Q1: Why do most input buffers have unity gain ? Q2: Noise

[Vivek]

Q1 : Why do most input buffers have unity gain ?

For example:



Why not a slight gain like maybe 2 or 3, to help a bit with the signal/noise ratio ?

Given that an Opamp can satisfy all necessary criteria like input impedance more than 500K, output impedance less than 1K, frequency response 1Hz to 30kHz, and still deliver gain with those constraints ?

Q2: Can most Opamps of today deliver lower noise than BJT and FET for input buffer designs ? Is it still wise to design discrete input buffers of unity gain or so much easier and better just to use almost any modern Opamp ?

[Rob Strand]

It's trade-off between headroom and signal to noise.    If operating on battery you might want the battery to run down to 7V.
With the level of signal on a guitar/bass you might stretch the gain to 3 but considering battery you might go for 2.  On non-distortion pedals you can do better by using pre-emphasis/de-emphasis.

For fully-bypassed you won't gain much signal to noise improvement so you might as well just go for a simpler unity gain.   It also means you don't need to have issue with matching input and output gains of x2 and x1/2 and ending-up with 1dB gain due to tolerances on a pedal.   The 0dB comes by nature from the Buffer (although a JFET buffer can end-up with a noticeable loss).

[Phoenix]

In the example shown, you don't really want your buffered bypass to have gain now do you?
I think this observation mainly applies to this situation. Most designs that do not have buffered bypass tend to have an input stage with some gain.

Regards your second question, a good discrete jfet buffer or amplifier is still the king of low noise for medium-to-high impedances.

[Vivek]

Oh, maybe poor choice of example posted above. Sorry !!

Please consider the case of a true bypass pedal

with a 18V charge pump

Would it be better for input buffer to have some gain, to help improve signal to noise ratio ?

[Phoenix]

Would it be better for input buffer to have some gain, to help improve signal to noise ratio ?

Yes, though you're only going to confuse people by calling it a "buffer with gain", most would call it an amplifier (of which yes, a buffer is a type). All amplifiers buffer; i.e. provide impedance transformation, but the average person is hyper-focused on the idea of a buffer being unity gain.

[Steben]

Oh, maybe poor choice of example posted above. Sorry !!

Please consider the case of a true bypass pedal

with a 18V charge pump

Would it be better for input buffer to have some gain, to help improve signal to noise ratio ?

If a distortion circuit follows, having gain in stage 1 is more than useful. If not, gain might be a burden If what follows is very sensitive. I've come to appreciate a good opamp will never add annoying noise levels if the signal is big enough. In case of a guitar they usually are. On top of that, analog delay based circuits add more noise than any input buffer can.

[antonis]

It's trade-off between headroom and signal to noise.

This..!! 

By implementing a buffer with gain, you come to terms with "first amplify, then attenuate" resulting into possible final S/N ratio loss due to "extra" noise arisen fron attenuation configuration..

[Vivek]

Thanks everyone

A) So would you say for a new design with true bypass and +18V supply, with first stage that is followed by 2-4 stages that distort

Is it better to have stage 1 with gain 1
or
Stage 1 has gain of 3 - 5 x ?

The idea I am trying to understand is "Will first stage with gain help in improving S/N ratio of the pedal since signal levels will be higher ?"


B) I heard that interstage "amplify then attenuate" is acceptable concept in tube amps, when the designer wants to distort the signal by overdriving an earlier stage, but he does not want to overdrive the next stage too much. Can't that concept be used advantageously in SS pedals and AIAB ?

[GibsonGM]

My 2 cents:

A) I think it depends what you're after. If you want high input impedance, you buffer first (unity gain), then on to the circuit.   If your input section has 'good enough' input impedance and it's appropriate to what you're doing, use a gain stage.        Starting with a VERY high input Z may help you get more high end; more 'chime'.  Or, 'regular' input Z of say, a FET opamp, might be more than enough for you.   It depends on taste and what the topology is...there is no right answer every time!   Some people love a buffer first for what it gives the overall effect; others don't use them much at all.

B)  In a tube amp, the signal voltage after the input stage is extremely high when compared to a 9 or 18 volt powered pedal.  Even after attenuation, the SNR is still quite good because of this, so you can get away with some pretty brute-sounding techniques like heavy attenuation, or using loading to achieve a particular tone and so on.  It's a little bit different world  

[Steben]

It's trade-off between headroom and signal to noise.

This..!! 

By implementing a buffer with gain, you come to terms with "first amplify, then attenuate" resulting into possible final S/N ratio loss due to "extra" noise arisen fron attenuation configuration..

Indeed.
Brings the question what really needs amplification of a guitar signal besides distortion circuits?

[Vivek]

TC ELECTRONIC integrated preamp and many other pedals amplify signals but do not distort.

[temol]

TC ELECTRONIC integrated preamp ........... amplify signals but do not distort.

Not entirely true. TCE Integrated pre with  controls maxed.

[Steben]

Well, you know.... what do most use preamp and booster pedals for? 

[ElectricDruid]

Q1 : Why do most input buffers have unity gain ?

Why not a slight gain like maybe 2 or 3, to help a bit with the signal/noise ratio ?

Depends a bit what the rest of the circuit is. OTAs are famously noisy, for example, but there wouldn't be much point adding gain to the input stage in that case since you'd have to reduce the signal level for the OTAs anyway. That affects several phaser designs, some tremolos, and many envelop filters! BBDs don't have a lot fo headroom either, and the distortion they produce is unpleasant. A 3207 run at 5V can only cope with about 1.2Vpp. The PT2399 is slightly better, but still only a couple of volts. So adding gain of x2 or x3 is certainly likely to clip with some guitars/pickups.

For your specific situation of a distortion pedal, yes, I'd say more signal level will help and some boost on the way in is sensible. This is in fact exactly what I did in the Hard Bargain pedal design - the classic op-amp follower gets tweaked to become a non-inverting amp with x4.3 gain (12dB). And incidentally I *did* call it an "input buffer with gain"!! That's how I think about it too. The "buffer" function for me is about impedance matching, providing a high impedance input and a low impedance output. Whether that is done at unity gain or something a bit higher is therefore irrelevant to the buffering action.

https://electricdruid.net/wp-content/uploads/2019/01/HardBargain5SchematicPg1-scaled.jpg

[Vivek]

Thanks Tom !


I did some analysis on the Rat which has rail saturated Opamp feeding hard clip diodes.

Please teach me what does adding hard clipped diodes do after rail saturated Opamp ?

[Phoenix]

It's perhaps worth pointing out that the example circuit you originally posted Vivek, the Ibanez Super Tube, is actually a good example of an input circuit with gain!
Once you strip away the electronic switching - down to the bare effect - the input stage does in fact have a variable gain of between 1 (unity) and ~107.

If you were to build this true bypass you'd just up the value of R3 to provide a more sensible input impedance, lower the value of C2 to suit the new value of R3, and add an input pulldown resistor. If you wanted to keep the output impedance lowish, I'd lower the value of the level control to say 10k, R12 to 100R and up the value of C8 to compensate for those changes.

[Steben]

Thanks Tom !


I did some analysis on the Rat which has rail saturated Opamp feeding hard clip diodes.

Please teach me what does adding hard clipped diodes do after rail saturated Opamp ?

little at very high gain except cutting the max amplitude to 0.7V and lengthen the sustain even more

[R.G.]

We keep running into questions that seem like they should have simple answers,  but don't.   

The whole issue of whether it's better to buffer with no gain or amplify first, etc. gets back to the eternal question of objectives - what are you trying to do? Several of the comments in the thread get at this, noting the size of the typical guitar signal, the size of the power supply available to a pedal, whether distortion is wanted or avoided, and so on. They're right. And they correctly note that the requirements may work well with each other or compete. It's always a trade-off, isn't it?

For high performance amplification, you have to take into account the nature of your signal source and the nature of your load. For guitar electronics, the source comes in only a few flavors. These are single coil pickups, multicoil humbuckers and the like, and active pickups. I'll ignore active pickups. Single and multicoils have some combination of signal voltage and an R-L-C source impedance. These all matter to the design of an effect front end. The high inductance, typically 1/2 to four henries leads to most of the complications. It causes a resonant peak with the pickup's self-capacitance, and interacts with the cable capacitance and the input impedance of anything the guitar is plugged into. The coil resistance and any resistance of the tone/volume controls contribute thermal noise as a guitar input also.

The solution effects enthusiasts have settled on is to make the input impedance of pedals and amplifiers be "high" compared to the impedance of the pickup/controls/cable combination. a quick hand-wave at inductive impedance shows that an input impedance of 1M or so is good enough for rock and roll and that's where we generally leave it. This makes the pedal/amp not contribute to the treble loss that the guitar/cable combination has naturally.

In the discrete-component era, amplifiers were mostly bipolar transistors, and these have to be fiddled with to get them up to a meg or so input impedance, generally by using either a follower or gain-plus-feedback. The emitter follower has other issues, but is reasonably good at getting up to a high input impedance and low distortion for signals at 1V or so with a 9V supply, so it got used a lot. To get more performance, multi-transistor circuits with gain let designers use feedback to trade gain for higher input impedance and more predictable gain, so it was possible to do a bipolar buffer-with-gain, a high input impedance amplifier. Noise performance was ignored a lot, as early bipolar transistors were themselves so noisy that the pickup noise was negligible. Today, we have miracles-on-a-chip so the input impedance and gain-feedback questions are largely moot. Just pick what you want, and be careful with the biasing structure so it doesn't mess things up. As an aside, the low input impedance of  the early bipolar distortion pedals worked well with the practice of filtering highs before distortion to cut down on some ear-bleeding sounds, and also promoted the idea of having to use "true-bypass" to disconnect the treble-sucking effect input in bypass.

For high performance and low noise design, the maxim is to use an input impedance that interacts with the source impedance to produce the lowest RMS combination of thermal noise of the source, thermal noise of the biasing network, internal voltage noise of the amplifier and internal current noise of the amplifier all mixed together and to limit bandwidth to cut out noise in frequencies you aren't interested in. I have two textbooks on low noise design that rant on about this. In the big leagues, this leads to the idea that the noise of the first stage dominates the noise of the whole amplifier, and that you should always use gain in the first stage, and match the noise characteristic of the first stage to the source impedance of the signal source for the desired bandwidth. Guitar source impedance is moderately wide band (making a wide window for noise to get in) and primarily inductive. It's a similar set of requirements to design of a low noise phono preamp, although bigger signal, narrower bandwidth, and much higher inductance. And now we're down into the tradeoffs that make unity gain buffers versus buffers with gain a question of objectives.

The output side of the nominal effect gets into this a bit, although it's more tenuous. A typical tube-input guitar amp has the grid of a 12AX7 and a 1M resistor connected to one or two 68K resistors going to the input jack(s). That's the input out ears expect from a tube amp, historically. A 12AX7 has an input range of about 1.2V from cutoff to the zero volts from grid to source that makes its input impedance change from near-infinite to about 5K. A 12X7 grid usually hard-clips a high impedance input signal that pushes it over the Vgk=0 level, while squash-clipping signal peaks that drive it to about -1V. This is all baked into what we think a guitar amp sounds like, and gets much ado about compression, tube sound, touch sensitivity, yada, yada, yada. What's not widely recognized is that triodes (well, and pentodes, etc. too) don't just quit amplifying or changing plate current when the grid voltage equals the cathode voltage. There's more current left to move from cathode to plate, but the grid impedance changes to be radically lower. If you can drive the grid with a low-ish impedance signal, the grid conducts all right, but the signal at the plate continues to move around instead of being hard-clipped. It is in fact possible to run a tube with the grid biased with a very high resistance to the plate supply and have it work; this is (was...) called "running in clamp".

So a low impedance amplifier driving a guitar input can in some circumstances drive the grid positive relative to the cathode, and get a squashed signal instead of a hard-clipped signal from grid overdrive. Low impedance drive of tube grids can make them sound different for certain signals. I suspect the term "overdrive" may have originated from someone noting this effect, although I can't prove it. In any case, some pedals with good reps deliberately have big output voltages and probably get some effect from putting big signal voltages intl a 12AX7 grid. It might be that 18V pedal supplies have this lurking behind them. A 9V supply is entirely adequate to feed a typical guitar amp input more than the ~ 1.5V pk-pk that would cause the input tube to distort, 18V even more so. The first bipolar effects used 3V or even 1.5V batteries.

But I'm wandering around again. My take on buffers versus buffers with gain is that you need to know the source impedance driving the buffer stage, the load impedance after it, the signal voltage level of the input signal, and the signal level tolerance of the stage after it, and pick whether you need gain or not. That is - it depends. Sorry. 

[Mark Hammer]

Q1 : Why do most input buffers have unity gain ?

For example:



Why not a slight gain like maybe 2 or 3, to help a bit with the signal/noise ratio ?

Given that an Opamp can satisfy all necessary criteria like input impedance more than 500K, output impedance less than 1K, frequency response 1Hz to 30kHz, and still deliver gain with those constraints ?

Q2: Can most Opamps of today deliver lower noise than BJT and FET for input buffer designs ? Is it still wise to design discrete input buffers of unity gain or so much easier and better just to use almost any modern Opamp ?

If a musician had one single pedal, feeding an amp with lots of headroom.  You might be right.  But let us say that we had 4 pedals, whose buffer - still in the signal path, even in "bypass" - had a gain of 2x.  Our "bypass" signal from the 4th pedal would be 16x greater than the input from the guitar.  I used to have a simple preamp in a guitar with a gain of around 4x.  It would overdrive the amp easily (a good thing), but I could NEVER get a pleasing sound from a Tube Screamer.

Buffers are unity gain because the manufacturer/designer has absolutely no idea what you're feeding the pedal, what that pedal is feeding, and what the amplifier is eventually getting.  Unity gain is the most neutral kind of buffer.

[Steben]

Quick sidenote. My dad is an electronic engineer. I went for engineering as well, but with architecture and structure as specialisation. Lets say the apple fell close, but just on another property.
My dad enjoyed music as well. I enjoyed making some noise, he enjoyed making his own hifi set.

Now, what I am aiming at. At one point in growing up I started discussing guitar and electronics.
One of the first things he said - I will never forget that - was: "Guitar? That is a HIGH voltage (amplitude) source!"

Bottom line: turntable (ceramic) outputs in the early days have ten to hundred times as low a voltage. Just imagine the singal to noise questions we can think of as in guitar circuits and compare them with that.
 
He is still is a true solid state fan by the way.