Paul G. wrote:
On Sun, 09 Nov 2008 08:07:26 +0000, Ian Bell
wrote:
Phil Allison wrote:
"Ian Bell"
Phil Allison wrote:
"Ian Bell"
One of the potential sources I have discovered is excess noise in the
anode resistors. This could account for 20uV or more of the total noise.
The solution seems to be to replace metal film resistors with wire wound
types.
** Where does this LUNATIC get these wacky ideas from ??
Metal film resistors are near as damn it free of voltage noise
Are they? Most metal film resistor suppliers that actually bother to spec
this usually put a figure of about 0.1uV/V for excess noise.
** What measurement bandwidth - eh ??
What resistance value - eh ??
Got no ****ig clue - have you.
You ****ING **** WIT
Sounds like you are back pedalling, Phil. You of all people will know
exactly how excess noise is specified (which includes the bandwidth). I
happen to have a 100K resistor in one anode which, from one
manufacturers graph, shows its excess noise is 0.1V/V.
As a said before, show me a manufacturer that quotes less.
CHeers
Ian
....... Phil
Vishay sells bulk foil resistors
http://www.vishay.com/company/press/...08/080507foil/ , and
according to "JIS C5202 5.9" method of measurement, they are at -40db,
where 0db is 1 uv/v, and I assume it is for 1Hz bandwidth, since I
can't find any details of what this standard uses (without paying for
it). That means the -40db would give you about .01uv/v (10 nv/v)
voltage noise. My guess is that this resistor is as good as you'll
get. I've been told that most wirewounds are typically -38db. Metal
film resistors often use this spec, and typical values for good
resistors (100kohms) are about -20 db using JIS C5202 5.9.
Thanks for the link. The data sheet itself says little more than the
current noise is -40dB, no graphs or further explanation.
From looking at other data sheets I get the impression that this type
of noise increases with resistance and decreases with increasing power
rating. The data sheet for Multicomp's MF series for example shows a
0.125W 1K resistor has a current noise of 0.01uV/V (-40dB) but the 100K
only achieves 0.1uV/V (-20dB). However the 0.5W version achieve -40dB
for values up to about 50K and the 100K is only a couple of dB worse.
From an article promoting Vishay resistors:
http://www.planetanalog.com/features...leID=177105460
"Current noise is the bunching and releasing of electrons associated
with current flow. The amount of current noise (or lack thereof)
depends largely on the resistor technology employed, and it is
measurable and is expressed as a function of the input voltage. The
magnitude is microvolts per volt applied. A noise index is expressed
in decibels, and the equation converting µV/V to dB is:
dB = 20 x log (noise voltage [in µV]/DC voltage [in V]).
For example, 0 dB equates to 1.0 µV/V, and 15 dB equates to 5.6 µV/V."
(end of quotation)
Read the article.... it answers a lot of questions, and poses some
possible audio nuttyness.
Yes, plenty of myth masquerading as science in that article.
Current noise has a 1/f frequency distribution. Bear that in mind when
you read the above article.... The frequency of the artifacts is very
low, and the oscilloscope pictures suggest a problem that is really
not as audibly bad as presented.
Carbon resistors have "contact noise", which is an additional very
significant noise. You should avoid carbon resistors, especially the
composition carbon ones (cheapo resistors).
The Johnson noise for the 100K resistor is about 41 nv (bw=1Hz),
That's 41nV per root Hz presumably?
or about 128 nv for the decade
Doesn't that depend on which decade?
(I don't know why manufacturers specify
a decade when 1 Hz is more specific).
Probably because it is 1/f noise.
It is wide band noise, not 1/f
noise. The different frequency distributions (1/f and wideband) make
judgements difficult, especially with the ear's sensitivity vs.
frequency. It's going to take a lot of 1/f noise to be noticed
compared to the wideband Johnson noise.
Good point.
Wirewound and bulk film resistors are considered best for noise,
even by Phil Allison http://sound.westhost.com/project66.htm "Metal
film resistors are about the best only bettered by wire wound which is
a bit impractical"
You must also be careful of wirewound resistors, since they can
inductively pick up AC magnetic fields, especially the precision
bobbin types of resistors.
The resistor noise voltage generated, is on the anode circuit for
your tube, if you use the input noise voltage for a typical low noise
tube (12AX7), you have about 450nv (bw=1Hz) effectively at the grid.
Considering typical gains of 10-100, that gives a noise (bw=1Hz) at
4.5uv to 45uv at the anode. The tube noise will completely swamp the
noise of the resistor! Make sure to use the correct bandwidth when
making the comparison.
That's a rather noisy tube. I currently see 65uV at the preamp output in
a 20KHz measurement bandwidth, which by my reckoning is about 460nV/rootHz.
At the very low level signal levels you seem to want, you will
probably get more satisfaction by using some of the more recent low
noise op-amps, and then amplify to get the voltage levels required by
your tube circuits. Vacuum tubes DO have considerable 1/f noise
(Johnson, John B, "The Schottky effect in low frequency circuits,"
Physical Review, July 1925, pg 71.). That's 1925! Not that I have much
against tube circuitry.... but they are not appropriate for low level
signal amplification (IMHO).
There we'll just have to disagree. ;-)
Cheers and thanks for the input
Ian
Paul G.