I’ve asked a similar question in the past. I’m working on an amplifier that has 8 outputs on each channels heatsink. At the moment I am unsure of the voltage rails for the outputs as the schematic is pretty unreadable. The originals are MT200s 2SC2773 and 2SA1169
Of course one channel has multiple failures, so I’m going to do what I’ve done in the past and replace them with modern transistors. Generally if I had them I would just go with TO-264 devices, but recently in a difference thread I replaced all of the outputs in a high end Kyrocera (A-910) and it wasn’t possible to use the TO-264s. I had to use smaller TO-3P and it’s been running hard for at least a year now. So now I’m thinking that the TO-3P are up to the job. The NJW0218/0302 have the same specs or better as the originals, 150w dissipation, higher voltage ratings.
The Kyrocera has 6 outputs per channel and is 150w at 8 ohms and 200w at 4 ohms. So I’m guessing the Kyrocera may be a little less demanding on the outputs.
So basically in this amp I’m working on each pair of outputs will be responsible for 62.5 watts at 8 ohms and 82.5 watts at 4 ohms if driven to rated output. I have a few hundred of each (NJW0302/0281) which is why I’d like to use them if possible, I already have them. Amp is a Yamaha M-80
Dan
Of course one channel has multiple failures, so I’m going to do what I’ve done in the past and replace them with modern transistors. Generally if I had them I would just go with TO-264 devices, but recently in a difference thread I replaced all of the outputs in a high end Kyrocera (A-910) and it wasn’t possible to use the TO-264s. I had to use smaller TO-3P and it’s been running hard for at least a year now. So now I’m thinking that the TO-3P are up to the job. The NJW0218/0302 have the same specs or better as the originals, 150w dissipation, higher voltage ratings.
The Kyrocera has 6 outputs per channel and is 150w at 8 ohms and 200w at 4 ohms. So I’m guessing the Kyrocera may be a little less demanding on the outputs.
So basically in this amp I’m working on each pair of outputs will be responsible for 62.5 watts at 8 ohms and 82.5 watts at 4 ohms if driven to rated output. I have a few hundred of each (NJW0302/0281) which is why I’d like to use them if possible, I already have them. Amp is a Yamaha M-80
Dan
Last edited:
It also seems that any combination of the Onsemi semiconductors are out of stock. Be it MJL, MJW, NJW, ect, so the NJW0302/0281 are my only options that way as of now. I could go with Toshiba if I wanted a TO-264 in the 2SC5200 and 2SA1943, but they are rated at 150w dissipation which the TO-3P I have are already rated the same.
Dan
Dan
Single pair of 2SC5200 is good for 100W output stage as stated in a datasheet. It can stand even a bit more but better be on a safe side.
It is also good to look at SOA and derating curves from both old and new transistors.
It is also good to look at SOA and derating curves from both old and new transistors.
Last edited:
The M-80 is a class H design. Maths aren't that easy as you think. The inner devices have to cope with the full power the outer devices are able to "pump" into them. Take a close look into the SOA charts in the datasheets of your new outputs.
The peak power dissipation that occurs in a typical class H design is just as high, and at just as high a vce into a resistive load as a normal class AB amp. Peak reactive dissipation is much lower and at lower Vce, and the conduction angles at which the peaks occur are much shorter. If you were being truly conservative in the design, staying within the continuous DC SOA under all conditions, you need just as many outputs as you normally would - plus the commutators. In practice, you can use a 10 millisecond SOA curve, properly derated for temperature for the audio peaks, but still use the DC curve for the reactive/short circuit current limit. This happens at the lower Vce in class H, so short circuits aren’t nearly as hard on the amp. To reduce the peak dissipations, you need at least 3 steps in the power supply, so the switcheroo doesn’t happen exactly half way up the rail where dissipation normally peaks. If you don’t have that transformer, with enough taps for three rails, you’re screwed.
Right now the only high fT OnSemi pair you can find in stock is the Fairchild version of the C5200/A1943 - in both TO-264 and 3P. All the others are out of one or the other, and it looks like they’ll start trickling back into stock around end of March/Mid April if delivery dates hold. There is reason to believe they might. A few days ago Digikey got in the 1500-ish MJL21194’s that were on order with a late January expected date. They hit this. Now DK has the MJL21193/4 pair and a handful of 21195/6 pairs in stock. If you can stand using 4 MHz devices, of course. I ordered a bunch more for a 4 step class H monster I’ve been developing since 2009. Building 7 of them to drive a rave PA rig, along with a couple smaller versions, replacing all the store-bought amps with DIY. Bucket list project.
Right now the only high fT OnSemi pair you can find in stock is the Fairchild version of the C5200/A1943 - in both TO-264 and 3P. All the others are out of one or the other, and it looks like they’ll start trickling back into stock around end of March/Mid April if delivery dates hold. There is reason to believe they might. A few days ago Digikey got in the 1500-ish MJL21194’s that were on order with a late January expected date. They hit this. Now DK has the MJL21193/4 pair and a handful of 21195/6 pairs in stock. If you can stand using 4 MHz devices, of course. I ordered a bunch more for a 4 step class H monster I’ve been developing since 2009. Building 7 of them to drive a rave PA rig, along with a couple smaller versions, replacing all the store-bought amps with DIY. Bucket list project.
I’m not going to pretend to know exactly how class H works, I appreciate you taking the time to walk me through it.
I can’t find an SOA graph for the original devices, not sure if they exist, but here is a comparison of the 2SC5200 (Fairchild) and NJW0281G, both 150w NPN.
2SC5200
NJW0281G
I am very new to interpreting these graphs so I can certainly be reading them wrong. Looking at the 10 ms line it looks as though the 5200 starts dropping off in current at a slightly higher voltage while the 0281 can have a bit more current handling at the peak voltage (250v). To me they look pretty close, but as I’m a novice as looking at them there could be a huge difference between the two.
Yeah, it’s a real shame about the shortage of outputs. I saw it coming and glad I grabbed a few hundred 0281/0302 pairs when I could. When I grabbed them they were pretty much the only devices in stock, other than the Toshibas.
I hadn’t thought about going to a slower 4 MHz device. You grabbed some of the 4 MHz devices for a class H you’re building? That’s a great bucket list item to have. I have several Pass boards and the Wolverine boards and will slowly be starting those projects. I have a similar bucket list goal, though with all commercial speakers I have. I’ve just started 3 speakers for my HT/music front and centers. Each speaker uses all JBL drivers, 045be tweeters, 435be 3” compression drivers with H9800 waveguides (both drivers being beryllium), 2x 251j 10” midbass, and 2x 1500sub 15” woofers. Should hopefully be pretty killer.
Back to the project. Were it you changing the outputs in this M80 would you think the 0281/0302 pair would be up to the task? With the original outputs being 10 MHz and 20 MHz would you use the MJL21193/94? I’m going to go check my stock here pretty soon and let you all know what I have on hand. Seems if I act quickly and the thought is the MJL would be the better way to I could still grab them from DK. If the faster TO3p are good I’d like to go that route of course, but if I would be flying too close to the sun I get it.
Dan
Looked through my stash of parts and on top of the NJW0302/0281 I also have plenty Onsemi 2SC5200/2SA1943 (TO-264, 150w dis, 30 MHz), enough Motorola branded MJL21193/MJL21194 (TO-264, 200w dis, 4 MHz), and plenty of Onsemi NJW21193/NJW21194 (TO-3p, 200w dis, 4 MHz).
So that’s what I have to play with. While I don’t need to buy more MJL21193/94 I likely will if I use them for this project.
Dan
So that’s what I have to play with. While I don’t need to buy more MJL21193/94 I likely will if I use them for this project.
Dan
MJL21193/94 are quite powerfull but also slower than orginal transistors (frequency and output capacitance). After such change it is worth to check stability with scope.
I've also tried to find orginal (full) datasheet and it seems to be difficult to find other than that from Savantic and ISC.
I've also tried to find orginal (full) datasheet and it seems to be difficult to find other than that from Savantic and ISC.
Oh yes, I would absolutely look out for any oscillating or anything once done, but obviously the speed is literally the only thing preventing me from using the 21193/94. If they were faster, like say 15 MHz I would use them without question. Which is why I ask about the 30 MHz devices. Too bad they’re difficult to get. The originals are 150w devices and the replacements I have are 150w devices (5200/1943 and 0302/0281) but are different packages thus having differently sized heat spreaders. I figure if you have good coupling to the heatsink it wouldn’t make a huge difference. I’d think regardless the heatsink will be saturated the same amount with any of those.
Dan
What you would normally do with N-type packages is use more of them, going 3 for 2 in place of MT200’s when building new. Even if the spec is the same the smaller package physically runs hotter, all things - including the mounting to the heat sink - considered. Going to two will always give an advantage over one big device.
Oh interesting, I hasdnt thought of that, obviously. So you’re saying that for instead of 4 NPN and 4 PNP per heatsink step up to 6 NPN and 6 PNP per?
Please excuse my crude drawings, it’s very hard to draw on a phone using a photo editor. I illustrated only one side, so basically edited the PNP, but it would be the same for the NPN.
So for the main set I would add one PNP for the pair and one NPN for the pair. Basically the collector would be wired in parallel with the other two, base through a 2.2 ohm resistor to meet with the other 2.2 ohm resistors coming off of the other bases, and the emitter to a .22 ohm emitter resistor that is wired to where the emitter resistors meet.
Here is the second set of outputs, again, collector in parallel with the other collectors, emitter through 0.1 ohm resistor to where the other 0.1 ohm resistors meet, and then lastly the base through a 2.2 ohm resistor to meet up with the other 2.2 ohm resistors. Does that look about right? It’s not often/never that I’ve had to add transistors to make it easier on the outputs. Any values need changing?
Dan