an earlier post of mine), and could not see any.
Cheers,
Bob
I like the approach of "doesn't matter until proven matters" 🙂
Seriously the reverse recovery may have some secondary impact on ringings, but is not an issue itself, because even some dosens of nano coulombs 120 times per second results in several microamperes reverse DC. As simple as it gets. Also, commutation time is mostly affected by reservoir capacitance and spread inductances of a transformer and mains line.
Seriously the reverse recovery may have some secondary impact on ringings, but is not an issue itself, because even some dosens of nano coulombs 120 times per second results in several microamperes reverse DC. As simple as it gets. Also, commutation time is mostly affected by reservoir capacitance and spread inductances of a transformer and mains line.
Attachments
- Status
- Not open for further replies.