Yes, but the problem is to find good matching pairs. I just can't find matching BD137/138 or BD139/140 used as output transistors in my headphone amp.john curl said:
I don't know what TO126 (or TO92) devices I should be looking at instead of these. I find the BD's from ONSemi slightly soft and dark, and I'd be glad if someone could point me to a better sounding device. The bias is 20 - 30 mA and supply voltage is 10 - 20 V. I've tried perfectly matched BC327/337, but they sound worse. BC639/640 from ONSemi is even worse.
I'm awaiting ELFA to get Philips BC639/640 (Dutch made) back in stock. They're hopefully better matched and better sounding.
nelsonvandal
Have you tried 2SA1930/2SC5171 in place of BD139/BD140 in your headphone amplifier ? These Toshiba 200MHZ devices usually have similar HFE in both polarities, and mine usually measure between 170-210 HFE. Just make sure they are genuine Toshiba devices, and not an Eastern clone. They work very well in the Silicon Chip headphone amplifier, which originally specified BD139/BD140.
However, you will need to rotate them because of their pinout, which may necessitate the heatsinks facing the other way around.
SandyK
Have you tried 2SA1930/2SC5171 in place of BD139/BD140 in your headphone amplifier ? These Toshiba 200MHZ devices usually have similar HFE in both polarities, and mine usually measure between 170-210 HFE. Just make sure they are genuine Toshiba devices, and not an Eastern clone. They work very well in the Silicon Chip headphone amplifier, which originally specified BD139/BD140.
However, you will need to rotate them because of their pinout, which may necessitate the heatsinks facing the other way around.
SandyK
i delive that the
best way to do this i provided by andrewT in one thread ....
meaning that specific circuits have to be made for npn npn pairs and also npn and pnp pairs configured as a bridge and also with common thermal junction .....
the same thing has to be constructed for various range of power transistors ....
there is a very fine working circuit at esp pages .... but then again its only made for npn and also doesnt work as a bridge ....
very painfull story to deal with but applyied in asmall P3A amp the results were breath taking
best way to do this i provided by andrewT in one thread ....
meaning that specific circuits have to be made for npn npn pairs and also npn and pnp pairs configured as a bridge and also with common thermal junction .....
the same thing has to be constructed for various range of power transistors ....
there is a very fine working circuit at esp pages .... but then again its only made for npn and also doesnt work as a bridge ....
very painfull story to deal with but applyied in asmall P3A amp the results were breath taking
Low voltage circuits, always the 2sb, no preference between the other two, depends which one I can get cheaper and faster, I like buying these sanyos because where i get them from supply me same batch numbers, they are shipped to him matched pnp/npn pairs, which is great, no other manufacturer does this, sometimes the hfe matching is only a percent or two off.
Those toshibas are very good too, but a hassle to get even close matching. I prefer the sanyo for their low cob, when used in cfp as driver in ef they are more stable, no need for big base resisters and what not which messes up the reverse biasing and causes crossconduction.
2sc3067 are discontinued but still many going around, lucky the european and american market never locked on to them, buy american they say its better,
2sc3067/2sa1240. There is another set of pairs too, identical basically but those ones can only get by pulling some strings at sanyo factory.
Those toshibas are very good too, but a hassle to get even close matching. I prefer the sanyo for their low cob, when used in cfp as driver in ef they are more stable, no need for big base resisters and what not which messes up the reverse biasing and causes crossconduction.
2sc3067 are discontinued but still many going around, lucky the european and american market never locked on to them, buy american they say its better,
2sc3067/2sa1240. There is another set of pairs too, identical basically but those ones can only get by pulling some strings at sanyo factory.
Hi nelsonvandal
I match as much as possible. Even with lots of NFB you can improve the distortion by matching. Linearity after NFB depends on linearity before NFB.
How to do - not difficult, but you have to measure several transistors. I use a 6V power supply (AKA lead-acid battery) and a resistor board with half-decade steps as a base limiter (1 ohm/3.3 ohm/10 ohm/33 etc) but you could use 1/2.2/4.7/10 etc for finer steps. Measure the collector current for each base resistor on the NPN and swap all the leads round for the PNP. Start with the high resistors (10k) and work down, but the other point is that at higher currents (>1A) you will need to bolt the transistor to a heatsink. It's time consuming for more than a few devices.
If you have an old Tektronix curve tracer you can measure the characteristics directly. If you want to make a virtual curve tracer out of a dual beam oscilloscope you can use a (lo-fi, please) amp as a variable AC supply into a transformer, rectify the output and connect the tranny in X-Y mode with the vertical connected across a series resistor to measure current. A 1 ohm resistor is handy to give a direct A to V conversion.
You don't need to calculate the gain using the resistor method, just note how many amps you get with a particular resistor and match the amps.
I'm wondering if it is worth it for the old style transistors like the 2N3055/MJ2955. These have a 3:1 (or worse) gain ratio by themselves over their current range. You need to have a high OLG and NFB to get this under control, by which time you could probably tolerate gain differences. But matching will still be better than not.
Matching the new devices (MJ21193/4, MJL1302/3821 etc) which are pretty linear should be worth the effort.
cheers
John
I match as much as possible. Even with lots of NFB you can improve the distortion by matching. Linearity after NFB depends on linearity before NFB.
How to do - not difficult, but you have to measure several transistors. I use a 6V power supply (AKA lead-acid battery) and a resistor board with half-decade steps as a base limiter (1 ohm/3.3 ohm/10 ohm/33 etc) but you could use 1/2.2/4.7/10 etc for finer steps. Measure the collector current for each base resistor on the NPN and swap all the leads round for the PNP. Start with the high resistors (10k) and work down, but the other point is that at higher currents (>1A) you will need to bolt the transistor to a heatsink. It's time consuming for more than a few devices.
If you have an old Tektronix curve tracer you can measure the characteristics directly. If you want to make a virtual curve tracer out of a dual beam oscilloscope you can use a (lo-fi, please) amp as a variable AC supply into a transformer, rectify the output and connect the tranny in X-Y mode with the vertical connected across a series resistor to measure current. A 1 ohm resistor is handy to give a direct A to V conversion.
You don't need to calculate the gain using the resistor method, just note how many amps you get with a particular resistor and match the amps.
I'm wondering if it is worth it for the old style transistors like the 2N3055/MJ2955. These have a 3:1 (or worse) gain ratio by themselves over their current range. You need to have a high OLG and NFB to get this under control, by which time you could probably tolerate gain differences. But matching will still be better than not.
Matching the new devices (MJ21193/4, MJL1302/3821 etc) which are pretty linear should be worth the effort.
cheers
John
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.