DMMT3904 matched pair

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

recnsci

Well-known member
Joined
Jun 29, 2006
Messages
163
Location
Lekas Hill State
Anyway, has anyone played with these:

http://www.diodes.com/datasheets/ds30311.pdf
http://www.diodes.com/datasheets/ds30293.pdf

They are cheap, available, lownoise (for some applications atleaset)
and claim rather nice matching (1mV for Vbe and 2% for hfe).

What I am wondering , if we asume that these are just like two
matched 2N3904/6 , what is worst-case lowest usable current for 39xx
parts? (never played with those, nor used them in a product, BC550
(and variants) is yer tipical NPN round here.)

cheerz
Urosh
 
> what is worst-case lowest usable current for 39xx parts?

I would not stop to worry running a 390? at 10 micro Amps, nor have many audio reasons to run any part much lower.

Hfe is shown near-flat 20mA to 100uA, so it is unlikely to fall far at 10uA. If you need high Hfe at low current, you want to check this.

Icex is claimed to be 50nA max (at 25C). If true, and Iceo is twice as high, then 10uA-20uA would be a safe lower limit.

If you are hand-building, I would bet 90+% of parts do much better than 50nA, so you could sort-out the leaky parts without much loss of time or money. (Of course if you need to build 1,000 per hour, you should buy a better-spec part; you don't have time to find and replace ten leakers.)

That assumes shirt-sleeve work. If you are dropping probes in hot oil wells, or building for my server-closet at work, you want to think about temperature effects.
 
I got some samples of these a while ago but never got around to trying them. The idea is a good one, which Philips/NXP may have had first, namely plucking adjacent devices from the wafer and making a two-chip dual out of them. You get likely-good matching and no diode-isolated substrate capacitance between---the "dielectric isolation" in this case is due to plastic.

3904s and 3906s have seemingly gotten better over the years---if you go back and look at ancient National Semi databooks with their helpful curves, the beta falloff with lower collector current is pretty foul. More recent curves like the Diodes Inc. and Toshiba parts are much flatter. The printed min beta specs remain the same, per convention.

But they are neither particularly high beta nor low rbb'.* From the noise curves in Motchenbacher and Fitchin** for a similar Motorola NPN part, one can deduce that rbb' is around 130 ohms, setting a lower limit to voltage noise. In many apps that's plenty good enough, but beaten soundly by other general-purpose devices like the 2SC1815, which also have higher and flatter beta.

Back when, I called Diodes Inc. about their parts and asked what the breakdown voltage between the two devices was. The guy initially said Well of course it's unlimited---they are isolated chips!

I said, So I can apply a few kilovolts without concern?

He was silent for a moment and then admitted that they hadn't considered what I was asking about, and didn't know.


*it may be that the newer parts are also lower rbb' than the old ones, but it's hard to tell without noise data.

** Low-Noise Electronic Design
 
I have been eyeing these parts but never got around to sampling them. why do they (Diodes) only offer these and one other mediocre (higher breakdown voltage) matched pair? why not matched 4401/4403? they have those in the same package, as duals. that would be a different story. those are just quiet enough to be more useful IMO. BCM847/850 etc is worth look. no stock anywhere last time I checked. pmp4201 too.
 
[quote author="bcarso"] More recent curves like the Diodes Inc. and Toshiba parts are much flatter. The printed min beta specs remain the same, per convention.
[/i][/quote]

Now, this is convention I knew nothing about. Beta vs Ic curve in datasheet
indeed doesnt match minimum beta specs.

[quote author="mikep"]
I have been eyeing these parts but never got around to sampling them. why do they (Diodes) only offer these and one other mediocre (higher breakdown voltage) matched pair? why not matched 4401/4403? they have those in the same package, as duals. that would be a different story. those are just quiet enough to be more useful IMO. BCM847/850 etc is worth look. no stock anywhere last time I checked. pmp4201 too.
[/quote]

These parts seem to have smallest Max Vbe offset of all cheap
Tx pairs. They seem BC550ish in specs which is good enough for
application I am interested in. I only wish I could reliably enough run them
down to microampere or two.

cheerz
urosh
 
[quote author="recnsci"]
Now, this is convention I knew nothing about. Beta vs Ic curve in datasheet
indeed doesnt match minimum beta specs.
cheerz
urosh[/quote]

When a part gets an industry standard number like the "2N" I think the minimum specs are supposed to be met. I don't believe the "typical" ones have to be.

So a manufacturer could tabulate some typical specs, but it might make the QA people less happy.

(Somewhat) like PRR says, you could probably extrapolate to few-microamp levels, but there would be no guarantees.

Sometimes even just as second-source parts, manufacturers will simply reprint the original part spec. The CA3046 array was fairly low beta and plagued with popcorn noise. STMicro made them and the datasheet was identical---but they had low noise and much higher beta. Then they discontinued them. I would have been happy to specify them for a variety of things had they kept making them---they typically had of the order of 10uV of offset voltage and were cheap.
 
My father used those by the bucketful (as the non-guaranteed-match CA3086) to simply substitute for the discretes on a particular board, because he said the discretes looked old-fashioned. This was in the 70's.

I think the last design I did with them was as part of a limiter, in about 1994.

I regretted seeing them phased out, particularly after discovering how much better the STM parts were. But needless to say there wasn't much money to be made with them, and they were taking up resources that could be used more profitably.
 
> curve in datasheet indeed doesnt match minimum beta specs.

The printed spec is guaranteed... if it does not meet that spec (with all footnotes), it is "bad" and you can ask to replace it. If you just have one bad part, it is not worth the effort; if you buy a million parts you save your duds and bring them to the negotiation for your next million parts, maybe get some allowance.

The curves are "Typical". They show trends. If the average performance is more important than the worst-case performance, they may be useful. Interesting, but any critical design must be based on the Guaranteed specs. Maybe they get Hfe=150 most of the time, but 55 on a bad day.... if the Spec says "50" then they can sell them under that spec.

> 3904s and 3906s have seemingly gotten better over the years

I think they have far fewer "55" days. Or they make so many, so cheap, they can afford to throw-out a low-Hfe batch. Or they have side-contracts who will buy low-Hfe parts if the price is low.

> the beta falloff with lower collector current is {was} pretty foul.

Reasonable for that to improve with improved processing.

> neither particularly high beta nor low rbb'.*

You been in the racket long enough to recall what we got just a few years before. Transistors were more like thermometers, and Noise Figure below 20dB-10dB was worthy of note.

The 2N39__ parts were a lot nicer to work with than the older cheap parts. Like the 2N2219/2222, only cheaper.

"Low Noise" must be qualified. For audio, for 1K-10K, these parts ain't bad. Often a bunch better than a tube.

> why not matched 4401/4403?

Because only microphone inputs need anything that beefy. OK, other things do too, but many-many more chores can use the smaller parts, get more product per wafer.

OTOH, ON Semi also makes big BJTs with onboard bias diodes, pretty much an audio-only thing, so they are not deaf to audio product needs. Maybe a fatter pair is on the schedule. Or maybe they feel THAT Corp and the LM394 sew-up the market. Product Marketing is not always rational. Or "taking up resources that could be used more profitably."
 
In a related story if there isn't a spec for it on the data sheet don't make ASSumptions about what the part should do.

I had a noise problem with one of the major IC companies on a pretty common 78/7915 3 terminal VR.

Theirs was noisy, so noisy it was a problem in one product (long story). They refused to even offer a typical noise spec and got dropped from the approved vendor list. I don't recall exact numbers but we could have been buying close to a million per annum.
------
I recall another company trying to get approved as a second source on TL074. They said theirs was just as quiet but refused to guarantee a noise spec... see ya.

Trust but verify.

JR
 
If your volume is high enough some semi companies will screen for you and give you a special part number. Unfortunately the required volumes tend to be quite large, so a mere mega-group-buy is not going to cut it.
 
Yup... but for the case of the noisy 3 terminal regulator I had multiple other vendors already selling usable parts "with" noise specs. My suspicion is that vendor M was hitting their 6 sigma targets by opening the window on what they considered good, rather than cleaning up their process.

We had house part numbers for a selected OTA (selected for input offset voltage to minimize control feed though) and a 5532 selected for noise floor spectra (a ratio between noise levels at different freqs, mainly a screen for I/F noise). We also had a nice selection of graded to-3 power devices for use in sundry power level, audio amp modules.

We didn't use house numbers casually because there is a cost associated with doing so. but in large volume production, you need to know what you're going to get to keep product performance between the ditches.

JR
 
I was just looking for NPN/PNP parts: 40-60V breakdown, 100-200mA Ice max, *low capacitances* and 500mW dissipation at 50C. must be SMD. so I settled on sot89 or sot223 package for the dissipation and size. on first search I find only 2 options. You can get 2222/2907 in sot89 or 3904/3906 in sot223. 3904/3906 are better capacitance-wise.

Im looking at the Fairchild datasheet and these specs are 20 years old at least. Like the duals, is it reasonable to assume that these minimum specs are probably being easily met by a newer device, branded with a familiar part number to aid marketing? the Fairchild supplied spice model shows performance exactly as I had hoped. looks like I have to buy a full reel to try these though.

anybody have a recommendation for other sot223 BJTs that meet my requirements?

mike p
 
[quote author="bcarso"]Peter, do you know what the corresponding NXP parts are to these Diodes Inc. ones?[/quote]
Hi Brad,

Not having toyed with these, I couldn't say which the corresponding types would be. 'Corresponding' to which extend ? Pin-compatible for instance ?For now let's have the selection tree:

http://www.nxp.com/#/homepage/cb=[t=p,p=/50805/50833/41772]|pp=[t=pfp,i=41772]

(please copy & paste the url; I guess it doesn't 'get blue' because of the the brackets inside)

Regards,

Peter
 
[quote author="clintrubber"]Not having toyed with these, I couldn't say which the corresponding types would be. 'Corresponding' to which extend ? Pin-compatible for instance ?For now let's have the selection tree:
Regards,
Peter[/quote]

Meaning that they are the same idea as the Diodes Inc. parts: two adjacent Q on a wafer are kept together after slicing and dicing. I thought perhaps Philips/NXP did this first, but it may have been Diodes Inc.

What is funny about that link is the "new part" that you get to readily says the transistors are isolated, but the datasheet that comes up and the diagram shows things tied together as a current mirror. It's a little harder to get to the isolated parts.
 
OK this is an example of some of the NXP fully isolated pairs:

http://www.nxp.com/acrobat_download/datasheets/PMP4201V_G_Y_3.pdf

Urosh, note that they do hazard showing a "typical" current gain of 250 at 10uA Ic.

Noise is not too bad at around 3dB NF for a 2k source Z at Ic of 200uA.
 

Latest posts

Back
Top