WTD: Small signal NPN, HFE >= 200 at Ic < 2uA...

[Fred Bartoli]

Joerg a écrit :

Wow, that would be cool. A bit pricey but if they really hold the line
under 1nv/rtHz that would be ok. Do you remember at which current you
had measured them?

I measured them with my 3470A, so it must have been between 1mA and 3mA.
Will measure them again tomorrow or the day after.

Also, if you're after ultra low noise, I observed that there's a sweet
VCE spot at around 2-3V VCE. Not much, but enough to be noticeable.

As I said they're not speced noise-wise but their ultra low Vcesat
mandates low Rbb' so...
Also the beta wasn't bad, but I don't remember clearly so I prefer not
to tell.

I compared them with some 2SB737 I have and there wasn't clear
difference between them. Maybe the 737 was a bit better at low
frequency, but I didn't have enough time to fully test them.

The down side is that they have pretty high capacitance.

 

[Fred Bartoli]

Greg Neff a écrit :

....surface mount, Vceo >= 30V. Low frequency.

That Ic is not a typo. I have to operate with an Ic in the range of
0.01 to 2 microamps. Not my idea, but I have been tasked to determine
the feasibility. Considering NXP PMBT6429 and Fairchild KST5089, but
my Ic is still way off the left side of the DC current gain curves.
Any recommendations?

The 5089 is indeed a good one. You might want to check the MPSA18
variant (same die as the 5089) which should be even better. Don't know
what the SOT23 number is though. Maybe MMBTA18 but I couldn't find it.

 

[[email protected]]

BC847C looks promising. Lots of gain at the low end, and at low
temperatures.

================================

Greg Neff
VP Engineering
*Microsym* Computers Inc.
[email protected]

are you running the low current to get the 1/f noise down, or to
reduce power consumption?

 

[Joerg]

Joerg a écrit :


I measured them with my 3470A, so it must have been between 1mA and 3mA.
Will measure them again tomorrow or the day after.

Please let us all know what you find. I bet there are lots of audio
folks here who need low 1/f.

Also, if you're after ultra low noise, I observed that there's a sweet
VCE spot at around 2-3V VCE. Not much, but enough to be noticeable.

It'll actually be running thereabouts.

As I said they're not speced noise-wise but their ultra low Vcesat
mandates low Rbb' so...
Also the beta wasn't bad, but I don't remember clearly so I prefer not
to tell.

Beta isn't a concern in most low noise apps. Has to be driven hard and
any resistors around it need to be low anyhow.

I compared them with some 2SB737 I have and there wasn't clear
difference between them. Maybe the 737 was a bit better at low
frequency, but I didn't have enough time to fully test them.

The down side is that they have pretty high capacitance.

Not a problem either in my case. There's a fat laser diode connected to
it and I don't have to care about anything beyond a few kHz.

 

[Robert Baer]

A FET would be great, but I don't have enough headroom to drive the
gate. I can barely accommodate Vbe.

================================

Greg Neff
VP Engineering
*Microsym* Computers Inc.
[email protected]

How about a depletion mode FET like the LND150 or the DN3545 or a JFET?

 

[Eeyore]

Joerg a écrit :

Yep, the 3904 isn't so great, noise-wise.

Have a look at Zetex (their low Vcesat eline, like FFMT618/718...)
They're not noise speced but I did measure them at about 0.5nV/rtHz and
1/f corner under 100Hz.
Lots of choice in this family.

Ohooo, listen Graham... they're much better than the 4400.

I haven't used the 4403 since 1980.

The easiest device to acquire for that application these days is the
2SA1083/4/5. Maybe some 2SA1316s are around still too ?

Graham

 

[Eeyore]

I compared them [Zetex FMMT619] with some 2SB737 I have and there wasn't clear
difference between them.

A very good sign.

Maybe the 737 was a bit better at low
frequency, but I didn't have enough time to fully test them.

The down side is that they have pretty high capacitance.

As does the 2SB737. Large die size and all...........

Graham

 

[Greg Neff]

are you running the low current to get the 1/f noise down, or to
reduce power consumption?

Low power consumption.

================================

Greg Neff
VP Engineering
*Microsym* Computers Inc.
[email protected]

 

[Greg Neff]

Greg Neff a écrit :

The 5089 is indeed a good one. You might want to check the MPSA18
variant (same die as the 5089) which should be even better. Don't know
what the SOT23 number is though. Maybe MMBTA18 but I couldn't find it.

I have to have surface mount, but thanks for the suggestion.

================================

Greg Neff
VP Engineering
*Microsym* Computers Inc.
[email protected]

 

[Ian]

"Fred Bartoli"

Greg Neff a écrit :

The 5089 is indeed a good one. You might want to check the MPSA18 variant
(same die as the 5089) which should be even better. Don't know what the
SOT23 number is though. Maybe MMBTA18 but I couldn't find it.

MMBT5089 is there.

Why should the MPSA18 be better than the 5089? Just curious.
The noise curves don't look promising, although the 1/f corner
is clearly wrong for the 100uA/200uA cases, so maybe there
are other errors.

Regards
Ian

 

[Joerg]

Fred Bartoli wrote:




I haven't used the 4403 since 1980.

The easiest device to acquire for that application these days is the
2SA1083/4/5. Maybe some 2SA1316s are around still too ?

Not too popular in the US and pnp. Digikey lists the 1084 but no stock.
What's your lowest noise npn favorite these days?

 

[Fred Bartoli]

Joerg a écrit :

Please let us all know what you find. I bet there are lots of audio
folks here who need low 1/f.

OK, checked it again.

Actually it was ZTX718 (PNP) not the ZTX618 (NPN)

Tested the same one I did some time ago, but when going at 10mA, the LF
noise appeared abnormally high, and somewhat poppy. I don't remember
this from previous test, so I also tested another one at 10mA and I
might have zapped a bit the BE junction from manipulation.
Also checked a 737 at 10mA for comparison.

Noise in nV/rtHz:

---------- ZTX 718/A ------------- ZTX718/B -- 2SB737
@IC= 3mA 1mA 0.3mA 10mA 10mA 10mA
F=10kHz 0.6 0.8 1.1 0.5 - 0.45
F=1kHz 0.6 0.8 1.15 0.6 0.55 0.45
F=100Hz 0.8 0.85 1.2 0.9 0.65 0.45
F=30Hz 0.9 0.9 1.25 1.6 0.65 0.45
F=10Hz 1.05 1.05 1.4 3.3 0.9 0.7

 

[Joerg]

Joerg a écrit :


OK, checked it again.

Actually it was ZTX718 (PNP) not the ZTX618 (NPN)

Tested the same one I did some time ago, but when going at 10mA, the LF
noise appeared abnormally high, and somewhat poppy. I don't remember
this from previous test, so I also tested another one at 10mA and I
might have zapped a bit the BE junction from manipulation.
Also checked a 737 at 10mA for comparison.

Noise in nV/rtHz:

---------- ZTX 718/A ------------- ZTX718/B -- 2SB737
@IC= 3mA 1mA 0.3mA 10mA 10mA 10mA
F=10kHz 0.6 0.8 1.1 0.5 - 0.45
F=1kHz 0.6 0.8 1.15 0.6 0.55 0.45
F=100Hz 0.8 0.85 1.2 0.9 0.65 0.45
F=30Hz 0.9 0.9 1.25 1.6 0.65 0.45
F=10Hz 1.05 1.05 1.4 3.3 0.9 0.7

Thanks, Fred. That's pretty darn good. Except that the ZTX718 doesn't
seem to come in SMT and Digikey doesn't carry the 2SB737 at all :-(

I guess we could fit in some thru-hole parts if there is a copelling
reason. And your table above sure is a compelling reason.

 

[Fred Bartoli]

Joerg a écrit :

Thanks, Fred. That's pretty darn good. Except that the ZTX718 doesn't
seem to come in SMT and Digikey doesn't carry the 2SB737 at all :-(

Yes they exist. The part number is FMMT718. ZTX is for their eline
thermally enhanced 'TO92', FMMT is for their SOT23 package and digikey
has them :

<http://www.digikey.com/scripts/DkSearch/dksus.dll?Detail?Ref=248038&Row=584937&Site=US>

but these are the PNP. I think you wanted the NPN:

<http://www.digikey.com/scripts/DkSearch/dksus.dll?Detail?Ref=249251&Row=584913&Site=US>

The 2SB737 has been obsoloted some times ago, so no surprise for
digikey. But is still available through the grey market.
Here, I bought mines in France at 20cts a pop, unit price.
Toshiba and hitachi/renesas still have some interesting low noise
replacements.
Digikey even have the renesas 2sa1084 in stock. Sorry that's TO92.

 

[[email protected]]

I am not at liberty to discuss details of the circuit, other than I
need low frequency (DC - 100Hz) linear operation, with Vce around 10V.

================================

Greg Neff
VP Engineering
*Microsym* Computers Inc.
[email protected]

Does your application have a self-test? When you deal with designs on
the edge, you are more likely to find defective parts. I never worked
anywhere that did discrete devices, so I don't know the test flow. If
there is any testing at all, I'd suspect IC in the mA range, not uA
range.

 

[Joerg]

Joerg a écrit :


Yes they exist. The part number is FMMT718. ZTX is for their eline
thermally enhanced 'TO92', FMMT is for their SOT23 package and digikey
has them :

<http://www.digikey.com/scripts/DkSearch/dksus.dll?Detail?Ref=248038&Row=584937&Site=US>


but these are the PNP. I think you wanted the NPN:

<http://www.digikey.com/scripts/DkSearch/dksus.dll?Detail?Ref=249251&Row=584913&Site=US>

Can you let me know which one that was? Digikey's search terms expire
within minutes. One of those non-sensical "features" of their search engine.

The 2SB737 has been obsoloted some times ago, so no surprise for
digikey. But is still available through the grey market.
Here, I bought mines in France at 20cts a pop, unit price.
Toshiba and hitachi/renesas still have some interesting low noise
replacements.
Digikey even have the renesas 2sa1084 in stock. Sorry that's TO92.

TO would not be nice but ok. Gray market stuff is a big no-no though.

 

[Fred Bartoli]

Joerg a écrit :

Can you let me know which one that was? Digikey's search terms expire
within minutes. One of those non-sensical "features" of their search
engine.

Ooops!
FMMT718 for the PNP, FMMT618 for the NPN.
2SA1084 for the renesas PNP (TO). The NPN is 2SC2545/2546/2547 but I
didn't see them.

 

[Joerg]

Joerg a écrit :


Ooops!
FMMT718 for the PNP, FMMT618 for the NPN.

Thanks, Fred. I guess the 618 must be the "similar noise specs"
companion to the 718 then.

2SA1084 for the renesas PNP (TO). The NPN is 2SC2545/2546/2547 but I
didn't see them.

Many Japanese transistors are very hard to find in the US. It was
similar in Europe when I lived there.

 

[Fred Bartoli]

Joerg a écrit :

Thanks, Fred. I guess the 618 must be the "similar noise specs"
companion to the 718 then.

Yep. I think I've measured one too. IIRC the PNP was marginally better,
but by a tiny by four cut hair.

 

[James Arthur]

...surface mount, Vceo >= 30V. Low frequency.

That Ic is not a typo. I have to operate with an Ic in the range of
0.01 to 2 microamps. Not my idea, but I have been tasked to determine
the feasibility. Considering NXP PMBT6429 and Fairchild KST5089, but
my Ic is still way off the left side of the DC current gain curves.
Any recommendations?

FWIW,
From an old notebook,
some crude tests:

.. +5.1V
.. -+-
.. |
.. 1M
.. | Figure 1.
.. .----+ Test Circuit
.. | |
.. 100M |
.. | |
.. | |/
.. '--|
.. |>e
.. |
.. |
.. ===
.. GND

Mfr type ic (uA) Hfe
--- --------- ------- ---
ITT 2n3904 #1 1.89 69
FSC 2n3904 0.18 4
ITT 2n3904 #2 1.01 27
ITT 2n3904 #3 1.01 28
ITT 2n3904 #4 1.01 56
Mot 2n3904 #1 2.30 92
Mot 2n3904 #2 2.92 150

Mot MPSA18 #1 3.93 374
Mot MPSA18 #2 3.95 383
(note that Vce is only 1.1V.)

KTC3198 #1 3.43 228
KTC3198 #2 3.45 231



Just for fun:

.. +5.1V
.. -+-
.. |
.. .----+ Figure 2.
.. | | Test Circuit
.. | 1M
.. 50G |
.. | |
.. | |/
.. '--|
.. |>e ib=93pA
.. |
.. |
.. ===
.. GND


The above was duly air-wired in
recognition of the small currents
involved. d(ic)/d(ib) rather than
static gain was measured to increase
confidence and reduce leakage
current effects.

Mfr Part ic d(ic)/d(ib)
--- ---------- ---- -----------
Mot MPSA18 #1 61nA 721
Mot MPSA18 #2 79 816

ITT 2n3904 #1 8.2 5
(dirty, leaky, used part, icbo=3.8nA)

KTC3198 #1 18 189
KTC3198 #2 18 195

Mot BC547B #1 18 189
Mot BC547B #2 16 174
Mot BC547B #3 18 189

Best regards,
James Arthur