Low low noise transistors
CC01...@brownvm.brown.edu
of the head coil is ohhh 200 ohms (ignoring for the moment inductive
rises etc.) I figured that if I went with a bipolar device I wanted
the base spreading resistance and the dynamic emitter resistance as
low as possible. I looked at the BC550 and BC560 transistors and to
my surprice the Motorola catalog states that rbb' is the same for the
pnp 560 and the npn 550. The graphs are identical -- I speculate
it's just the same generic graph but I'm not sure. At least in theory
the n-material in the base of the BC560 should have much higher mobility
(and less resistance) than the p material in the base of the BC550.
I'm thinking of running a bunch a parallel and want to know if anyone
can shed some light on this mystery and if there are other (cheap!)
suitable devices.
I'm aware of the LM394 and the matched MAT transistors
john
Donald J. Miller
: I want to build a low-noise preamp for a tape head. The resistance
: I'm thinking of running a bunch a parallel and want to know if anyone
: can shed some light on this mystery and if there are other (cheap!)
: suitable devices.
I remember some years ago in college I took a low-noise design course
which had an associated lab. We had to build a low noise amp, test it,
etc... Remarkable was the fact that cheapo 2N4401 and 2N4403 switching
transistors have remarkably small Rbb for their cost and are suitable
for some low-noise applications.
Have fun!
--
-------------------------------------------------
Don Miller Electronic System Products
dmi...@crl.com
-------------------------------------------------
John E. Lundgren
> I want to build a low-noise preamp for a tape head. The resistance
> of the head coil is ohhh 200 ohms (ignoring for the moment inductive
> rises etc.) I figured that if I went with a bipolar device I wanted
> the base spreading resistance and the dynamic emitter resistance as
> low as possible. I looked at the BC550 and BC560 transistors and to
> my surprice the Motorola catalog states that rbb' is the same for the
> pnp 560 and the npn 550. The graphs are identical -- I speculate
> it's just the same generic graph but I'm not sure. At least in theory
> the n-material in the base of the BC560 should have much higher mobility
> (and less resistance) than the p material in the base of the BC550.
So instead of a line, Motorola should have drawn a gray area, with the
gray darkest where the line now is, and lighter towards the edges.
Welcome to worst case design.
Run the transistors at low collector currents, and pick and choose for
lowest noise level. Use lots of metal film resistors.
> I'm thinking of running a bunch a parallel and want to know if anyone
> can shed some light on this mystery and if there are other (cheap!)
> suitable devices.
> I'm aware of the LM394 and the matched MAT transistors
> john
--
C.M. Hicks
You may like to look at the SSM2021 and SSM2022. These are dual
matched NPN and PNP transistors respectively, made by PMI, now part of
AD. As you suggest above the PNP is a little quieter; I don't have the
databook handy, but I recall input noise voltages of 800pV and
700pV. For a 200 ohm load you should be able to get a noise figure of
1dB or so with these devices.
Christopher
--
=====================================================
Christopher Hicks http://www.eng.cam.ac.uk/~cmh
c...@eng.cam.ac.uk Voice: (+44) 223 332767
=====================================================
CC01...@brownvm.brown.edu
pnp and npn base spreading resistances aren't that big after all
in complementary transistors.
As for using low-noise resistors, Audio Amateur just arrived
and in it is a reprint of Ben Duncan's "Piece de Resistance".
The original was a three part ( I think) installment in the
British HF&RR beginning in March 1987. Duncan looks closely
at different types of resistors, how they "work", mechanisms
of noise and distortion and how they sound.
john
Robert Everitt Heiss
>I'm thinking of running a bunch a parallel and want to know if anyone
>can shed some light on this mystery and if there are other (cheap!)
>suitable devices.
They might test to (or merely market) complementary specifications, but
there is really no such thing as a complementary pair of transistors.
For a 50 ohm preamp design I've been considering PNP medium power types such
MPS-A55 or A56. As I interpret the NS data book, they might be as good as
0.5 dB NF well below 1Kohm. Unusually large die area; cheap TO-92 package.
shum...@eisner.decus.org
In article <304j3c$k...@cat.cis.Brown.EDU>, CC01...@brownvm.brown.edu () writes:
> I want to build a low-noise preamp for a tape head. The resistance
> of the head coil is ohhh 200 ohms (ignoring for the moment inductive
I don't believe that you can safely ignore the reactance. A 200 ohm 100 mH
head will represent a source impedance over 12 Kohm -- nearly all reactive --
at 20 kHz and your amplifier must be able to work with this. You might want
to consider using a low-noise dual FET (like the Toshiba 2SJ109 or 2SK389) as
a zero-offset buffer/follower to transform the impedance and keep the input
bias current out of the head, followed by a really low-noise IC like the
Analog Devices AD797 (or OP-37).
> ... I figured that if I went with a bipolar device I wanted
> the base spreading resistance and the dynamic emitter resistance as
> low as possible. I looked at the BC550 and BC560 transistors and to
> my surprice the Motorola catalog states that rbb' is the same for the
> pnp 560 and the npn 550. The graphs are identical -- I speculate
> it's just the same generic graph but I'm not sure. At least in theory
> the n-material in the base of the BC560 should have much higher mobility
> (and less resistance) than the p material in the base of the BC550.
Toshiba specifies an rbb' of 50 ohms (at 10 V, 1 mA, 30 MHz) for the NPN
2SC1815(L), and 30 ohms for the 'complementary' 2SA1015(L). These are pretty
low-noise bipolars, 0.2 dB typical (3 dB max) NF at 6V, 100 uA, 10 Kohm, and
1 kHz (where the BC550 is 0.6 and 2.5 dB at 5 V, 200 uA, 2 Kohm, 1 kHz, and the
MPSA-18 is 0.5 and 1.5 dB at 5 V, 100 uA, 10KOhm, 1 kHz). If you can select
from many units, the Toshibas might be quieter that the others (they appear to
also have a lower 100 Hz noise, as well; not insignificant given the large
low-frequency boose required).
CC01...@brownvm.brown.edu
said regarding tape head inductance and noise:
>I don't believe that you can safely ignore the reactance. A 200 ohm
>100 mH head will represent a source impedance over 12 Kohm -- nearly
>all reactive -- at 20 kHz and your amplifier must be able to work with
>this.
This is a very good point. I have measured many high-output moving
coil phono cartridges with R_dc of 120 to 200 and found the inductance
to be negligible. I kind of extrapolated to a cassette deck head. It
dawned on me tonight that with more windings (R_dc = 271 ohms) and a
lot more iron the inductance could (maybe :-) ) be as high as indicated
by shumaker. I shall measure the inductance when I get time.
Now if the Zsource (Rs+jXs) indeed goes as high as 12k then clearly
I can't just thank people for their advice on bipolars and go for it.
Right now there is a dual op-amp performing the job as a head amp.
It's the 5220. A dual low-noise bipolar but of suspect performance.
I have the Matsushita catalog but most everything is hidden
behind assumptions of RIAA use, filters etc.
However, if the inductive rise is as high as 12k at 20kHz I
might just put a jfet diff. gain stage in front of it:
>You might want to consider using a low-noise dual FET (like the Toshiba
>2SJ109 or 2SK389) as a zero-offset buffer/follower to transform the
>impedance and keep the input bias current out of the head, followed by
>a really low-noise IC like the Analog Devices AD797 (or OP-37).
Is there any advantage to this approach as compared to my suggestion
of a jfet diff. stage with the op-amp feed taken from the drains,
i.e the front-end fets provide gain? It is possible that the fet
input capacitance is high but I currently (factory installed)
have a 220pF load capacitor over the head coil (and 16k to ground).
>Toshiba specifies an rbb' of 50 ohms (at 10 V, 1 mA, 30 MHz) for the
>NPN 2SC1815(L), and 30 ohms for the 'complementary' 2SA1015(L).
>These are pretty low-noise bipolars, 0.2 dB typical (3 dB max) NF
>at 6V, 100uA, 10 Kohm, and 1 kHz (where the BC550 is 0.6 and 2.5 dB
>at 5 V, 200 uA, 2 Kohm, 1 kHz, and the MPSA-18 is 0.5 and 1.5 dB at
>5 V, 100 uA, 10KOhm, 1 kHz). If you can select from many units,
>the Toshibas might be quieter that the others (they appear to
>also have a lower 100 Hz noise, as well; not insignificant given
>the large low-frequency boose required).
What a coincidence - I have scavenged a few of these. (I have too
look and see if I have the "(L)" ).
Also, someone suggested that the coil inductance and load
capacitance will form a resonant (albeit lossy) circuit.
Is there any reason to think that the load capacitance over
the head coil is in any way "tuned" ?
thanks
john
Robert Everitt Heiss
>
>I don't believe that you can safely ignore the reactance. A 200 ohm 100 mH
>head will represent a source impedance over 12 Kohm -- nearly all reactive --
Ouch. A bipolar voltage amp would have to run at a lower current,
compromising shot noise performance to get the higher input impedance.
With high enough beta and low enough 1/f corner, good performance may
still be possible over the impedance range. Maybe 2N4250A in parallel.
I haven't tried FETs. Is the inexpensive 2N5457 any good with that kind
of source impedance?
Charles Grosjean
>I want to build a low-noise preamp for a tape head. The resistance
>of the head coil is ohhh 200 ohms (ignoring for the moment inductive
>rises etc.)
How about Japanese FET's? The ones that come to mind are the Toshiba 2SJ72/73
and the 2SK146/147. These are single and dual p and n JFET's with very low 1/f
noise, 0.7nv/rtHz voltage noise (often better than this) and very low current
noise. They are almost impossible to get these days as they were discontinued
a little while back, but some places still might have some. They will probably
beat just about any bipolar you can get (including the ROHMs) for your use as
they have been used quite a bit in the past to make MC head amplifiers with an
equivalent noise of 10ohms or so.
If you can live with a little more voltage noise, Toshiba also makes the 2SJ109
and 2SK389 monolithic matched pairs with about 1.0nV/rtHz voltage noise and
they also have singles with similar performance.
Hope this helps,
Charles Grosjean
David Burton
> I have the Matsushita catalog but...
Say, could you do me a favor? I have an old (circa 1977) Technics
(Matsushita/Panasonic) SL-23 turntable, with malfunctioning motor drive /
speed-control circuit. The circuit is on a tiny little board, about
1" x 3", with just one IC and some discretes.
Well, Matsushita says they don't have and can't get a service manual,
a schematic, parts, part number cross-reference, or *anything*. They
basically say that they can't tell me anything about this unit, because
it is so old. :-(
The IC has a strange number on it: 1003C2. (There's also a tiny "E73E"
on the chip.) It is an 8-pin DIP, with a heatsink tab coming out one
end. Positive side of the (unregulated) supply is to pin 8. Negative
side seems connected only to the heatsink tab, but also through 680 ohms
to pin 2.
It drives a 12V DC motor, controlling the negative lead to the motor.
Could you please see if there is anything like this IC in your Matsushita
catalog? I really need a data sheet or something on this chip. Thanks!
___________
pin 8? --| |-- pin 1
--| 1003C2 |--
--| E73E |--
--|_________|--
____| |____
|_____________|
Note: I assume this is how they number the pins, but maybe they count
the heatsink tab as a pin or two?
-Dave Burton <dbu...@salzo.cary.nc.us>
For my PGP public key, finger dbu...@cybernetics.net
David Burton
in this unit has *three* wires coming from it. Red wire is connected to
the positive supply. Blue is coming from the motor drive / speed control
circuit, driven negative by the mystery IC.
This much makes sense. If I connect a few volts DC across the Red and
Blue leads, the motor turns; the higher the voltage, the faster it goes.
The third wire is white, and I guess that it is some kind of feedback
signal, to tell the control circuit how fast the motor is actually
turning. What, exactly, is the white wire, and where does it come from?
(If it is generating either a DC or AC voltage, it is quite low, because
I can't measure it on my 25+ year old Heathkit voltmeter.)
I'm thinking it'd be pretty easy to make a new motor speed control,
if I understood how it was supposed to work. This turntable has a neon
strobe, so getting the speed right should be no problem.
CC01...@brownvm.brown.edu
(David Burton) said:
>(Matsushita/Panasonic) SL-23 turntable, with malfunctioning motor
>about 1" x 3", with just one IC and some discretes.
You asked me in public so I'll reply in public.
If Matsushita can't help you I'm not sure if the parts distributor
Tritronics can help either but call 1-410-676-7300 and ask. Also,
most repair shops have loads of old service manuals.
I do not have any info on speed controllers. I assume this is a
belt driven table. There are lots of other belt driven tables out
there w/controllers. I saw one at the local salvation army for $6.
john
CC01...@brownvm.brown.edu
Mark Shumaker argued against a JFET diff. pair on the input.
I thought I would understand his arguments after some thinking
but I must confess I don't. A matched pair of fets have a totally
linear i-v transfer curve when hooked up in the diff. pair
configuration assuming they each obey a square i-v law.
The second point was the need for a low bias current. This is
satisfied with a JFET diff. stage except maybe during turn-on/turn-off
where possibly Vgs can be positive and the gate-source junction
acts like a forward biased diode.
While the head probably isn't in proximity to the tape during
power up/down there is still the danger of magnetizing the head
with such currents.
At the moment I do not know if this is more of a problem with
a diff. stage rather than just a source follower.
I wonder if Mark would spell out his arguments in more detail
to clarify these two points.
Charles Grosjean wrote
>How about Japanese FET's? .. Toshiba 2SJ72/73
>and the 2SK146/147.. single and dual p and n JFET's with very low 1/f
>noise, 0.7nv/rtHz voltage noise (often better than this).
>They were discontinued... They will probably
>beat just about any bipolar you can get (including the ROHMs).
>They have been used quite a bit in the past to make MC head
>amplifiers with an equivalent noise of 10ohms or so.
>If you can live with a little more voltage noise, Toshiba
>2SJ109 and 2SK389 monolithic matched pairs w/1.0nV/rtHz voltage
>noise and they also have singles with similar performance.
All are worthy candidates I think. In addition:
I think the 2SK170 is a candidate. The 2SK117 is not as good
but used in similar circuits. Mark V kits I think use the
2SK34 in phono inputs. The 2N3819 isn't quite as low in noise
or leakeage as the 2SK170 but for tape heads the leakeage is still
very low and it might work well with R_source from 300ohms and up.
As I posted, the typical tape head has an inductance of order 0.1H !
(I was totally unaware of this when I first posted).
That's way way more than any MC cartridge I've measured.
At 10kHz, the source impedance is of the order 10k ohms.
In light of what I have been told then I shall first
try out a few FETs. Right now there is a Mitsubishi
5220 low-noise bipolar op-amp there. The Mitsubishi databook
is totally useless so I don't know if anything can be gained
by just changing the op-amp. However, the well-regarded (dis-
continued of course) Sony 870ES has a 2SK170 diff. stage ahead
of a 5532. The current Sony flagship, the 909ES, uses the 5220
alone.
Due to the low source impedance (not!) I assumed I could do
better with bipolars. After now having been informed that
the source impedance shoots up, JFETs seem not only to be
a quick and simple solution but *the* solution.
I have only one source for Japanese FETs and that is MCM.
(Wellborne Labs has a very select few but at an exorbitant price).
Are there others who sell these FETs ?
Now a general question: Imagine the tape headamp as an op-amp.
All circuits I have looked at (four :-))
have the head feed directly the + input terminal. The output
is fed back to the inverting via Rfeedback and then there
is a 300ohms or so resistor from the inverting to ground.
This 300 ohms adds to the noise. Now knowing that the head
is 300ohms and higher would I be better off noise wise to
configure the thing in inverting mode ?
------------- Rfeedback----------
º (x15 smaller) º
-----------------say 20 ohms----º (-) º
º output-------------------->
300+jwL
(head) -----(+)
º º
º-------------------------ground
john
CC01...@brownvm.brown.edu
>would I be better off noise wise to
>configure the thing in inverting mode ?
>
>
> ------------- Rfeedback----------
> -----------------say 20 ohms----: (-) :
> : output-------------------->
>300+jwL
> (head) -----(+)
> : :
> :-------------------------ground
>
I realize that the effective load seen by source is only 20 ohms
as the inverting input is at virtual ground. This solution is
therefore not a solution at all.
john
Wout Serdijn
by: Ernst H. Nordholt
Elsevier Scientific Publishing Company, Amsterdam, 1983
>Noise matching with inductive source impedance.
>To keep the noise voltage small, the transconductance factor gm should have
>a large value. ... shows that the junction FET can be a favorable choice as
>an input device, however, usually at the expense of a large bias current.
>In the frequency region where the 1/f noise contribution is important, there
>is, as with capacitive sources, an essential difference between the noise
>performance of an FET and transistor input stage. ...
>... indicate that a bipolar transistor can be a better choice than an FET if
>the noise in the low-frequency region is of considerable interest.
More information can be found in:
E.H. Nordholt, "Comments on Improvement of the Noise Characteristics of
Amplifiers for Magnetic Transducers", Journ. Audio Eng. Soc., Vol. 27,
p. 680, Sept. 1979.
Hope this helps,
Wouter