EMRFD Message Archive 1814

Message Date From Subject
1814 2008-07-14 16:23:25 Glen Leinweber HF noise sources
Many of us want noise generators for noise figure test setups.
Calibrated noise generators preferably!
Have tried building up a simple hot-filament noise generator
using standard bulbs. A #12 bulb has desirably short leads,
with a filament very similar to the bayonet #47 bulb. Both of
these are rated at 6.3v and 150 mA.. This puts their "hot"
resistance of 42.5 ohms close to the desired 50 ohms. I run
my bulb a little hotter at 7.1 v DC giving DC resistance of
45.5 ohms. Running any hotter puts the filament at risk.
Tungsten resistivity from CRC HANDBOOK was used to
estimate the temperature of the hot filament. This requires a
room-temperature measurement of bulb resistance -
something not easy to do, since any ohmmeter applies DC
current that heats the filament a little. Room temperature
resistance was about 4.03 ohms.
<http://link.aip.org/link/?PTE/40/101/1>
This put the hot filament temperature in the 2200-2300
degree (K) ballpark, giving a noise source having a bit more
than 8 dB ENR. Could this simple noise source give
reproducible ENR? Without a calibrated source to compare
with, I can't tell, and don't have access to an optical pyrometer.

The bulb was fed DC through a few chokes and filter
capacitors to prevent power supply noise from influencing
the noise source.
I'd guess that this noise source has enough inductance
that it would mostly be useful for HF (not VHF) work.

For a more robust (but uncalibrated) noise source, a small
transistor base-emitter junction, reverse biased might be
substituted for a NOISE-COM noise diode, in Sabin's
circuit. I used a 2N3904 transistor, with a 27 ohm series
resistor to boost to 50 ohms, whereas Sabin's
diode circuit used 46.4 ohms. I estimate ENR noise
output around 31 dB. with no attenuator pad. Again, likely
useful for HF work, but still puts out almost as much noise
at VHF.
1817 2008-07-15 04:21:02 Ian White GM3SEK Re: HF noise sources
Glen Leinweber wrote:
>For a more robust (but uncalibrated) noise source, a small
>transistor base-emitter junction, reverse biased might be
>substituted for a NOISE-COM noise diode, in Sabin's
>circuit. I used a 2N3904 transistor, with a 27 ohm series
>resistor to boost to 50 ohms, whereas Sabin's
>diode circuit used 46.4 ohms. I estimate ENR noise
>output around 31 dB. with no attenuator pad. Again, likely
>useful for HF work, but still puts out almost as much noise
>at VHF.

Most avalanche diodes (including reverse-biased base-emitter junctions)
have an ENR in the mid-30s of dB. ENR versus current typically shows a
broad peak around 10mA, and the diode should be operated at that peak to
minimize sensitivity to long-term current variation.

It is important that the ENR does not vary with frequency, and
particularly that there are no wiggles due to component resonances.
Commercial noise sources quote ENR data at 10MHz, 100MHz, 1GHz and then
onward into the microwave region at 1GHz intervals. Those are the only
frequencies at which fundamental calibration standards are available.
Calibration between those frequency points - and at any frequency below
10MHz - must assume that the ENR characteristic of the source is flat
(or at worst, that it varies slowly and monotonically).

The major sources of variation in ENR are, in approximate order of
importance:

1. Choke resonances in parallel-fed circuits - these can be disastrous.

2. Reactance of the DC blocking capacitor: too low a value will make the
ENR roll off at low frequencies; but too high a value may have a rising
inductive reactance in the high VHF region.

3. Construction and layout: for precision at HF, build like it's
VHF/UHF; for precision at VHF/UHF, build like it's microwave; for
precision at microwaves... well, you get the idea!

4. Variation of ENR in the noise diode itself: if everything else about
the noise source is right, the diode package will probably be the least
of your worries. The b-e junction of a transistor in an SOT or other
high-frequency package makes an excellent noise generator, relatively
free of package parasitics that could cause wobbles in the ENR.

5. Variation in diode current. Use a high quality constant-current
source (not constant-voltage) with good temperature stability and good
prospects for long-term stability.

Taking all of these factors together, probably the best type of
construction for a home-built HF-VHF noise source is SMD on hand-cut
microstrip.

A home-built noise source can quite easily be cross-calibrated from a
commercial source - for example, there are several opportunities every
year at microwave gatherings in the US and Europe. Intercomparisons
require that all noise sources can be operated from a laboratory noise
figure analyser. The industry-standard interface is +28V DC, pulsed at a
low frequency. Note that the Sabin sources are non-standard in this
respect, and require a level shifter to interface with the rest of the
NF measurement world.

For accurate measurements, it is VITAL that the noise source has an
accurate 50-ohm impedance in both the "on" and the "off" state. That can
be tricky, because the impedance of the diode itself will change
dramatically when switched. For measurements on HF receivers, a suitable
ENR is about 15dB. With a diode ENR in the mid-30s, this can be achieved
by building an accurate 50-ohm 20dB attenuator into the noise source,
thus guaranteeing a minimum return loss of 40dB regardless of changes
due to the noise diode.

However, that isn't good enough for measurements on extremely low-noise
devices such as GaAsFET amplifiers, because their gain can be extremely
dependent on the impedance of the RF source; and this in turn can create
large errors in the indicated NF [1]. An ENR of about 6dB is required
for these measurements, and can be achieved by adding a further 7-8dB of
attenuation in the noise source, thus improving the variation in return
loss by a further 14-16dB. (At low values of ENR, the ENR does not
decrease dB-for-dB with attenuation, because of thermal noise generated
in the attenuator itself.) In practice, the best way to make a low-ENR
noise source is to build the 7-8dB part of the attenuation into the
source using chip resistors, and then bolt a precision 20dB pad
permanently onto the output.


[1] http://www.mrs.bt.co.uk/dubus/9004-2.pdf
There are many more articles on HB noise measurement in the archives of
the DUBUS quarterly, at:
http://www.dubus.org/


--

73 from Ian GM3SEK
http://www.ifwtech.co.uk/g3sek
1818 2008-07-15 07:06:19 Ian White GM3SEK Re: HF noise sources
Earlier, I wrote:
>There are many more articles on HB noise measurement in the archives of
>the DUBUS quarterly, at:
>http://www.dubus.org/
>

The link to an article on construction of a precision noise source is:

http://www.mrs.bt.co.uk/dubus/9003-3.pdf


--

73 from Ian GM3SEK

http://www.ifwtech.co.uk/g3sek