Matched transistors might not be matched?

[Flyback]

Hello,

Q6:A and Q6:B of the following schematic (page 2) show a dual PNP being used for high side LED current sensing. This method depends on the Vbe's being matched. This can't be guaranteed, thus it is an inaccurate method. Do you agree?
The MMDT5401 datasheet just says its a dual PNP transistor, not "matched " transistors. There is no reason to believe that their Vbe's will be matched? BJTs are manufactured with attention being payed only to achieving high hfe, and matching of parameters such as vbe, does not have any interest to the manufacturers?

Surely R40 should be 21K, because the vbe's of those transistors will not be the same if there Ic's are different, and in that cicuit, their Ic's are different?

schematic on page 2:
https://www.ti.com/lit/ml/slur512/slur512.pdf

Dual PNP datasheet (MMDT5401)
https://www.diodes.com/datasheets/ds30169.pdf

 

[audioguru]

The datasheet does not say a single word about "matching" so the dual transistor has two completely separate (and maybe very different) transistors inside.
Even their temperatures might be different.

 

[Flyback]

thanks, that's what I thought, and I cannot understand why they have used that circuit to measure led current?

 

[audioguru]

I cannot understand why they have used that circuit to measure led current?

Texas Instruments hires little kids (family?) too, you know.

 

[tvtech]

audioguru

Always ignore Flyback's posts. Or if you want....tell him he is talking Crap.
He will still give you a "like". No matter what you say. You can tell him to go away and stop "liking" you. You can even swear at him. He will still "like" you

Either one of three things:

1. A marketing genius
2. A bot
3. A lost soul.

If I want "likes", I tell him he is basically stupid. Still "likes" the post though ????

Marketing genius..../bot....../lost soul???

Nobody knows yet. ...but we shall surely find out. Only a matter of time.

Flyback is under investigation. There are other Forums he posts the same stuff...

tvtech

 

[4pyros]

Like me

 

[ronsimpson]

The data sheet is not clear. National used to include a picture of the die.
If there are two transistors on the same die they will be very close to the same. (maybe not matched but close) The temperature will be the same thus making the VBE track.
If there are two dies then they could be different.

 

[Flyback]

do you agree that because the collector currents are obviously not equal, there is no way the vbe's can be equal anyway, even if they do happen to be matched.?
Also, the fact that they haven't stated they are matched makes me feel sure that these are not matched, as otherwise they would surely state it so as to increase sales?
Also, Q6:B has its collecter and base tied together, is this to ensure it never saturates?..and if so, why should Q6:B never saturate?

Also, presumably R39 is chosen in value such that Q6:A doesn't saturate, and also so that Q6:A doesn't dissipate too much energy?

 

[MrAl]

Hello Flyback,

If you want matched transistors then you would at least in the past buy a set of matched transistors, or a package with two matched transistor inside. I dont know if you can get this anymore though. I had an older meter with a matched set inside used for measurements but unfortunately i discarded it a long time ago.
But these days there are usually other techniques used anyway, such as an op amp maybe FET input.

 

[ronsimpson]

do you agree that because the collector currents are obviously not equal, there is no way the vbe's can be equal anyway, even if they do happen to be matched.?
Also, the fact that they haven't stated they are matched makes me feel sure that these are not matched, as otherwise they would surely state it so as to increase sales?
Also, Q6:B has its collecter and base tied together, is this to ensure it never saturates?..and if so, why should Q6:B never saturate?

Also, presumably R39 is chosen in value such that Q6:A doesn't saturate, and also so that Q6:A doesn't dissipate too much energy?

OK so you have never studied current mirrors.
There are two or more ways to explain current mirrors. I could take a page with many numbers. I will use the method with little math.

See drawing. If we do not have Q2 then; current from Vcc passes through the resistor and into the base of Q1. (current gain = 100) The resistor limits the current to 100uA. 1uA in the base will allow the transistor to pass 100uA in the collector. Cruddy speaking 1uA in the base and 99uA in the collector. "Why the C is tied to the B". Well at first all current went into the base which turns on the transistor until the C steals away most of the current. (The C can pull down to a point it takes most of the current)

Now add Q2. If the Vbe is the same for Q1 & Q2, then Q2 will also get 1uA. Now Q1c=98uA. Q1B=1uA, Q2B=1uA and thus Q2 will turn on to the point Q2C will be 98uA.

This current mirror are every where. It is used in ICs. You will find many in a 741 op-amp. The transistors are made on the same die and thus will have matching Vbe and matching B.

I don't have time to explain well the circuit from TI. The Vbe does not have to match. (I think it is matched "OK" because the two transistors are on the same die) Even if they are not matched the TI circuit will not be effected too much. In the TI circuit there is a 1 ohm resistor between the two emitters and current flow through the LED will cause 1 volt difference between the emitters. There also is a 499 ohm resistor in one of the emitters. With one volt across the 1 ohm resistor the transistor pair will force a equal 1 volt across the 499 resistor to match. (any time you add emitter resistor to a 'current mirror' you decrease the need for Vbe to match) With out the resistors a 1mV mis-match will have a effect. Now with 1 volt across the resistors the 1mV error will be about 1% error. I don't have time to do the math but the Vbe effect is greatly reduced by adding resistors.

 

[Flyback]

thanks,
I have to say I don’t recognise the MMDT5401 (dual PNP current monitor) circuit as a current mirror, because the current in the reference branch is not equal to the current in the “mirror” branch?
I am familiar with adding resistance to effect “emitter degeneration” in current mirrors that are attempting to equalise, for example, parallel LED currents, but the useage seen here (in the top post circuit) does not seem to tally with that use.
MMDT5401 is just 8 cents (quantity = 3000) on digikey.
Diodes.com also make a high side current monitor IC called ZXCT1082…The ZXCT1082 costs 68 cents on digikey (quantity = 3000).

ZXCT1082 current monitor IC DATASHEET: (admittedly only does to 60V)
https://www.diodes.com/datasheets/ZXCT1082_87.pdf
If the Dual PNP (MMDT5401) circuit of the schematic in the top post really is “bona fide”, then why do diodes.com also make the ZXCT1082, given that it costs eight times as much? This tends to make me believe that the dual PNP (MMDT5401) current monitor sub-circuit of the top post is very poor in comparison to the ZXCT1082?
I think it might be preferable to use matched transistors as follows ? (please see below schematic) –at least they would be matched transistors, and they only cost 15 cents. (I had to use two due to their 45V voltage rating)
PMP5201 DUAL PNP DATASHEET: (15 cents @ 3000 pieces on digikey)
https://www.nxp.com/documents/data_sheet/PMP5201V_G_Y.pdf

(LT spice sim also attached)

 

[large_ghostman]

Hello Flyback,

If you want matched transistors then you would at least in the past buy a set of matched transistors, or a package with two matched transistor inside. I dont know if you can get this anymore though. I had an older meter with a matched set inside used for measurements but unfortunately i discarded it a long time ago.
But these days there are usually other techniques used anyway, such as an op amp maybe FET input.

I have seen, new/old stock matched transistor's for sale on ebay, they were still in the packets(little box's) I didnt buy them though because the cost was high!

 

[MrAl]

Hi,

Oh ok. Well i forgot to mention the ones in my meter were matched FET's, not bipolar, but Flyback managed to find some cheap matched PNP's that might be interesting.

 

[Flyback]

Even if the bjts are perfectly matched, like in this simulation, the circuit has inaccuracy...here the 350mA of current is measured as 322mA....and that's with the perfectly matched transistor models of the ltspice simulator (schem and sim attached).
In fact , you can see that because of the slightly different Ic's, there is 25mV of difference in their Vbe's.
I am dubious about this method of measuring high side current.

 

[ronsimpson]

here the 350mA of current is measured as 322mA

It really does not matter what the current transfer ratio is as long as it is consistent.
Every one made probably needs to read the same within 10 or 20% not 1%.
If the transistor pair read 322mA in 100mA out but all transistors did that it will be fine.

 

[MrAl]

Hi,

My question would be is the 'reading' a constant ratio of the supposed 'true' reading?
In other words, if we expected to see 100ma but see 90 ma instead, then we expect to see 200ma but see 180ma instead, then we expect to see 400ma but see 360ma instead, then that means that the transfer gain is constant and we can make up for the incorrect gain some other way. It's only when it is non linear that it becomes more of a problem.

So what you have to do is take more data. You have to apply different levels of currents and measure the output and see if there is a constant ratio or if it varies, and if it varies then by how much. Sometimes it will vary a little but it ends up being acceptable anyway. But one single measurement isnt enough you need at the very very least two measurements but even better three or more different data points: low, mid, and high.