Best discrete output buffer for a pedal

[JFace]

Suppose you have a high impedance tone control on the output of your pedal design. You need to follow this with a buffer with a moderate to high input impedance, and the ability to drive the input to the next pedal/amplifier in the chain. What discrete semiconductor would give the best performance in this role? The usual suspects being BJT, JFET, MOSFET, Darlington. I suspect the BJT would usually be best, but because of the high input impedance requirement, the supply current would need to be very small, potentially limiting the drive ability. Thoughts?

[midwayfair]

What's "moderate to high"? A BJT is fine for 1M impedance and will have low output impedance and is cheap. If you're following anything that needs more than 1M input impedance, you may want to rethink the previous stage, or use an amplification stage if your source impedance is THAT lossy.

[JFace]

What's "moderate to high"? A BJT is fine for 1M impedance and will have low output impedance and is cheap. If you're following anything that needs more than 1M input impedance, you may want to rethink the previous stage, or use an amplification stage if your source impedance is THAT lossy.

The output impedance of the tone control is 50k, so an input of >500k would be ideal. To get a 1M input impedance on a BJT, one would need a large emitter resistor, like 100K, to limit the current. Would this not limit the drive capability of the BJT?

[amptramp]

The schematic on page 2 shows how to use four transistors (two of each gender) and four resistors to get a buffer that has 400K input impedance and an output impedance of 10 ohms or less:

https://www.jameco.com/Jameco/Products/ProdDS/23051NSC.pdf

[midwayfair]

The output impedance of the tone control is 50k, so an input of >500k would be ideal. To get a 1M input impedance on a BJT, one would need a large emitter resistor, like 100K, to limit the current. Would this not limit the drive capability of the BJT?

What kind of tone control are we talking here? It's kind of hard to give better advice without seeing a schematic.

I can't speak for the current limiting resistor, but I suspect that 100K will not have an audible difference or lose you more than a couple decibels compared to a 10K or 5K ... I suspect even the most basic BJT buffer is adequate. Just use a 1M and tie it to a Vb and you should have an input impedance over 500K. If your output from the previous stage isn't over about 3.5V, you could probably use a 2N3819 FET instead (or just measure a bunch of other FETs to find one with a Vgs off of over 3.5V). Or use a MOSFET (though I think they sound kind of lousy as source followers).

Once you get to the point of investigating using more than one transistor for a single buffer, you're probably better off using an op amp (and I say this as a guy who loves discrete devices to death and only uses an op amp when forced).

[Gus]

You can bootstrap the input of a BJT.  Look for some of my posts

[bool]

just use a darlington ... cheap, works, simple, plenty of drive for such a low part count ..

[R.G.]

From almost two years ago:
http://geofex.com/FX_images/Onboard_Preamp.pdf

A preamp with high input impedance, current source loading for high drive out, and low quiescent current drain. Easily rebiased for 9V.

Frankly, an opamp would be easier (and why aren't you interested in an opamp for this??) but would draw more power supply current.

[blackieNYC]

4 transistors?  I'm missing something, but ready to learn - Why aren't we talking about something like a big muff pi output stage?

[PRR]

> and the ability to drive the input to the next pedal/amplifier in the chain

Which is WHAT? Key design number here.

Generally in guitar-chain work, we may assume the "next" is 50K or higher.

> The output impedance of the tone control is 50k, so an input of >500k would be ideal.

Agree.

And unity voltage gain.

To get from 500K to 50K impedance, we need a Current Gain of 10.

A few-cent BJT offers current gain of 50 to 500; but this will be reduced by bias components. Two such devices suggest Current Gain *squared*, 10,000 or more, for very little added cost.

JFETs and MOSFETs (and vacuum tubes) appear to offer "infinite" current gain.... the low frequency input impedance is hundreds of Megs or more. However there is capacitance, and at the top of the audio band this will be lower-Z than the device's input Z at DC. You can figure a tube gain stage input is coming below 200K by 20KHz. Here we do not need voltage gain, nor true 20KHz response, so a *small* device will be fine (a TO-220 power MOSFET may suck).

And "chip opamps" are typically 10 to 30 BJTs or maybe a few JFETs, stacked for current gain in the millions.

So first-crack, ANY of these can probably serve the purpose.

> Best discrete output buffer for a pedal

"Best"? What is the best guitar? What is the best guitar cord? What is the best lover? In many situations, many alternatives are "perfectly good, blameless". And implementation often trumps "an ideal device".

> To get a 1M input impedance on a BJT, one would need a large emitter resistor, like 100K, to limit the current. Would this not limit the drive capability of the BJT?

Work it out.

Assume transistors with hFE=500 are readily available.

First crack: for 500K input, we can use an emitter resistor of 500K/hFE or 1K. This will clearly drive any guitar-chain load. But we have not allowed for any bias scheme.

Since 1K is "way low" for driving 50K loads, let's go up to 10K emitter resistor.

Naked, the input Z is now hFE*10K or 5,000K or 5 Megs.

Loaded with 50K, the "10K" now looks like 8.3K, so 4.17Meg.

The simplest way to bias this is with a collector-base resistor hFE bigger than the emitter-ground resistor. This is "flawed" because hFE is not known exactly. However the bias-shift against hFE is small, and this scheme is widely used because it does work.

This collector-base resistor calculates as hFE*10K or 5,000K or 5Megs.

The input Z, with bias and load, is 5Meg||4.17Meg or 2.27Meg, well in excess of requirement.

The output can drive a 50K load to about 50K/60K or 83% of the no-load condition. Assuming 9V supply and 0.6V lost in the transistor, 3.5V peak. This is very good for "9V" systems.

> The output impedance of the tone control is 50k

This will reflect-through the transistor as 50K/hFE or about 100 ohms. To this we must add the emitter impedance of the transistor. At 4V bias and 10K emitter resistor we have 0.4mA of current. 26Ohms/0.4mA is 65 Ohms. Total output Z is 100+65 or under 200 Ohms. This is plenty low to absorb any crap trying to corrupt the signal on the output cable (assuming stage-length; stadium-long runs need more oomph).

2N5089 is 30 cents at Small Bear; cheaper in quantity or from less reputable suppliers everywhere. hFE at 0.4mA is minimum 400-420, max about 3X higher, so very-near "500".

[PRR]

> 4 transistors?  I'm missing something

Yes-- a real need for a LH0002.

It gives super-good performance to a thousand times higher frequency than we can hear. It gives an output Z that we don't need. It will drive 50 Ohms, a thousand times smaller than we typically need.

It may not be amenable to DIY construction. The bias scheme assumes you select device sizes in a ratio. If you use same-type devices the output stage will be low on current, and may get into crossover distortion. (It does anyway-- it is meant to be used inside an opamp's loop to reduce crossover crap.)

[garcho]

Frankly, an opamp would be easier (and why aren't you interested in an opamp for this??)

any reason?

[JFace]

The big disconnect for me was calculating the input impedance of a BJT. Thanks to PRR I now know a standard BJT buffer with 10K emitter resistor is more than adequate. What threw me off was the datasheets giving an input resistance in the range of about 10k (albeit with a operating current of 2.0 mA at higher voltages), so it made me question the effectiveness of a BJT following a tone control.

Yes, an opamp is ideal for the situation. Often times a smaller, cheaper component can provide the "just as good" performance given the requirements and make life easier regarding PCB layout, and costs if they are multiplied out in large quantities.

[PRR]

> the datasheets giving an input resistance in the range of about 10k

Show me. (It is not often given on modern datasheets.)

Datasheets usually cite the Common Emitter connection. Here we want unity gain so the Common Collector (Emitter Follower) connection is of interest.

> a operating current of 2.0 mA

At 2mA the emitter impedance is 14 Ohms. (26 Ohms at 1mA, declining as current rises-- true for all happy BJTs near room temp.) Yes, if the hFE were 770, the input Z would be 10K. But if we stuck in a 10K collector load, the voltage gain would be well over 500, which is insane for a buffer. Also 2mA in a 10K load is 20V drop right there, plus some for the transistor, it aint gonna work with a 9V supply.

Taking the load in the emitter rather than the collector ruins voltage gain, increases input Z, decreases output Z.

Ah-- Fairchild's 2N2222 sheet gives "Real Part of Common-Emitter High Frequency Input Impedance" as 60 Ohms; but this is at 20mA and _300_MHz! Using low-frequency approximation, 20mA gives 1.3 Ohms in the emitter, time hFE which is at least 80 at low frequency but falls-off to unity at 300MHz (fT is 300MHz at 20mA). So the LF approximation says 1.3 Ohms. However there is about an Ohm of parasitic resistance in the emitter, perhaps 10 Ohms in the base and perhaps more, then there are inductances in the leads which we can ignore at low freqs but are very significant at 300MHz. So this data-point gives us much information that we just do not care about.

2N140 data gives 700 Ohms at 0.6mA at 1MHz as a converter. rE should be 44 Ohms, hFE is 48, so we expect 2K, except the hFE may be far down at 1MHz. Ah-- GBW is 16MHz, so at 1MHz the hFE may be 16, and 44*16 is 704 Ohms, spot-on. (This is a VERY old part with spec we would not expect today.)

[Transmogrifox]

+1 PRR suggestion for 2N5089.  Simple, tried & true.

[Gus]

Best makes no sense

What do you want to spend to build it?
How many parts do you want?
Buffers should not be heard EXCEPT for the interaction with the guitar and cable to the input.

Some things I posted in the past


And an over designed buffer

[JFace]

Why are we getting hung up on the word "best"? I think it is a dirty word in the guitar community due to kids asking what is the best guitar/effect/amp... I was not asking a subjective question; I laid out my technical requirements. In other words, I was asking which discrete buffer best fit those parameters.

[bool]

Reply #6