NSFW What's up with the buffers in cornish designs?

No sweat man. It was fresh in my mind because I was working on it today and this seemed the most logical place to post it.

What I actually found interesting were a few things most of us would see as “kludges” or “fudging it”.

R1 and R8 come to mind. Obviously Cornish was targeting a specific input impedance he couldn't get with the bootstrapped emitter follower alone.

It seems that emitter voltage around 50% of source voltage was important to him, so R1 acting as an artificial reducer of input impedance made more sense than trying to get blood from a stone.

Furthermore, it seems that 100 ohm output impedance was a design criteria so he simple added R8 in series to get there.

If anything, I appreciate him a bit more because of these little practical things.
What do you mean by source voltage?
 
What do you mean by source voltage?

For the purposes of calculations, I assume a raw source voltage, in this case 9v.

In practical terms it would be the whatever the filtered source voltage is. Vf is usually what I call it in my schematics.

After the diode and filter caps.
 
For the purposes of calculations, I assume a raw source voltage, in this case 9v.

In practical terms it would be the whatever the filtered source voltage is. Vf is usually what I call it in my schematics.

After the diode and filter caps.
Ah ok. Source is a bit of an overloaded term given that it indicates one of the pins of a FET. I think it’s relatively common to want the emitter voltage at half Vcc because that maximizes the dynamic range (from 0 to Vcc) of the output signal before clipping.
 
Last edited:
Ah ok. Source is a bit of an overloaded term given that it indicates one of the pins of a FET.

Its a semantics things. I don't use JFETs for anything so have never run into that contradiction. I'll remember that going forward.

I think it’s relatively common to want the emitter voltage at half Vcc because that maximizes the dynamic range (from 0 to Vcc) of the output signal before clipping.

Which was one of the interesting insights I found trying to break the circuit down:

1721997411709.png

In Cornish's Design, we have the following:

R1 = 10M
Rb1 = 120k
Rb2 = 200k
Rbs = 120k
Re = 7.5k

In the absence of R1, Zin = ~2.99M. Adding in R1 reduces this to ~2.3M, but not close to the 1M he quotes.

It could be my calculations are boogered up but I verified them against a few known Simple/Full bootstrap derivations and they match those papers so I am pretty confident in the implementation. I did end up implementing the "Full" calculation because the "Simple" type seemed to be about 40k-50k higher in the 3 different derivations I tried:

1721997883383.png

I added in a dropdown that lets me add R1 in to show the effect:

1721997990607.png

The first thing that jumps out when I wanted to adapt this was some of the non-standard resistor values. Not a huge deal in the Through Hole world, but Basic Parts availability in off values at JLCPCB can present challenges. So I set out just tweaking a few values.

Using the Standard Cornish values. Note: I have R1 "turned off", analytically speaking:

1721998201850.png

My first stab at changing stuff:

1721998281578.png

Ve jumps up a bit. Assuming Cornish was targeting Vcc/2 for Ve, this presents a challenge. Ve factors into Zi pretty significantly so using that as a lever to change Ve is sort of a dead end, i.e. once you lower it enough to impact the emitter voltage, you have dropped the input impedance significantly.

I imagine the values he picked were a compromise between impedance and voltage.

I went to JLCPCB and searched a few parts to see what basic parts I'd have at my disposal. 150k and 220k were readily available in my preferred 0805 case:

1721998560545.png

At this point, I sized R1 to give me about 1M input impedance:

1721998624657.png

One thing I still don't understand, beyond the obvious effect it has on the effectiveness of the "bootstrapping", is if there are any frequency shaping impacts of Cbs1. Cbs2 is obviously providing some passive RF shunting, but I'm not sure if there are any impacts to increasing or decreasing the size of Cbc1.
 
Its a semantics things. I don't use JFETs for anything so have never run into that contradiction. I'll remember that going forward.



Which was one of the interesting insights I found trying to break the circuit down:

View attachment 79114

In Cornish's Design, we have the following:

R1 = 10M
Rb1 = 120k
Rb2 = 200k
Rbs = 120k
Re = 7.5k

In the absence of R1, Zin = ~2.99M. Adding in R1 reduces this to ~2.3M, but not close to the 1M he quotes.

It could be my calculations are boogered up but I verified them against a few known Simple/Full bootstrap derivations and they match those papers so I am pretty confident in the implementation. I did end up implementing the "Full" calculation because the "Simple" type seemed to be about 40k-50k higher in the 3 different derivations I tried:

View attachment 79115

I added in a dropdown that lets me add R1 in to show the effect:

View attachment 79116

The first thing that jumps out when I wanted to adapt this was some of the non-standard resistor values. Not a huge deal in the Through Hole world, but Basic Parts availability in off values at JLCPCB can present challenges. So I set out just tweaking a few values.

Using the Standard Cornish values. Note: I have R1 "turned off", analytically speaking:

View attachment 79117

My first stab at changing stuff:

View attachment 79118

Ve jumps up a bit. Assuming Cornish was targeting Vcc/2 for Ve, this presents a challenge. Ve factors into Zi pretty significantly so using that as a lever to change Ve is sort of a dead end, i.e. once you lower it enough to impact the emitter voltage, you have dropped the input impedance significantly.

I imagine the values he picked were a compromise between impedance and voltage.

I went to JLCPCB and searched a few parts to see what basic parts I'd have at my disposal. 150k and 220k were readily available in my preferred 0805 case:

View attachment 79119

At this point, I sized R1 to give me about 1M input impedance:

View attachment 79120

One thing I still don't understand, beyond the obvious effect it has on the effectiveness of the "bootstrapping", is if there are any frequency shaping impacts of Cbs1. Cbs2 is obviously providing some passive RF shunting, but I'm not sure if there are any impacts to increasing or decreasing the size of Cbc1.
I don’t think that in guitar pedal applications 5V makes a huge difference compared to 4.5V. I guess it depends what your application is but a guitar signal that’s 9V peek to peek is pretty high and will clip in the amp anyways. Why did you lower the input impedance to 1M? The higher the better. I don’t know the answer to your capacitor question. I would need to run a simulation to figure that out.
 
Why did you lower the input impedance to 1M? The higher the better.

Well, I guess that depends on your opinion. There is a threshold after which it doesn't matter how much higher you go.

500k is typically high enough. Ultimately it doesn't really matter to me.

It's worth noting that Cornish himself artificially reduces the input impedance with R1. Ultimately it doesn't really matter to me. I was mostly just playing with the parameters to try and match what he quotes on his website.

Again though, even with his R1 set to 10M, the input impedance is around 2.3M ohm, not the 1M he quotes on his site.
 
Well, I guess that depends on your opinion. There is a threshold after which it doesn't matter how much higher you go.

500k is typically high enough. Ultimately it doesn't really matter to me.

It's worth noting that Cornish himself artificially reduces the input impedance with R1. Ultimately it doesn't really matter to me. I was mostly just playing with the parameters to try and match what he quotes on his website.

Again though, even with his R1 set to 10M, the input impedance is around 2.3M ohm, not the 1M he quotes on his site.
Got it. Does he say 1M or over 1M? 😉
 
Back
Top