Resistor Excess Noise

[ricardo]

Resistor Excess Noise is caused by DC current flow.  It is dependent on the type of resistor and has a 1/f, pink spectrum.

http://linearaudio.net/userfiles/file/letters/Volume_1_LTE_FB.pdf
Frank Blohbaum reply to David Kimber re “A New, Low-Noise Circuit Approach for Pentodes” in Vol 0

says Excess Noise is dependent only on the voltage across the resistor and  independent of its value.
He quotes a value of 0.06uV/V/decade for metal film resistors.

http://books.google.com.au/books?id=PvKPEFu2PVkC&pg=PA46&lpg=PA46&dq=resistor+excess+noise&source=bl&ots=YiN4vyZH-C&sig=IN81Jku14oWRHAFOV4F50XRi0qg&hl=en&sa=X&ei=Z0swT-rOMPGYiAfAmJDxDg&ved=0CGIQ6AEwBg#v=onepage&q=resistor%20excess%20noise&f=false
Douglas Self in "SMALL SIGNAL Audio Design" p46 says

“.. low ohmic resistor has lower excess noise .. rises approximately with the square root of resistance”.  Fig 2.4

I have 3 questions

1  Is Frank correct about Excess Noise being independent of resistor value?

2  What is the source resistance of this "voltage noise source"?  Is it the value of the resistor?

3  Is Self's uV/V the same as Frank’s uV/V/decade?
[edit]  Thanks to Sam, the answers are

1  Frank is wrong.  Self is correct

2  Source resistance is the value of the resistor

3  Yes.

See Sam's post for more detail & good references.

 

[trobbins]

ricardo - is this generally just about current-related noise, which is dependant on the resistor material, and not Johnson noise at all ?

 

[ricardo]

ricardo - is this generally just about current-related noise, which is dependant on the resistor material, and not Johnson noise at all ?

Yes.  Johnson noise is well documented. 

But Current related Excess Noise in resistors seems to be smoke & mirrors; especially the dependency on resistor value.

 

[JohnRoberts]

Resistor Excess Noise is caused by DC current flow.  It is dependent on the type of resistor and has a 1/f, pink spectrum.

http://linearaudio.net/userfiles/file/letters/Volume_1_LTE_FB.pdf
Frank Blohbaum reply to David Kimber re “A New, Low-Noise Circuit Approach for Pentodes” in Vol 0

says Excess Noise is dependent only on the voltage across the resistor and  independent of its value.
He quotes a value of 0.06uV/V/decade for metal film resistors.

http://books.google.com.au/books?id=PvKPEFu2PVkC&pg=PA46&lpg=PA46&dq=resistor+excess+noise&source=bl&ots=YiN4vyZH-C&sig=IN81Jku14oWRHAFOV4F50XRi0qg&hl=en&sa=X&ei=Z0swT-rOMPGYiAfAmJDxDg&ved=0CGIQ6AEwBg#v=onepage&q=resistor%20excess%20noise&f=false
Douglas Self in "SMALL SIGNAL Audio Design" p46 says

“.. low ohmic resistor has lower excess noise .. rises approximately with the square root of resistance”.  Fig 2.4

I have 3 questions

1  Is Frank correct about Excess Noise being independent of resistor value?

2  What is the source resistance of this "voltage noise source"?  Is it the value of the resistor?

3  Is Self's uV/V the same as Frank’s uV/V/decade?

The mechanism is basically noise caused by non-uniform current flow in a resistive substrate, so the more uniform the resistive material the lower this current flow irregularity caused noise. While it seems it should scale with current flow, apparently it tracks with bulk resistance in many resistor types such that even though a lower resistance draws more current for a given voltage, the noise caused by the current remains proportional with the terminal voltage.

I seem to recall some brands of carbon comp claiming very low excess noise back in the '80s, while carbon comp is generally considered inferior. IIRC it can literally vary between different brands. I have never found this to be significant for modest voltage solid state designs, I think it's mostly a tube thing with high voltage bias strings, I recall looking into it wrt the phantom power resistors in the front end of mic preamps with tens of volts across them and it wasn't a factor there (IIRC).

It seems the noise caused by this would be a current source, so even though there is more current noise created by a lower value resistor per volt, that current noise is terminated by the lower value resistor in shunt, so therefore the tracking more proportional with voltage than current.

or not.....  Last time I looked at this was decades ago and I quickly lost interest when it didn't appear to impact my area of design effort.. For most audio paths this creates a microvolt/per volt level modulation noise, that I know i can't hear, or measure.

JR

 

[PRR]

> It is dependent on the type of resistor

Right. And since today resistors with very low excess noise are quite affordable, why would you care?

(I think some "tube sound" IS the 1/f noise of tubes and composition resistors.... in which case you copy existing artifacts and hope for similar euphonics.)

 

[ricardo]

Right. And since today resistors with very low excess noise are quite affordable, why would you care?

The application is a tube mike preamp similar to Frank's article.  This is probably the only modern (?) audio application where this is relevant.

Note the noise figures he quotes which compare nicely with 5534 etc.

Anyone have answers to my 3 questions?

 

[Samuel Groner]

1  Is Frank correct about Excess Noise being independent of resistor value?

For typical resistors, excess noise increases for values above a few kOhm (see www.vishay.com/docs/28714/melfpre.pdf). Sometimes, an increase for very low values (below 100 Ohm) is observed too.

2  What is the source resistance of this "voltage noise source"?  Is it the value of the resistor?

The voltage noise source is in series with the resistor. So yes, in a certain sense its source impedance is the resistor value.

3  Is Self's uV/V the same as Frank’s uV/V/decade?

Yes.

Good reading here: https://dcc.ligo.org/public/0002/T0900200/001/current_noise.pdf

Samuel

 

[JohnRoberts]

Good link...

JR

 

[riggler]

Are modern carbon comp resistors generally less-noisy than carbon comp resistors of the tube era?

 

[JohnRoberts]

I would hope so, but hope is not a reliable design discipline.

The link Samuel provided shows a lot of comparisons between types and brands of current resistor types.

Carbon comp is generally not considered a good performer in this aspect, while I recall one brand claiming otherwise back in the day, that may have been marketing bluster or compared to other carbon comps.

If excess noise is important for your particular application it may be worth investing in a premium R, and perhaps even comparing a few brands if you don't have advice from the report Samuel linked us to.  I wouldn't lose sleep over this for typical low voltage solid state design but for some applications parts can matter.

JR

 

[ricardo]

For typical resistors, excess noise increases for values above a few kOhm (see www.vishay.com/docs/28714/melfpre.pdf).

I think the noise levelling off below a few kOhm is simply the noise floor of their gear.

I'll edit the original post to reflect Sam's input.  Many thanks.

 

[trobbins]

There was a lot of development work in the 1950-60's on resistors to improve on the normal carbon type that was being used.  A few technical papers indicate the dramatic changes that were taking place, which generally led to standard available resistors in the 60's onwards having resistor excess noise levels that were not an issue from most applications.

Two papers I have come across are:
http://dalmura.com.au/projects/Resistor%20noise.pdf
http://dalmura.com.au/projects/Resistor%20noise1.pdf

 

[Samuel Groner]

I think the noise levelling off below a few kOhm is simply the noise floor of their gear.

Doesn't make sense to me--why should the MMB0207 then reach lower noise?

Samuel

 

[ricardo]

I think the noise levelling off below a few kOhm is simply the noise floor of their gear.

Doesn't make sense to me--why should the MMB0207 then reach lower noise?

My guess is that the larger resistor allows them to use a larger current and detect lower levels of uV/V/decade.

Self in "Small Signal Audio Design" says it is measurement noise floor in his similar Fig 2.4 curves for Carbon film

 

[Jean Clochet]

(I think some "tube sound" IS the 1/f noise of tubes and composition resistors....

I think so too.  I also suspect there are tiny non-linearities depending on current and voltage differential through those old resistors - although there are magnitudes more non-linearities from a bog-standard ECC83 yet I see some people worried about what resistors and caps to use rather than which active device!
An old trick for reducing some noise in sensitive valve pre stages when using comp. resistors for anode loads etc. was to use a larger part than called for - 1W vs 1/4W.  I can't say I've worried about it in quite some time but if you feel that carbon gives you some valve mojo, there are modern carbon parts available without having to use some cruddied-up old Allen Bradley . 
Just MHO.

p.s. thanks for the link Samuel, a good one 

 

[Jean Clochet]

A wee follow up on my post above: 
I went looking through a few books hoping for some examples of non-linearity measurements of carbon-comp resistors.  The only one I have on hand that had figures was from Doug Self -  Small Signal Audio Design.
He quotes voltage coefficient of carbon composition @ around 350 ppm (parts per million) and, looking @ his SPICE calculated figures (closest listed being 320 ppm), he shows .034% THD @  +15dBu and .060% THD @ +20 dBu. 
For reference, a metal film of 1 ppm being listed as .00011% THD and .00019% THD for the same respective levels.
Elsewhere, I see carbon-comp stated as being up to 700 ppm which would, of course, give higher THD figures.

Now back to our regular scheduled program...
Cheers.

p.s.  Given that thick-film surface mount resistors also exhibit the same high (and variable) ppm (this worsening at high ohmic values and small power ratings) - they may be an alternative for adding in some of that real valve mojo to your op-amps!!