Matched transistors

[Mikebits]

Greetings; I have application where I need to make a few differential amps, so Beta match is important. What I am wondering, while not guaranteed, how closely matched do you think 2 transistors are if they are on same die such as in a transistor array? For example: https://www.mouser.com/ProductDetai...GAEpiMZZMshyDBzk1/Wi7dklbCN21lXxt0C%2bj5He94=

One would think they should be pretty close, right? You may be asking "Hey! Mike, what is the reason you don't just purchase matched transistor arrays?"
Well, I have about $8.00 reasons. Device mentioned above $.37, matched array $8.00 Also would diode arrays be a close match for Vf as well.

Mucho Gracias

 

[AnalogKid]

They will be better matched that two random TO-92 or SOT-23 devices, but the word "match" is not on the data sheet anywhere. Also, they will temperature track very well, which usually is more important since that cannot be adjusted out of a circuit the way offsets can.

ak

 

[spec]

Greetings; I have application where I need to make a few differential amps, so Beta match is important. What I am wondering, while not guaranteed, how closely matched do you think 2 transistors are if they are on same die such as in a transistor array? For example: https://www.mouser.com/ProductDetail/NXP-Semiconductors/BC817DS115/?qs=sGAEpiMZZMshyDBzk1/Wi7dklbCN21lXxt0C%2bj5He94=

One would think they should be pretty close, right? You may be asking "Hey! Mike, what is the reason you don't just purchase matched transistor arrays?"
Well, I have about $8.00 reasons. Device mentioned above $.37, matched array $8.00 Also would diode arrays be a close match for Vf as well.

Mucho Gracias

Hi Mike,

Here are some low cost, small signal matched NBJT pairs:

ON (Motorola) NST450
ON NST650
NXP (Phillips) BCM57B
NXP BCM61B
NXP BCM847

spec

 

[schmitt trigger]

I Agree with analog kid, any differences between the pair of integrated devices may be trimmed at room temp.
But tracking over your expected temperature range is far more difficult.

But these two transistors, even if they are not hfe-matched, should temperature track very closely.

 

[Mikebits]

Hi Mike,

Here are some low cost, small signal matched NBJT pairs:

ON (Motorola) NST450
ON NST650
NXP (Phillips) BCM57B
NXP BCM61B
NXP BCM847

spec

Thanks Specster, I will do the Google.

 

[spec]

I Agree with analog kid, any differences between the pair of integrated devices may be trimmed at room temp.
But tracking over your expected temperature range is far more difficult.

Yes, it would be much better if the two matched transistors were on the same substrate, like the more expensive matched transistor pairs.
But as the individual transistors are in the same pak, and a very small pak at that, the temperature of the junctions should be pretty close.

The other aspect of tracking is to ensure that both transistors have the same VCE, the same IC and see the same input and output resistances and impedance. Also the dynamic heating effects need to be sorted. The only practical method is to make the modulation index (dV and dI) as close to zero as possible.

But these two transistors, even if they are not hfe-matched, should temperature track very closely.

With the NST450, for example, the transistors are hFE matched to 10% and the VBE, which is the more important, is matched incredibly closely at 2mV.

One approach is to put the circuits in a small oven and keep that at a constant temperature, say 60 deg C. This is quite easy to do, even if it does sound crazy. This has the advantage that hFE and VBE would be constant as would be the characteristics of all the other components in the difference amp.

spec

 

[spec]

Thanks Specster, I will do the Google.

No probs Mike,

If cheap matched NBJTs and PBJTs turn out to be any good, they may save me some dosh in the future too.

From RS UK the prices range from £0.09 UK to £0.15 UK.

spec

Here are the data sheets:
**broken link removed**
NST 650 data sheet not available and RS UK are awaiting stock
**broken link removed**
**broken link removed**
**broken link removed**

 

[spec]

Also PNP matched pair: ON NST300

https://www.onsemi.com/pub_link/Collateral/NST30010MXV6-D.PDF

 

[BobW]

I had a similar requirement for a pair of well matched transistors, both NPN and PNP. I tried inexpensive NXP unmatched BC847BS (dual NPN) and BC857BS (dual PNP) transistors, and they worked beautifully.

I had previously tried hand matched separate transistors, and glued them together in order to keep them as close together in temperature as possible, but they still drifted terribly with temperature. The dual transistors mentioned above stayed so well balanced that there was no need to spend money on specially matched ones.

 

[Mikebits]

I want to use the transistors to build a Gilbert Cell mixer, and the guy presenting the You tube tutorial showed how Δβ between transistors greatly affected circuit operation.

 

[spec]

I had a similar requirement for a pair of well matched transistors, both NPN and PNP. I tried inexpensive NXP unmatched BC847BS (dual NPN) and BC857BS (dual PNP) transistors, and they worked beautifully.

I had previously tried hand matched separate transistors, and glued them together in order to keep them as close together in temperature as possible, but they still drifted terribly with temperature. The dual transistors mentioned above stayed so well balanced that there was no need to spend money on specially matched ones.

Thks for info Bob- that is encouraging.

spec

 

[spec]

Couple more low cost transistor pairs; theses are 'intrinsically matched' and are 'built with an adjacent die from a single wafer' :
NPN: DIODES DMMT5551 **broken link removed**
PNP: DIODES DMMT5401 **broken link removed**

spec

 

[spec]

I want to use the transistors to build a Gilbert Cell mixer, and the guy presenting the You tube tutorial showed how Δβ between transistors greatly affected circuit operation.

Hy Mike,

I had to Google to find out what a Gilbert cell mixer is: https://www.radio-electronics.com/info/rf-technology-design/mixers/gilbert-cell-mixer-multiplier.php

Being a non RF person, I know them as a four quadrant multipliers (FQMs). They are ingenious devices that can do all sorts of things. You can also get into all sorts of pickles with them; good layout and decoupling are the secret, as you no doubt know.

I'm intrigued though; presumably you are rolling your own Gilbert cell mixer- rather than using one of the excellent integrated circuit versions- for the hell of it.
https://www.analog.com/media/en/training-seminars/tutorials/MT-079.pdf
https://www.analog.com/media/en/technical-documentation/data-sheets/AD834.pdf
**broken link removed**
**broken link removed**
https://www.intersil.com/content/dam/Intersil/documents/ha-2/ha-2556.pdf

What frequency are you dealing with?

spec

 

[BobW]

FYI, my application was also an analog multiplier circuit (used in a wattmeter circuit). With the original separate transistors, I would adjust the DC zero pot when the unit was first switched on, and then you could see it immediately start to drift just from junction heating due to quiescent current. When I switched to the dual transistors, the drift disappeared.

Transistors fabricated on the same die should be inherently well matched. If they aren't, there has to be something seriously wrong with the process.

 

[AnalogKid]

I had to Google to find out what a Gilbert cell mixer is:

Analog design god: https://en.wikipedia.org/wiki/Barrie_Gilbert

ak

 

[spec]

Amalog design god: https://en.wikipedia.org/wiki/Barrie_Gilbert

ak

Thnks AK,

As you say Gilbert is a design god. People like him always amaze me- how do they do it? The surprising thing is that they seem to be able to apply their genius to such a wide range of areas.

spec

 

[spec]

FYI, my application was also an analog multiplier circuit (used in a wattmeter circuit). With the original separate transistors, I would adjust the DC zero pot when the unit was first switched on, and then you could see it immediately start to drift just from junction heating due to quiescent current. When I switched to the dual transistors, the drift disappeared.

Nice

Transistors fabricated on the same die should be inherently well matched. If they aren't, there has to be something seriously wrong with the process.

I am not sure about that, but then I am no expert in semiconductor fabrication. From a human perspective the transistors should be identical if, as you say, they are fabricated on the same die, but at a molecular level they are miles apart. Some of the variables are, penetration, local doping levels, local impurities and mask tollerances.

In view of the fantastic things that the semiconductor manufacturers do inside chips, you would think it would be simple for them to fabricate two transistors with very close matching of all parameters on the same substrate, which would give excellent thermal tracking too. But this seems to be a very expensive process. The cheap transistor pairs, although in the same tiny pak, are all individual (discrete).

If you take a typical small signal BJT, the BC546, there is a wide spread of parameters, especially hFE and VCEmax, between individual transistors even those from the same die (I think). Some transistors families are produced from the same die but are just selected by the manufacturer. This is possibly the case with the ubiquitous BC546 to BC550 family of NBJTs and the complementary BC556 to 560 PBJTs.

The beautiful TI (National) LM194/LM394 super match pair shows what can be done with transistors on the same substrate, but they were always very expensive and are obsolete now- sadly.
**broken link removed**

The Analog Devices MATxxx series of matched transistors are still available but, even though AD are the best analog chip makers, in my humble opinion, their stuff is pricey.

spec

 

[spec]

The Intersil CA3045/CA3046 chip may be worth considering and is not too expensive ($1 US). It has five NPN transistors on a common substrate. Two of the transistors are matched and have their emitters joined, as for a long tailed pair. Texas Instruments also seem to make the CA3046 as the LM3046.

Data sheet
https://www.intersil.com/content/dam/Intersil/documents/ca-3/ca-3046.pdf
**broken link removed**

Application note
https://www.intersil.com/content/dam/Intersil/documents/an52/an5296.pdf

Vendors
https://www.jameco.com/1/1/560-ca30...-14for-more-about-transistors-click-here.html
**broken link removed**

Also CA3083 comprising five independent transistors- two matched- on one substrate:

Data sheet
https://www.intersil.com/content/dam/Intersil/documents/ca-3/ca-3083.pdf

Vendor
https://uk.farnell.com/intersil/ca3083z/transistor-array-5npn-3083-dip16/dp/1201540?ost=CA3083&selectedCategoryId=&categoryName=All+Categories&categoryNameResp=All+Categories

spec