Noise of 0 ohm resistor?

[Svart]

We all know about the different types of noise from resistors.  My question is: Is anyone familiar with noise using a 0 ohm "resistor"?  It seems that they actually have a small amount of resistance but I can't find a lot of information on noise on them.  I'm trying to squeeze out 1db more SNR to meet a spec and I'm running out of other ideas besides changing out series resistors in the signal path for lower noise types.  Most of my signal path has 0 ohm parts stuffed because we used to have pads and things stuffed in those spots on a previous iteration of the circuit.  Respinning the PCB is not an option at this point.

Any thoughts?

 

[sahib]

I do not have a micro-ohm meter and I did not carry out a measurement, but  since they are zero ohm links I would assume that their resistance would be equal to that of a same length of wire, or even a copper track. Hence their noise.

 

[Svart]

I guess my question is are they really zero ohm links and is the conductor really metal?  I don't have a micro-ohmmeter either but I seem to remember looking at a 0 ohm "resistor" a long time ago and it wasn't really zero ohms, it was some small amount that wasn't zero, like .1 ohms or something.  The other part of this question is can we trust the conductor material to not introduce noise into the signal?  There isn't much DC so shot noise isn't much of an issue but there is a good amount of broadband AC signal going through them and Johnson/Nyquist noise is of concern.

 

[JohnRoberts]

-1 dB compared to what?? The last 1 dB of NF is a killer while getting 20 dB NF down to 19 dB is not so much.

If the Johnson noise of circuit board traces, and jumpers is significant, you are dealing with some seriously low Z stuff...  you might consider use of Peltier cooling on devices/components. While that is generally not practical.

To answer you question specifically plug the actual resistance into the equations, but small fraction of an ohm will not be much noise.

JR

 

[sahib]

0.1 ohm is a significant resistance and zero ohm links certainly don't have resistances like that. My last batch which I still use were rated as 0.005 ohm maximum. I dont think you would be bothered very much with that even if you had so many of them. Unless as John said you are dealing with a really low impedance stuff.

 

[pstamler]

0.1 ohm is a significant resistance in some contexts, but not in this one. Its thermal noise will be about -162.6dBu. That's pretty quiet.

And, as was just pointed out, those 0 ohm "resistors" are way lower than 0.1 ohm.

Peace,
Paul

 

[Svart]

There are about 30 of them in the signal path which is a 50ohm transmission line.  Oh well, I've never actually paid attention to how they affect things and was concerned more about my lack of knowledge than anything.. Cause someone is going to ask me what I'm going to do to fix it..  :

-1 dB compared to what??

 

Compared to what I have now.  ;D

It's complicated because what we are measuring is relative to the signal at the input of the unit, the frequency and how it is impaired.

The piece I'm working on is a broadband tuner.  It's used in a device that measures signal levels as well as MER/BER in QAM systems.  I'm working on degraded (M)odulation (E)rror (R)ate (MER for short) at 0dBmv input levels to the tuner.  I inject noise until I get a certain MER and then plot (B)it (E)rror (R)ate (BER for short).  I'm following the theoretical curve for BER at a specified MER and I find that I deviate too much at low MER signals.  During testing I've found that the noise floor of the unit impacts this signal more at lower MER and causes higher bit errors for signals that are degraded beyond a certain point.  We misjudged the curve during design and figured we had a little more room for error in our noise figure budget and designed the board around that to save some time and BOM costs.  Now that the board is done(and we are not allowed any more PCB changes) we've found this issue by accident and rather than let it go to the customers like this, we are going to fix it.  Since we are very close to design spec, we only need a dB or so SNR/MER to meet the spec and pass the design on to manufacturing.

I'm looking into some other things too but I just wanted some opinions.

 

[CJ]

what kind of solder?

 

[Svart]

ROHS..  :

I don't know exactly what solder our plant uses.  I suppose I should though eh?

 

[CJ]

don't know if solder has much noise

interesting topic,

i bet it is like the guy who wanted resistor noise values for SOP stuff,
no known data exists at this time...

 

[Andy Peters]

There are about 30 of them in the signal path which is a 50ohm transmission line.  Oh well, I've never actually paid attention to how they affect things and was concerned more about my lack of knowledge than anything.. Cause someone is going to ask me what I'm going to do to fix it..   :

Depending on the frequency of operation, I'd be more concerned with all of the impedance discontinuities between the SMT (I assume) pads and the parts.

-a

 

[Svart]

yes they are 0402 SMD.  I wish we could have gotten rid of them but they are there to stay.  using an rtl

Frequencies are 1mhz to 1200mhz.

 

[CJ]

do you have one of those machines that you hook up, say,
a sender at one end, and a sensor at the other,
to take delay figures for single traces?


just set up a straight trace,
then a trace with resistors,
measure the response time,
then ask somebody to verbalize the results for us.

 

[jdbakker]

Depending on the frequency of operation, I'd be more concerned with all of the impedance discontinuities between the SMT (I assume) pads and the parts.

+1. I would be rather worried about the S11 of thirty resistors in what is supposed to be a transmission line. That's going to hurt you at the higher end of your spectrum. From your description it's not impossible that what you're seeing is (a form of) inter-symbol interference.

we are not allowed any more PCB changes

That's a polite way of saying "we're screwed"... If a customer came to me with this problem re-spinning the PCB would be one of the first things that come to mind.

Try different brands of 0R resistors. Some might have a geometry which lead to a lower VSWR, but few manufacturers offer this data (mostly because people often go through great lengths to avoid having lots o'jumpers in their transmission lines...). You may even find that low-value resistors minimize reflections and give better overall performance than 0R links.

Is there any way you can hook up a vector network analyzer to see what the traces look like?

At the end of the day the technical reality may be that changing the PCB is cheaper than the engineering hours that are spent trying to patch the current design. I fully understand that the corporate/political reality may make this impossible, though.

JDB.

 

[Svart]

We used a TDR and a network analyzer to tune the transmission lines.  We took into account the mismatches in the line and tuned the ground around them to keep things fairly smooth.  I know it's not the *right* way to do it but we had to pick the lesser of two evils available to us at the time.

I wish we could respin.  I have some other changes I want to make but it is exactly that corporate/political crap that keeps us from doing it.

 

[JohnRoberts]

More questions, not that I could help if I knew.

I still don't know 1 dB out of how many?

Not sure what 0dBmv is. 0dBm is 1 mW, so I guess this is the voltage that makes 1 mW into 50 ohms?

milli Hertz or Mega Hertz, I ASSume Mega. In which case transmission line effects will matter.

From your description of the problem it sounds like the only option available is trying different brands of smt jumpers.

Good luck..

JR
 

 

[Svart]

sorry John its Megahurts (megahertz)  ;D .

Since I work pretty much exclusively in the Mhz region, most people around here are used to being lazy and typing it mhz.

The dB scale of MER is relative to the input signal vs. signal impairment(distortion, jitter/phase noise, drift, etc.) It's more of a goal than an absolute number.  My spec goal was that if I input a signal that is known to be 37MER at 0dBmv RF input to the tuner, that my IF @ 36.125Mhz is also 37MER.  Our design showed 36.2MER on the IF coming out of the tuner.

Anyway, we found some things that got us to the goal this morning.  One was biasing a monolithic gain block differently and some was removing a LPF in our 1st IF path(1db of insertion loss) and then removing a small pad before the first mixer for slightly better conversion loss(at the detriment of slightly lower 1db compression point).  This way we moved more gain before loss that is before an AGC IC which compensates for changes in the input signal levels during calibration.  We then studied the decoupling of certain ICs and decided that we would increase the decoupling.  All of this helped enough to get us about 1.5db better MER/SNR.

I would still like to study more about thermal noise in coupling caps though.

 

[JohnRoberts]

Thermal (johnson) noise in a cap would be associated with real resistance term. Noise current [from active devices] would also work into the cap's effective impedance.

JR

 

[SSLtech]

What does Ohm's law tell us about VOLTAGE noise across ZERO ohms?

Keith

 

[Svart]

but that's my point, it's not exactly zero.  The other part is how does the jumper look in a transmission line (like an inductor) and how does the material react to temperatures, etc.  It's not exactly transparent if you know what I mean.