Impedance scaling, low-noise, and does it matter?

[sault]

I'm reading Small Signal Audio Design, and one of the big concepts that Mr. Self brings up numerous times is impedance scaling and how lowering resistances helps lower noise (this made my ears perk up - I basically hate noise with a passion).

Example - say the schematic calls for a 47k and 470k resistor in an inverting op-amp. Instead, put a 47k (or whatever) shunt resistor in front, then use 1k and 10k resistors instead. Your noise (and op-amp distortion) drops considerably.

It's kinda like "noiseless biasing"... using two 10k resistors to get a 1/2 supply voltage, and connecting via 470k or 1Meg resistors to wherever it needs to go, versus using two 10Meg resistors and connecting directly. (if this isn't making sense I can post a schematic to illustrate what I mean)


Is this kind of effort worthwhile for guitar pedals, or is this something that stays in the realm of hi-fi?

[R.G.]

It's a question of practicalities. All active devices have some internal noise. All amplifying devices amplify their own noise, in addition to any noise contributed by the resistances around them and the noise of the signal source itself. If your active devices contribute a lot of noise, or have a poor match to the source impedances (in the noise sense, not in the power-transfer sense) then diddling with resistor values doesn't help enough to make a difference. If the gain of the active devices is so high that the thermal noise of the input resistors is amplified to more than the amplified noise of the active device, then yes, messing with resistors and resistor scaling makes sense.

None of which answers your question. As a practical matter, I've found that trying to get resistances as low as possible is generally not fruitful except for cases where the signal source impedance is quite low (that is, not your typical electric guitar pickup) or where the amplification is very high. Warp-9 distortion pedals present an interesting situation, because the signal is clipped off, but the noise is not, so the noise gets amplified more than the signal. It may be worth tinkering with in that case.

A real answer is that sure, it would be useful to compare the noise contribution of the active devices and resistors and find out where the noise is coming from. But in practice, the humble TL072 is so good in most cases that only for very high gain pedals do you get much out of messing with lowering resistances. There's a good sendup of the kind of calculations needed in Henry Ott's "Noise Reduction for Electronic Systems", where the idea of a noise budget is presented. To get less noise, you have to find out what's producing it, and what's even possible to change. Messing with resistor values only lowers resistor thermal noise. If the other contributors are the dominant noise sources, resistor scaling won't help.

[earthtonesaudio]

Also, other types of passive components have their own noise issues, so if you aren't careful you could trade one form of noise for another.  For example inductors can pick up hum, and capacitors can transduce mechanical vibration, and both of these issues may get worse as you increase their values. 

[PRR]

> lowering resistances helps lower noise

Lower noise voltage.

Noise _power_ is a constant. The only true improvement in S/N is More Power.

Note that noise voltage changes as square-root of resistance. It takes a BIG change of resistance (and circuit power) to make an audible change of hiss.

Note that gain masks hiss faster than resistance scaling can. If your source hiss is 1uV and you can find low-noise gain of 10, then you have 10uV of amplified source hiss and quite large resistors won't add more.

> worthwhile for guitar pedals

The rub is that guitar is 5K-250K impedance. In the hiss range you can approximate 100K of hiss resistance. This makes 2uV hiss across the guitar band.

Or in other words: the POWER available from gitar pickup is the limit.

There is another way to do it. Instead of winding-up for max voltage until capacitance hammers our highs, we could wind-down to 200 ohm impedance and use a transformer if needed to suit available preamp devices. Les Paul did this on some recording guitars. It has never caught-on, because of cost but mostly (IMHO) because winding-up to cable capacitance (and loading lightly) adds a resonance which is musically useful.

> It's kinda like "noiseless biasing"...

No. That fine scheme clobbers power supply "noise" (usually hum/buzz). Battery designs split the raw battery for bias, putting half the battery hash on the input, which was fine because batts are low hash. Replace battery with a wall-wart, you have half of huge ripple going right to the input. The RRCR plan clobbers the buzz-path from the power supply.

> say the schematic calls for a 47k and 470k resistor in an inverting op-amp

NEVER use that type inverting amp as a low-hiss input. It just can't work. (That said: I have used such schemes and nobody complained.)

> put a 47k (or whatever) shunt resistor in front, then use 1k and 10k resistors instead.

Well that gives a 0.979K input impedance. 47K is already low for guitar, <1K will SUCK all the voltage off the pickup and S/N will be awful just because the _S_ is so low.

> op-amp distortion.. drops considerably.

Depends on the opamp; but generally 10K-100K resistors will not induce the kind of THD you care about for guitar. Yes, audiophiles can find nonlinear CMRR and nonlinear input currents which raise THD from 0.000,1% to 0.01%..... what me worry?

[Earthscum]

I can second the "amplifying existing hiss" vs. resistor scaling. I got that lesson this week while playing with filters. I kept thinking how all these LP filters using a variable HP T filter in feedback are always so hissy, and here I'm running 4 in series. I had scaled the resistances to 10k's all the way through, boost in front, and was trying to use various feedback to increase the effective filtering to compensate for my losses.

Well, the hiss went down. Two days ago I decided to turn the input stage into just a buffer and add gain at each stage (guess what... 47k and a 100k feedback resistor give just about enough for unity at the output!). When I "opened up" the filter, a ton less hiss than I had previously. If I wasn't so frustrated with it, I'd bump up the resistances some more and have the nice filtering I wanted in the first place, but anyways... that is one of the things I learned this week. I think next week I'll teach myself that OPA's really are quieter than discrete transistors when I design stuff. 

[sault]

Well that gives a 0.979K input impedance. 47K is already low for guitar, <1K will SUCK all the voltage off the pickup and S/N will be awful just because the _S_ is so low.

This is what I meant - it was the example given. Presumably they needed to present 47k impedance in this particular situation. For guitar it would definitely need to be higher, sure.

Note that noise voltage changes as square-root of resistance. It takes a BIG change of resistance (and circuit power) to make an audible change of hiss.

The examples he gave were reductions of approximately an order of magnitude or two.... going from 47k down to 500-1000 ohms.

Warp-9 distortion pedals present an interesting situation,

And this is pretty much exactly why I became so interested. Sometimes I take a break from the soundboard and sub in on guitar for the worship team, and if I go straight into the board (guitar -> pedals -> preamp -> soundboard) I get a fair bit of noise. Looking at ways to cut down the noise, especially with gain. I have an onboard buffer preamp in the guitar, so I'm wondering if its worth it to take the soldering iron to it... (TLxxx opamp, so perhaps not)

[R.G.]

Warp-9 distortion pedals present an interesting situation,

And this is pretty much exactly why I became so interested. Sometimes I take a break from the soundboard and sub in on guitar for the worship team, and if I go straight into the board (guitar -> pedals -> preamp -> soundboard) I get a fair bit of noise. Looking at ways to cut down the noise, especially with gain. I have an onboard buffer preamp in the guitar, so I'm wondering if its worth it to take the soldering iron to it... (TLxxx opamp, so perhaps not)

I would guess that unless the preamp on the guitar is really bad in some way, no, it wouldn't help a lot. A subtle noise gate and maybe a quieter distortion pedal might. There is also the possibility that the guitar controls themselves have excess noise. I've seen really badly noisy resistors. Could be a bad contact inside the guitar as well. These aren't likely, but possible.

Electric guitar presents some unusual challenges. Because of the inductive nature and the need to keep treble in the signal, you're really stuck with having to present it an input impedance of more than 100K, and 1M is what everyone's ears expect for historical reasons. As a result, scaling down the resistors that a guitar drives is impractical for tonal reasons.

As a general principle, for lowest noise, you want to concentrate on the first amplification stage the signal touches. You want to select the amplification device to work well (in a noise sense) with the signal impedance, and you want to get a big slug of gain in that first stage. The noise of the first stage is amplified by everything else, so you want whatever internal noise the first stage contributes to be amplified as little as possible by the rest of the circuit. Hence, get your gain early.

An onboard buffer is a good idea for the normal situation of needing to drive a capacitive cable and effects that may eat treble by loading, but it is often a low-gain device. And properly so, because if it drives pedals, you usually don't need volts of signal driving pedals that expect to be given signals in the 100mV range. So you get low impedance but also low gain, which means that a pedal with high gain amplifies the added noise of the onboard buffer too. Generally you get away with this.

You might want to try bypassing the onboard electronics for testing if you can. If this makes the whole chain a lot quieter, then it's worth the effort to dig into the onboard electronics. If not, check out the guitar controls, solder joints, etc. Once you eliminate the possibility of there being a problem on the guitar, you can move to the next issue of whether the distortion pedal itself has noise problems.

[Earthscum]

And this is pretty much exactly why I became so interested. Sometimes I take a break from the soundboard and sub in on guitar for the worship team, and if I go straight into the board (guitar -> pedals -> preamp -> soundboard) I get a fair bit of noise. Looking at ways to cut down the noise, especially with gain. I have an onboard buffer preamp in the guitar, so I'm wondering if its worth it to take the soldering iron to it... (TLxxx opamp, so perhaps not)

Ooooh... playing bass I run into this quite often. I avoid running to the board directly at all costs because of this. No hiss from my amp, right? I breadboarded a speaker sim circuit, started with the one from Runoffgroove and played around with it a little, but... I discovered the hiss is always there, and tolerable THROUGH my amp. The speaker sim did an awesome job of making the signal sound fairly close to what my amp puts out, and cut down the hiss to the board by about the same amount. You may give it a try... I'm just going to build one into my Pedalboard, incorporating it into the DI out so I don't have to argue with sound guys anymore (argument always ended when I turned on the power and jammed with my drummer... we play at the same level, so the sound guy only has to worry about guitars and vocals... BW's are the bomb!)

[sault]

so the sound guy only has to worry about guitars and vocals...

I'm a little jealous! I'm working with a relatively small room, and its a constant effort to balance sound. I have to implore the drummer to play quietly - even with those clear baffle things its very easy to overpower everyone else, especially when the crash signal gets a good whack. Its kind of a nightmare... 90% of the sound comes from the monitors   , so all I can do is use the mains to pick up the slack and fill in the sound.    I have to ride the faders, especially when I get a very dynamic singer... it's nice from the perspective that I've got people who *can* sing and have very resonate voices... except when it comes time to keep them from blowing up the mix...

A subtle noise gate and maybe a quieter distortion pedal might.

I messed around a bit last night with guitar -> Behringer V-Amp Pro -> Tubeworks mosfet power amp -> PA speaker. It actually worked quite well, I was very happy with it. I only got noise, this time around, when I selected the very high gain amp models. However, guess what I like to play with! The other piece of gear I use is a Tubeworks tube preamp, and that's got noise issues for sure - I have to turn up the gain to get a positive overall volume boost, and that means the noise comes back...

you're really stuck with having to present it an input impedance of more than 100K, and 1M is what everyone's ears expect for historical reasons. As a result, scaling down the resistors that a guitar drives is impractical for tonal reasons.

My thought was to go pickups -> buffer (with shunt Rin of 470k-ish) -> tone controls (impedance scaled) -> out

This way I can avoid using a 1Meg pot. I have two Seymour Duncan pickups on this guitar... an Alnico II Pro in the neck and an Alternative 8 in the bridge. They sound great together, the leads are rather Slash-ish and the bridge can still represent with a bit of chunk without losing too much clarity (a la the JB that its loosely based on). The Alt-8 has a 1Meg pot, though, and its the noisier pickup of the two. So, to make a long story a little shorter, the grand idea would be to suitably buffer the pickups, show them a decent impedance, but still get the benefit of a lower impedance (ie not having to use a 500k tone pot).

I *think* that the onboard buffer as it is only comes after the tone pot.

...

You know, after some thought, I just realized that I used to play in a rock band, with this guitar, into a Mesa Boogie Recto Head, and never had noise problems like this. Dang, this would have been a valuable insight like a month ago when I start futzing with this, or even a few paragraphs ago.

See, this is what having kids does to you - you forget things.   

Looks like I need to focus on that tube preamp and maybe get or build a power conditioner. Ahem. Sorry.   

[R.G.]

ES has a point. Guitar amps and speakers don't have much treble response, so they wipe off some of the hiss that's there. A board tries to preserve all the treble it can, so it ensures you hear that there is. A speaker sim or just a high cut filter may do what you need if there are not other bugs in the path.