There are also these interesting senseFET devices that look promising:
http://www.nxp.com/documents/application_note/AN10322.pdf
http://www.nxp.com/documents/data_sheet/BUK7C10-75AITE.pdf
These devices have a 75V/114A/10mohm rating and a SOA that allows 250A/10us at 60V Vds
Will be looking into these....
http://www.nxp.com/documents/application_note/AN10322.pdf
http://www.nxp.com/documents/data_sheet/BUK7C10-75AITE.pdf
These devices have a 75V/114A/10mohm rating and a SOA that allows 250A/10us at 60V Vds
Will be looking into these....
Last edited:
Well, farnell USA doesn't have buk-7905 nxp's.
As far as quick rail FET shutoff , I think I'll parallel the photovoltaic optoisolator (FET driver) to the gate of the FET shutting off the rails , with a regular optoisolator driven by the other side of the 74hc74 flip flop that remembers there was a fault. Thus a 4n29 conventional optoisolator turns on gate to source when the FET driver APV1122 shuts off the drive. This should cut the time to eliminate the gate charge by 100:1, while maintaining the lack of a special power supply above the upper rail voltage to control the shutoff FET. I've got a batch of 4N29 laying around with not enough CTR (current transfer ratio) to be useful for most things. A 3 ma current discharging the 3900 pf gate capacitor should finish in a hurry. I can just put the input to the 4N29 in series with the red fault LED I've already installed, and cut the resistor size from the driving transistor. I somehow think cutting the time of shutoff of a ~1000 amp fault current would be important to FET survival, without any good way to test it.
I thought of Hall effect sensors for sensing the speaker line, but am vague how they would handle DC fault current (failed output transistor) versus AC fault currents (1/4 phone plug part way out) ie would the same hall effect sensor sense both faults? Newark does have the ACS712 hall effect sensor in 20 A and 30 A sizes, but you have to buy a reel, and they are surface mount, which is a problem for me. My amplifier is good to 22.5 amp. I scanned the datasheet, ??? Peavey uses Current sense transformers in the protection circuit on the speaker line in the CS800s, and I have some of those salvage from VFD motor drives, but that doesn't work on DC faults.
As far as quick rail FET shutoff , I think I'll parallel the photovoltaic optoisolator (FET driver) to the gate of the FET shutting off the rails , with a regular optoisolator driven by the other side of the 74hc74 flip flop that remembers there was a fault. Thus a 4n29 conventional optoisolator turns on gate to source when the FET driver APV1122 shuts off the drive. This should cut the time to eliminate the gate charge by 100:1, while maintaining the lack of a special power supply above the upper rail voltage to control the shutoff FET. I've got a batch of 4N29 laying around with not enough CTR (current transfer ratio) to be useful for most things. A 3 ma current discharging the 3900 pf gate capacitor should finish in a hurry. I can just put the input to the 4N29 in series with the red fault LED I've already installed, and cut the resistor size from the driving transistor. I somehow think cutting the time of shutoff of a ~1000 amp fault current would be important to FET survival, without any good way to test it.
I thought of Hall effect sensors for sensing the speaker line, but am vague how they would handle DC fault current (failed output transistor) versus AC fault currents (1/4 phone plug part way out) ie would the same hall effect sensor sense both faults? Newark does have the ACS712 hall effect sensor in 20 A and 30 A sizes, but you have to buy a reel, and they are surface mount, which is a problem for me. My amplifier is good to 22.5 amp. I scanned the datasheet, ??? Peavey uses Current sense transformers in the protection circuit on the speaker line in the CS800s, and I have some of those salvage from VFD motor drives, but that doesn't work on DC faults.
Last edited:
Ciao Jason.
Look at Allen Newby's work.
http://www.diyaudio.com/forums/pass-labs/96737-fail-safe-power-supply.html
Look at Allen Newby's work.
http://www.diyaudio.com/forums/pass-labs/96737-fail-safe-power-supply.html
Looks like Mr Newby is using the FET as a linear regulator, which is not very efficient and limited to low power amps, not 1300 watt ones (25 amp rails) like I am working on.
Reading through the FDP52N20 nfet datasheet (I bought these), they produce 1.5 v DrainSource at 52 A. That would mean they produce about 1.2 v @ 40 A which is just a bit over where I want the FET to trip the power rails off, 30 A. 1.2 v is the conduct voltage of an optoisolator input. The output of this optoisolator could go in parallel to the open collectors of the other optoisolators that pull down the Set input of the flip flop that remembers I had a fault, back at ground level.
So without a separate power supply above the rail voltage, you could roughly detect over current and shut off both rail FETs.
The problem that I see, is that I can't limit the current to the input of the overcurrent detect optoisolator after the rail FET shuts off and is holding off 85 V. I'm using a J111 JFET to limit current to the optoisolators that sense the DC detector going off, but those will only limit current to Igss of 20 ma up to about 30 v.
Jason suggested TO252 packaged NXP FET transistors, and somebody on the other page suggested surface mount package only Allegro Hall effect sensors. How do you guys handle surface mount? Do you own wave soldering machines? Do you etch your own PCB's? I have a terrible time getting the solder not to bridge between legs of a DIP with 0.1" spacing, and have been using $2 DIP project boards from mcmelectronics.com to package my through hole DIP projects. The surface mount to dip converter boards from the big distributors are about $7 a package, which is way beyond my budget for this hobby.
Reading through the FDP52N20 nfet datasheet (I bought these), they produce 1.5 v DrainSource at 52 A. That would mean they produce about 1.2 v @ 40 A which is just a bit over where I want the FET to trip the power rails off, 30 A. 1.2 v is the conduct voltage of an optoisolator input. The output of this optoisolator could go in parallel to the open collectors of the other optoisolators that pull down the Set input of the flip flop that remembers I had a fault, back at ground level.
So without a separate power supply above the rail voltage, you could roughly detect over current and shut off both rail FETs.
The problem that I see, is that I can't limit the current to the input of the overcurrent detect optoisolator after the rail FET shuts off and is holding off 85 V. I'm using a J111 JFET to limit current to the optoisolators that sense the DC detector going off, but those will only limit current to Igss of 20 ma up to about 30 v.
Jason suggested TO252 packaged NXP FET transistors, and somebody on the other page suggested surface mount package only Allegro Hall effect sensors. How do you guys handle surface mount? Do you own wave soldering machines? Do you etch your own PCB's? I have a terrible time getting the solder not to bridge between legs of a DIP with 0.1" spacing, and have been using $2 DIP project boards from mcmelectronics.com to package my through hole DIP projects. The surface mount to dip converter boards from the big distributors are about $7 a package, which is way beyond my budget for this hobby.
That's the major issue I have at the moment - the high Vdiff when the FETs are OFF. This means (as you have pointed out) that the detector, along with sensing the mV when operational, also has to handle the rail voltage when the FET is OFF. The other way is to have a sense resistor separate to the FET, but that starts to make a mockery of designing a low impedance power supply.
If you do SMT (I do a LOT of hand SMT - including 144 TQFP's etc.) make sure you have a flux pen. This is invaluable and you cannot properly do SMT without one.
I have used a PIC based solution a few times.
The PIC controls a relay and hold it off for 4 seconds on power up.
I used the fact that a PIC input changes state at around 2 volts to monitor the output signal. If it stays above 10 volts or below -10 volts for 20ms I shut off the relay.
This wont save the amp but it does save speakers. Must have saved a fortune on speakers when testing new designs.
The PIC controls a relay and hold it off for 4 seconds on power up.
I used the fact that a PIC input changes state at around 2 volts to monitor the output signal. If it stays above 10 volts or below -10 volts for 20ms I shut off the relay.
This wont save the amp but it does save speakers. Must have saved a fortune on speakers when testing new designs.
One nice simple approach noted in this thread is the BJT base-emitter across the FET. This would be great if the Vbe was less than 0.6V. For a 10mohm FET, I = 60A. Still have the OFF Vds to deal with as well.
I've got DC detection and rail shutoff, I think, built totally in through hole parts under $2 each. http://www.diyaudio.com/forums/pa-systems/206058-peavey-1-3k-versus-pv-2000-a-4.html post 39.
My turn on delay is a resistor, ceramic capacitor, and the reset input of a 74HC74 flip flop.
I'd like to see a microprocessor operating at 10 volt below a 170 Vpeak flying rail. The cool thing about the photovoltaic FET drivers, they don't need a power supply above postive rail.
The reason for the complications in my DC detect schematic in the link is the PV-1.3k has flying ground which is different for the two channels if not paralleled.
This thread is about early detection of massive speaker current before the $5 output transistors start popping. The goal is to shutoff the same rail FET's.
JasonP, if you use a slightly different rail FET you could get to 0.6 v trip across it to energize a transistor at lower than 60 amps. I'm pretty stuck at FET selection to .05 ohm, my PV-1.3k rails have to go to 170 v normally, and FETs at that voltage (200 V) are limited to .05 ohm for TO220 (FDP52n20) and .08 ohm if I want a TO3-P package.
I'm looking for a PTCR (positive temperature coefficient resistor) that allows a rail FET to drive an optoisolator at > 10 ma for shutoff trip @ 1.2 V, but <30 ma steady state after the FET shuts off to 170 VDC (if OT is shorted). PTCR "fuses" would be okay, the amp has to be turned off to reset this circuit, but most of PTCR fuses are for currents bigger than 30 ma and voltages under 72 v.
Last edited:
There is no reason the PIC circuit couldn't be modified to break the power rails in the event of a fault. It just needs double pole relay added to break both supply rails.
You could keep the speaker relay too for power up hold off.
You could keep the speaker relay too for power up hold off.
An externally hosted image should be here but it was not working when we last tested it.
Nigel,
I am actually using NXP LPC89Cxxx parts, one for each amp side (2) and also one for the main power supply. They talk to each other over I2C. These devices have comparators/ADC and some have DAC's.
If an amp uC detects DC or over-current, it de-activates the speaker relay and can tell the main PS to shut down. At the moment, I have a speaker relay, but was trying to eliminate it and instead have fault conditions turn the whole amp "side" off via FET rail switches .
I am now re-thought detecting rail current and instead use dual slope SOA in the OPS. This is because detecting current in the rail FET's will not be enough to protect the OPS from reactive loads - since current in a rail may not represent the true Pd in the OPS.
You can then use the SOAp(rotect) to directly control the OPS, and/or use it to control the rails in the event of over-current. Having both options available to use if needed is probably better - especially if the OPS blows shorted. This can also remove the need for the speaker relay.
I am actually using NXP LPC89Cxxx parts, one for each amp side (2) and also one for the main power supply. They talk to each other over I2C. These devices have comparators/ADC and some have DAC's.
If an amp uC detects DC or over-current, it de-activates the speaker relay and can tell the main PS to shut down. At the moment, I have a speaker relay, but was trying to eliminate it and instead have fault conditions turn the whole amp "side" off via FET rail switches .
I am now re-thought detecting rail current and instead use dual slope SOA in the OPS. This is because detecting current in the rail FET's will not be enough to protect the OPS from reactive loads - since current in a rail may not represent the true Pd in the OPS.
You can then use the SOAp(rotect) to directly control the OPS, and/or use it to control the rails in the event of over-current. Having both options available to use if needed is probably better - especially if the OPS blows shorted. This can also remove the need for the speaker relay.
With possible 1200 amp short circuit current, (85 v / .02 ohm equiv resistance of capacitor) I gave up on relays, also. Anything that can break that current, ie huge copper contacts, is not going to be sonically invisible at 1/8 watt. I'm designing rail shutoff FETs driven by photovoltaic optoisolators like R. G. Keen suggested.
Jason, you can detect 0.3 v across the rail FET with a germanium transistor. They still sell them in the USA @ Vox Organ Germanium Transistors or RS in the UK has them.
That way you could detect maybe 30 amp on your 0.1 ohm rail fet, without fuses.
I drew up a 1n278 circuit because 1. I have some and 2. they will take 100 ma on the base line. The smallest PTCR fuse I found was a 150 ma trip current one from Murata, a PTGL04AS560K6B5180 that has cold resistance of 56 ohms. In series with a 47 ohm resistor, this could put significant current in the base lead of a NPN Ge transistor when overcurrent occurs. Then the Ge transistor collector can energize an optoisolator to cause pull down of the fault current set pin back at speaker ground, I'm using a 74hc74 flip flop. A resistor-zener arrangement between speaker ground and minus rail is necessary to provide limited current to the optoisolator diode driven by the Ge transistor collector.
With the rail FET off, the voltage on the base lead of the Ge transistor builds to 85, the PTCR will trip about 150 ma. Schematic attached. The MJ15025 transistor shown is supposed to represent the pnp output transistor driving the speaker ground. Of course another whole transistor optoisolator and power supply zener is required up at the positive rail.
**** the schematic won't upload. Invalid file. The world is addicted to microsoft, which you must update with your credit card biannually. I had microsoft Win 7 working for a whole 5 weeks, I'm back on lubuntu 13 op system which apparently DIYA doesn't want to read in .sch format.
What is an OPS? I understand a microprocessor can calculate output transistor SOA, but without a current transformer around the speaker line or something, I'm not sure what one is using for data. I'm not using a microprocessor anyway, the software is all windows, see the above rant. The VisualBasic 5 package I bought is probably as destroyed by incompatible .dll files as the Office 97 Microsoft obsoleted for me with Win XP. Peavey calculated SOA on the CS800x using op amps and current transformers, that is about my speed.
Jason, you can detect 0.3 v across the rail FET with a germanium transistor. They still sell them in the USA @ Vox Organ Germanium Transistors or RS in the UK has them.
That way you could detect maybe 30 amp on your 0.1 ohm rail fet, without fuses.
I drew up a 1n278 circuit because 1. I have some and 2. they will take 100 ma on the base line. The smallest PTCR fuse I found was a 150 ma trip current one from Murata, a PTGL04AS560K6B5180 that has cold resistance of 56 ohms. In series with a 47 ohm resistor, this could put significant current in the base lead of a NPN Ge transistor when overcurrent occurs. Then the Ge transistor collector can energize an optoisolator to cause pull down of the fault current set pin back at speaker ground, I'm using a 74hc74 flip flop. A resistor-zener arrangement between speaker ground and minus rail is necessary to provide limited current to the optoisolator diode driven by the Ge transistor collector.
With the rail FET off, the voltage on the base lead of the Ge transistor builds to 85, the PTCR will trip about 150 ma. Schematic attached. The MJ15025 transistor shown is supposed to represent the pnp output transistor driving the speaker ground. Of course another whole transistor optoisolator and power supply zener is required up at the positive rail.
**** the schematic won't upload. Invalid file. The world is addicted to microsoft, which you must update with your credit card biannually. I had microsoft Win 7 working for a whole 5 weeks, I'm back on lubuntu 13 op system which apparently DIYA doesn't want to read in .sch format.
What is an OPS? I understand a microprocessor can calculate output transistor SOA, but without a current transformer around the speaker line or something, I'm not sure what one is using for data. I'm not using a microprocessor anyway, the software is all windows, see the above rant. The VisualBasic 5 package I bought is probably as destroyed by incompatible .dll files as the Office 97 Microsoft obsoleted for me with Win XP. Peavey calculated SOA on the CS800x using op amps and current transformers, that is about my speed.
Last edited:
> With possible 1200 amp short circuit current, (85 v / .02 ohm equiv resistance of capacitor)
If your output zobel is only 1µH, 1µs after short circuit event, current has risen only to a very manageable 85 amps over the allowed max value.... By that time a decent protection circuit has already finished its job...
If your output zobel is only 1µH, 1µs after short circuit event, current has risen only to a very manageable 85 amps over the allowed max value.... By that time a decent protection circuit has already finished its job...
- Status
- Not open for further replies.