Thank you so much for this info, I was looking at random sites trying to find this type of info, it's hard to find. I didn't think to put an NTE temp device or MOS surge surpressor. Those are great tips, I have a dim bulb tester already, I had the amp running on a +/- 36V supply that I had from another amp. I was able to set the bias at around 18mV and I tested it at around 46 Watts RMS before clipping with a 1KHz sin wave as input into 8 ohms.
Had a little problem while testing my hb today, I built the new psu and soft start board today. I hooked everything up and got +/- 60 VDC out of the supply, I hooked up one ch of the HB with everything on a dim bulb tester, I set the bias and let it warm up everything looked ok, so I took the dim bulb out of the circuit.
The bias kept climbing quickly, so I turned the bias pot clockwise to turn it down. It stops and then started shooting up again. Eventually it took out 2 resistors and almost blew the fuses but I unplugged it before it blew. The two resistors were 22 ohms 1/2watts by the fuse, R53 and R54. Any idea what might have caused this? Too late at night to test all the components at this time.
The bias kept climbing quickly, so I turned the bias pot clockwise to turn it down. It stops and then started shooting up again. Eventually it took out 2 resistors and almost blew the fuses but I unplugged it before it blew. The two resistors were 22 ohms 1/2watts by the fuse, R53 and R54. Any idea what might have caused this? Too late at night to test all the components at this time.
Last edited:
Turns out it did blow both rail fuses, and q21 is also testing shorted while in circuit between all pins. That's also the one that was running hotest out of all of them during testing.
R24 is the feedback resistor, not unusual for it to blow if an output transistor blows.
R23 is a limiter for predriver Q10, pretty weird for it to blow if the driver Q14 didn't blow. Also Q21 is on the opposite side from R24, the npn side.
Q21 is an output transistor. You did buy these 2SA1943 2SC5200 from an authorized distributor like digikey, mouser, farnell, arrow didn't you? there is another in Vancouver, buddy's I think I remember? Not real up on sources of real transistor in Canada. Fake output transistors are a good way to blow everything all the way back to the input transistor.
Another thing people do is match the Vbe of the output transistors. That means you have to have some extras to put in the junk box. Or match the Vce at gain 10 like I did, more difficult.
The light bulb box is a useful device. Getting off it is like training wheels on a 20" bike. Lots of skinned knees in the beginning.
An amp with 5 pairs OT, I used a 1200 W room heater element as the AC line ballast. Had convenient spade lug terminals inside a grounded case. Room heaters blow the tip over switch all the time, rarely the heating element.
Once rail voltage gets loose from an output transistor, you have to test everything ahead of it at Vce 12 at least. Double diode tests at 2 v don't tell you anything. Leakage current at 12 v helped me spot the transistors that had been overstressed by excess voltage. Vceo test, 12 from collector to emitter (or -12 for pnp) with the base open. Measure current with a 200 ma scale dvm limited by a 10k or 47 k resistor series the collector. Leakage current should be zero.
Guys in 3rd world countries where the border guards/customs workers switch real transistors in shipments for fakes, they test every transistor received at the real Vceo voltage. Dangerous over 25 v though. One hand at a time over 24 v.
Honeybadger has .22 ohm emitter resistors on the output transistors. That is a recipe for thermal runaway IMHO. Peavey used .5 ohms in their nineties 10 OT models. Maybe you better try that for emitter resistors until you get it stabilized. gives you more time to react. I put cinch type solder terminal strips off the edge of the board to measure emitter current of each transistor, don't seem to cause oscillation in my PV-1.3k. Poking around the transistor emitters themselves is a good way to blow up things, IMHO.
Having mentioned the word oscillation, you do have a scope don't you? Ultrasonic oscillation caused by a bad solder joint can heat transistors up really fast. There is a way to check for it with an ANALOG VOM on AC scale if you have one of those instead of a scope.
R23 is a limiter for predriver Q10, pretty weird for it to blow if the driver Q14 didn't blow. Also Q21 is on the opposite side from R24, the npn side.
Q21 is an output transistor. You did buy these 2SA1943 2SC5200 from an authorized distributor like digikey, mouser, farnell, arrow didn't you? there is another in Vancouver, buddy's I think I remember? Not real up on sources of real transistor in Canada. Fake output transistors are a good way to blow everything all the way back to the input transistor.
Another thing people do is match the Vbe of the output transistors. That means you have to have some extras to put in the junk box. Or match the Vce at gain 10 like I did, more difficult.
The light bulb box is a useful device. Getting off it is like training wheels on a 20" bike. Lots of skinned knees in the beginning.
An amp with 5 pairs OT, I used a 1200 W room heater element as the AC line ballast. Had convenient spade lug terminals inside a grounded case. Room heaters blow the tip over switch all the time, rarely the heating element.
Once rail voltage gets loose from an output transistor, you have to test everything ahead of it at Vce 12 at least. Double diode tests at 2 v don't tell you anything. Leakage current at 12 v helped me spot the transistors that had been overstressed by excess voltage. Vceo test, 12 from collector to emitter (or -12 for pnp) with the base open. Measure current with a 200 ma scale dvm limited by a 10k or 47 k resistor series the collector. Leakage current should be zero.
Guys in 3rd world countries where the border guards/customs workers switch real transistors in shipments for fakes, they test every transistor received at the real Vceo voltage. Dangerous over 25 v though. One hand at a time over 24 v.
Honeybadger has .22 ohm emitter resistors on the output transistors. That is a recipe for thermal runaway IMHO. Peavey used .5 ohms in their nineties 10 OT models. Maybe you better try that for emitter resistors until you get it stabilized. gives you more time to react. I put cinch type solder terminal strips off the edge of the board to measure emitter current of each transistor, don't seem to cause oscillation in my PV-1.3k. Poking around the transistor emitters themselves is a good way to blow up things, IMHO.
Having mentioned the word oscillation, you do have a scope don't you? Ultrasonic oscillation caused by a bad solder joint can heat transistors up really fast. There is a way to check for it with an ANALOG VOM on AC scale if you have one of those instead of a scope.
Last edited:
After some sleep, with fresh eyes, you said R53, R54. Yeah, those power the rails after the rail fuse blows, I don't know why they are there but any shorted output transistor would take them out.
Another bullet proofing idea, R20 the bias pot, the wiper can lose contact on some pots as you adjust it, running the voltage across it up to rediculous values. You can clamp that to a maximum voltage with a parallel diode or something. I don't know what voltage a honeybadger is expecting on R20, but you could start with a 1n4148 for maximum .7 v and see what the voltage really is on R20 for 10-20 ma on the emitter resistors. If you need less most schottky diodes are .4 v and some diodes salvaged from PCAT power supplies have measured as low as .16 v on my DVM. The line goes down as R20 is installed on the print. For salvage diodes without a line on the body, the negative lead of the diode as tested on the dvm goes down.
If the proper voltage on R20 ends up being bigger than 0.7 v, you can stack up more than one diode for 1.4 v, 2.1 v clamp, whatever. At 2.1 you start getting into voltages you can buy in zener diodes. Zener diodes the line goes to the plus side (top of R20).
Another bullet proofing idea, R20 the bias pot, the wiper can lose contact on some pots as you adjust it, running the voltage across it up to rediculous values. You can clamp that to a maximum voltage with a parallel diode or something. I don't know what voltage a honeybadger is expecting on R20, but you could start with a 1n4148 for maximum .7 v and see what the voltage really is on R20 for 10-20 ma on the emitter resistors. If you need less most schottky diodes are .4 v and some diodes salvaged from PCAT power supplies have measured as low as .16 v on my DVM. The line goes down as R20 is installed on the print. For salvage diodes without a line on the body, the negative lead of the diode as tested on the dvm goes down.
If the proper voltage on R20 ends up being bigger than 0.7 v, you can stack up more than one diode for 1.4 v, 2.1 v clamp, whatever. At 2.1 you start getting into voltages you can buy in zener diodes. Zener diodes the line goes to the plus side (top of R20).
Last edited:
Hey Indianajo, Thanks for your reply. I am North of New York State. I ordered all components from Digikey and Mouser. However, I couldn't get matched pairs of the output transistors because of crazy backorders, one goes in stock the other goes out. I ended up using NJW0302G/NJW3281G pair for output instead of NJW0302G/NJW0281G because the MJW0281G was backordered at the time. Not sure if I should just use the complement for the NJW3281G trans. According the Digikey they are a direct replacement.
I have to go through everything and see what got fried before I make any orders. Not sure if I should just replace all the NJW0302's with something that complements the NJW3281G?
There was one NJW0302G that always got hotter than the rest, it didn't have good contact with the heatsink at first but I thought I took care of that. I am using those little grey fiberglass sil pad things in insulate, I hope they don't need heat grease like the mica pads, or that'd be a real new guy mistake. haha. I have a scope, but I don't even have a chance to hook it up without making sure the amp is stable. I was able to measure a sin wave with no distortion when I had it hooked up with a lower voltage test supply, didn't notice any oscillation at that time.
Last edited:
MJL, NJL, MJW NJW, I don't know what the differences are. Not very much. The datasheets show what the complement is supposed to be. You can download datasheets from digikey or mouser either one, using the link. Do it the day you buy, and read it before you place the order. If a distributor doesn't have the complement to the transistor on the day you are buying, don't buy it. The distributors charge less for orphan transistors they don't have a match to. Love to sell those off to $100 a year customers.
On via mouser substituted TO220 instead of TO247 for TIP147 one day with a stupid package code substitution without mouser putting TO220 in the description. Mouser did not give me my money back. Haven't bought from them since, either.
I got a 50 of NJW21194 from Newark with a 250 mw power rating on the selector table, and it wasn't a misprint. They blew up immediately. Selling off the rejects for $1.80 is what they were doing. At least the defect wasn't hidden.
On doesn't care, they feel like they are doing us a favor by still selling packages with leads. All the big boys that On makes a profit from are buying nothing but surface mount for their $400000 soldering machines. On is not making a lot of money, their stock is toast. Who is making money are cell phone parts suppliers, people buy a new one of those every year and for $700 too.
We small lot builders can't get entire rails of 25 transistors built on the same day. You have to order 25 to do that. So you had better order extras and do Vcb matching anyway. I hooked MJ21193/21194 up on a heat sink with 100 ohm resistor to base & 10 ohm resistor to collector, then measured Vce after 10 second warmup. Tedious, I don't know if that matches them any better than Vbe or not. I hadn't read about Vbe matching yet, which is a lot quicker.
Soft silicon rubber pads should not need heat sink compound.
Not oscillating once is a good sign. Not proof perfect.
.5 ohm emitter resistors on output transistors will give you more time to react to problems. Cut your max power by a half percent or .01 db since ears respond logrithmically. And clamp your R20 pot in case the wiper is losing contact. It will in 10 years anyway, oxygen never sleeps.
On via mouser substituted TO220 instead of TO247 for TIP147 one day with a stupid package code substitution without mouser putting TO220 in the description. Mouser did not give me my money back. Haven't bought from them since, either.
I got a 50 of NJW21194 from Newark with a 250 mw power rating on the selector table, and it wasn't a misprint. They blew up immediately. Selling off the rejects for $1.80 is what they were doing. At least the defect wasn't hidden.
On doesn't care, they feel like they are doing us a favor by still selling packages with leads. All the big boys that On makes a profit from are buying nothing but surface mount for their $400000 soldering machines. On is not making a lot of money, their stock is toast. Who is making money are cell phone parts suppliers, people buy a new one of those every year and for $700 too.
We small lot builders can't get entire rails of 25 transistors built on the same day. You have to order 25 to do that. So you had better order extras and do Vcb matching anyway. I hooked MJ21193/21194 up on a heat sink with 100 ohm resistor to base & 10 ohm resistor to collector, then measured Vce after 10 second warmup. Tedious, I don't know if that matches them any better than Vbe or not. I hadn't read about Vbe matching yet, which is a lot quicker.
Soft silicon rubber pads should not need heat sink compound.
Not oscillating once is a good sign. Not proof perfect.
.5 ohm emitter resistors on output transistors will give you more time to react to problems. Cut your max power by a half percent or .01 db since ears respond logrithmically. And clamp your R20 pot in case the wiper is losing contact. It will in 10 years anyway, oxygen never sleeps.
Last edited:
I de-soldered all the output transistors which was no easy feat because my de-soldering tool tip didn't fit around the leads. As far as a go/no go test with a transistor testing china special tool, only the one device tests bad, dead short on all leads. The others all test ok at least on low voltages. The smaller transistors all test ok too, but I still have a few to test.
Looking at the bad transistor, it seems it was a thermal conductivity issue. I can see where the back of the package was touching the heat sink with some witness marks from the pad where as the rest had none. I don't have experience with testing each trans under load with higher voltage, I'd probably end up damaging them in the process. I am going to reassemble the ones that test ok and keep the whole thing in a really tight leash. I am confident that the lack of heat sink contact caused this, but I don't know how much damage it caused to the other components in the process.
Looking at the bad transistor, it seems it was a thermal conductivity issue. I can see where the back of the package was touching the heat sink with some witness marks from the pad where as the rest had none. I don't have experience with testing each trans under load with higher voltage, I'd probably end up damaging them in the process. I am going to reassemble the ones that test ok and keep the whole thing in a really tight leash. I am confident that the lack of heat sink contact caused this, but I don't know how much damage it caused to the other components in the process.
Per post 44, 2 v tests are garbage on overstressed possibly damaged transistors. You need to assemble a 12 v power supply, a 47k resistor (or 22 or 10) some alligator clip leads, and the 200 ma scale of a dvm. I use a motorcycle battery charger followed by a filter capacitor. Look for no leakage current at 12 v c to e with the base open. Opposit polarity on pnp. You probably don't have to pull the drivers to Iceo test them if the output transistors are out.
You need to prove the drivers at least are not damaged. If they are not then the ones further back won't be. That is Q14 & Q15. Limit from e to b backwards is only about 7 volts.
I use the squeeze bulb desoldering tool from parts-express.com . I can quickly pull the teflon tip out and run a pick through it if solder gets stuck in the hole. Costs $3.
Yeah, assembly error could have caused a heat problem but pro techs nearly always replace all output transistors as a set. I wouldn't worry much about njw0302 versus 3281 but I would Vbe check them as they are out to see if they match.
You need to prove the drivers at least are not damaged. If they are not then the ones further back won't be. That is Q14 & Q15. Limit from e to b backwards is only about 7 volts.
I use the squeeze bulb desoldering tool from parts-express.com . I can quickly pull the teflon tip out and run a pick through it if solder gets stuck in the hole. Costs $3.
Yeah, assembly error could have caused a heat problem but pro techs nearly always replace all output transistors as a set. I wouldn't worry much about njw0302 versus 3281 but I would Vbe check them as they are out to see if they match.
Last edited:
Thanks again, I'll give that a shot. Does it matter if the resistor is on the emitter or the collector? As long as it's in series with the 12V supply and the current meter should be ok? I have a lab power supply, I will be using that for the 12 volts.
I finished testing all the transistors with the 12v ce test. All were good except q21 obviously due to the short. If only q21 is bad which transistors behind it are most likely to fail?
Thank you, I really appreciate all your help indianajo. I don't have enough transistors to match right now, but I do have a home made curve tracer that I built based on a design by Mr. Carlson's lab, who's a youtuber. I'm going to dust it off and watch his video's on how to check transistors with it. He uses +/- 9 volts to check power transistor curves.
Your curve tracer could explain differences between NJW0302 and nJW3281 (npn vs pnp).
Vbe is a 1 second test with DVM ohms scale, takes 50 x longer to take them out & return than it does to do the test. If you don't have a vbe mismatch, problem solved. If you do, either increase emitter resistors or match the transistors. Everybody senior here matches transistors.
Vbe is a 1 second test with DVM ohms scale, takes 50 x longer to take them out & return than it does to do the test. If you don't have a vbe mismatch, problem solved. If you do, either increase emitter resistors or match the transistors. Everybody senior here matches transistors.
How do I do a Vbe test, I tried searching the forum. It always comes up with random threads without how to info. Seems easy enough but how exactly do I do that. Does anybody have any links that are helpful for annoying noob type questions? haha
You put the DVM on ohms scale x2000. for mine. Others you have to use the diode scale.
You put plus on the p leg and negative on the n leg. On TO247 parts base is on the left and e is on the right, numbers up.
the number you get, 680 or 450 or something, is millivolts. That is the number you match up. I write the first two digits on the top with a sharpie marker.
You put plus on the p leg and negative on the n leg. On TO247 parts base is on the left and e is on the right, numbers up.
the number you get, 680 or 450 or something, is millivolts. That is the number you match up. I write the first two digits on the top with a sharpie marker.
Last edited:
Ahhh I see, so the forward voltage between base and emitter. Ok great, yeah my meters all do that in diode test mode. Thanks again, I think I even did that just while testing and they are all very similar. Great tip with marking them with a sharpie.
Good news, I got everything running off of the dim bulb tester. I made sure all the transistors had perfect heatsink contact. The build guide says to set bias for 15 - 20 mA. I set it for 15 and after 45 min of warm up it was around 22 mA. The guide then goes to say 20 - 25 is cool and 30 to 40 is running hot. Should i let it run for an hour before final adjustment of bias, or how does that work? I am happy that i don't have smoke escaping, that is a frustrating and costly experience. By the the way the output trans were almost identical as far as the Vbe, except for the one i replaced it with that was off by 0.01 .572 vs. .580 for the other 2.
I wouldn't worry about .57 versus .58.
I'd be happy with 15 ma warm in case it creeps up after a loud party which causes hot transistors. Or do a full wattage 1 hour check into load resistors, then measure the bias with hot transistors.
Again, pots as permanent bias setters are a recipe for disaster. Peavey doesn't use them. The wipers oxidize over the years, the bias goes to high level as r20 is the full value of the pot.
Measure the pot resistance at the perfect idle bias, then replace the pot with a fixed resistor. And or clamp the pot with a suitable diode combination. You didn't measure voltage across R20 at the proper idle bias did you?
I'd be happy with 15 ma warm in case it creeps up after a loud party which causes hot transistors. Or do a full wattage 1 hour check into load resistors, then measure the bias with hot transistors.
Again, pots as permanent bias setters are a recipe for disaster. Peavey doesn't use them. The wipers oxidize over the years, the bias goes to high level as r20 is the full value of the pot.
Measure the pot resistance at the perfect idle bias, then replace the pot with a fixed resistor. And or clamp the pot with a suitable diode combination. You didn't measure voltage across R20 at the proper idle bias did you?
Last edited: