EMRFD Message Archive 7654

Message Date From Subject
7654 2012-08-02 08:36:48 Glen Fet Noise specs
Audio and RF applications of FETs seem to be at odds. Audio noise is specified as nv/sqrt(Hz), and sometimes plotted against frequency.
You never see current noise specified. I guess that's because at low
frequencies, input impedance is "very high" - so high, that voltage
noise dominates for common-source amplifiers.
As frequency rises to the RF region, noise voltage generally decreases to some floor, then rises again at very high frequencies. At some point, input impedance must decrease enough that current noise begins
to affect overall noise figure. For FETs that show optimum source Z for
achieving best noise figure, you rarely see data extend below 100 MHz.
Even at 100 MHz., optimum Z is often kilo-ohms. We can only guess that for RF frequencies below 100 MHz., optimum Z is even higher than kilo-ohms.

That leaves RF design in the low-RF region spec-less. I've never seen a
design that takes pains to optimize noise figure. A low-RF design would
likely transform input impedance w-a-y up. Such an amplifier overloads
easily. Designers consider dynamic range to be more important than
noise figure, and are reluctant to transform up to such high Z's .

Curious about MOSfet noise, I found an old data sheet for a discrete
device that showed noise figure for audio. Yikes! it wasn't pretty,
especially at low frequencies. At the top-frequency end of the graph,
noise figure was still dropping - I could only guess that you'd hit
the noise floor somewhere in the RF region.
Has anyone seen noise voltage or noise figure or optimum Z specified in the 10KHz - 100 MHz. region for FETs or MOSfets?
7655 2012-08-02 09:55:28 Thomas S. Knutsen Re: Fet Noise specs
Hello.
I'l break it up a bit in order to make it clearer:

> Audio and RF applications of FETs seem to be at odds. Audio noise is
> specified as nv/sqrt(Hz), and sometimes plotted against frequency.
> You never see current noise specified. I guess that's because at low
> frequencies, input impedance is "very high" - so high, that voltage
> noise dominates for common-source amplifiers.
>
At Low frequencies 1/f noise is the dominant noise. Depending on the
semiconductor type, the noise can be dominant up to low VHF. Bipolar
transistors have low 1/f noise, while some types of fets, like MESFET and
GaAsFets have high 1/f noise.


> As frequency rises to the RF region, noise voltage generally decreases to
> some floor, then rises again at very high frequencies. At some point, input
> impedance must decrease enough that current noise begins
> to affect overall noise figure. For FETs that show optimum source Z for
> achieving best noise figure, you rarely see data extend below 100 MHz.
> Even at 100 MHz., optimum Z is often kilo-ohms. We can only guess that for
> RF frequencies below 100 MHz., optimum Z is even higher than kilo-ohms.
>
On higher frequencies the dominant noise is white noise or Johnson termic
noise.
Data below 100MHz is not often used by professional designers, and taking
unneccesary measurments is expencive. Doing noise figure impedance match
measurments are expencive and require a lot of time. It's not unusual to
find measurments that takes more than 50 hours in order to get satisfactory
frequency coverage.

I have done several experiments with both J-fets and dual gate mosfets in
order to find some kind of optimum impedance match, and it's a bit tedious.
Most of the dual gate mosfets have input impedances around 2K ohms.


> That leaves RF design in the low-RF region spec-less. I've never seen a
> design that takes pains to optimize noise figure. A low-RF design would
> likely transform input impedance w-a-y up. Such an amplifier overloads
> easily. Designers consider dynamic range to be more important than
> noise figure, and are reluctant to transform up to such high Z's .
>

Dynamic ranges is most of the time more important than NF, except when
doing EME. The temprature of the ground sets the noise temprature of an
range and that's seldom below 5dB, so for terestial use, going below 3 dB
NF don't realy make sence.


> Curious about MOSfet noise, I found an old data sheet for a discrete
> device that showed noise figure for audio. Yikes! it wasn't pretty,
> especially at low frequencies. At the top-frequency end of the graph,
> noise figure was still dropping - I could only guess that you'd hit
> the noise floor somewhere in the RF region.
> Has anyone seen noise voltage or noise figure or optimum Z specified in
> the 10KHz - 100 MHz. region for FETs or MOSfets?
> __
>
There are dual gate MOSfets like the BF992 that have exccelent noise figure
and dynamic range. Based on past experience, with an good noise match (this
is not S11 match!) you should be able to get an acceptable noise figure
with most fets. Most of the time lowest NF is not at the point where you
get max gain.
I'l do some experiments with some regular MOSfets (2N7000 or 2N7002), in
order to measure the NF, but I expect it to be below 3dB.

Having access to an direct reading NF meter while doing such experiments
are nice, and there have been several described that uses common tools and
an PC or microcontroller to do the measurments.

73 de Thomas LA3PNA.

--

Please avoid sending me Word or PowerPoint attachments.
See <http://www.gnu.org/philosophy/no-word-attachments.html>
PDF is an better alternative and there are always LaTeX!


[Non-text portions of this message have been removed]
7656 2012-08-03 08:32:55 JamesDuffey@comca... Fet Noise specs
I hope that this is not too far off thread. I wonder if anyone has measured the noise figure of a typical computer sound card, or knows what it is? - Duffey KK6MC





[Non-text portions of this message have been removed]
7657 2012-08-03 12:28:04 Gary, WB9JPS Re: Fet Noise specs
7658 2012-08-03 12:31:12 Alberto I2PHD Re: Fet Noise specs
7659 2012-08-03 15:50:01 kb1gmx Re: Fet Noise specs
I've used them in Feedback amplifiers with good gain and
reasonable noise of 5db (for resistive feedback) to over 30mhz.

They are also good to VHF but the input impedance needs to
low impedance as the gate is largely a capacitor to ground.

As a class a linear device they will net about 20dbm output
power easily with heatsinking.

They also work as HF switches in passive mixers.

Very useful devices.


Allison



7662 2012-08-05 20:45:03 Reginald Beardsle... Sound Card Noise specs
I suspect that the quality of antialias filtering in sound cards varies from poor (single pole RC filter) to very good (e.g 8-12 pole active filter) w/ prices to match. On top of that there's the problem of EMI inside the computer.

I've got a 6-7 year old digital audio workstation ($1300 at the time) which has a lot of noise w/ the inputs shorted. Every time it writes to the hard drive there's a little "brzzzzp". Only noticeable if you take the time to measure it. On the other hand a $100 Zoom H1 is quiet as the grave at 24 bit 96 KHz.

It's actually quite easy to measure the performance of the sound card:

Feed a fast rise time 1 KHz square wave in and look at the spectrum. That will characterize the antialias filtering.

Terminate the input w/ an appropriate resistor and you can measure the EMI & other noise.

WAV files are easy to bring into Matlab, Octave, Scilab etc. So the quality of the measurement depends almost entirely on how sophisticated your understanding of noise analysis is.

Have Fun!
Reg
7664 2012-08-06 01:32:59 d29602960 Re: Fet Noise specs
Hi Allison and the group,

Allison wrote:
I've used them in Feedback amplifiers with good gain and reasonable noise of 5db (for resistive feedback) to over 30mhz.
**********

I have a few 2n7000,s here to play with and the configurati
7668 2012-08-07 14:09:21 kb1gmx Re: Fet Noise specs
7669 2012-08-11 06:32:52 d29602960 Re: Fet Noise specs
Hi again Allis