I am trying to design a discrete headphone amp with BJT differential input, but it's hard to find matched pair transistors models which are still in production, I find that Nexperia is still producing quite a few models like BCM856BS, BCM62B, but their Vce-Ic curve isn't that flat, SSM2212 is an expensive choice which seems to perform really well, but I cannot find its Vce-Ic curve in the datasheet.
SO actually my question is: Is a flat Vce-Ic curve needed for matched pair transistors used in differential input stage? My guess is that since Ic is fixed in a BJT differential input stage with constant current-source on the emitter side, Vce-Ic curve doesn't matter that much. Or did I miss anything?
SO actually my question is: Is a flat Vce-Ic curve needed for matched pair transistors used in differential input stage? My guess is that since Ic is fixed in a BJT differential input stage with constant current-source on the emitter side, Vce-Ic curve doesn't matter that much. Or did I miss anything?
When it is not more or less flat, that means the output resistance of the differential pair is not very high. What impact that has on the rest of the circuit depends on the rest of the circuit. It could have almost no impact or it could cost you some loop gain.
BCM56DS and BCM53DS are also still in production, or at least they were a few years ago.
Last edited:
These bjt's appears to me as rather high current devices for an input diff pair, ranging from 40 up to 200mA where one would expect only 1/10 of those values. Why these specific, not very suitable bjt's for this input job?
If you do want to proceed with them, the output impedance ranges from approx 450Ω (lowest curve) to130Ω (top curve), so as transconductance devices, from an input voltage to an output current, the load should be better then less of 1/10 to remain the 'current source' behaviour at the collectors. Very challenging.
And then the rest of the circuit (#3).
If you do want to proceed with them, the output impedance ranges from approx 450Ω (lowest curve) to130Ω (top curve), so as transconductance devices, from an input voltage to an output current, the load should be better then less of 1/10 to remain the 'current source' behaviour at the collectors. Very challenging.
And then the rest of the circuit (#3).
I mentioned them because they are in production, they are cheap, they match well, they can handle reasonably high voltages and they have a reasonable hFE, which doesn't drop much at 0.1 mA according to the Nexperia BCM56DS datasheet. The output resistance will increase when you reduce current.
They are supposed to have the same dies inside as BC639 and BC640. Some time ago, I measured 6.1 ohm base resistance for the BC639, and 3.5 ohm for the BC640, which are really low values. That only matters when input noise voltage is critical, like in a microphone or a MC amplifier.
They are supposed to have the same dies inside as BC639 and BC640. Some time ago, I measured 6.1 ohm base resistance for the BC639, and 3.5 ohm for the BC640, which are really low values. That only matters when input noise voltage is critical, like in a microphone or a MC amplifier.
Last edited:
I misinterpreted @Citizen124032 's post, I thought the comment referred to the BCM56DS/BCM53DS, but it related to the BCM856BS. For BCM856BS, output resistance will also increase at lower current, it's roughly inversely proportional to current.
Matching individual transistors to match beta is a pain. You also have to do it a certain way or what you think is a match - isn't even close. So purchasing a dual transistor intended for differential pair use is a great idea, even though they are expensive. I normally buy parts in 100 lots to get a good yield for matched devices. So compare that to the time and the cost of the jig and equipment.
lol!, the 2N2915 is an old part for sure. Transistors became better as time went on, so use something that is current production. Also, noise is important so pick a quiet pair. Also pay attention to breakdown voltage, you might want to cascode them.
lol!, the 2N2915 is an old part for sure. Transistors became better as time went on, so use something that is current production. Also, noise is important so pick a quiet pair. Also pay attention to breakdown voltage, you might want to cascode them.
I gave a short talk at Burning Amp called BJT Simple Matcher -- test jig presented at 2024 Burning Amp Festival
It presented one way to test individual thru-hole transistors and find matched pairs. The whole slide deck is available at that link; I've copied a couple of the figures below.
This approach might be preferred when you believe you absolutely MUST use transistors whose Ic-Vce curve is extremely flat (---> whose Early Voltage is extremely large), flatter than any dual transistor you can find easily. To name one example, perhaps the ZTX857 shown below. Its measured Early Voltage is pretty high: 540 , but it's not available as a dual. Then purchase a stockpile of ZTX857 parts and perform your own tests to find pairs which are excellent matches.
_
It presented one way to test individual thru-hole transistors and find matched pairs. The whole slide deck is available at that link; I've copied a couple of the figures below.
This approach might be preferred when you believe you absolutely MUST use transistors whose Ic-Vce curve is extremely flat (---> whose Early Voltage is extremely large), flatter than any dual transistor you can find easily. To name one example, perhaps the ZTX857 shown below. Its measured Early Voltage is pretty high: 540 , but it's not available as a dual. Then purchase a stockpile of ZTX857 parts and perform your own tests to find pairs which are excellent matches.
_
Attachments
Hi Mark,
Is that one similar to the matcher I designed? It holds the pair at the same temperature and simply shows a match in collector current, a CCS determines the current through the pair. Exactly like a diff pair in an amplifier. The meter in my design is simply a null detector. You could use it to figure out beta if you wanted by measuring other points provided on the PCB.
You cannot determine a match unless the two parts are at the exact same temperature, and your jig appears to do that.
Is that one similar to the matcher I designed? It holds the pair at the same temperature and simply shows a match in collector current, a CCS determines the current through the pair. Exactly like a diff pair in an amplifier. The meter in my design is simply a null detector. You could use it to figure out beta if you wanted by measuring other points provided on the PCB.
You cannot determine a match unless the two parts are at the exact same temperature, and your jig appears to do that.
Why don’t you just cascode the diff pair since that will fix Vce?
The BCM devices are readily available and cheap so make an ideal diff amp pair. The other option is to go for a JFET input - the LSK389 or JFE2140 (available at Mouser) but these are not cheap - especially the LSK devices. The Vos in a completed amp is surprisingly tight indicating good matching.
The BCM devices are readily available and cheap so make an ideal diff amp pair. The other option is to go for a JFET input - the LSK389 or JFE2140 (available at Mouser) but these are not cheap - especially the LSK devices. The Vos in a completed amp is surprisingly tight indicating good matching.