MJL3281 Vbe matching

Status
Not open for further replies.
I just tested over 50 MJL3281. I set up to run about 194mA through the transistor with Vcb=3V. I ground the base and use a 100ohm resistor to pull the emitter to -20V.

I bought from different three different vendors so they are likely not from the same lot. I got deviation from 0.636V to 0.639V in all 50 of them!!!

majority of them are group into 0.637V and 0.638V with a few 0.639V and 0.636V.

Wow!!!! This is tight matching. Even 4mV difference in the 50 is not bad at all. I saw 7 to 8mV difference in my stock Acurus power amp. I assume they had tried to match the transistor before putting in. I can easily match to 1mV with the MJL3281.

I first did about 20 with 100mA bias, they are just as tight. I decided to up the current to 200mA as this is the bias I will run my transistors.

I am going to do the MJL1302 next as I have about 40 of them. I'll report back.
 
I tested 45 of the MJL1302 at 196mA. They are from 0.622V to 0.626V. So it's still 4mV.

I don't remember are they two different badges of single. More likely two different badges. they are really good!!!

FYI, I try to do it as consistent as possible.

1) I keep the power supply on through out. I measure direct between the Base and Emitter pin of each transistor. This means I eliminates the contact resistance between the transistor pins and the socket. This turned out to be a big problem, I had to redo over 20 of the transistors as I notice if I measure at the socket, I got about 20mV drop and vary when I jiggle the transistor to change the contact surface. so I finally had to probe directly on the pin of the transistor, then it's much more consistent and repeatable.

2) I even make sure the time between plugging in the transistor to measurement is roughly the same so the transistor do not heat up differently from part to part. Of cause it is only accurate in terms of " one seasamy, two seasamy " counting from the heart only!!!

This is pretty amazing. I remember back in the days when I design BJT IC, we can only count on 3mV matching between transistors on the same die even when they are reasonably close together. I got 4mV matching on all of them. I shouldn't even buy so many of them as I thought I need a large quantity to find 5 good matching transistors badge. Now, I have enough transistors to make a few amps!!!

Now my next step is to find matching input transistor for the LTP input stage. I bought at least 100 of each for matching!!!

Also, don't take for granted they have those MJL in stock. Digikey did ran out of MJL3281, the lead time was over a month. That's the reason I bought extras from other places to end up having over 60 of them.
 
Last edited:
I just tested over 50 MJL3281. I set up to run about 194mA through the transistor with Vcb=3V. I ground the base and use a 100ohm resistor to pull the emitter to -20V.

I bought from different three different vendors so they are likely not from the same lot. I got deviation from 0.636V to 0.639V in all 50 of them!!!

majority of them are group into 0.637V and 0.638V with a few 0.639V and 0.636V.

Wow!!!! This is tight matching. Even 4mV difference in the 50 is not bad at all. I saw 7 to 8mV difference in my stock Acurus power amp. I assume they had tried to match the transistor before putting in. I can easily match to 1mV with the MJL3281.

I first did about 20 with 100mA bias, they are just as tight. I decided to up the current to 200mA as this is the bias I will run my transistors.

I am going to do the MJL1302 next as I have about 40 of them. I'll report back.
Can you post a draft schematic?
 
Can you post a draft schematic?

The measurement MUST be measured directly on the Base and Emitter pin of the transistor. When I use breadboard that I can plug in transistors, too much voltage drop across the contact between the pin of the transistor and the socket. I measure up to 30mV just across the contact and it changes when I jiggle the transistor. So it is a must to measure directly on the pins.

I pull 194mA as this will be my bias point, increase the resistor to lower the current.

Keep the -V at 20V to null out the contact drop.
 

Attachments

  • MJL match test.jpg
    MJL match test.jpg
    22.4 KB · Views: 1,005
The measurement MUST be measured directly on the Base and Emitter pin of the transistor. When I use breadboard that I can plug in transistors, too much voltage drop across the contact between the pin of the transistor and the socket. I measure up to 30mV just across the contact and it changes when I jiggle the transistor. So it is a must to measure directly on the pins.

I pull 194mA as this will be my bias point, increase the resistor to lower the current.

Keep the -V at 20V to null out the contact drop.
2 separate power supplies ?
 

Attachments

  • MJL match test.jpg
    MJL match test.jpg
    26.3 KB · Views: 959
Last edited:
Your results look very good.
4mVbe variation across a batch of 50 seems exceptional.

But it makes me wonder if the feedback across the 100r of Re is masking some of the variation.
Looking at the two extremes:
20V across {100r+622mVbe} compared to 20V across {100r+626mVbe} indicates a change in current of only (193.78mA-193.74mA) 40uA (0.02%) and seems to be inconsequential. The method has ensured the power dissipated is substantially the same and thus Tj is substantially the same.

Would it be too much to ask that you assemble a selected pair that appear to be identical (both Vbe and hFE) using your matching and set up two in parallel.
Use a small value base resistor (1r0) and a zero value emitter resistor.
Add a pair of collector resistors so that you can measure the current down through each device. 1r in each collector allows good resolution using the 199.9mVdc scale of your DMM. (match all your 1r0 resistors before the jig assembly)
Apply a Vbe from a voltage source to both of those base resistors. 4mA with an hFE of 100 gives your 2times 200mA outputs. That 4mA could be from your 3V supply via a single resistor (3-0.62}/0.004 = 595r Use either 560r or 620r. You don't need a dual polarity supply. A trimmer either in parallel, or in series, with the 560r can be used to adjust until the total collector current is exactly the same as in your first matching method.

If the previous matching is as good as the results indicate, then the collector currents should be very close between the parallel pair.
 
Last edited:
I don't think the collector current is necessary matching. That, really depends on the beta of the transistor.

In most of the cases of power amps, it is more important to match the idle current of the transistors and develope 26mV across the emitter resistor of each of the power transistor. It is my understanding that matching Vbe is the important part of ensuring voltage drop of each of the emitter resistor of all the transistors is closely matched and be adjusted to 26mV to achieve lowest crossover distortion.

Maybe there is advantage of matching beta ( collector current), I just have not run across any article talking about it.

It might be interesting to test the collector current. But I have been very busy. My former company that I retired from 10 years ago called me back for a contracting job two months ago. I have been working day and night lately. All my pcbs, chassis and parts of my dream audiophile amp has be just sitting there collecting dust in the past months. I just sent the pcb for the company out two days ago and have a few days to catch my breath, so I did the matching to pick transistor to solder onto the OPS board. It's going to be a while before I can get to testing the collector current.

It's a shame seeing all my parts, pcb and all sitting in the room idling. BUT, when people offering me money, it's hard to refuse. I'll be back to the forum, I still have passion in the power amp, just the greens has a stronger pull for now!!!:D
 
I don't think the collector current is necessary matching. That, really depends on the beta of the transistor.
it's the Vbe that drives the output. Once the output is determined the hFE determines the base current required, not the other way around. The low value base resistors, similar to what might be used in a circuit, allow base current to be checked, after checking the collector current for reasonable match.
In most of the cases of power amps, it is more important to match the idle current of the transistors and develope 26mV across the emitter resistor of each of the power transistor. It is my understanding that matching Vbe is the important part of ensuring voltage drop of each of the emitter resistor of all the transistors is closely matched and be adjusted to 26mV to achieve lowest crossover distortion.
I think you have read it correctly, Vbe matching is important in paralleled devices, yet many match by hFE instead. I would argue that both be matched at the chosen idle current.
Maybe there is advantage of matching beta ( collector current), I just have not run across any article talking about it.

It might be interesting to test the collector current. But I have been very busy. My former company that I retired from 10 years ago called me back for a contracting job two months ago. I have been working day and night lately. All my pcbs, chassis and parts of my dream audiophile amp has be just sitting there collecting dust in the past months. I just sent the pcb for the company out two days ago and have a few days to catch my breath, so I did the matching to pick transistor to solder onto the OPS board. It's going to be a while before I can get to testing the collector current.

It's a shame seeing all my parts, pcb and all sitting in the room idling. BUT, when people offering me money, it's hard to refuse. I'll be back to the forum, I still have passion in the power amp, just the greens has a stronger pull for now!!!:D
The emitter resistor creates feedback. A collector resistor does not create that feedback. It becomes the load that can be used to check matching and not complicate the comparison by generating a "hidden" feedback that papers over differences in the devices being compared.

The emitter resistor is useful in the completed paralleled output.
It creates that feedback that helps with matching of device currents when the output current requires the device current to be different from the idle current.
The emitter resistor is also useful in the completed circuit to stabilise the temperature gain. R.Cordell explains this in his interviews Thread.
 
Last edited:
first, my transistors are all TO-264. It is MJW, not MJL

I did beta testing. I run 387mA through the transistor with Vcb=5V. I put a 2.2ohm in series between the base of the transistor and ground. I measured the voltage drop across the 2.2ohm right on the resistor leads to avoid any drop between the resistor lead and the breadboard socket.

I finished the MJW3281. most of them are between beta of 97 to 100. I can easily pick a set of 5 that match Vbe to 1mV and beta to 1%.

Only a few with beta as low as 96. So the worst spread is less than 5 or 5% for over 60 of them from different vendors.

I need to do the MJW1302 later.


Question is what is a good match? How many mV matching of Vbe is consider matched? how many % difference of beta is consider matched?

Also, do I need to match between NPN and PNP in the push pull pair. I don't really see the need.
 
Just finished the MJW1302. variation is a lot greater.

I have 4 with beta only 96 to 101

A few 116 to 120

A lot of them are 125 to 130.

So the questions are:

1) How close do I have to match for a set?

2) All the MJW3281 are between 96 to 100. Is it ok to match with MJW1302 with beta 125 to 131 in the push pull stage?

that's disappointing.
 
for a push pull output stage you don't need a Vbe match across the N & P sides.
There is a benefit in getting fairly close for hFE. That determines the draw on the drivers and the drivers need to see the same or similar load to perform to their design brief.

For parallel sets, you need very good Vbe match and just good hFE match.

I would aim for Vbe matching to better than+-1mVbe, but preferably +-0.5mVbe (i.e the same millivolt).
hFE in the N of +-2% and similar in the Pdevices.

Your difference between N & P will result in a different driver load.
This will ripple back through previous stages.
Maybe by using a slight mismatch in the drivers' hFE could cancel the error, so that the previous stage sees an almost equal N & P current draw ?
In the extreme, a big mismatch in base currents will severely affect the way the VAS and it's sink pass current. Look very carefully at what this VAS current effect is when the downstream devices are not matched.
 
Last edited:
unmatched -

There are two on this amp I am tempted to replace , but it does not
seem to be too bad.
Stays close to these readings cold startup to after heavy use.

ON is pretty consistent , devices are njw1302/3281 -72V rails/.22R re.
Edit - same outputs were about 1mv spread @ 45V rails.

OS
 

Attachments

  • slew200.jpg
    slew200.jpg
    123.6 KB · Views: 453
Last edited:
There are two on this amp I am tempted to replace , but it does not
seem to be too bad.
Stays close to these readings cold startup to after heavy use.

ON is pretty consistent , devices are njw1302/3281 -72V rails/.22R re.
Edit - same outputs were about 1mv spread @ 45V rails.

OS

From my experience, watch out the PNP. There were 4 of them that the beta are way off. The beta was only 96 to 100 compare to the rest that mostly are over 125. That's a big difference. I bought in two separate batches, meaning they are mixed into the the batches, it's not like I have 20 of those that you can say one batch is 125 and the other is 97. So even if you buy one batch, there might be a few that are way off.

Now I have 4 of them, not enough to make up a set of 5 for my amp.

The Vbe do match very tight.
 
From my experience, watch out the PNP. There were 4 of them that the beta are way off. The beta was only 96 to 100 compare to the rest that mostly are over 125. That's a big difference. I bought in two separate batches, meaning they are mixed into the the batches, it's not like I have 20 of those that you can say one batch is 125 and the other is 97. So even if you buy one batch, there might be a few that are way off.

Now I have 4 of them, not enough to make up a set of 5 for my amp.

The Vbe do match very tight.

Yes , that bottom row is the P's notice the 13.8/11.4mv readings.
The P's were the worst on the other amp , as well.

OS
 
8 x Matched Genuine Sanken 2SC6011A 2SA2151A Transistor | eBayHallo OS.
I'm on working Prozess to finish the Slewmaster OPT- Board.
There for i am searching good matched BJT's.
Is it possible to use These- ebay devices. Not shure, can i trust tubeshunter- you only
can see same Produktion Code.
What do you ore other mean?
Transistors have good technical data with low Cob- 450pF.
Thanks and Regards.
 
Status
Not open for further replies.