Fixing a big old Peavey PV-8.5C amp for a friend. I located the shorted NPN and PNP outputs in one channel. They're 70473180 and 70483180 Peavey house numbers. Reading various posts, it seems a suitable replacement is the MJ-15024 and 25. There are four NPNs and four PNPs wired in parallel groups, with a gain of about 100 (20 mA base gives 2A collector current @ 10V). What are the chances that just replacing the single shorted devices will be OK; how sensitive do you think the groups will be to Hfe will be using 0.33 ohm emitter resistors? Obviously there's a huge cost difference if I have to replace all of 'em. 
well you can spend the time trying to beta match them to the old ones (unlikely with out a whole groupings of parts), or most would replace a rail at a time from a "lot" with same date codes. the age old adage "time is money" to most. If it's easy to service you could take a chance and check the balance of Vre's at elevated idle temps and decide from that.
Last edited:
I used to do warranty work for Peavey and asked that question many times, always told just replace the bad ones. This was on warranty and non-warranty repairs. I don't particularly agree but I never had any returns because of it. To me the outputs are cheap enough to just shotgun them all. Also I used to order all my Motorola/ON semiconductors thru them as they were by far the least expensive but don't think that is the case anymore. You may also consider the PITA factor to replace outputs in the whatever amp you are repairing. If it truly is a PITA replace them all.
Craig
Craig
Interesting range of replies! Time isn't a big deal here, so I think I'll order a few and put them on the curve tracer with the originals. If they aren't close, I'll just order more. I wonder if Peavey even bothered to match them in the original production. BTW, I've measured a fair number of power devices and can reliably say that coming from the same batch means nothing. If you truly need matched devices, you have to match them. It can also take a remarkably large number, same batch or not, to come up with say, 8 or 10 that match. Other times you just get lucky.
Being somewhat a-retentive, once I decided to replace all the devices in one channel I'd have a hard time not replacing the other channel, as I don't think the voltage or current margin is all that great on the original parts, nor do I like steel cases from a thermal standpoint.
Thanks!
CH
Being somewhat a-retentive, once I decided to replace all the devices in one channel I'd have a hard time not replacing the other channel, as I don't think the voltage or current margin is all that great on the original parts, nor do I like steel cases from a thermal standpoint.
Thanks!
CH
so that blows up some peoples (mine) ideas. seems you should/can get a fairly tight groupings with a mature semi process right off the factory floor, but OTOH I'm also sure a weeks production at ON Semi includes lots of different wafers not to mention how the distrib's hand them out. Be interested in any data you can post, esp currents and instrument used.
No data on recent devices, so maybe things have improved over the years. I have a Tek 575 122C mod curve tracer (OK, boat anchor) that works well and I've always used it to build up power amps. That a-retentive thing makes me test every singe component I use. Back in my Tiger amp days I discovered the THD could be quite dependent on matching of the devices and it's really tough to get the PNP and NPNs close. A friend and I kept buying devices hoping for a decent match. We sorted both flavors by hfe at moderate currents. Most of the tested devices were MJ802, 4502, the similar TI devices and various TIP series for some blameless amps I built later. Quantities were typically 10-20 devices. I had about as much luck with different batches as I did working within one batch.
One thought that occurs to me is that getting the NPN and PNP to match required working near the edges of the distributions. The usual scenario was to blow one or the other, then have zero luck finding a similar one to replace it, so we'd order another batch and hope for the best.
My guess, and it's only a guess, is that devices are not the same across a given wafer, much less wafer to wafer. I also wonder if they cherry pick devices to meet certain companies special order hfe requirements, as I often found devices in amps and equipment that couldn't be replaced even from large batches of supposedly the same jedec number- way off the bell curve. Or, I could just be remarkably attractive to bad luck.
One thought that occurs to me is that getting the NPN and PNP to match required working near the edges of the distributions. The usual scenario was to blow one or the other, then have zero luck finding a similar one to replace it, so we'd order another batch and hope for the best.
My guess, and it's only a guess, is that devices are not the same across a given wafer, much less wafer to wafer. I also wonder if they cherry pick devices to meet certain companies special order hfe requirements, as I often found devices in amps and equipment that couldn't be replaced even from large batches of supposedly the same jedec number- way off the bell curve. Or, I could just be remarkably attractive to bad luck.
Last edited:
I recently bought 50 ea of MJ15024/25 same date code. but my home made beta tester is buried out there somewhere, mostly broken. so... if I get the gumption I might use a cheap DMM and rig a TO-3 socket. I might trade for some ON drivers and Fairchild TO-216 and small signal stuff.
Last edited:
What counts for the sharing is not the Hfe (not directly anyway), but the collector current resulting from a fixed and imposed Vbe.Fixing a big old Peavey PV-8.5C amp for a friend. I located the shorted NPN and PNP outputs in one channel. They're 70473180 and 70483180 Peavey house numbers. Reading various posts, it seems a suitable replacement is the MJ-15024 and 25. There are four NPNs and four PNPs wired in parallel groups, with a gain of about 100 (20 mA base gives 2A collector current @ 10V). What are the chances that just replacing the single shorted devices will be OK; how sensitive do you think the groups will be to Hfe will be using 0.33 ohm emitter resistors? Obviously there's a huge cost difference if I have to replace all of 'em.
Not something usually measured.
Obviously, due to to the technology of semiconductors, there will generally be a correlation between this parameter, the Vbe and the Hfe, but correlation is not the same as equivalence.
But with 0.33R emitter resistors, each mV will translate into only 3mA unbalance; not alarming in my opinion.
yes there s/b an inverse correlation between value of Re used and Vbe matching. ie local feed back given good thermal tracking of all paralleled devices. Besides the low betas at higher currents and when things go wrong IE SOA fails, and all static beta testing at several mA are not sure to track in any way. not sure, does this make any sense?
Elvee, that makes a lot of sense and is a better way of looking at it.
Once you exceed SOA and start getting hot spots, my guess is anything could happen, and usually something bad. Hopefully with good matching or high emitter R values one can stay in the better behaved area. My wife has been trying to keep me in the better behaved area for years.
Once you exceed SOA and start getting hot spots, my guess is anything could happen, and usually something bad. Hopefully with good matching or high emitter R values one can stay in the better behaved area. My wife has been trying to keep me in the better behaved area for years.
I'm working on a PV1.3k. (same generation) In the cheapskate department, what do you think of ON MJ21195 for the PNP and NJW21194 (TO247) for the NPN? I've got it working but my biggest load is 10 ohm @ 400 watts, which is not much of a test. Was a PITA to put the TO247 in but they have copper frame even if only rated 200W per. Also $2.50 each cheaper. I put a 22 pf on each B-E line on the output frame, in case the new stuff is too fast & wants to oscillate.
What do you think of the same for the TO3 driver. Cross ref says MJ15020 and 21 but I didn't buy any of those. I'm a little vague on reason to buy a lower power driver except to save money, and the 20 & 21 cost more now.
What do you think of the same for the TO3 driver. Cross ref says MJ15020 and 21 but I didn't buy any of those. I'm a little vague on reason to buy a lower power driver except to save money, and the 20 & 21 cost more now.
Last edited:
I thought TO-247 was supposed to be trivial to put in a TO-3 location, but I've never done it. As for what to use, I looked at the published specs of the amp, then measured the supply voltages. I like a healthy margin on the Vce to account for high line. Figuring somebody would put a low impedance load on it, say 2-3 ohms, and dividing the current amongst the number of outputs, I decided on a current rating. IMO, many commercial amps go a bit light on outputs and only survive due to power supply sag and luck. There are some statements on the 'net about the current and voltage ratings of the Peavey devices and what the numbers mean, but I really doubt they're correct. If they are, then the devices are seriously inadequate IMHO. Others may have better methods but I like no less than 1.5X the rail-to-rail voltage under no load, and no less than 1.5X of the peak expected current under the worst case load. Using those numbers and assuming no supply sag, the margin ends up pretty good. BTW, be careful slowing down the devices to prevent oscillation, as they also need to turn off quickly to prevent a rail-to-rail short. You can get outside the SOA very quickly that way!
Thanks for the tip. Consider the 20 pf BE cap removed until oscillation does occur.
NJM21194 and MJ21195 seemed to have same 95 V SOA compared to MJ 15024 and 25, although the 21194 had less advertised watts (200 instead of 250). All had rediculously high Vceo, 250. I decided to go with soa equivalence instead of absolute watt rating, to save some $.
I had to cut the corners off the pad under the TO247 NJW21194 to get them to not bump the leads of the emiter resistors. They are mounted diagonally in a PV1.3k or an 8.5 probably. This corner wasn't touching the heat sink anyway, so no great loss I hope.
I was hoping the .33 ohm emiter resistors allowed one to not match transistors, since I don't have any setup to do that. As 4 of 5 were blown in two places, I did them all of the top or the bottom. Triac was melted, one diac was blown, the lead to triac was melted. Isn't the snap action thermostat put in series with the output on top of the farthest output from the fan, funny? they did a lot better job of protection on the CS800s. I put 25 amp fuses in series with the O.T's to try to prevent 4 of 5 O.T.'s from cooking if one overheats, next time the triac goes off.
1000 W/ch rating @ 2 ohms gets 22 amps. With 5 O.T.'s per side top and 5 bottom, that is still a bit above a 2.5a soa at 95 v rail these O.T's seem to be rated at. Bit safer at the 650 W 4 ohm rating.
Here is a tip for those with sixtyish vision like me- don't plug the harness from the driver to the OT frame in one pin off. Fits fine, but blows about 25 parts on the driver PCB, mostly diodes, but a driver to92 transistor, a cap. a 3w resistor, a 4558. Hope I didn't get the 13080 driver IC(U201) , no way to buy those. I think it has something to do with quiescent bias current, but I'm not sure. Been thinking LM4562 has enough slew rate and drive current to replace the 13080 if I have to, but the pinout is all wrong.
NJM21194 and MJ21195 seemed to have same 95 V SOA compared to MJ 15024 and 25, although the 21194 had less advertised watts (200 instead of 250). All had rediculously high Vceo, 250. I decided to go with soa equivalence instead of absolute watt rating, to save some $.
I had to cut the corners off the pad under the TO247 NJW21194 to get them to not bump the leads of the emiter resistors. They are mounted diagonally in a PV1.3k or an 8.5 probably. This corner wasn't touching the heat sink anyway, so no great loss I hope.
I was hoping the .33 ohm emiter resistors allowed one to not match transistors, since I don't have any setup to do that. As 4 of 5 were blown in two places, I did them all of the top or the bottom. Triac was melted, one diac was blown, the lead to triac was melted. Isn't the snap action thermostat put in series with the output on top of the farthest output from the fan, funny? they did a lot better job of protection on the CS800s. I put 25 amp fuses in series with the O.T's to try to prevent 4 of 5 O.T.'s from cooking if one overheats, next time the triac goes off.
1000 W/ch rating @ 2 ohms gets 22 amps. With 5 O.T.'s per side top and 5 bottom, that is still a bit above a 2.5a soa at 95 v rail these O.T's seem to be rated at. Bit safer at the 650 W 4 ohm rating.
Here is a tip for those with sixtyish vision like me- don't plug the harness from the driver to the OT frame in one pin off. Fits fine, but blows about 25 parts on the driver PCB, mostly diodes, but a driver to92 transistor, a cap. a 3w resistor, a 4558. Hope I didn't get the 13080 driver IC(U201) , no way to buy those. I think it has something to do with quiescent bias current, but I'm not sure. Been thinking LM4562 has enough slew rate and drive current to replace the 13080 if I have to, but the pinout is all wrong.
Last edited:
I got the whole amp for $55. A transformer of that power is $80-$100. The heat sink and fans are nice, too. 15 replacement O.T's were $40 (5 PNP @ $4 and 10 NPN @ $2), and the diodes & transistor etc on the driver board were $2, mostly salvage from a PCAT power supply. The triacs & diac were $3. A cheap education in solid state, more modern than the dynakit ST120. Thanks for the tip on the 3080 if I end up needing it.
Last edited:
Well, I finally got back to this project and replaced just the bad outputs. Setting up the output board on the bench with power supplies and checking the emitter resistor voltages for various currents showed that the currents match within 5%. Seems good enough to me.
There were various other problems including a bad ground connection where the factory crammed a wire shield into an IDC connector with slip on tubing that the IDC couldn't really cut through. There were also two shorted diodes that are involved in setting the bias, so the crossover distortion was huge (12%+). Somebody had also replaced a 4558 with a 1458 during previous service. The difference isn't huge but the GBP is, I think, about a factor of two less. I fixed all that but didn't have a 4558 handy.
I tried several different dual opamps to improve the performance, but none were really successful. They either accomplished nothing or increased the low power distortion measurement due to slight oscillation or an increase in noise. This is not a circuit that benefits from "opamp rolling" without other circuit modifications; it needs the slow and benign 4558. (Tried MC34072, SSM2135 and LT1013. Higher performance opamps are way too fast.)
Anyway, my question- does anybody know the THD specs for units made around 1992? This one has rising distortion at 20kHz, to somewhat more than the 0.1% spec. But, it also has all the output devices wired in parallel, unlike the 1995 schematics that show the first of four output devices being used as a driver for the other three. It seems like that might account for the performance difference.
There were various other problems including a bad ground connection where the factory crammed a wire shield into an IDC connector with slip on tubing that the IDC couldn't really cut through. There were also two shorted diodes that are involved in setting the bias, so the crossover distortion was huge (12%+). Somebody had also replaced a 4558 with a 1458 during previous service. The difference isn't huge but the GBP is, I think, about a factor of two less. I fixed all that but didn't have a 4558 handy.
I tried several different dual opamps to improve the performance, but none were really successful. They either accomplished nothing or increased the low power distortion measurement due to slight oscillation or an increase in noise. This is not a circuit that benefits from "opamp rolling" without other circuit modifications; it needs the slow and benign 4558. (Tried MC34072, SSM2135 and LT1013. Higher performance opamps are way too fast.)
Anyway, my question- does anybody know the THD specs for units made around 1992? This one has rising distortion at 20kHz, to somewhat more than the 0.1% spec. But, it also has all the output devices wired in parallel, unlike the 1995 schematics that show the first of four output devices being used as a driver for the other three. It seems like that might account for the performance difference.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.