An onboard buffer for guitars

[R.G.]

An onboard buffer for guitars
http://www.geofex.com/FX_images/Onboard_Preamp.pdf

[Cliff Schecht]

So this has gain up in the hundreds of MHz.. I guess that's just a caveat of using a really low biasing current, with modern discrete devices the gain will definitely be way up there. Why not bandwidth limit this to prevent potential RF interference issues? Did you play around with bootstrapping some capacitance across the 2N5087 C-E junction in your Sziklai pair or something like this? I guess I would need to run some sims before I guess too much but I tend to bandwidth limit things like this at 40kHz-100kHz (depending on slew rate needs, which should be insignificant here even for super hard string slaps) to prevent any potential RF issues without killing needed gain in the audio spectrum.

[Mark Hammer]

Here you call it a buffer.  The document describes it as an onboard preamp.  I'm confused.  Does the circuit have gain or not?  And if it does, how would I adjust how much gain it has...especially if I don't want very much?

[earthtonesaudio]

@Mark:
It has unity voltage gain, big current gain.  So if you plugged into a high-z input it would not be much louder, but if you plugged into a low-z input it would be.

[R.G.]

So this has gain up in the hundreds of MHz.. I guess that's just a caveat of using a really low biasing current, with modern discrete devices the gain will definitely be way up there. Why not bandwidth limit this to prevent potential RF interference issues? Did you play around with bootstrapping some capacitance across the 2N5087 C-E junction in your Sziklai pair or something like this? I guess I would need to run some sims before I guess too much but I tend to bandwidth limit things like this at 40kHz-100kHz (depending on slew rate needs, which should be insignificant here even for super hard string slaps) to prevent any potential RF issues without killing needed gain in the audio spectrum.

I commonly do that too. I have spent some time with a simulator trying to cleanly kill the response between 20kHz and 100kHz, which is what I usually try for if there are not specific requirements. Did I mention that this circuit is quirky?  

I messed with limiter caps on B-E, C-B and C-E on both transistors. In many of these cases, the bandwidth "reduction" resulted in a nasty peak in the AM band. I also messed with the usual base stoppers and other tricks. The most promising of these was a C-E cap on the PNP. Even that can be touchy with some values of load capacitance.

Frankly, the cleanest response - no funny RF peaks and such - was just to let it run out as far as the junction parasitics would let it. I don't trust that, of course.  

So, I don't yet fully understand where the sensitivity is coming from, and I noted as much. I feel pretty sure it's not going to sing with variable cable loading with the series resistor on the output. Without that, it was prone to singing with some values of cable capacitance, like many followers do.

No, it wasn't because I was ignorant of the issues.  

Here you call it a buffer.  The document describes it as an onboard preamp.  I'm confused.  Does the circuit have gain or not?  And if it does, how would I adjust how much gain it has...especially if I don't want very much?

Sorry. I'll go clear that up. It's a buffer. Gain is about 0.999, thanks to the high gain of the complementary feedback pair and current source loading of the pair's "emitter".

It could be made to have gain, but that's not what it started as. This started life as the solution to a question on another forum about how to get a particular tone from a pickup that only existed with the volume control full up, but was muddied by turning the volume down. I speculated that the volume control's varying loading was interacting with the pickup to steal some treble. Hence a buffer, to stop the loading, and the "magic load" to hold the pickup at the right place in the multidimensional tone curves.

[artifus]

@Mark:
It has unity voltage gain, big current gain.  So if you plugged into a high-z input it would not be much louder, but if you plugged into a low-z input it would be.

so could a low/high imp switch be simply added with appropriate r/c choices? pot?

[R.G.]

I believe what he meant was that an amplifier would respond differently to the buffer on the high impedance versus low impedance inputs.

[artifus]

ah, yes - started to consider that possibility exactly 33 milliseconds after clicking post... ho hum... 

[R.G.]

Through  the magic of revisionist history, it's now a ... buffer!

[Gus]

R.G. have you tried a series input resistor to help?  This should have the added benefit of reducing the HF end of the bandwidth.
Followers can look simple but they can be tricky

So it looks like the Sziklai pair has a active current circuit in the "emitter" leg  that gets some extra drive from the "collector' from the 100 ohm collector node to drive the bottom right 5088 base.  

I am guessing you did that to not use a constant current circuit that would draw more current, where this one has the lower constant current modulated by the signal off the "collector"

[R.G.]

R.G. have you tried a series input resistor to help?  This should have the added benefit of reducing the HF end of the bandwidth.
Followers can look simple but they can be tricky

I did, expecting it to be simple. It moved the RF peak a bit. I also tried a stopper in the PNP's base. No help.

So it looks like the Sziklai pair has a active current circuit in the "emitter" leg  that gets some extra drive from the "collector' from the 100 ohm collector node to drive the bottom right 5088 base.  

I am guessing you did that to not use a constant current circuit that would draw more current, where this one has the lower constant current modulated by the signal off the "collector"

Yes. The signal from the collector of the CFP corrects the CCS in certain instances and prevents gain droop.

[Cliff Schecht]

Ok, the fact that this is a buffer and not an amplifier with voltage gain (didn't notice that initially!) makes me much less worried about the bandwidth being so high.

Sort of.

The collector is most sensitive to capacitance (with regards to stability in a C-E amplifier), the source of a C-C amplifier is most sensitive to stray inductance (of which there is PLENTY in guitar cables). I am not sure about the sensitivity of the Sziklai pair to stray inductance but my gut tells me that this will hurt stability as well. How about the addition of a single bandwidth-limited stage with a known specific loading that keeps the Sziklai stage happy (i.e. optimal loading) while driving the cable/volume pot with a C-C/C-D follower?

[PRR]

Why can't I figure this out?



Seems to me the top PNP is off, doing nothing.

If it did work: yes that connection does ring at a dozen MHz. This causes little problem in small-signal work. It can get real nasty in a class B power stage; yet many are built and few have serious problem. Yes, compensation is confounding, anti-intuitive, and often fruitless.

> pretty sure it's not going to sing with variable cable loading with the series resistor on the output.

I don't see series resistor on the output? (A few hundred ohms would be ample.)

[Thecomedian]

Why can't I figure this out?



Seems to me the top PNP is off, doing nothing.

If it did work: yes that connection does ring at a dozen MHz. This causes little problem in small-signal work. It can get real nasty in a class B power stage; yet many are built and few have serious problem. Yes, compensation is confounding, anti-intuitive, and often fruitless.

> pretty sure it's not going to sing with variable cable loading with the series resistor on the output.

I don't see series resistor on the output? (A few hundred ohms would be ample.)

input signal is going to base of 2N5088, which means voltage/current will be bigger for base of 2N5087. Right? How much more negative is the collector of 2N5087 (because the PNP is upside down) compared to the base? That's all that matters for turning it on, right?

[R.G.]

Why can't I figure this out?
Seems to me the top PNP is off, doing nothing.

Oh. Doh.
Because I can't type. I inserted a decimal point where there should be none. That should be 33K. Schemo updated and on line.

If it did work: yes that connection does ring at a dozen MHz. This causes little problem in small-signal work. It can get real nasty in a class B power stage; yet many are built and few have serious problem. Yes, compensation is confounding, anti-intuitive, and often fruitless.

I had a forgotten component in my sim, and left off part of the modelling of a pickup when testing input impedance.

I neglected to put in the self-capacitance of the pickup, and the inductance was angry enough to resonate with the input capacitance at the 5088. Inserting the self capacitance cured that. There's no obvious peaking on a Bode plot, and only a cycle or so of ringing with 10k square waves in.

I don't see series resistor on the output? (A few hundred ohms would be ample.)

Corrected. 47R seems to fix any possible cable interaction. Once again, I didn't clean up my sim schemo.

It's much better now. Thanks for pointing those out, Paul.

[EATyourGuitar]

hey RG. once again you make a great post that is a total game changer. I was trying to design a low power preamp for a customer (not for production) using a low power 3v CMOS opamp. never got it off the ground cause it looked like battery sag would put me right out of my minimum power needs. this looks great. can you tell me a bit more about what watch batteries you selected?

on a side note, there are some white papers for designing 1.5v CMOS opamps but I'm not paying money to get access to those websites. I also don't have a IC factory in my back pocket.

[R.G.]

I tried opamps first, but this is a situation where the right compromise opamp is not easily available.

The fundamental problem with onboard electronics is powering them. Sticking in a 9V battery is possible, but often requires carving a cavity in the body for it. I wanted a different solution.

I chose two 3V lithium coil cells in series for 6V, primarily to be able to swing at least 4V for the output of a humbucker. Opamps that swing rail to rail would do the job. But they also need to have low noise, low operating current, wide frequency response, and low distortion. Those requirements seem to be contradictory in this particular application.

The biggest issue was low current versus low noise. The low noise opamps were not low current. I set a goal of 100uA operating current, and found no opamps with low noise there. I could get to low noise at about 1ma per amplifier, but that gave ten times worse battery life.

An emitter follower could do the job, if the input impedance got high enough and the output swing was wide enough. I started with a darlington, didn't like the voltage available, then went for the CFP with current source loading.

Again, the central issue is power. Coin cells will give around 200-300 hours of on-time at 100uA if I calculated/guessed right. That's not good enough not to use a power switch on it. But much lower than 100uA means going to Class AB, with crossover distortion, or Class A with doubtful ability to drive cables well.

I had an idea about how to put batteries in. Imagine a coin slot in the pickguard of a guitar.

[Labaris]

... This started life as the solution to a question on another forum about how to get a particular tone from a pickup that only existed with the volume control full up, but was muddied by turning the volume down. I speculated that the volume control's varying loading was interacting with the pickup to steal some treble. Hence a buffer, to stop the loading, and the "magic load" to hold the pickup at the right place in the multidimensional tone curves.

That's seem to me like a great concept to work on. Thanks for sharing.
What criteria do you use for picking the 1M resistor at the output? I always choose 10k and a 1u capacitor before that, because I assume the input of the next circuit will be 10k or higher, hence the high-pass filter would be formed mainly by the 10k and 1u. Anyway, I only see it as a filter, maybe I should see it like something else too.

[bool]

I use a fairly similar (cfp based) buffer since quite some time (since early 00's) and it works better than expected. I have it at a higher current and at 9v, but it's not that far off (sans the "white" thing on a ccs).

[R.G.]

What criteria do you use for picking the 1M resistor at the output? I always choose 10k and a 1u capacitor before that, because I assume the input of the next circuit will be 10k or higher, hence the high-pass filter would be formed mainly by the 10k and 1u. Anyway, I only see it as a filter, maybe I should see it like something else too.

It's there just to ensure that the output is firmly pulled to 0Vdc. The actual value is immaterial. I tried the circuit (in a simulator, mind!) with loads from 4.7K to 1M, with generally good results. If the load gets too low, the CCS doesn't have enough current flow to pull it down fast enough on negative going signals. But the typical application would be into another pedal with modern designs at 1M or more, and even vintage designs at 100K or more. None of those are a problem.

The only problematic pedal I can think of is the fuzz face. Expecting a raw pickup as it does, a buffered guitar of any kind will cause harder clipping distortion. But reinserting a pickup impedance in the form of a resistor, an inductor, and a cap will allow one to reset the impedance and get the softer clipping back.

The real choice on the output is the 1uF cap. That was chosen to be able to drive full audio bandwidth (16Hz in this case) into a 10K load resistor. With a 1M amp or pedal load after the guitar, that could be cut back to 10nF for the same 16Hz. On the input, the DC-blocking cap only needs to be 4.7nF to get full (and way over) guitar and bass bandwidth because of the high input impedance.

Notice that there are no caps over 1uF. The actual target of this design is SMD, where normal electrolytics are too big. A quick and dirty SMD with 0805 parts and SOT-23 transistors will fit entirely on a USA quarter, that being about 24mm diameter, using two electro caps. I'm actually looking at making this in 0402 SMD parts with smaller SMD transistors. I expect that will be the size of a thumbnail.