Desperately Seeking Low Noise PNP

Help Support GroupDIY:

swpaskett

Member
Joined
Jun 2, 2021
Messages
14
Location
Mesa Arizona
It really shouldn't be that hard to find a PNP BJT that doesn't hiss like a snake...... I have searched my favorite vendors (Digikey, Mouser, Tayda) and they do not seem to think NF is worthy of their search tools. Nor does ONSemi, my goto for reliable data sheets and a broad spectrum of parts but lousy search tool; also TI. I can find NPNs with NF data, but no PNPs. Is it not possible to make a low noise PNP?
I am looking for a small signal device, VCEo about 45V, Ic about 150-200mA, Hfe in the 150+ neighborhood, power 500-750mW, and NF<2. Most datasheets do not even bother to tell me the NF. I presume this means it must be pretty high, as in >50 (I learned long ago, if you are not proud of it, don't print it). Am I mistaken?
I found mention on this site of a number of low noise devices. They are either obsolete or made of unobtainium. I am willing to pay a 10% premium for what is otherwise a 2N3906, 10,000% is not in my budget.

Sorry you had to read my rant; I just didn't want anyone to think I hadn't done my due diligence before posting my plea: What is your favorite small signal, low noise PNP that is not obsolete and readily available?
 

john12ax7

Well-known member
Joined
Oct 14, 2010
Messages
1,933
Location
California, US
It's possible ON Semi no longer does the thru hole, but they do still make the SMT MMBT5087.

For thru hole you can get the Central version as listed above.
 

JohnRoberts

Well-known member
Staff member
Moderator
Joined
Nov 30, 2006
Messages
20,971
Location
Hickory, MS
my favorite low noise PNP was the 2sb737s but it was obsoleted last century.

NF (actually constant NF curves) for bipolar devices are relative to source resistance and collector current. You haven't mentioned you application.

ein calculations involve device noise voltage (typically specified as nV/rt Hz) and noise current. The total calculation is beyond a simple post.

There are suitable substitutes for my old 737s but these days most mic preamp designers just use canned IC solutions.

JR
 

MisterCMRR

Well-known member
Joined
Jul 1, 2019
Messages
119
Location
Ventura, CA
It might be worth looking at THAT's selection of low-noise arrays. And, if your source impedance is really low, these parts are matched closely enough to connect in parallel for lower e-noise. Back in the day, one of my favorite low-noise PNP parts was the 2N4250 - but haven't checked their availability. It does seem that even the reliable sources of guaranteed low-noise discretes are eliminating the noise spec on the modern data sheets for the old parts. I suspect the new versions may actually be another part, but because it meets all the original specs except noise, they make fewer different parts but still use the old numbers. I guess it's the result of lower volumes in the parts will guaranteed noise specs - what used to be a guaranteed limit is now just a "typical" spec. It's no way to run an airline!!
 

swpaskett

Member
Joined
Jun 2, 2021
Messages
14
Location
Mesa Arizona
Thanks for all the suggestions/thoughts. Actual Ic is only a few mA. Application is a Schoeps electret front end. SMT is a possibility if I have to go there (I found a couple SMT parts, but was hoping for through-hole). I loathe soldering SOT-23s, my failure rate is about 50% counting the ones I drop on the floor and melt because I forget to turn down the iron :)
 

sdimond

New member
Joined
May 20, 2021
Messages
1
Location
Manor, Texas
Have a look at the 2N4403 also available as MMBT4403 and a dual MMDT4403. It has a larger emitter-base area and lower rbb than the 2N4250 or 2N5087. If you check the 1978 National Discrete Handbook you will find the 4250 and the 5087 are derived from the same die.
 

MisterCMRR

Well-known member
Joined
Jul 1, 2019
Messages
119
Location
Ventura, CA
I almost forgot about the SSM2220 dual PNP with typical en of 0.7 nV/rt Hz. Sold by Analog Devices but now listed as "lifetime buys" which means they'll stop making it soon.
 

Winston OBoogie

Well-known member
Joined
Jun 3, 2004
Messages
2,647
Location
UK.
It does seem that even the reliable sources of guaranteed low-noise discretes are eliminating the noise spec on the modern data sheets for the old parts.

Absolutely! Remember when noise curves and rbb were part of the specifications? 😉

John's favourite, the 2SB737, was great for low source z from microphones etc., but we have IC's that do all of that mic pre stuff now so...

All depends on your source resistance, the beta you want, at the current required for low noise at that source resistance etc.
 

swpaskett

Member
Joined
Jun 2, 2021
Messages
14
Location
Mesa Arizona
Thanks @sdimond for the 4403 suggestion. I guess I have to re-learn how to read data sheets. In my searches I never went beyond the tables; one would think if a transistor were low noise the mfr would tout it as a defining characteristic, but then there are a number of variables as @JR pointed out, so maybe it is hard to do. <sigh> Even the lowly 3906 is not that bad when I dig into the datasheet a bit deeper.
The THAT chip may be a bit big for the present application but i will keep it in mind for future. An array would be ideal but those things have gotten really expensive, I suppose due to integrated solutions.
 

Bo Deadly

Well-known member
Joined
Dec 22, 2015
Messages
2,622
Location
New Jersey, USA
one would think if a transistor were low noise the mfr would tout it as a defining characteristic

That's like asking why carburetor manufacturers don't list performance numbers for leaded gas.

The vast majority of electronic devices manufactured today are simply support circuits (power and I/O) for digital. They don't teach EE students how to make an analog amplifier. EE courses are all about material science and GPU computing and stuff that has almost nothing to do with amplifiers. There are no mass produced analog amplifiers that use transistor amplification that I can think of. The closest would be music gear that is supposed to be some clone of an old circuit but there are only a handful of companies that make anything that would be considered "mass produced" at this point. How many Boog synths and Mackie mixers are sold every year? Not enough for transistor manufacturers to justify mass production of some low noise through hole PNP transistor.
 

abbey road d enfer

Well-known member
Moderator
Joined
Jan 22, 2008
Messages
12,996
Location
Marcelland
It really shouldn't be that hard to find a PNP BJT that doesn't hiss like a snake...... I have searched my favorite vendors (Digikey, Mouser, Tayda) and they do not seem to think NF is worthy of their search tools. Nor does ONSemi, my goto for reliable data sheets and a broad spectrum of parts but lousy search tool; also TI. I can find NPNs with NF data, but no PNPs. Is it not possible to make a low noise PNP?
I am looking for a small signal device, VCEo about 45V, Ic about 150-200mA, Hfe in the 150+ neighborhood, power 500-750mW, and NF<2. Most datasheets do not even bother to tell me the NF. I presume this means it must be pretty high, as in >50 (I learned long ago, if you are not proud of it, don't print it). Am I mistaken?
What application?
Do you want to run them at 50uA or at 2 mA. What source impedance?That makes a huge difference.
NF data are published for RF applications most of teh time, and they don't correlate well with audio usage.
If your PNP BJT "hiss like a snake", your circuit/usage/choice/expectations may be inadequate.
For low impedance sources, Rbb' is a very important factor.
You may want to look at ZTX951. That's the current favorite replacement for 2SB737.
For medium source impedances (5-20k), the operating current is paramount.
For Hi-Z sources, nothing beats an FET.
Do you know the basics of OSI (Optimum Source Impedance), and how the choice of operating current impacts the input noise voltage and current?
 

abbey road d enfer

Well-known member
Moderator
Joined
Jan 22, 2008
Messages
12,996
Location
Marcelland

swpaskett

Member
Joined
Jun 2, 2021
Messages
14
Location
Mesa Arizona
Ic is about 2.5mA, Rx is about 5K. But I think I have some acceptable answers. I don't want to argue, because I did say NF < 2 and didn't state the conditions. My bad. I did assume that somethign that is low noise with Rs=1K and Ic=100uA woould also be low noise with Rs=10K and Ic=2mA. I believe that to be true, but NF will not be the same, obviously. I have been away from analog design for nigh on 40 years and just picked it up again to try to keep what is left of my brain active.

A question for those familiar with the Schoeps topology, which looks to me like a couple emitter followers hung off pins 2 and 3 of the XLR -- a neat trick to harvest power from phantom. The circuit I started with had Vb (relative to gnd) about 40V with 48v phantom; too close to the rail for my taste, so I upped Ic significantly to bring Vb closer to the midpoint. It's now resting around 28 to 30V, but Ic is now more than doubled (but I can use 30V phantom, which is kind of important to me for one field mixer which is all it offers). Otherwise, Vb was sitting around 28V with 30V phantom. That did not seem anywhere near where I wanted to be.
Back to the noise curves -- I see that as Ic increases, so does noise (ref 2N4403). Also there is an ideal Rs, the farther you get from ideal, up or down, the worse NF gets. That one seems counter-intuitife. Can someone explain?
At this point in the signal chain I are dealing with mV signals on average, but I also expect some performers to drive that mic into the multi-volt range at peaks. Is that a realistic expectation (the reason I shifted the bias to begin with)? Am i hopelessly lost in the weeds? <sigh> It is beginning to look like I need to get smarter on how I bias this thing if I want it to perform well with 24V phantom, also a possibility in the future. Or maybe stop caring :)

Finally, in response to @Bo Deadly's comment on engineers. You're on target there. I went through tech school in the 70s, engineering school in the 80s. As a tech I learned how to bias a transistor in the linear range (wish I could remember it); in engineering school I learned to use them as switches -- way easier -- but neither school taught how to bias a transistor for performance; that you had to learn at the bench, after hours at work, so as to not embarrass yourself in front of the old hands. Simulators, if you had them, were crude things run as batches a night, you might get lab time on your own hours, but CPU was billed to your project and unauthorized runs meant lots of 'splainin' to do. It turns out that if you wanted to learn how to use transistors, you took RF courses, something I didn't know at the time and wouldn't have done anyway.
My EE books went to the Goodwill long ago; I still have the tech books. Forgive the trip down memory lane and the way too long post. I can't fill PRR's shoes, but I can take up his space :)

@Abbey - thanks for the link. I'll have a look.
 
Top