Re: Lowest-Noise Device
- From: Patrick Turner <info@xxxxxxxxxxxxxxxxxx>
- Date: Mon, 02 Jan 2006 15:12:40 GMT
NonDigital@xxxxxxxxx wrote:
> I am trying to figure out whether vacuum tubes or JFets have the lowest
> noise for a given source impedance.
The lowest source impedance is a short circuit, so noise tests
of a device should be with the device input taken to 0V with a short lead.
The noise that comes from the first device sets the majority part of the
snr
for the amp concerned.
The noise at the output is due to what is called the
equivalent noise resistance of the device; ie, the device is deemed
perfectly non noisy, but with some imaginery R between 0V and the active
input grid or gate.
for a triode, ENR = 2.5 / gm in mA/V, from RDH4.
for a j-fet ENR = 0.7 / gm in mA/V, from the british amateur handbook.
So a 12AX7 with gm = 1.6mA/V has far more ENR and noise than a
2SK369 with gM = 40mA/V.
But with tubes, there isn't only input grid noise, but
plate shot noise and flicker noise, so that when a phono stage or mic amp
is turned up loud with input grounded its like listening to a sputtery BBQ.
Jfet noise is more hissy, less LF, and mosfet noise is like tube noise,
maybe worse.
At the end of the day you have to pick through the tubes for the best.
But while a 12AX7 may be fine to use with a MM cart such as Shure V15,
the same amp would be hopelessly too noisy for a Denon 103D MC
with 20 times less signal output.
Then you will find the 2SK369 a far quieter device.
Since the signal is lower, the jfet thd, ( mainly 2H, like a tube )
will be non polluting of the sonic outcome.
>
> JFets have shot-noise in the drain current (Id) and this is actually
> lower as the drain current is increased. JFets also have input
> gate-current-noise which flows back into the input impedance and
> creates noise. The input current-noise also happens to be the lowest
> for higher Id.
try using a jfet such as the 2SK369 or 2SK147, don't worry about the
theoretical noise figures one should get with the devices; usually tubes
are noisier than theory
suggests, but the high gm jfets I mention are much quieter than tubes.
The 2SK369/147 have a number like a mosfet, but are not mosfets,
which are no good for front ends. These jfets have their 40mA/V
at only 5mA of drain current.
See Allen Wright's site at http://www.vacuumstate.com for more about
jfet use, also Earno Borbelli has a lot to say.
> JFets may be stacked to form a low-noise gm-multiplier so the gm of a
> JFet is never a problem - the gm may be increased by a factor of 10x or
> 100x at little noise penalty (1%).
>
> I know little about the noise mechanisms of tubes however a vacuum tube
> operates somewhat like a JFet so I figure the tube must have similar
> noise mechanisms. The problem is tubes do not have noise specs like
> JFets.
Well yes, but then jfets are only to be trusted with low level signals
since the
2H % per volt of output can be 100 times greater than a typical signal
triode.
>
>
> Question: Do tubes have higher or lower shot noise (plate-current
> related) and do they have higher or lower grid noise-current?
Try testing a few of these dvices to find out the facts.
set up an SE jfet or triode, and then use a low noise amp with gain =
1,000 and with bw = 20kHz to amplify the output of the device under test.
Say you measure 0.28Vrms of noise at the test amp.
Then there must be 0.28mV at the input, and as your measuring amp
would have maybe only 4uV of noise, the measuring amp
isn't the source of the noise. Nor will the plate resistance and Ra of the
12AX7.
If the 12AX7 you are testing has a gain of 70 at its output at its anode,
then the equivalent noise at its input is 280uV / 70 = 4uV, a figure
for a mediocre sample.
A cascode circuit using a 2SK369 jfet and any triode like
12AT7, 6CG7, 6SN7, 12AU7, 6DJ8 etc and with say 22k as the triode load RL
will all give a gain of the fet gm x RL, = 0.04 A/V x 22,000 ohms = 880
gain.
Using a bias R of say 18 ohms in the fet source reduces gain a little,
but doesn't increase noise much because Rs is such a low value.
Beware of RF instability using cascode circuits.
Patrick Turner.
>
>
> Charles Gilbert
> NonDigital@xxxxxxxxx
.
- References:
- Lowest-Noise Device
- From: NonDigital
- Lowest-Noise Device
- Prev by Date: Re: Lowest-Noise Device
- Next by Date: Re: Tube audio books in PDF format !
- Previous by thread: Re: Lowest-Noise Device
- Next by thread: Re: Lowest-Noise Device
- Index(es):
Relevant Pages
|
