NJW0302G/NJW0281G vs NJW1302G/NJW3281G

Hey all, I’m hoping you can help me out by answering a couple of questions for me. Comparing these two set, and I guess comparing other TO3P like NJW21193G/NJW21194G. I know the 21193/94 have a much slower Ft.

Would the 0302/0281 pair work in most instances where the 1302/3281 pair is called for? Spec wise they look to be identical other than the 0302/0281 pair being rated for 150w dissipation vs the 200w of the 1302/3281 pair. And then the 21193/21194 pair have that same 200w rating, but also rated for higher voltage and higher current.

I’m guessing for instance when building this STK-0050 replacement board that was wonderfully designed by great members here and audiokarma
https://audiokarma.org/forums/index.php?threads/stk-0050-replacement-for-sx-780-and-others.721181/
That the 0302/0281 outputs would work just fine in place of the 1302/3281. Are there many designs out there where the extra 50 watts of dissipation would absolutely be necessary?

And then comes the SOA of the devices, I’m guessing that the 1302/3281 is better in this regard, is that correct?

Here is why I’m asking. Parts are starting to get harder to source. The 1302/3281 packages are out of stock and have been out of stock. I still have a couple dozen of each though. I was going to buy in quantity.

Do I buy the 1302/3281 at $3.80 each and wait for them to come into stock which looks like it will be at least 9-10 months?

Or do I buy the 0302/0281 at a much lower cost of $1.50 each, but I have to buy in much larger bulk, minimum of 166 pieces of each device? Plus I get them now.

If there are many instances known where the 0302/0281 devices won’t work where the 1302/3281 do, then maybe it doesn’t make sense to buy so many and just wait until parts are back in stock over at Mouser/Digikey. But, if they will work in most instances (example being the STK-0050 board replacement I linked above) then maybe I should grab them while I can.

Dan
 
I did just find out that the 0302/0281 pair will indeed work in the STK-0050 package replacement. Looks like that has been covered, but curious about the viability of replacing in other popular designs and comparisons of the SOA. Considering that even at Mouser the 1302/3281 devices are a little over a dollar more for each device than the 0302/0281.

Dan
 
Refer to datasheet for SOA

plenty of more detailed math used for derating and thermal junctions

Depends on Voltage , Current and Case temperature
And there is Only so much thermal exchange between the actual Die to the case, then case to heatsink.

Or in other words without specific calculation, the case is always hotter than the heatsink, and the die is always hotter than
the case.

Fig 1 in Data sheet shows derating for case temp and keep in mind the actual die temp would be hotter.
Fig 2 shows how quickly the die or wire bond fails at 150c, depending on current / voltage

if you can maintain case temperature 60 to 100c then approx 60 to 100 watts

within reason at 80 to 100 volts @ 150c they fail at around 1 amp or less current.

way better than many vintage transistors where SOA at 150c would not be more than 30 to 40 volts.
Depends, datasheet will show what is expected.

I wouldn't wait 9 or 10 months to save a few dollars
Since the better transistors or best of the best tend to be 2 bucks more.
Cost per day to be human, transistors are cheaper...much cheaper

1691818523441.png
 
The 150w bjts are fine for a pioneer sx-780 or equivalent
The 200w are used for higher supply voltage units or in cases where you want to drive lower impedance speakers or multiple pairs
Kits use toshiba 150w (2sa1943n/2sc5200n) for outputs and (ksa940/ksc2073) as drivers, instead of onsemi 150w devices since the toshiba parts have better stocking and shorter lead times
 
Last edited:
Refer to datasheet for SOA

plenty of more detailed math used for derating and thermal junctions

Depends on Voltage , Current and Case temperature
And there is Only so much thermal exchange between the actual Die to the case, then case to heatsink.

Or in other words without specific calculation, the case is always hotter than the heatsink, and the die is always hotter than
the case.

Fig 1 in Data sheet shows derating for case temp and keep in mind the actual die temp would be hotter.
Fig 2 shows how quickly the die or wire bond fails at 150c, depending on current / voltage

if you can maintain case temperature 60 to 100c then approx 60 to 100 watts

within reason at 80 to 100 volts @ 150c they fail at around 1 amp or less current.

way better than many vintage transistors where SOA at 150c would not be more than 30 to 40 volts.
Depends, datasheet will show what is expected.

I wouldn't wait 9 or 10 months to save a few dollars
Since the better transistors or best of the best tend to be 2 bucks more.
Cost per day to be human, transistors are cheaper...much cheaper

View attachment 1201866

The 150w bjts are fine for a pioneer sx-780 or equivalent
The 200w are used for higher supply voltage units or in cases where you want to drive lower impedance speakers or multiple pairs
Kits use toshiba 150w (2sa1943n/2sc5200n) for outputs and (ksa940/ksc2073) as drivers, instead of onsemi 150w devices since the toshiba parts have better stocking and shorter lead times
Thank you, I figured the SOA portion wouldn’t be as easy as “A is significantly better than B”

Maybe I didn’t explain it well, but I’d be waiting 9-10 months to spend more. Since I’d have to wait longer for the better more expensive 1302/3281. I can get the other, cheaper devices right away.

I have another example, I saw someone use those package replacement boards in a Fisher receiver. So am I right in thinking that a pair of the 150w devices would be okay in this situation where the output of the amp is 100w at 8 ohm (want to say he measured around 110w after his replacement) and it runs on +/- 57-60v rails. Manual says 57v, but I think measured voltage was closer to 58-59.

Oh man, you made me aware of some Toshiba devices that I didn’t even know were in stock. They look to be very similarly spec’d as well, just slightly higher voltage and gain on the Onsemi’s. Numbers that aren’t significant I’m guessing real world.

Any opinion on which of the devices would be considered the better, which had the better die? I can get the Onsemi for less , but we’re only talking a dime less or just over at 13 cents per device. At the same time if I went with the Toshibas I could buy less, say 100 PNP and 100 NPN where with the Onsemi I would have to buy 166 of each at a time. Not a big deal though, I don’t mind buying 166 of each if it’s a better device. And other than OCD I see no problem in using the Onsemi drivers with Toshiba outputs?


Thank you for that link, I watched it right when the video was released, it never really helped me with this question though.

I just don’t want to buy a whole bunch of a pair of devices that will be so limited that I won’t be able to use them very often. Whereas the 1302/3281 are basically some of the highest power rated devices for their size so you don’t really have more options for the size.

Dan
 
Last edited:
For an amp that runs on +/-57 volts, it would depend on the number of devices in parallel. With 150W devices, TWO are required. If the amp uses only one, you need the biggest version you can get your hands on. That would be the 4281 or 21194, BTW. Even a 3281 would be asking for failure down the road. Do the math and read SOA charts, it runs right on the limits with 8 ohms. That’s why Toshiba called it an output device “for 100 watt amplifiers”. You can’t go higher.

Three or four would be better but I doubt you’ll see that, unless it’s a 2 ohm rated PA amp. Yes, and when you bridge them into 4 ohms and run them hard even they die after a couple of years use.
 
For an amp that runs on +/-57 volts, it would depend on the number of devices in parallel. With 150W devices, TWO are required. If the amp uses only one, you need the biggest version you can get your hands on. That would be the 4281 or 21194, BTW. Even a 3281 would be asking for failure down the road. Do the math and read SOA charts, it runs right on the limits with 8 ohms. That’s why Toshiba called it an output device “for 100 watt amplifiers”. You can’t go higher.

Three or four would be better but I doubt you’ll see that, unless it’s a 2 ohm rated PA amp. Yes, and when you bridge them into 4 ohms and run them hard even they die after a couple of years use.
When you say TWO are required I’m guessing you’re meaning two PNP and two NPN, so 4 outputs per channel? I don’t know of any other way of interpreting that lol, since I’ve never seen an output with one device haha. So at that point they’d only be good for 25w output each device.

This is the specific amp I was talking about. It was a Fisher RS-2010 and used STK-0105 packs, but at some point in production must have switched to STK-0080 packs. He was replacing the STK-0080 packa using this same board made by the forum members here
IMG_7551.jpeg


It clearly only has space for a single complementary pair of outputs and he used the NJW3281/NJW1302 pair in that video. Was that a poor choice for 100w or am I reading what you said wrong?

But definitely using this board with the 0302/0281 pair in each channel for 100w isn’t a wise choice?

Dan
 
The more I look it seems like the less options there are. If I’m in need of some 200w devices I still have a whole bunch of NJW21193/NJW21194 and MJL devices. Also, if there is room on the heatsink and I’m using this board, I could always parallel a second pair of NPN/PNP correct? Is that what X-ray Tony should have done?

From what I’ve been reading on the forums, the 0302/0281 seem to be a superior pair over the 1302/3281 with the exception of the power dissipation. I can get 200 of each and have them delivered “now” who knows when the rest will come into stock. Parts are getting scary low. I just spent quite a bit of money not too long ago since Nichicon announced they’re discontinuing a bunch of their electrolytics.

IMG_7552.png



I’m guessing in most instances that the 150w devices will do well, it seems most amps rated at 100w have more then two outputs unless they are TO264 or something similar. Even the larger devices are pretty much gone at mouser/Digikey.

Other option is to go with NJL 5 pin devices, but those would hardly get used.

Just so I know, i have trouble interpreting those graphs. With adequate heat sinking and figuring on 35 to 50v rails how much power can I expect to safely get from a pair (one NPN and one PNP), two 150w devices. Could I safely get 75-80w? It was said above that the 200w 1302/3281 were made for 100w amplifiers and again I’m guessing that they are saying one pair (NPN/PNP) make 100w, so that’s one half the rating of the device. Going from the same thought would the 150w devices give me 75-80?

Regardless I don’t think I have an option unless I want to wait and hope parts come in to vendors, so I’ll be ordering them anyways. I would just like to know where I could safely run them.

Thank you,
Dan
 
Edit: I just measured the output of SX-780 and I’m seeing right at 58w before clipping on the waveform and I see that the 0302/0281 has been used in diy amps rated at 60w per channel, and that’s for a single output pair. So I know they’ll be good for at least 60w. I guess if I ever run into something like a 70w amp I could figure that out as a case to case.
Thank you,
Dan
 
Please don't mix the 21194,21193 and the 0281,0302 options.
21193,4 are slow devices when 0281,0302, 3281,1302 are fast.
Using slow devices in place of fast you are looking for troubles...
I know this the hard way 😉

I definitely understand in some situations they are not interchangeable, but in some instances they are. That’s one thing Tony said in one of his videos about swapping some of the slower 4 MHz devices in for the 30 MHz and to his ears couldn’t tell a difference. There are few threads here where people said they preferred the sound of MJL21193/94 over the MJL1302/3281 in the specific DIY amp they were cloning. As long as the amp is stable I would say. If there are no other options and I need to use 200w devices I’d rather use 21193/94 devices than nothing at all of the 30 MHz devices are all out of stock.

Dan
 
An amp that’s putting out 58 watts per channel will be running on a lot lower than +/-57 volt supplies, and would be safe with a single 150 watt pair. It really is all about the voltage you’re running at.

No, the 2119x and 3281-derivatives are not in general interchangeable. In the typical simplest circuit - an EF2, especially. The higher fT parts will result in audibly less crossover distortion, regardless of bias. But you can beat on the lower speed parts with less fear of retaliation (Burnout). It is just a matter of which one is more important to YOU. In class A operation where crossover artifacts are nonexistent, you will never be able to tell them apart listening. That’s why people can get away with sub 1 MHz hometaxial 3055’s making those cute little 10-20W class A amps. Rugged beats fast there, and a class A amp isn’t going to wait for you to turn it up to war volume before letting you know you don’t have enough SOA. You find out in about ten minutes of idling.

I have found that in EF3 configuration, high speed outputs are completely unnecessary. If the driver and pre driver are high speed and biased at sufficient current to be able to quickly suck all that charge out of a bank of 4 MHz outputs at switch-off.
 
  • Like
Reactions: 2 users
An amp that’s putting out 58 watts per channel will be running on a lot lower than +/-57 volt supplies, and would be safe with a single 150 watt pair. It really is all about the voltage you’re running at.

No, the 2119x and 3281-derivatives are not in general interchangeable. In the typical simplest circuit - an EF2, especially. The higher fT parts will result in audibly less crossover distortion, regardless of bias. But you can beat on the lower speed parts with less fear of retaliation (Burnout). It is just a matter of which one is more important to YOU. In class A operation where crossover artifacts are nonexistent, you will never be able to tell them apart listening. That’s why people can get away with sub 1 MHz hometaxial 3055’s making those cute little 10-20W class A amps. Rugged beats fast there, and a class A amp isn’t going to wait for you to turn it up to war volume before letting you know you don’t have enough SOA. You find out in about ten minutes of idling.

I have found that in EF3 configuration, high speed outputs are completely unnecessary. If the driver and pre driver are high speed and biased at sufficient current to be able to quickly suck all that charge out of a bank of 4 MHz outputs at switch-off.

Indeed the amp does run at lower voltage. The measured voltage is +/- 40v, of course the higher voltage is how they get the higher wattage from the Fisher as most everyone here knows.

I’m most instances I wouldn’t swap the fast and slower devices. Though I didn’t know if it would make a difference with the STK replacement boards I’ve brought up in this thread a few times. One of the amps I was speaking about that had people interchanging them was the Elliot P3A which even in the build instructions says you can use MJL3281/1302, NJW3281/1302, or MJL21193/21194 and I believe the bigger MJL4281 pair. So I read that when building that amp people tried both the slower and faster devices and there was a mix of what people liked most. I’m guessing that’s an EF3 design? Thanks for the help!

Dan
 
The STK replacement is not an EF3. 4 MHz outputs will work well enough in an EF2, but measurements and many people’s ears say faster ones work better. In a poorly designed or executed circuit, other distortion mechanisms may be so bad it makes no difference.

The P3A has enough other limitations besides the speeds of the outputs that it may be a wash. CFPs often do no like for you to use transistors which are too fast (problems develop, like internal ringing).

How bad the crossover artifacts are when using slow output transistors depends on the GAIN in between them and the current source driving them. In an EF3, with fast drivers and predrivers, you have plenty of that gain. With sustained-beta drivers and 4MHz outputs in an EF2 there is often enough to get rid of the low level fuzz, but there is usually enough improvement with 20+ MHz outputs to be worthwhile. Stick crap-o 4MHz TIP drivers in there, and it’s bad enough to drive you from the room.
 
  • Like
Reactions: 1 user
Can someone explain me what am I missing here. The DS of njw1302 gives very smooth DC Hfe both for 20v CE and 5v to be about 100, that is independent from Vce. The simulator says another thing. Bellow is Ib Ic curves of triangular wave for A1943 the draft1 and the njw1302 the draft2. But why?
a1943vs1302.JPG