Which dual npn transistor would be the best choice for differential input stage ?
I have a set of MAT02 already but need more and those are hard to find.
If buy from chinese seller, will metal cans probably be fake ?
What parameters are most important ? Hfe, hfe matching, ft, Vbe match ?
2N2920 has only 60 MHz, would that be enough ?
How much current is to much ? At 6mA per leg and with heatsink there is no problem. Hfe goes up with current, right ?
I have a set of MAT02 already but need more and those are hard to find.
If buy from chinese seller, will metal cans probably be fake ?
What parameters are most important ? Hfe, hfe matching, ft, Vbe match ?
2N2920 has only 60 MHz, would that be enough ?
How much current is to much ? At 6mA per leg and with heatsink there is no problem. Hfe goes up with current, right ?
These days the best matched transistors are inside opamps...
Any integrated transistor pair is likely to be well matched (same die), and have good thermal tracking.
High gain and low noise are important things whatever, as is chosing the optimal current for lowest noise operation. This is a compromize if you don't know the source impedance. The bandwidth isn't an issue for an input pair as there's no Miller effect.
Any integrated transistor pair is likely to be well matched (same die), and have good thermal tracking.
High gain and low noise are important things whatever, as is chosing the optimal current for lowest noise operation. This is a compromize if you don't know the source impedance. The bandwidth isn't an issue for an input pair as there's no Miller effect.
Obviously. 100% chance of it these days, the crooks have undercut the honest traders and made it impossible for them.
Thanks
MAT02 and LM394 have ceramic on the bottom of the case, I don`t know if they fake that too.
Overview of metal can transistors
2N2060 60V 500mA 1W 100 60MHz 5mV 0,9
2N2223 60V 500mA 1W 100 60MHz
2N2917 45V 30mA 1W 200 60MHz 10mV 0,8
2N2920 60V 50mA 0,5W 600 60MHz 5mV 0,9
2N3043 30mA 150 30MHz
2N3838 600mA 60 100MHz
BFY81 45V 50mA 0,5W 200 60MHz 10mV 0,8
BFY82 45V 0,1A 0,5W 120 250MHz 15mV 0,8
BFY91 45V 100mA 0,6W 150 60MHz 5mV 0,9
LM194 35V 20mA 0,16W 700 200MHz 50µV
MAT01 45V 25mA 0,6W 300 450MHz 100µV
MAT02 40V 20mA 0,6W 600 200MHz 50µV
2N2060 60V 500mA 1W 100 60MHz 5mV 0,9
2N2223 60V 500mA 1W 100 60MHz
2N2917 45V 30mA 1W 200 60MHz 10mV 0,8
2N2920 60V 50mA 0,5W 600 60MHz 5mV 0,9
2N3043 30mA 150 30MHz
2N3838 600mA 60 100MHz
BFY81 45V 50mA 0,5W 200 60MHz 10mV 0,8
BFY82 45V 0,1A 0,5W 120 250MHz 15mV 0,8
BFY91 45V 100mA 0,6W 150 60MHz 5mV 0,9
LM194 35V 20mA 0,16W 700 200MHz 50µV
MAT01 45V 25mA 0,6W 300 450MHz 100µV
MAT02 40V 20mA 0,6W 600 200MHz 50µV
Last edited:
You can still buy the really excellent Analog Devices NPN SSM2212 from here:
Search results for: SSM2212 Bipolar Transistors - BJT – Mouser
Original for sure.
Unfortunatelly PNP SSM2220 gone EOL.
Search results for: SSM2212 Bipolar Transistors - BJT – Mouser
Original for sure.
Unfortunatelly PNP SSM2220 gone EOL.
Looks good but is SMD, would need adapter because at the moment I use sockets. And difficult cooling ?
How much current for input stage??
George Erdi got excellent performance by using a tail current of 0.24 milliamperes. But please remember that he was forced to use an NPN differential pair, because the PNPs available on monolithic ICs, at the time, were much worse. However that was forty years ago and technology marches on. Today, with Zetex discrete devices, you can get PNPs that are measurably better than the best discrete NPNs.
Attachments below.
_
George Erdi got excellent performance by using a tail current of 0.24 milliamperes. But please remember that he was forced to use an NPN differential pair, because the PNPs available on monolithic ICs, at the time, were much worse. However that was forty years ago and technology marches on. Today, with Zetex discrete devices, you can get PNPs that are measurably better than the best discrete NPNs.
Attachments below.
_
Attachments
1mA is ideal for these duals. Parameter stability is much more important, than high hfe.
You can buy the Metal Can 6-Lead TO-78 MAT12 here:
https://www.mouser.com/ProductDetai...GAEpiMZZMsjJi7B1kCaqenKWtpV4eP8tt1V3Olx%2Bak=
(Price is high, but original for sure.)
I used to use BD139 for CCS. BD139 may be better than BC550 if original and not fake chinese. For simple but decent sound I would use a cascoded OnSemi MJE340G.
If you stick with TO-92, use OnSemi MPSA06 instead of BC550.
Here are three of the NPN transistors mentioned in post #15.
I measured them on my "Intelligent Curve Tracer by Locky-Z" at 8 milliamps of collector current. These aren't simulations, they are real devices that I purchased myself from Mouser and DigiKey, measured by me in my home workshop. Collector current (in milliamps) on the vertical axis, collector voltage (in volts) on the horizontal axis.
When building constant current sources, flatter curves are more desirable, since a textbook ideal constant current source has Iout = (a constant), regardless of Vce. In other words a perfectly horizontal, flat line.
(Of course there exist very well known circuit design stunts, which increase the flatness (== "output impedance") of constant current source circuits tremendously. But why not do both? Pick a flat transistor AND ALSO improve its flatness some more by using clever circuitry.)
_
I measured them on my "Intelligent Curve Tracer by Locky-Z" at 8 milliamps of collector current. These aren't simulations, they are real devices that I purchased myself from Mouser and DigiKey, measured by me in my home workshop. Collector current (in milliamps) on the vertical axis, collector voltage (in volts) on the horizontal axis.
When building constant current sources, flatter curves are more desirable, since a textbook ideal constant current source has Iout = (a constant), regardless of Vce. In other words a perfectly horizontal, flat line.
(Of course there exist very well known circuit design stunts, which increase the flatness (== "output impedance") of constant current source circuits tremendously. But why not do both? Pick a flat transistor AND ALSO improve its flatness some more by using clever circuitry.)
_
Attachments
Many Thanks Mark for your measurements!
For myself I've built the curve tracer adapter that can be seen in the attached pdf. Possible benefit of these curve tracers are that you can observe thermal loops like on Fig. 7 in the tracer article or on the left photo in my post here: https://www.diyaudio.com/forums/solid-state/5928-comments-thermal-distortion-2.html#post6339797 (BC560C on the left vs Hitachi 2SA1081 on the right) [You need to click on the photos and look at them in a larger size to see the thermal loops on the left.]
Not only Ic vs. Vce is not constant for BC550, but even its thermal distortion is high. Inherently brings more thermal distortion, no matter how high is its hfe.
For myself I've built the curve tracer adapter that can be seen in the attached pdf. Possible benefit of these curve tracers are that you can observe thermal loops like on Fig. 7 in the tracer article or on the left photo in my post here: https://www.diyaudio.com/forums/solid-state/5928-comments-thermal-distortion-2.html#post6339797 (BC560C on the left vs Hitachi 2SA1081 on the right) [You need to click on the photos and look at them in a larger size to see the thermal loops on the left.]
Not only Ic vs. Vce is not constant for BC550, but even its thermal distortion is high. Inherently brings more thermal distortion, no matter how high is its hfe.
Last edited:
By "better" do you mean noise figure ? or some other figure of merit involving noise, e g nf * Cin or power or something like that? Is this because the n doped base has higher e- mobility than the p base in an npn ?
P.S. I greatly enjoyed your guest lectures many years ago to my EE371 class, iirc you talked about ecl logic ...
Thanks for the curves, is there a explenation why MJE340 is better than BD139 ?