High 1/f noise in SM resistors

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

bcarso

Well-known member
Joined
Feb 20, 2005
Messages
4,055
Location
San Fernando Valley, Los Angeles
I have been working on an amplifier with a discrete component input stage. I did some noise estimates and figured that the design would be adequate for the application.

When samples began arriving I found a ton of excess low frequency noise. After many red herrings it was traced to excess noise in surface-mount 0805 resistors. A breadboard with axial parts was 25dB (!) quieter in a roughly 5Hz-400Hz noise bandwidth, compared to substituting an SM part in what was the most sensitive point in the circuit.

I then took four pieces of the SM part and connected them in series-parallel to get the same value but cut the current density through and voltage drop across each in half. The noise dropped by 6 dB. By the way, the power dissipation is small (about 7.5mW for a 1/10W part).

These resistors are standard 1% Ruthenium oxide parts from R*hm. I tried some alternate parts in 1206 that were just as bad. I have Yaeg* and P*nasonic parts on the way to test.

I wanted to share this experience and solicit comments, especially if any of you have had similar problems. The datasheets I've looked at so far for SM resistors have no information on an excess noise index.

This effect may account in part for the perception that many surface-mount boards in products have problems with sound quality. Even if there is little or no d.c. running through the parts, you will get excess noise riding on the signal when there is a voltage drop across the parts.
 
Thanks Brad, that's very interesting. Who would have thought that resistors make such a big difference. I'm curious if the other samples will be different.
 
Pah!

What we need is for someone who actually knows what they're talking about to give this some thought! :wink:

J/Kidding!

-Seriously though, Thanks for sharing the line of investigation and keep us posted! -It seems that a lot of the time there's no real penalty for using SM parts, but perhaps there's a particular circumstance in which certain types might be a bad choice...

I'm glad it's your brain pointed towards the problem and not mine!!! :wink:

:guinness:

Keith
 
There's very little about it on the web that I can find so far, although I was marginally entertained by a flame-ridden exchange about mic attentuators in another room, in which the possibility of 1/f noise occurring in a.c.-driven zero-average-d.c. parts was considered, but with no data. The offending party was amazingly insulting, and at one point another poster was sufficiently angry as to suggest someone cutting the flamer's fingers off! The contention and insulting were not related to the 1/f remarks however.

The usual discussion about 1/f noise in pre-SMD days was usually that it arose from flaky end cap connections to the deposited-film glass rod that was the resistive part. Now, from a simple-minded perspective one might think that without leads you would be better off, but so far my experience says anything but.
 
bcarso said:
I wanted to share this experience and solicit comments, especially if any of you have had similar problems

when I was working in the factory on analog tape machines, I remember the engineers were trying to go to surface mount parts but they were much noisier, which baffled them. theory is theory, reality is all-together differrent.
 
I have been wondering all day what the noise could be due to, the solder connection part or the makeup of the resistor body? This is interesting.
 
I have read that the mini-melf resistors should be good for low-noise apps. As far as I remember you can get them in 0805 and 1206 "compatiple" sizes. The Philips/BC/whatever the name is resistors are available in 1% and 0.1%, and I think they are metal film.

Best regards,

Mikkel C. Simonsen
 
If you have trouble getting the MELFs (I don't know where you are), I can send you some. Just ask.

Best regards,

Mikkel C. Simonsen
 
Thanks Mikkel. In the long run I will have to find a source that either is in, or ships to, China, as that is where this product will be manufactured.

I found some data on MELFs and for the first time in a SMD part there are specs on noise and voltage coefficient, so clearly they are thinking about this parameter. Koa Spe*r make some of them that will fit the 0805 footprint, which is good because we are running out of time to make board changes. Vish*y Beyschlag also have some parts.

Probably Garrett is the most promising local distributor to me for samples.
 
Hey Brad...

I've got a very large inventory of tantalum nitride melf resistors in 805 and 1206. I've used the Beyschleg/Philips resistors in the past. They are thin-film nichrome (I think). There are other thin-film, non-melf 805 resistors as will (Digi-Key has them to 0.1%). I'll send you some of the melf resistors to try out if you want them. Send me an email (not PM).
 
[quote author="FredForssell"]Hey Brad...

I've got a very large inventory of tantalum nitride melf resistors in 805 and 1206. I've used the Beyschleg/Philips resistors in the past. They are thin-film nichrome (I think). There are other thin-film, non-melf 805 resistors as will (Digi-Key has them to 0.1%). I'll send you some of the melf resistors to try out if you want them. Send me an email (not PM).[/quote]

Thanks Fred. Have you had good results with these? I may take you up on your offer.

Last night I substituted some Ko* Speer 1/8W axails on the actual PCB and the improvement was less dramatic. So it may be that for truly low noise the parts have to be pretty good-sized. Simply going to a mini-MELF may not be a huge improvement, although to keep it all in perspective the noise is getting to be within 5 dB of the noise downstream, so it is tractable. One of the problems with this design is the line level source is rather low level, necessitating a lot of system gain---the 0 dB digital output at max gain of the source is only about 160mV rms. If a had a few volts to work with life would be easier.

I just got a series of shipments in of 0805, 1206 and 1210 values from other vendors to try. It is going to be a long and tedious day...
 
[quote author="bcarso"]It is going to be a long and tedious day...[/quote]
But please report your tedious results when you're done :wink:

Best regards,

Mikkel C. Simonsen
 
Brad;

I was going to suggest you contact Fred Forssell
but he gave his 2 cents.

Quantech had a box that would measure resistor
noise. This meter was called out in the mil spec. that
discussed excess resistor noise. In my town
a resistor manufacturer has one of these machines.
I might be able to go there and use it to take measurments.

Example of instrument
http://www.atecorp.com/Equipment/Quantech/315B.htm
 
Thanks---yes, I remember Quantech now that you mention them.

My preliminary results:

The R*hm 0805 13.3k parts are by far the worst. In the test circuit the noise was 25dB above the reference standard of a 1/4 metal film axial part, probably from K*a Speer by the looks of the body.

I got P*nasonic and Y*geo 0805, 1206, and P*na 1210 parts today, as well as some D*le RN60 axials.

Ideally there should be many more tested to get any kind of decent statistics, but life is short. Here are the results, as noise excess compared to the reference 1/4W:

P 0805: +15.1dB

Y 0805: +15.1dB

P 1206: +14.5dB

Y 1206: +12.7dB

P 1210: +6.8dB

D 1/8 axial: +1.6dB

Noise Bandwidth was about 5Hz - 470Hz; The average voltage was eyeballed over many seconds of observation from the voltmeter function of an HP 334A---the analog meter being much easier to observe than a DVM. The test circuit (a replica of the SMD PCB with leaded parts except for the devices under test in the most susceptible location) output fed a preamp with a gain of 800 and negligible self-noise, and there was a one-pole lowpass filter with a -3dB point of 300Hz, hence the noise bandwidtrh was 1.57 times this. The noise is predominately 1/f-ish with probably most of the energy below 50Hz from the look of it on a scope. For the measurements the test circuit was shielded to reduce hum pickup.

I have yet to get any MELF parts.
 
> excess noise in surface-mount 0805 resistors. A breadboard with axial parts was 25dB (!) quieter in a roughly 5Hz-400Hz

This is disgusting. It is 1929 all over again, when resistors were made of coal-dust and clay-mud packed together in different proportions. I don't think the last carbon-composition resistors were anywhere near 25dB excess LF noise. I've never noticed excess noise in carbon-film.

How did the SMT folks screw this up so bad????
 
It's pessimistic to say this, but I fear the same thing is happening to analog electronics that happened to vinyl records. They hastened the demise by using crappier materials and poorer mastering and meanwhile introduced the compact disk, "perfect sound forever." (sorry Ph*lips)

But---I could be wrong.
 
Contact noise is dependent on both average DC current and resistor material/size. The most significant contributor to noise in guitar amplifiers is the use of low-wattage carbon composition resistors. Since the noise is proportional to resistor size, the use of 2W carbon comp resistors will improve the performance over that of 1/2W resistors. Studies have shown a factor of 3 difference between a 1/2W and a 2W carbon comp resistor operating at the same conditions. -aikenamps.com

He is talking about carbon comp but the principle applies to film resistors to some extent. I would expand that to say that 1/4w are more noisy than 1/2w and 1/8w more noisy still.... size does make a difference.

regards, Jack
 
Back
Top