Matched PNP/NPN dual transistors

[thanh]

I' using a symmetrical amp with differental VAS. Using a cascode LTP input will increase distortion.

 

[MikeB]

Using a cascode LTP input will increase distortion.

That shouldn't be, did you use fixed ref-voltage for cascode? Floating cascodes are better here, can you show schematic?
BTW, my experience was that symetrical differential VAS is inferior to asymetrical one.

Mike

 

[len_scanlan]

Digikey too.

I found that the LM devices and MAT devices are fairly widely available, and the SSM ones are there with one or two well-known retailers. Where do you get the Japanese high-Vceo devices? I too wanted to use them in the input LTP of amps instead of 2N5551/5401, but they're so impossible to find.

Check out Fairchild's website - I found they do dual matched versions of 2N5551 and 2N5401. Their part no.s are FMBM5551 and FMBM5401.

The only drawback is that they are SMD parts in a *very* small SSOT-6 package.

I'm going to get samples of these and try them anyway, it's all SMD components at work so I'm used to soldering things the size of ants!

regards,

Len.

 

[thanh]

Hi MikeB!
It seem that cascode LTP is good in asym topo. I used floating cascode. I will post my schematic

 

[thanh]

Hi ! this is my schematic.
- Run transient simulation and view .log file
- Remove "cascode part", adjust resistors to have 2mA current source

 

[CutRock]

Matched NPN Transistors

First, sorry for re-hashing a dead thread, but I figured this was most appropriate here.

A little background:
My end goal is to build Rene Schmitz' VCO3, which calls for a 2SC1583 (Matched NPN Transistor), which I believe is used to translate linear V/Oct inputs to an exponential increase in oscillator frequency. Ideally, I would like to use a component that will be easy to find for the next 5 to 10 years as I plan to turn this into a bit of a lifelong hobby. After quite a bit of research, I've found a couple options for NPN transistors, but none that are still in production and that require purchasing second hand or digging up reserve stockpiles. It seems the only sustainable method is to match transistors by hand measuring components and using thermal adhesive to bind them together. To me this sounds time consuming and likely unreliable.

Finally, I stumbled upon this component, the STD845DN40, on digikey. It is listed as a dual npn transistor and was brought to market last year. I'm not well versed in cross referencing parts or knowledgeable enough to know what values are important for these applications and was hoping someone from this community could weigh in. I'm not sure what the "matched" qualifier adds, but it seems that it being on a single chip would ensure that the two transistors be on the same wafer and therefore quite similar in design (please correct me if I'm wrong in assuming this).

If this chip is usable in some applications as a matched NPN, I believe it would provide a reliable source for modern audio enthusiasts so I figured I'd try my luck with the pros before dismissing it.

Thanks for all your help

 

[Bigun]

my 2c: Modern components evolve quickly. There are vacuum tubes from a long time ago that will be easier to obtain than the majority of modern components. In most cases you don't need exactly the same part - it's often a poor circuit design that relies on such things - but in such cases I suggest you buy a lifetime supply of these parts once satisfied that they are what you need. Otherwise, just allow for the fact that you may need to replace parts with nearest equivalents.

 

[jackinnj]

Mouser carries Thatcorp matched P and N channel arrays -- and their stuff is specifically oriented towards audio.

 

[dchisholm]

2SC1583

"Suburban Electronic Wholesalers" (Suitland Maryland) claims to have the 2SC1583. I have never dealt with them, and know nothing of their reputation.

Like Jack said, "THAT Corporation" (see < http://www.thatcorp.com/300-series_Matched_Transistor_Array_ICs.shtml >) probably has the highest performance matched devices specifically intended for audio applications. "Linear Systems" has quite a few monolithic duals covering a wide range of general application areas (see < http://www.linearsystems.com/products.html#GlossI >), but their parts don't seem to be easy for mere mortals (who don't buy a wheelbarrow-full at a time) to obtain. "Analog Devices" also offers a few parts (see < http://www.analog.com/en/special-linear-functions/matched-transistors/products/index.html >).

. . . I stumbled upon this component, the STD845DN40, on digikey. It is listed as a dual npn transistor . . .

That part doesn't impress me as likely to be an acceptable substitute, on two counts.

1. It is described, and specified, as a low duty-cycle switching transistor - not an amplifier. By small-signal audio amplifier standards its gain is VERY low: the specified maximum is less than the guaranteed minimum for, say, a 2N3904. It is also rather slow: The Ft isn't listed, but the resistive switching times are roughly an order of magnitude worse than the 2N3904.
   
   It's not clear whether the C-E diodes shown in the pinout drawing are intentional (a desirable feature for a switch), or a parasitic byproduct of the fabrication process. In either case, I can't imagine that they would be desirable in an audio application.
2. While this is a DUAL transistor, there is no indication that the two sections are MATCHED - or even tightly thermal-coupled. They may be two entirely separate transistor die, encased in a block of epoxy. In fact, there may be a case (considering the intended application) for keeping the two transistors as thermally-isolated as possible within the constraints of a common package.
   
   As already mentioned in this thread, the best matching (and thermal tracking) occurs when both transistors are constructed on the same silicon die, rather than separate devices. Data Sheets often describe this as a "monolithic pair", though "monolithic" isn't always mentioned, even when the devices are constructed that way.
   
   I believe the LM394 (circa mid 1970's) took the construction of "matched monolithic transistors" one step further: each "transistor" in the LM394 is actually the parallel interconnection of several dozen small transistors. Consequently, the usual spread of performance characteristics due to manufacturing tolerances are "averaged" to give the final device a well-defined, tight-tolerance, set of electrical specifications. I don't know if the later-generation MATxx and SSM2xxx device families continued this design technique, or relied on tight manufacturing controls to create inherently well-matched individual transistors on a common die.

Dale

 

[ingenieus]

The THAT arrays look promising. They even have SPICE models on their website.

Device Models for THAT ICs

According to its datasheet, the STD845DN40 is intended for use in fluorescent lights.

 

[CutRock]

Thanks Dale, for outlining what characteristics are important and providing a little more depth to this term, "matched."

The THAT arrays look promising. They even have SPICE models on their website.

Device Models for THAT ICs

According to its datasheet, the STD845DN40 is intended for use in fluorescent lights.

Thanks for the heads up, ingenieus. Yes, I realized that they intended use was far from audio applications, but I figured I'd bring it up as I knew of no other currently produced dual transistor components. After reading through the datasheet for the THAT devices, they seem much more suitable for my intended use.

 

[CBS240]

THAT 3XX series matched arrays are large geometry die devices. The dies are made individually and then matched, and placed in the same pakage. They typically measure DC Hfe ~90. They are very good for input stages and can be cascode for higher voltage operation. I have used them and they work very well but they are a bit pricey.

 

[twest820]

Well, after 13 years it's probably fair to put a twist to the OP's question. Parts selectors have, perhaps, evolved a bit from earlier discussion and finding matched NPN+NPN or PNP+PNP BJT pairs or a complementary NPN+PNP is not particularly difficult. Matched (as opposed to complementary) NPN+PNP is rather trickier. I'm well aware normally one would recast the circuit to use an NPN+NPN matched pair but I've a quirky application where it's hard to resolve the side effects of converting a PNP to an NPN. I'm also well aware of the ability to buy third party matched pairs but would prefer to simply pick up a single package solution.

So I've been looking around for NPN+PNP complementaries where the two transistors approach matched and was curious what all folks might be aware of for line level, analog signal BJTs just needing to handle a few mA. After hitting up most major parts selectors and few more obscure ones the BC847 and MMDT4413 are among the more promising of what I've come across. Some larger complementaries like the NSS40302 are well matched but rather bigger than I need.

 

[spookydd]

Nowadays it seems more and more difficult to find proper matched pairs.
Most available are either obsolete or far too expensive.
Plus they're also not always available as through-hole parts.

I've again been searching for matched pairs, but I think now it's probably more effective, both financially and for performance, to buy big batches of regular low noise, fast and high gain trannies, and then match them.

It's probably a fool's errand to try to match perfectly the npn to the pnp, but we can always try to get as close as possible, and with big batches on hand, perhaps this can somewhat be possible.

The key then, is to have a proper rig to match those trannies properly, at least for Vce and hfe.

A transistor matcher project is a must...

 

[Mark Tillotson]

Every opamp contains a matched pair of input transistors of course...


Matching NPN to PNP isn't achievable like with 2 NPNs or 2 PNPs, as the dopants are different for the two devices. But the techniques for thermal gradient cancellation should be applicable, as is laser-trimming.

 

[spookydd]

Making an effort, even if not entirely successful, to match pnp and npn is better than not doing so.
And with large enough batches, there is a chance to get somewhere near a good match.
In any case, far better than just not trying to match.
So, having a matching rig is a must, regardless. One of my projects on the wishlist.
It would need to be simple, easy and quick to use. Connect to a computer so everything can be driven via software and all data recorded and viewed on screen, and the saved data can be used to do the matching via software, which will always be far faster than manually doing so.
Pairs of npn and pnp can be made, and then try to match the pairs to each other again, so fairly good matches should be somewhat feasible.
With many pairs, they can be used for more than just the ltp pairs, they can be used for the current mirrors and even cascodes...
There is a pretty good base to start from as far as computer linked device, which is locky-z's curve tracer. It does have one limitation, it was only aimed at windows, and the curves themselves aren't the only thing of importance. The data collected needs to be used as a database for matching..

 

[Bensen]

Hi Spookydd,

The 2SC3381 / 2SA1349 are some nice dual monolythic low power bjt's.
I've used these in a diff pair before.



Don't know if these are still available. I've quite some in my partsbin, if interested...PM.

Regards

 

[spookydd]

I haven't found any of those things available. At least not from any of the well known suppliers. I have seen some obscure sources, probably not to be trusted, and they're quite expensive there. So I don't think those are a good choice from that point of view.

I'll have to find some spice models and take a closer look anyway..

But in any case, matching is still a necessity, even if it's not done on such dual types.

So a good matching rig is a must at any rate.