Best DIY buffer pedal

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Frankly, having read the manufacturer's descriptions of the two pedals a schematic ain't going to help you. You need a very high input impedance/low noise op-amp (discrete or IC) and some seriously good noise management techniques (they're claiming -100db noise floors). If you really want schematics, look at appropriate IC datasheets or Horowitz and Hall.
 
I think I'm gonna build this veroboard, I want to try it after my wah
22d2685d35205fec6e95c9818320c2d0.jpg
 
Member
Joined 2011
Paid Member
1N5401 is a VERY LARGE diode, rated for 3 amperes of continuous forward current. It's not going to fit in the 0.2" zone allocated for it on the stripboard layout in post #4. Check out the mechanical dimensions on the 1N5401's datasheet. The leads are larger diameter than the drill holes on a typical stripboard!
 
Member
Joined 2011
Paid Member
... You need ... and some seriously good noise management techniques (they're claiming -100db noise floors).
It's not that frightening.

The Cornish LD-1 claims Output Noise = -110dBm (10 kHz bandwidth)

Converting dBm to volts, Output Noise = 2.45 microvolts (10 kHz bandwidth)

Converting to noise spectral density over a 10 kHz bandwidth, Output Noise Density = 24.5 nV/rt.Hz

A low cost NE5532 IC, operating at unity gain, will achieve this easily. Since the input and output will be AC coupled, you can trim out any offset voltage that may arise from the 1Megohm input impedance and the opamp's input bias current, without the need for a DC servo.
 
I'd also found this schematic, but it's different
Yup. Not sure either are what Mr Cornish had done (although the BC549 is historically likely). The schematic won't get the Zin >1M that Mr Cornish insists on. I've not had a chance to reverse engineer the vero board version. I do suspect the original has been through a few variations over the decades


It's not that frightening.
Agreed, it's all down to layout, power supply and noise management. Which I usually see done badly. (And I'm sure Mr Cornish gets this right)

Normally I'd reach for a JFET opamp if I'm looking for high impedance. It's been a few years (ok, more) but I seem to recall having noise problems with the 5532 when the impedance values go up. What trick did you have in mind here?
 
I see you want to go for super-low noise, but along the journey, you could try a very simple jfet source follower as well, like this or similar, at least for comparison:

buffer1.png


Its very high input impedance (better than the bjt design above) and very transparent and clear sounding, with negligible noise though that can be improved further (eg run the bias from a voltage divider of two 100k resistors with the mid'point grounded by a cap, then use a 1M metal oxide resistor to bias the gate)
 
Last edited:
Normally I'd reach for a JFET opamp if I'm looking for high impedance. It's been a few years (ok, more) but I seem to recall having noise problems with the 5532 when the impedance values go up.
Yeah. JFET input is quieter at impedances this high. I would probably just throw a TL072 at it. An oldie, but well suited to something like this, and the noise spectral density is more than low enough for the job, too.

Incidentally, I have never come across an electric guitar with a 110 dB signal-to-noise ratio, or anything anywhere near that! Between hum, buzz, thermal noise from the large-value pots inside the guitar, and fairly limited inherent dynamic range (hardly likely to exceed 30 - 40 dB at most), do we really need an extraordinarily low-noise buffer when dealing with such a rough-n-ready signal source? :)

-Gnobuddy
 
I chose to build the Cornish buffer, because I saw some videos on youtube and I liked more how it sounds. I think many other buffers increase trebble too much, for my taste. And another thing I liked is how it sounds with distortion, it purrs more like a rolls royce (maybe it boosts signal a little).

But I know nothing about electronics and I dont know what makes a Buffer faithfull to the original guitar signal or brighter
 
I chose to build the Cornish buffer, because I saw some videos on youtube and I liked more how it sounds.
It's your project, so of course you should pick the one you like best! :)

The tricky part is that it is hard to know from a Youtube video exactly which parts of the audio chain are affecting the sound.

I built a JFET buffer years ago, and what it did was make the guitar sound the same whether I used a 2-metre guitar cable, or a 10-metre one.

That means, if you like the (brighter) sound of the guitar with the 2-metre cable, then you'll like what the buffer does - it will preserve that tone even when you use the longer 10-metre cable.

But if your ears find the guitar too bright with the 2 m cable, and just right with the 10 m cable, then adding the buffer will make it sound too bright!

So it's all about your personal taste in sound, as well as the particular guitar and guitar cables you like to use.

By the way, when I built my buffer years ago, I didn't have any distortion or overdrive pedals to try with the buffer, though, so I can't offer any useful suggestions about that.

These days I usually use a Danelectro Fish-n-Chips graphic EQ pedal as my buffer/clean boost, if I need one.

(maybe it boosts signal a little).
The schematic I found online (attached) is what is called an "emitter follower". This is a type of one-transistor circuit that passes a signal through unchanged - it doesn't make it bigger, it doesn't make it smaller. (Technically, it has a voltage gain that is just a hair less than 1.0 .)

I don't know if this is really the correct circuit for a Cornish Booster, but if it is, then it doesn't actually boost the signal at all. But it does "buffer" it, which means the guitar signal won't become weaker - loaded down - when you feed it into a pedal with a somewhat low input impedance, like a Big Muff, for instance.

In the same way, the guitar won't be loaded down by a long cable's capacitance, so you won't lose treble even with a fairly long cable.

Compared to the germanium transistors that came before, that BC549 transistor was a marvel when it was first introduced on the market - quiet, fast, reliable, very high current gain.

-Gnobuddy
 

Attachments

  • cornish_buffer.png
    cornish_buffer.png
    55.2 KB · Views: 628
Quickly tracing through the vero board it's also an emitter follower - but with a whole bunch of clever filtering going on (yes, I'm too lazy to work out the corners) and additional power filtering. Plus a rather clever bias scheme.

And is likely to providin the claimed 1M input impedance. ( beta x 30K ish)

So I'd say it' more likely to be derived from a Cornish design than that schematic
 
Since there is continued interest in this circuit, I took a few minutes to turn the Veroboard layout back into a schematic, and then run it through LTSpice to see what, if any, mystical magical musical mojo-matic properties the circuit might have. :D

The transistor I used in the simulation is from the same family, and has very similar characteristics, to the specified one.

...additional power filtering.
Yes, via a 100 ohm resistor (R8 in my attached LTSpice schematic), and 220 uF capacitor (C2).

A good idea, and also entirely typical. Every guitar pedal has something similar to make the power supply rails behave, and ensure low power supply impedance at high frequencies.

..with a whole bunch of clever filtering going on
Sadly, I don't see any clever filtering. There is a 1nF cap across the transistors base and emitter, which is a poor place to put it, because there are nearly equal signal voltages at both ends of the cap.

That means the Miller Effect, in this case, makes the capacitor behave much smaller than it really is. As a result, that capacitor is essentially useless at any frequency we care about.

I ran a quick frequency response simulation in LTSpice, and sure enough, the bandwidth goes out beyond 10 MHz. You could (and very likely, will) pick up shortwave radio with this thing!

To put it plainly: building an audio circuit with a 10 MHz bandwidth is either incompetent, or irresponsible. Especially when it's a guitar circuit, which will have a length of cable and magnetic pickups connected to the input. That stuff dangling off the input is likely to feed unwanted RF signals into it, so the proper thing to do is ensure that any guitar electronics rejects frequencies that are substantially higher than, say, 15 kHz.

As it turns out, there are other problems with the circuit. Bootstrapping via a feedback capacitor looks very clever, but has a nasty side-effect: it almost always creates a huge hump in the frequency response at very low frequencies. This is caused by the combination of positive feedback, and increasing phase shifts at low frequencies in the input and bootstrap capacitors.

In this case, LTSpice confirms that we have a 4 dB peak at about 3 Hz (!) Thump on the guitar body or on the low "E" string and this booster circuit might experience blocking distortion as a result of large subsonic signal swings at 3 Hz.

A guitar circuit has no business having any sort of response down to 3 Hz; all that does is open a window for excessive low-frequency (flicker) noise, low frequency instability, et cetera. A six-string guitar in standard tuning only goes down to 83 Hz, so there is no logical reason for a guitar buffer to go lower than, say, 50 Hz.

Lastly, there is at least one outright mistake in the circuit, which is unlikely to have come from Mr. Cornish, but from someone who failed to notice R5's function. So he/she added the utterly useless 50k resistor (R6). It does nothing that is not already done by R5.

At this point, having traced out the schematic, and simulated its workings, I can say with confidence that this particular booster circuit has absolutely nothing special or magical about it. It's just an entirely typical BJT buffer, with a few sloppy design mistakes thrown in. The mistakes are not show-stoppers, so the circuit will certainly work.

I'm attaching the LTSpice sim file (.asc) as well as the screenshot. If anyone wants to, feel free to run the simulation yourself, and/or tinker with the schematic to see if you can fix the poor frequency response characteristics.

-Gnobuddy
 

Attachments

  • Cornish_Booser_From_Vero_Layout_w_mistakes.asc
    2.9 KB · Views: 144
  • Cornish_03_Sim.png
    Cornish_03_Sim.png
    76.6 KB · Views: 425
Thanks for posting the analysis. I got interested and set it up myself, using 5Spice. I got virtually identical results though my transistor was nominally a BC848C.

But I think there is some cleverness going on here!, in the context of trying to get the best performance out of a single bjt buffer stage. I decided to figure out the input and output impedances.

My inputs were relative to a 1V signal, so x1 gain is 0db.

Output impedance,

I tested by negating that extra 50k resistor, and noting the output signal rose from -0.1300db to -0.0954 db at 1khz. No much difference, but work that back into a voltage change, and figure the output z as the top half of a voltage divider based on that 50k. and the output impedance was 120 to 130 ohms. ie, nice and low and a characteristic of this type of circuit, but no surprise yet.

input impedance

Then to the input impedance. If you change the 1k input resistor to 101k (like as if the guitar has a 100k output impedance), the output drops about -0.43 db, or about 1/1.05 on voltage (10^(0.43/20) = 1.05), This suggests that the input z is about 2M, quite good, and very good for a bjt stage.

The key to it is the 4.7uF cap. Take that out and two things happen.

There is a roll off below 50 hz, all fine

But with the extra 100k input z, the output signal drops from -0.168db to -3.971db, a drop of -3.8db. Work that back and the effective input z is only about 150k, not good at all!

So that 4.7uF feedback resistor, albeit with its quirky boost at sub bass frequencies, is actually significantly reducing load on the guitar in a useful way. Its clever, if you don't have a jfet instead.

(Note: the image below is without the 4.7uF cap and with extra 100k input. i see i omitted c5)
 

Attachments

  • cornish buffer.gif
    cornish buffer.gif
    36.2 KB · Views: 296
Last edited:
Sadly, I don't see any clever filtering.

No. I made a mistake when I transcribed it. :-(

So that 4.7uF feedback resistor, albeit with its quirky boost at sub bass frequencies, is actually significantly reducing load on the guitar in a useful way. Its clever, if you don't have a jfet instead.

Yup, straight from the textbook (H&H fig. 2.65 in the 2nd edition)

Thanks both for posting schematics - saved me from scanning my scribbles
 
There must be some cost though, increasing input z maybe increasing noise? i dont know. it still looks to me tbat its a solution to not having a jfet. The one i posted first with one jfet is a lot simpler and Ive built it many times. its linear, low noise and low current.
 
I got virtually identical results though my transistor was nominally a BC848C.
Nice, independent confirmation is always nice to have!

...the output impedance was 120 to 130 ohms.
LTSpice is a little bit in error here, on the low side. The output impedance is dominated by R7, which is 150 ohms.

There will be a small additional contribution from the transistors own output resistance; this is approximately given (in ohms) by:

re = 0.026/Ic

Where Ic is the collector current in mA, and re the small-signal output resistance at the emitter. With the transistor running at roughly 0.63 mA, re will be about 40 ohms.

So, my back-of-the-envelope calculation yields about 200 ohms for the output resistance of the circuit, at frequencies high enough to ignore the impedance of the 22 uF cap. Nice and low, as you said, which is what we expect from these wonderful little high-gain silicon epitaxial transistors!

So that 4.7uF feedback resistor, albeit with its quirky boost at sub bass frequencies, is actually significantly reducing load on the guitar in a useful way. Its clever, if you don't have a jfet instead.
Whoever invented bootstrapping was clever, no doubt about that! But in that era long before personal computers and LTSpice, it took considerable mathematical savvy - or a careful frequency response measurement over an extended frequency range - to realize that bootstrapping had some poor side-effects.

Using bootstrapping in this booster was pretty much a given - it's about the only way to get a BJT to achieve that level of input impedance, though a Darlington pair run at low collector current is an alternative without the same bad side-effects.

Keep in mind that the concept of bootstrapping was very much in the mind of any audio electronics designer of that era - bootstrapping was routinely used on the driver stage of just about every transformerless transistor power amplifier of the time. (Nowadays we're likely to see a constant current source in that location instead of a bootstrapped collector load resistor.)

So the use of bootstrapping was a reasonable design decision. The mistake was in the choice of the input and bootstrap capacitor values - 0.1 uF and 4.7 uF, respectively. The first of those is a bit too big - but the second one is way, way, way too big! You see, using a too-large bootstrap capacitor is the cause of that big subsonic bass peak.

To demonstrate that point, I tweaked the cap value down - a lot - and you can see the results in the attached screenshot. No more subsonic peak! Now the bootstrapping is well-behaved.

I also tweaked the input coupling capacitor to create a more appropriate bandwidth for guitar use. And I removed the useless 1 nF cap, and added in a smaller cap in a better location to roll off the frequency response above 10 - 15 kHz.

I added the 33k between the input voltage source and the buffer circuit to simulate the source impedance of the guitar - unfortunately, this is very variable, depending on where the guitar volume pot is set. And it does affect where the HF rolloff occurs.

-Gnobuddy
 

Attachments

  • Cornish_04_Sim.png
    Cornish_04_Sim.png
    72.8 KB · Views: 369
No. I made a mistake when I transcribed it. :-(
Easy to do, especially when we're pre-disposed to expect something magical and wonderful.

I can't speak for you, Thoglette, but with the huge amount of adulation this pedal has received over the years, some primitive part of my brain was anticipating seeing some clever design trick in the circuit.

I wanted the circuit to be clever and magical, even though my more rational left-brain kept saying "There is no magic in the pedal, the magic is in David Gilmour's astonishing musicality!"

Well, this time mathematics and the left-brain turned out to be right! :)

-Gnobuddy
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.