I received 100 pcs each of On Semi MPSA42 & MPSA92. I measured the hfe of the devices using a Mastech MS8230B. I also had some older National Semi MPSA42’s so I tested them too. I’m not certain of the exact circuit this digital meter uses internally for this test but here are the results. My obvious purpose was to match NPN and PNP pairs but the results were not as good as I expected.
I am very interested in hearing your comments as this is the first time I have done this type of measurement.
Cheers,
Shawn.
On Semi MPSA92 PNP
hfe - QTY
120 - 2
118 - 3
117 - 5
116 - 5
115 - 2
114 - 3
112 - 9
111 - 11
110 - 8
109 - 24
108 - 13
107 - 12
105 - 2
On Semi MPSA42 NPN
Hfe - QTY
101 - 5
97 - 6
95 - 13
94 - 2
93 - 12
92 - 17
91 - 7
90 - 5
89 - 12
87 - 6
85 - 4
65 - 2
64 - 1
53 - 2
48 - 1
47 - 1
46 - 1
38 - 1
35 - 1
National Semi MPSA42
hfe - QTY
274 - 1
252 - 1
246 - 1
226 - 1
129 - 1
114 - 1
97 - 1
87 - 1
85 - 1
74 - 1
57 - 1
56 - 1
50 - 1
47 - 1
I am very interested in hearing your comments as this is the first time I have done this type of measurement.
Cheers,
Shawn.
On Semi MPSA92 PNP
hfe - QTY
120 - 2
118 - 3
117 - 5
116 - 5
115 - 2
114 - 3
112 - 9
111 - 11
110 - 8
109 - 24
108 - 13
107 - 12
105 - 2
On Semi MPSA42 NPN
Hfe - QTY
101 - 5
97 - 6
95 - 13
94 - 2
93 - 12
92 - 17
91 - 7
90 - 5
89 - 12
87 - 6
85 - 4
65 - 2
64 - 1
53 - 2
48 - 1
47 - 1
46 - 1
38 - 1
35 - 1
National Semi MPSA42
hfe - QTY
274 - 1
252 - 1
246 - 1
226 - 1
129 - 1
114 - 1
97 - 1
87 - 1
85 - 1
74 - 1
57 - 1
56 - 1
50 - 1
47 - 1
Attachments
hfe continued
I didn't stop there as I ordered 15 pcs each of On Semi MJE15032
& MJE15033. The results:
MJE15032_ _MJE15033
hfe - qty _ _ _hfe - qty
119 - 1_ _ _ _139 - 5
112 - 1_ _ _ _138 - 4
109 - 1_ _ _ _137 - 6
104 - 1
99 - 1
80 - 2
79 - 2
78 - 1
77 - 3
76 - 2
So when you match bipolars, if you do match them at all, do you use hfe in the selection process? Do you pair them as close as you can? Or do you match them for exact specs? Do you buy your semiconductors by the thousands when you only need 10?
It's new to me, I like the concept but geez this could be expensive to go all out 100% everytime?
Your thoughts and opinions are appreciated.
Shawn.
Edit: Better looking table.
I didn't stop there as I ordered 15 pcs each of On Semi MJE15032
& MJE15033. The results:
MJE15032_ _MJE15033
hfe - qty _ _ _hfe - qty
119 - 1_ _ _ _139 - 5
112 - 1_ _ _ _138 - 4
109 - 1_ _ _ _137 - 6
104 - 1
99 - 1
80 - 2
79 - 2
78 - 1
77 - 3
76 - 2
So when you match bipolars, if you do match them at all, do you use hfe in the selection process? Do you pair them as close as you can? Or do you match them for exact specs? Do you buy your semiconductors by the thousands when you only need 10?
It's new to me, I like the concept but geez this could be expensive to go all out 100% everytime?
Your thoughts and opinions are appreciated.
Shawn.
Edit: Better looking table.
TomWaits said:
It is not often we can get a perfect match of any NPN PNP.
In a symmetrical input pair,you can match Vbe, for minimum offsets due to Vbe differences.
When you do this, you usually end up with different HFE for the npn/pnp.
This means you get input bias current offset
as the gain is different, and so one pair will need more base current for same collector current.
There are 2 ways to compensate for this bias current offset.
1. Keep the same collector currents in both pairs.
And provide a cancelling little current to the input.
2. Change the current in one pair, until the the base currents are the same.
This will also give ZERO current in input resistor = no offset.
Now also when matching NPN to NPN in a pair
you may end up with different HFE in these 2 transistors,
if you have matched for Vbe.
--------------------
This can happen especially if your transistors are NOT from same BATCH.
This is why it is good to buy for example 100 small signal transistors
in one package, and so try to make sure,
they were made from the same silicon substrate
on the same day and by the same machine!
What will determine the quality of a transistor
is the quality and pureness of the raw material used.
Like with diamond and pearls ... there can be imperfections in the pearl.
Same with the silicon mineral sheets that are used for semiconductors.
When you order you can even ask if you can get all your BC550C from the same original package.
If your supplier can do this, it is good.
Of course this goes for Power Devices, too.
There is a better chance they have similar parameters
if they come from same production unit in a specific day.
Buying a few transistors here and a few there and in different years/months
and maybe even from different brands, manufacturers,
will make it more different to find 2 good matching BC550C.
Regards
lineup
When you buy for example BD139
you can try to find transistors that are sorted.
I think BD139 can be found in these 3 classes
order by HFE:
BD139-10 gain 63-160
BD139-16 gain 100-250
BD139-25 gain 160-400
10 means average gain ~100
16 means average ~160
And if you buy one marked only BD139 it is unsorted.
And the variations can be maximal: 63-400
Compare this to BC547, BC547A, BC547B, BC547C
One BC547, unsorted, can have a gain of 110-800.
It is not like that, that the manufacturer decide:
- Now we will make some BC547C from this silicon piece.
No!
They take the silicon sheet and start making small signal transistors.
Then they check the result and the data of the finished transistors.
If they match the 'BC547C specifications' .. they label them BC547C
Other transistors will have lower hFE and so they are labelled BC547A
It comes from the quality of silicon material used.
What the resulting transistor will be called.
And each marked, labelled transistor
should match the datasheet specifications and tolerances for it.
When it comes to parameters.
lineup
Lineup Audio Transistor Lab
you can try to find transistors that are sorted.
I think BD139 can be found in these 3 classes
order by HFE:
BD139-10 gain 63-160
BD139-16 gain 100-250
BD139-25 gain 160-400
10 means average gain ~100
16 means average ~160
And if you buy one marked only BD139 it is unsorted.
And the variations can be maximal: 63-400
Compare this to BC547, BC547A, BC547B, BC547C
One BC547, unsorted, can have a gain of 110-800.
It is not like that, that the manufacturer decide:
- Now we will make some BC547C from this silicon piece.
No!
They take the silicon sheet and start making small signal transistors.
Then they check the result and the data of the finished transistors.
If they match the 'BC547C specifications' .. they label them BC547C
Other transistors will have lower hFE and so they are labelled BC547A
It comes from the quality of silicon material used.
What the resulting transistor will be called.
And each marked, labelled transistor
should match the datasheet specifications and tolerances for it.
When it comes to parameters.
lineup
Lineup Audio Transistor Lab
lineup said:
Do you bother with hfe or is Vbe more important? I see Leach suggests matching current gain with a little circuit to lower DC offset. Also setting up a curve tracer circuit with a scope is an option too.
I want to match as best as possible, not perfect but good matching. I admit to be a little confussed by the different methods that can be used to match.
Jack's double humps of hfe just throws a wrench in the machine.
Shawn.
I have had the privledge to visit an Intel wafer fab in the 90's. We got to wear "space suits" to visit the production area. It was an intersting experience to say the least. Foggy memories for a young lad on a marketing/training conference.
TomWaits said:
There is no good rule I use.
For a single LTP NPN pair input, which I use in 9 cases out of 10
it is easy:
Match Vbe .... and deal with eventual offset compensation due to hFE, if needed.
See this Topic by lineup
which tells about the best method for provide bias current compensation:
Offset Correction using Bias Compensation
For symmetrical NPN/PNP input, it is a bit more tricky.
1. Try to find complementary that are not too different in Vbe.
Then you have to use one of the 2 methods I already told about.
There is also other ways, methods to cancel any in-symmetricals in these amplifiers.
But those that are more experienced, may tell you more of this.
They who have chosen to build amplifiers, that ARE NOT Very SYMMETRICAL, at all.
In fact, it is somewhat easier to get a normal, un-symmetrical amplifier
to be electrically symmetrical!
This is the paradox
If looking Nelson Pass amplifiers, we can see he prefers non-symmetrical designs, with a few exceptions.
Myself is also a follower of the normal left-right logically better symmetry concept.
Aleph has long been the Top Of Line of Pass amplifiers.
The symmetry used, is more left-right symmetry that is provided by the input pair.
And not any vertical, NPN-PNP, so called symmetry attempt.
I attach an Aleph-mini amplifier schematic to show this.
lineup
Attachments
Symmetrical differential topology is a kind of topology which can kill offstet voltage well.
If load of input stage is about 3K and more, offset will be very small. Before, I used to do like that. At that time, I never matched tst but I never met "offset voltage problem".
Near day, I don't use big load for input stage longer and offset is a problem. I chose and matched tst 2N5551/2N5401, and I feel very comfortable although I don't use any extra component.
Yes! Offset voltage can destroy our speakers. You also should study a DC protection circtuit to sure your speakers which never "burn"
If load of input stage is about 3K and more, offset will be very small. Before, I used to do like that. At that time, I never matched tst but I never met "offset voltage problem".
Near day, I don't use big load for input stage longer and offset is a problem. I chose and matched tst 2N5551/2N5401, and I feel very comfortable
although I don't use any extra component. Yes! Offset voltage can destroy our speakers. You also should study a DC protection circtuit to sure your speakers which never "burn"
TomWaits said:
My Understanding is that it is more important to match like devices with like devices. match all the NPN's to each other and all the PNP's to each other.
I have been asking the same questions related to this recently and and found more questions then answers actually. I have a much better understanding of the process but now i have even more questions! Isnt that always the case? the more you learn the more you need to know.
Zc
Zero Cool said:
Certainly and as thanh says, the different topologies may influence the method used to match devices or even no matching at all? I would like to understand why like devices are matched and not complimentary sets of NPN and PNP. I wish I could get my head around that.
As you can see from my measurments it may be a good idea to test the devices anyway to sort out the odd ones?
Cheers,
Shawn.
IMO the important thing is to get a symmetrical output signal(including clipping) from the "black box" that is the amp. If the amp is symmetrical or not in circuit topology make no difference to me. The last circuit I built is asymmetrical(I guess this puts me in the 'Nelson' group.), Av is 65. The Vgs of the input transistors are matched near perfect(took a large sample to find these ) and the DC offset is +18mV with no offset adjustment even in the topology, only 1 pot for Vbe mult. I made a second channel but used a very slightly different Vgs(still within 5% or so) and offset is +29mV. This offset is
OK, but it does illustrate the importance of matching certain transistors depending on how they are used. No doubt that if I just grabbed two from the package, the offset would be significant given the typical variance in jfet Vgs.
), Av is 65. The Vgs of the input transistors are matched near perfect(took a large sample to find these ) and the DC offset is +18mV with no offset adjustment even in the topology, only 1 pot for Vbe mult. I made a second channel but used a very slightly different Vgs(still within 5% or so) and offset is +29mV. This offset is OK, but it does illustrate the importance of matching certain transistors depending on how they are used. No doubt that if I just grabbed two from the package, the offset would be significant given the typical variance in jfet Vgs.
CBS240 said:The Vgs of the input transistors are matched near perfect(took a large sample to find these
OK, but it does illustrate the importance of matching certain transistors depending on how they are used. No doubt that if I just grabbed two from the package, the offset would be significant given the typical variance in jfet Vgs.
Those are great results and no pot to adjust offset.
That is a dream come true. What set up did you use to match the Vgs of your devices. What devices were you using and how many pieces did you start and end with, I'm very curious.Thanks much,
Shawn.
thanh
I also noticed, that these transistors are nearly identical just as you buy them. Hfe variations are around 10-15% and values are very similar between pnp and npn.
P.S. It is sometimes more important to match Vbe. I do it by diode-testing a b-e diode and multimeter displays voltage drop.
I also noticed, that these transistors are nearly identical just as you buy them. Hfe variations are around 10-15% and values are very similar between pnp and npn.
P.S. It is sometimes more important to match Vbe. I do it by diode-testing a b-e diode and multimeter displays voltage drop.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.