matching 2N4401

Status
Not open for further replies.
Jump to Latest
R

robmil

Member
Joined 2004
#1
Hi all

I'm in the process of matching 8 x 2N4401s for an MC input stage I'd like to try out. These 4401s will be in parallel.

The questions are:

1. how closely matched do they have to be
2. if not closely matched I have read somewhere that putting a 1ohm resistor on the emitters will negate ultra close matching.

Any advice or small pictures would be helpful on this one.

rgards Rob
 
#2
robmil said:
Hi all

I'm in the process of matching 8 x 2N4401s for an MC input stage I'd like to try out. These 4401s will be in parallel.

The questions are:

1. how closely matched do they have to be
2. if not closely matched I have read somewhere that putting a 1ohm resistor on the emitters will negate ultra close matching.

Any advice or small pictures would be helpful on this one.

rgards Rob
Click to expand...

18mV will cause ~2:1 (.0259*ln(2)Volts) current hogging, 1 Ohm won't help much in an input stage since you need 50mV or so drop to do much of anything (50mA per device).

.0259*ln(how-good-you-want-the-current-shared) is the Vbe match you want. You need to be very careful not to heat them up with your fingers while matching them, only a couple of degrees C will obviate the exercise.
 
Last edited:
#3
Hi Scott

Thank you for the reply.
That went right over my head!

In lay-mans terms please. I'm using 4 devices per channel but maybe 8 devices per if I feel brave. Why 50mV or 50mA per device? What is 0.0259* In ??

thanks
 
#4
robmil said:
Hi Scott

Thank you for the reply.
That went right over my head!

In lay-mans terms please. I'm using 4 devices per channel but maybe 8 devices per if I feel brave. Why 50mV or 50mA per device? What is 0.0259* In ??

thanks
Click to expand...

OK, the short story is that 1 Ohm won't help. If you connect the transistor like a diode (collector shorted to base) and run say 1mA through it the Vbe's should match to better than 5-10mV.

25.9mV is the thermal voltage, every doubling of current is about 18mV change in Vbe so the relationship is logarithmic. .0259*ln(2) =~ .018

A fresh 9V battery, 6.8k resistor, a three pin socket, and a DVM should be all you need (use pliers to change parts, the Vbe changes rapidly with temperature).
 
#6
AndrewT said:
Note, that where devices are in parallel, the important parameter that benefits from matching, is Vbe, not hFE.
Click to expand...

The hFE mismatch would make a microscopic contribution in the case of diode connected transistors probably less than 1% (250uV delta Vbe). You could probably use the diode check feature of a handheld DVM and get close enough, I think they are around 100uA test current.
 
#8
Or, if you can use matched pairs consider using the MMDT4401 in a 6 lead package. There are quads in a single package if you can use 2n2222, 2n3904 or my favorite, MAT14 from AD. MAT14 are very well matched 40V, 30 ma devices.
 
#10
stoc005 said:
Or, if you can use matched pairs consider using the MMDT4401 in a 6 lead package. There are quads in a single package if you can use 2n2222, 2n3904 or my favorite, MAT14 from AD. MAT14 are very well matched 40V, 30 ma devices.
Click to expand...

I don't as a rule push our parts, but I agree paralleling the 4 in the MAT14 would be a good choice, Avnet has the lowest price I think right now still a little pricey. You could check THAT Corp. and see if they have an equivalent.
 
#12
20mVbe is a big difference.
Set up a pair that far apart.
Link them like an LTP with the bases connected together.
Measure the collector currents. They will be massively different.
When you find a pair that have very close collector currents then Vce and Pq and Tc and Tj and Vbe are all the same. Now you have a pair @ at that one collector current.
Now try to find pairs that match over a range of currents.
Then the big one, find triplets and/or quads.
 
#14
robmil said:
Well I've just tested 8 x 2n4401s with my DCA55 transistor tester. The VBE measurements were 0.70v to 0.72v. across all 8 transistors. And that was from me just picking out the first 8. Seems like a good batch 🙂
Click to expand...

When you parallel two with ~20mV offset one will be about 2/3 of the total current and the other 1/3. If the total distribution is 20mV It should be easy to group them by +-5mV bins which should be good enough to benefit from paralleling.
 
#16
Vbe matching

Ian Fritz' pdf describes beautifully the DUT and REF comparison method.

I use this LTP comparison method all the time. I believe it is the only way I and similarly resource restricted amateurs can achieve accuracy.

Ian locates his measuring resistors in the emitters rather than the collectors where I place them. So I have copied the pdf to my reference files.
I shall use his exact method on my next batch to see how it compares with mine.

As a further development.
The measuring resistors stay in the Fritz location but they are loaded with a switchable CCS connected to the -ve rail (for the NPN testing, for PNP the CCS is swapped to the +ve rail as Fritz shows in the alternative diagram).

This allows measurement of the "Vbe diff" at a range of fixed currents. Maybe switchable from 100uA to 20mA in E3 steps (1 2 5 10) just 8 positions for that 200:1 current range to suit many lower current devices. 0.1" sil pins with a 0.1" shorting plug allow quick and repeatable switching.
A "big" version could be built up for 20mA to 500mA measuring of Power devices.
 
Last edited:
Status
Not open for further replies.
Top Bottom