Matched bipolars and OTAs

[R.G.]

Those of you who read the forum closely will remember that I shot off my mouth about building a discrete-transistor version of the CA3080. Having done that, I felt obligated to tell people how to do it.

The only real problem with this is that you need at least one pair of matched NPNs for the front end diffamp, and having two each matched pairs of NPNs and PNPs makes the four current mirrors work great instead of only OK.

So I went haring off to find transistor matching setups, and found something more useful.

Diodes Inc makes and both Mouser and Digikey sell the DMMT3904 and the DMMT3906 dual matched pairs. These are two devices of the obvious type number where they have simply cut the wafer into pairs of transistors instead of individuals. You actually get monolithic matching per pair. They are about $0.40 a pair. You need five duals for an OTA, so you get an OTA for $2.00, which isn't all that bad compared to a real 3080.

The only drawback is that these are not beginner parts. They're a six pin SMT package, and require a good SMT PCB to put them on and a good hand soldering. However, it is entirely possible to build a "3080" out of these parts and have it stick into a 3080 socket and work.

The more alert readers here will have already have tumbled to the fact that matched dual transistors are useful for much more than OTAs. I'll let you fill in the blanks.

[puretube]

free information block:

[Rob Strand]

I'm pretty sure you could get an OTA going with one match transistors, or perhaps a transistor array.  I remember doing this ages ago with a CA3046(? or one of those arrays).  These days I suspect the matching on those integrated units to be be fairly good.

DMMT... sounds like a nice alternative for the LM394,

The LM394 has been special designed for low noise (especiall for low source impedance circuits).  Inside it is actually a large number of devices in parallel, this helps reduce noise.  It can also reduce the statistical spread.  I think the devices RG quotes are matched but not necessarily low noise in the same sense - so they shouldn't be used as a general sub for the LM394.

You can parallel devices yourself to improve things.  Don't create parallel units from the same device unless *all* the units come from that chip.

[puretube]

free information block:

[R.G.]

I'm pretty sure you could get an OTA going with one match transistors, or perhaps a transistor array. I remember doing this ages ago with a CA3046(? or one of those arrays). These days I suspect the matching on those integrated units to be be fairly good.

That was my first impulse. It turns out that like the 3080, the 30xx transistor arrays are getting hard to find. No particular surprise - the main industrial use they were intended for has dried up, like the 3080. They're still available, but hard to find.

The LM394 ... I think the devices RG quotes are matched but not necessarily low noise in the same sense - so they shouldn't be used as a general sub for the LM394.

Yeah, I looked at the LM394. Also the MAT-0x series from Analog Devices. While they are in current production, they are also hard to find. I have no doubt that they would be **great** in this application, but I was after cheap, easy and simple for beginners with commonly available parts.

The DMMT items will not be perfect or optimum, but they are cheap and available.

Actually, I'm still looking at bipolar matching. I think you could buy a batch of $0.25 2N5088's and 2N5087's and then match up a few pairs for some dynamite low noise discrete OTAs. You could use the LM394 or MAT-0x series for premium stuff. I think TI may still make integrated current mirror parts (gotta look for those...) so that might make it even easier.

Still exploring.

[puretube]

the "synth-DIY-scene" got amazing solutions for subbing obsolete chips by discretes... (and also have to fight with the usual thermal coupling issues...)

[Rob Strand]

30xx transistor arrays are getting hard to find

Hmmm, didn't realize that - but not surprised either.

I was after cheap,

I can understand that - the prices of LM394's suck.

The DMMT items will not be perfect or optimum,

They should be good for most OTA applications (I wouldn't go sub'ing them for LM394's in microphone preamps etc).

2N5088's and 2N5087's and then match up a few pairs

I think the degree of matching required is *very* hard to achieve.  I've found fairly good matching in the same batch but definitely not good enough for OTA's.  Maybe if 4 devices are put in parallel on each side it might work.

Still exploring.

One idea is to use MOSFETs, perhaps monolithics devices would be OK.  The resulting OTA wouldn't not sub for 3080's though since the control function isn't the same.


I know there are other transistor array devices around.  There are dua transistor devices around too.  I have a feeling the ones you have found already will be hard to beat from a cost/availability perspective

[puretube]

free information block:

[puretube]

free information block:

[johngreene]

That was my first impulse. It turns out that like the 3080, the 30xx transistor arrays are getting hard to find. No particular surprise - the main industrial use they were intended for has dried up, like the 3080. They're still available, but hard to find.

Usually Digikey means pretty easy to find to me.

HFA3046B
HFA3096B

--john

[R.G.]

Usually Digikey means pretty easy to find to me.

Me too. I missed those. Good to know they're still there.

[A.S.P.]

once you`ll ask for quantities, their true face will show...