What's a buffer, how do they work and why would we want one?

Chuck D. Bones

Circuit Wizard
What's a buffer?
In the simplest terms, a buffer is a circuit that amplifies the current, but leaves the voltage unchanged. The perfect buffer would have a voltage gain of 1.000, flat freq response, infinite input impedance and zero output impedance. Such a thing doesn't exist, but we can get close enough.

Why would we want one?
A buffer is desirable when we want to prevent one circuit from loading down another one. A simple example is a long cable. Cables have capacitance and the longer the cable, the more capacitance. A shielded cable can have a capacitance around 25pF per foot. A 10ft cable would have 250pF capacitance, which is not a big deal. A 100ft cable would be 2.5nF and that will affect the brightness of a single coil pickup. A buffer at or near the guitar will prevent that long cable from affecting the tone.

Another example: The Parenthesis has two buffers, one at the input and one at the output of the Rat circuit. Here's the input buffer. It gets its input from the guitar and drives the octave up and octave blend circuits. The octave up & blend pot would load the guitar down and the guitar's high output impedance would interfere with the operation of the octave blend control. This is a very simple buffer, only three parts. There is an identical buffer after the Rat filter which prevents the Booster or the next pedal in the chain from loading the filter circuit.
1673029896290.png
We can also build a buffer with a regular BJT. Here's the input buffer from the LGSM. Pretty much the same as the JFET buffer, above. The only difference is we have to bias the transistor from Vref instead of GND. Some JFET buffers use Vref to bias the gate.

1673063225497.png

We can also make a buffer with a MOSFET. This one is from the Pussy Melter. Very similar to the BJT buffer above.
1673063410050.png
Finally, we can use an opamp to make a buffer. This is from the Diezel VH4 (Valhalla). Note that the opamp is biased from GND because the opamp is powered by ±9V.

1673063597027.png

Which one is better? How do we decide which buffer circuit to use? In most cases, it comes down to the designer's personal preference. Within limits, they all perform the same. The opamp has the lowest output impedance and is capable of driving the most current. Whether you need that capability depends on what comes after it. If you have a spare opamp on the board, then it's a no-brainer.

Two of the circuit examples above have buffers and don't even need them. The opamp stage following the buffer in the LGSM and Pussy Melter have a high input impedance, so putting a buffer there doesn't help. It doesn't hurt, but it's not needed. In the LGSM, the buffer is a leftover from the Tube Screamer's soft switching. It's anyone's guess why Steel Panther put a MOSFET buffer in front of a MOSFET opamp.

How do they work?
Let's look at the JFET buffer above. It's wired as a source follower. When we raise the the voltage on the gate, the drain current increases, causing the source voltage to increase. The source voltage follows the gate voltage, hence the name "source follower." The nominal voltage gain is not exactly 1.00, but it's close enough. The input impedance of a JFET is over a gigaohm. R3 sets the input impedance of the circuit at 1M. The output impedance is 1 / gm where gm is the transconductance. Gm varies from one JFET to another and also depends on drain current. I measured a bunch of JFETs and their gm varied from about 0.5mS to 2mS at the kind of drain currents we run in pedals. That gives us an output impedance between 500Ω and 2K. We can go lower by dialing up the drain current, but this is low enough for most pedal applications.

The BJT & MOSFET buffers work exactly the same way. The output impedance is 1 / gm, just like the JFET. The gm of a BJT is given by gm = Ic / 26mV. If we run the BJT at 0.5mA, then gm = 19mS and the output impedance is 52Ω. Much lower than the JFET. A BJT's input impedance is also lower than a JFET, but one of the advantages of an emitter follower is the transistor's input impedance is multiplied by the HFE. So we end up with an input impedance in the 2M range if we use a BJT with HFE = 200. The biasing resistor, R3, is in parallel with that 2M, so we end up with around 400K input impedance. That's high enough to have little or no affect on a guitar's tone.

MOSFETs have an extremely high input impedance, therefore the buffer's input impedance is determined by the biasing resistor (R2 in the Pussy Melter). A BS170 has a transconductance around 30mS at the 2mA drain current in the example circuit above. That translates to 33Ω. At the same collector current, the BJT buffer above would have an output impedance around 13Ω.

Which one is best?
Opamp voltage followers use feedback to produce an input impedance over a gigaohm and an output impedance of a few milliohms. Clearly, they are superior on paper, but in many cases it would be difficult to hear the difference between the various buffer circuits described above.

Bottom line, any of them will work. The JFET & opamp buffers offer the lowest noise (provided you use a low-noise opamp). The opamp buffer is a freebie if you have a spare opamp on the board, otherwise the BJT buffer is the least expensive. MOSFETs are noisier than the rest, so I would not use a MOSFET as an input buffer in a high-gain pedal. As an output buffer, it's fine.
 
maybe there is a real audible difference?

with the circuits i’ve built, i always found i prefer the OD808 (opamp buffers) over TS808 (BJT buffers), so i kinda just assumed that opamp buffers might be superior to BJT
 
Some good pedal circuits are biased with a voltage divider, such as the Fuchsia's 2nd & 3rd stages and the BMP's last stage.
The advantage of biasing with a filtered Vref is that power supply noise doesn't get coupled into the input like it would with a simple voltage divider off of Vcc. If more than one stage needs to be biased to 1/2 Vcc, then Vref can be shared as long as everything connected to Vref does not disturb Vref by pumping too much current into it. I've seen some marginal pedal designs where one or two of the things connected to Vref will couple noise or DC offset into Vref.
 
Some good pedal circuits are biased with a voltage divider, such as the Fuchsia's 2nd & 3rd stages and the BMP's last stage.
The advantage of biasing with a filtered Vref is that power supply noise doesn't get coupled into the input like it would with a simple voltage divider off of Vcc. If more than one stage needs to be biased to 1/2 Vcc, then Vref can be shared as long as everything connected to Vref does not disturb Vref by pumping too much current into it. I've seen some marginal pedal designs where one or two of the things connected to Vref will couple noise or DC offset into Vref.
How is Vref usually implemented? And if Vref is a reference voltage why is there a resistor between it and the base?
 
How is Vref usually implemented?
Look at any of the PedalPCB boards that use an opamp for examples.

And if Vref is a reference voltage why is there a resistor between it and the base?

Vref has a bypass cap and is therefore at AC ground. If you connect Vref directly to the base of a transistor, then that transistor's base is also at AC ground and the signal is shorted to GND.
 
Look at any of the PedalPCB boards that use an opamp for examples.



Vref has a bypass cap and is therefore at AC ground. If you connect Vref directly to the base of a transistor, then that transistor's base is also at AC ground and the signal is shorted to GND.
Ah that makes sense. I read about that for opamp biasing recently but didn’t put the two things together. Thank you Chuck!
 
Some pedals do not like to be driven by a low impedance source. Try driving a FuzzFace with a DS-1 in bypass mode and you'll hear what I mean. The problem isn't with the FuzzFace or with the DS-1's buffered bypass, the problem is with the combination of the two. Switch the order and everything's fine.

Rather than blame oneself for misapplying a piece of equipment, a fool will blame the equipment.
 
Some pedals do not like to be driven by a low impedance source. Try driving a FuzzFace with a DS-1 in bypass mode and you'll hear what I mean. The problem isn't with the FuzzFace or with the DS-1's buffered bypass, the problem is with the combination of the two. Switch the order and everything's fine.

Rather than blame oneself for misapplying a piece of equipment, a fool will blame the equipment.
Of course. A fuzz face with a buffer in front of it can be super fun, if you like that kind of stuff. Not the most versatile tone, but I’ve found it to be desirable in a handful of occasions
 
I always thought of certain fuzzes as being “broken” by design. They are designed to load the guitar pickup, but that means that they don’t fit with low impedance signals. But that’s what makes them magical.
 
Certain circuits out the have buffers in between stages rather than or in addition to input and output. Is there a general rule of thumb for circuit designing when and if a buffer should be used?
 
Certain circuits out the have buffers in between stages rather than or in addition to input and output. Is there a general rule of thumb for circuit designing when and if a buffer should be used?
usually they're inserted where there is a large impedance gradient
(high impedance circuit block/segment --> low impedance circuit block/segment) or vice versa

which is also why buffers are used on inputs + outputs
 
Last edited:
Good answers.
There is one more: Some circuit designers throw them in willy-nilly where they are not needed.

One example is the DS-1. The original MIJ DS-1 used a single opamp (TA7136). When that opamp went out of production, Boss switched over to the M5223AL dual opamp. What to do with that spare opamp? Boss put a buffer in between the 1st & 2nd stage. Wasn't needed in the old circuit and still isn't needed.

The VFE Ice Scream (Frost Drive) is another example. IC1.1 is not needed because IC1.2 has a high input impedance, but there was a spare opamp. This was actually wasteful because that spare opamp could have been used in place of Q1.
 
For transistor stages, consider that there are only three configurations possible for each transistor type (BJT, FET, MOSFET) and the formulas to compute input and output impedance are well known. Similarly for the two main op amp configurations. Next you should consider passive stages that may affect impedance, typically loading either the previous or next stage or both. Let me think of an example…
 
Example: a BJT in common emitter with degeneration has a pretty high input impedance, but the output impedance is Rc. That may be low enough for many applications but if Rc is high for whatever reason, it may be too high compared to the input impedance of the next stage and thus require a buffer in between.
 
Is there a way to eyeball a schematic to determine if one stage will load another?
You have to do some analysis. In many cases it's not obvious how much one stage loads another and more importantly, the effect of that loading. Sometimes the loading is critical to getting the correct sound.

Here's an example of a fairly simple circuit where the impedances and loading effects are not obvious. This is a slightly modified version of the Knight-Kit Fuzz from the '60s. What's the output impedance of the 1st stage? Less than 100K, maybe a lot less depending on the FUZZ setting and the output impedance of the device (guitar or another pedal) driving the Knight Fuzz. What's the input impedance of the 2nd stage? Let's assume Q2 is an MP38A, its HFE is 60 and it's biased so the collector is around 4V. Q2's input impedance is 4.6K with zero signal, but because it's non-linear, the impedance will vary as the signal varies. Q2 is being current-fed by Q1. Putting a buffer between Q1 & Q2 would completely change the tone. The impedance of the D1-D2 clipping diodes is much lower than the output impedance of Q2. Should we put a buffer between Q2 and the clipping diodes? No, because the clipping diodes are supposed to load Q2.

Knight & Day Fuzz.png
 
Back
Top