Noise Reduction in high Gain circuits - Metal Film?

[Steve Conner]

1. As a general principle, for any amplifier of significant gain (i.e. over about 10) the first stage noise dominates the noise of the whole unit unless there is a serious flaw in the design...

You mean like attenuating the signal by a factor of 20 to drive a series effects loop, then amplifying it back up again? Or running it through a reverb tank whose pickup coil always has full gain to the speaker, irrespective of what the volume controls are set to?

Classic amps are full of WTF moments like these that would have "real" EEs beating their heads against a wall. In fact, even using vacuum tubes is a WTF moment for many of them. They're certainly noisier as a first stage than your average 50 cent op-amp or JFET.

Precisely why I pull all the cathode resistors out, and use fixed bias. Then, the preamp gain can be set at any level since there is no noise floor to contend with.

Ah can nae break the laws ae physics, Cap'n! The cathode resistor doesn't contribute significantly to start with, because:

Its value is low, and the level of Johnson noise is proportional to the resistor value. (The square root of it to be precise)

It's bypassed by a capacitor that shorts the noise voltage. So the relevant resistance for the noise calculation is the capacitor's ESR. Or maybe the tube's internal cathode resistance of 1/gm, I wouldn't be surprised if that generated thermal noise just the same as a "real" resistor.

In high-gain amps I've built, it seemed that most of the thermal noise voltage came from the resistance of the guitar volume control. The hiss level would decrease as the volume was turned towards zero. Or maybe it was some internal noise voltage in the tube, feeding back out of its grid, that got shorted.

This is one reason why amps have shorting input jacks: so you can't hear the noise voltage of the first stage grid leak resistor when nothing is plugged in.

The absolute (Kelvin) temperature also appears in the equation for thermal noise voltage, so if you really want to get rid of noise, you could do a lot worse than dunking your first stage in liquid helium. People still do this in really high-performance preamps for radio telescopes and the like, but so far nobody has got Eric Johnson interested.

[mooreamps]

I've studied a few doctoral thesis's on the subject, when doing the research when I was writing my own. I am a real EE, and a rocket scientist so I think I have a pretty good idea what I am talking about. and for what it's worth, no I don't bang my head playing with guitar amps, actually I do it more as a recreational activity because the concepts are quite elementary...



As far as using a shorting plug on the input jack, might be a good question for someone like Kevin O'Conner of London Power. But if I were to guess, I would guess it is used to keep the amp quiet when nothing is plugged in, from picking up noise from the environment since the input circuit is a high impedance circuit.

-g

[R.G.]

I am a real EE, and a rocket scientist so I think I have a pretty good idea what I am talking about.

Kewl. So we're alll bozo... er, EEs on this bus.


Sadly, in my experience the average EE is from modestly to highly clueless about analog anything these days. No time in the schedule for that.

But I really, really did always want to work with things that had explosive bolts in them.

[mooreamps]

Kewl. So we're alll bozo... er, EEs on this bus.


Sadly, in my experience the average EE is from modestly to highly clueless about analog anything these days. No time in the schedule for that.

No time in the college class schedules either, as I have noticed. They just don't teach much of the basic electrical theory any more. Mostly very heavy in math, and abstract theory. It's doubtful if any of these kids could draw the basic schematic, for a circuit like say, a multivibrator. They may understand how it works as in a chip, or how to program the symbol into something like inside an ASIC, but not know how to draw one up from scratch.

But I really, really did always want to work with things that had explosive bolts in them.

We call them "fragmentible bolts". I know this because I did the pyrotechnics for the Orion program when I was working in Denver last year.


-g

[R.G.]

We call them "fragmentible bolts". I know this because I did the pyrotechnics for the Orion program when I was working in Denver last year.

Yeah. "Explosive" sounds so ... terrorist...

Probably the terms will eventually get to be something like "bolts which may be opened with extreme prejudice, remotely" or the like.

[mooreamps]

It's the Flight Software that fires the charges that blows the bolts and blows the cable cutters when it does it's separation sequence of the crew capsule from the launch vehicle.


-g

[mooreamps]

Carbon comp resistors will hiss like a snake. Apparently the most popular approach is to go with the metal film. Myself, I don't use any resistor in the cathode circuit, instead using fixed bias.


-g

yes I did. But I'll elaborate my point, just one more time on this thread, as I have wrote the same comment on other threads, on other forums, which they didn't get it either. The thermal noise generated by the resistor is amplified by the high input impedance of the grid to cathode circuit. Further, if you read theads that suggest putting two triode gain stages in parallel gives you a noise reduction of 3 db, it's not because of the two gain stages in parallel. If you notice, when putting the two gain stages in parallel, the typical value of the cathode resistor drops from 1.5 K to 820 ohms. This is where the 3 db comes from. Everybody missed that one, even Kevin....

-g

[Robert M. Martinelli]

Hi all,
Back on topic (though I find rocket science is a very interesting subject), I'd like to highlight that we have to distinguish between the various "kinds" of noise. Carbon comp resistors are noisier because their "shot" noise figure is worse. Their "thermal" or "Johnson-Nyquist" noise figure is the same of the best metal film resistor of the same value, because, as Steve said, this "kind" of noise is proportional to the resistors' value, and has nothing to do with the material they're made of.

As we all know, the "thermal" noise within a given bandwidth is : En=Sqrroot(4KTR*DeltaF), where K is the Boltzmann constant, T is the resistor's temperature in degrees K, R is the resistor's value in Ohms, and DeltaF is the frequency bandwidth of interest.

This means that, once we have chosen to move to metal film resistor to minimize "shot" noise, the only other factors we can try to alter to lower the thermal noise figure within the same bandwidth are the resistor's value and its temperature, therefore, when possible, it's good to try to keep the resistors' value as low as possible (compatibly with the design needs), and to lower their temperature (without using liquid Helium ). To achieve this, when noise is a main concern, oversized resistors can be used (1 or even 2 W where a 1/2 W would be enough).

This said, I tend to agree with Steve that the cathode resistor in a "normal" gain stage is far less significant, "noise-wise", if compared to the anode or grid leak resistor, whose typical values are hundreds or even a thousand times greater. (things change in an LTPI, where the "tail" resistor value is higher, say 10 to 47 KOhm depending on the tail "length"). If the stage needs to be bypassed, when possible I would tend to use poly capacitors to get rid of the ESR (and low frequency distortion) normally associated with electrolytic caps.

I also agree about the importance of keeping noise as low as possible in the very first stage to achieve the best possible S/N ratio.

JM2CW

Best regards

Bob

[raiken]

Further, if you read theads that suggest putting two triode gain stages in parallel gives you a noise reduction of 3 db, it's not because of the two gain stages in parallel. If you notice, when putting the two gain stages in parallel, the typical value of the cathode resistor drops from 1.5 K to 820 ohms. This is where the 3 db comes from. Everybody missed that one, even Kevin....

-g

This is utter horseshit. Nobody missed anything except you, obviously. The noise level is 3dB lower for paralleled triodes because the noise sources of the two triodes are uncorrelated and add as the square root of the sum of the squares of the individual noise sources, while the signal is correlated and adds directly, giving a 3dB improvement in signal-to-noise ratio. It has nothing to do with the lowered cathode resistor value.


Randall Aiken

[d95err]

if you read theads that suggest putting two triode gain stages in parallel gives you a noise reduction of 3 db, it's not because of the two gain stages in parallel. If you notice, when putting the two gain stages in parallel, the typical value of the cathode resistor drops from 1.5 K to 820 ohms. This is where the 3 db comes from.

The noise level is 3dB lower for paralleled triodes because the noise sources of the two triodes are uncorrelated and add as the square root of the sum of the squares of the individual noise sources, while the signal is correlated and adds directly, giving a 3dB improvement in signal-to-noise ratio. It has nothing to do with the lowered cathode resistor value.

So, if a parallelled input stage has separate cathode resistors for each half. Would this reduce noise?

[mooreamps]

It has nothing to do with the lowered cathode resistor value.


Randall Aiken

You can measure it on an o'scope.


-g

[wizard333]

TO clarify a point made earlier,

Leaving carbon comps aside, can one expect much of a noise difference between carbon films (what date range qualifies as "modern" here?) and metal film, given the same value/rating? Again, talking about a high gain circuit.

[raiken]

You can measure it on an o'scope.


-g

Measure what? The 3dB improvement in signal-to-noise ratio? Please detail your procedure.

Better yet, since you are an EE and rocket scientist, how about detailing the math behind how the smaller cathode resistor gives you a 3dB improvement in signal-to-noise ratio in paralleled triodes.

Also, if you get a 3dB improvement by halving the cathode resistor, does this mean that you get a 6dB improvement if you reduce it by a factor of four? Enquiring minds want to know!

Randall Aiken

[raiken]

TO clarify a point made earlier,
Leaving carbon comps aside, can one expect much of a noise difference between carbon films (what date range qualifies as "modern" here?) and metal film, given the same value/rating? Again, talking about a high gain circuit.

Thermal noise will be the same, as it is a function of the value of resistance, not the construction, and other factors, such as temperature. Contact noise will be higher for carbon film than metal film, so the metal film will be quieter.

Note that different brands and types of film resistors vary in noise performance, so they are not all equal.

Here's a link to a book on the subject:

Discrete electronic components - Google Books


BTW, take a look at Fig 3.6 at the bottom of page 60 - it details a "ballpark" relative comparison of noise (and other parameters) on a scale of 1 to 6 for potentiometers.


Randall Aiken

[wizard333]

Is there a particular brand/type that you would recommend for best noise performance?

Thanks!