TL072 bias resistor/input impedance

[Mick Bailey]

With a typical 9v single-rail supply you may have a couple of bypassed 100K Ohm resistors connected to the supply and their node feeding the 4.5v input bias via a 1M Ohm resistor, which effectively sets the input impedance. What is the practical limit to this value? I've used 10M Ohm without problems, but does increasing this to 22M Ohm present any issues in an audio circuit?

[nickb]

The full range bias current is 7nA. For an effective 11Meg source that a 77mV offset that so you'd need the same impedance on the other input.

The full range offset current is 2nA i.e 11mV. Whether that is significant depends on the closed loop gain and whether the output goes to something that is is sensitive to DC offsets

At 25C the numbers are a lot loss and 0.2nA and 0.1nA so if your temperature is stable then 2.2mV and 1.1mV are probably not going to be a problem.


A concern is the introduction of poles by virtue of these large impedances and stray capacitance that could lead to poor HF response or instability.

PS: I wonder at the reason behind this. Is it just for power saving? If so the saving would be tiny compare to the TL071. Better to go for a low power amp amp.

[Mick Bailey]

I was thinking about where you have an input buffer where the bias network appears in parallel with the source impedance (in this case a piezo disk)

[nickb]

Do you have the specs of the disk? What frequency response requirement is specified? How much gain? What is the signal output voltage of the disk?

PS: What I'm trying to get to is what is the lowest acceptable resistance that meets the specs. Anything higher that that is just introducing noise.

[Enzo]

What is the practical limit to this value? I've used 10M Ohm without problems, but does increasing this to 22M Ohm present any issues in an audio circuit?

Just an idle thought: I know piezos like high impedance, but does 22M really offer any benefits over 10M? OK, potential benefits? At some point does't it become like extending the bottom end of a guitar amp from 10Hz to 5Hz?

[SoulFetish]

With a typical 9v single-rail supply you may have a couple of bypassed 100K Ohm resistors connected to the supply and their node feeding the 4.5v input bias via a 1M Ohm resistor, which effectively sets the input impedance. What is the practical limit to this value? I've used 10M Ohm without problems, but does increasing this to 22M Ohm present any issues in an audio circuit?

So, there's the noise increase as resistance does. What's the capacitance of the piezo disk you're using?
You can get really high input impedances, but there might be a point where it makes sense to bootstrap the input, or bootstrap a buffered input using a FET.

I was thinking about where you have an input buffer where the bias network appears in parallel with the source impedance (in this case a piezo disk)

It's interesting, I was just reading a new article by Rod Elliot on High Impedance Input Stages dealing with this subject here:
High Impedance Input Stages
For audio, you probably don't need to achieve this level for input impedance, but here is an example of a bootstrapped Jfet buffer as an example.

[Mick Bailey]

The disc is unknown and part of a circuit I experimented with a very long time ago. It measures 80pf and is bonded to a thick aluminium disk and encapsulated with epoxy. The entire thing is 18mm diameter, so the actual disk is something like 15mm - quite small. I had it originally blocking a strat trem - fitting between the sustain block and the guitar body. The original circuit was just an opamp buffer followed by a simple fixed EQ. The sound was something I wasn't expecting - it sounded like a metal bodied resonator and when I first plugged it in my Wife said "that sounds just like that Dire Straits song", so I immediately learned Romeo and Juliet. I was just having a tidy up and came across the transducer assembly and just got me thinking what the practical limitation is of a simple opamp buffer with its impedance determined by resistance. I understand bootstrapping - I used this in my Ovation and it works very well.

My question isn't that I need to build a high-Z preamp, but what the practical limitation is for a unity gain non-inverting buffer. I never gave it much thought in the past, but with a buffer does resistance noise become more of an issue than the pickup source noise and general EMI?

[SoulFetish]

Ah, in that case there is a great resource you might be interested in. There’s a website based on the A book written called the ‘physics of the electric guitar‘ (or something like that). The text was written in German originally but much of it’s been translated into English and you can download chapters for free. In any case they dedicate a chapter which deals with Piezo transducers and magnetic pick ups in guitar and stringed instrument applications.
I have it bookmarked but I’m on my phone right now so I’ll grab it, and post it tomorrow. Maybe Helmholtz can chime in on this, as he’s the one who introduced the book snd website to me.

edit: I actually found a PDF of the particular chapter.

Physics of the Electric GuitarCh6Piezoelectric Pickups.pdf

[nickb]

The disc is unknown and part of a circuit I experimented with a very long time ago. It measures 80pf and is bonded to a thick aluminium disk and encapsulated with epoxy. The entire thing is 18mm diameter, so the actual disk is something like 15mm - quite small. I had it originally blocking a strat trem - fitting between the sustain block and the guitar body. The original circuit was just an opamp buffer followed by a simple fixed EQ. The sound was something I wasn't expecting - it sounded like a metal bodied resonator and when I first plugged it in my Wife said "that sounds just like that Dire Straits song", so I immediately learned Romeo and Juliet. I was just having a tidy up and came across the transducer assembly and just got me thinking what the practical limitation is of a simple opamp buffer with its impedance determined by resistance. I understand bootstrapping - I used this in my Ovation and it works very well.



My question isn't that I need to build a high-Z preamp, but what the practical limitation is for a unity gain non-inverting buffer. I never gave it much thought in the past, but with a buffer does resistance noise become more of an issue than the pickup source noise and general EMI?

For an 80pf disk, -3db @ 50hz requires a 40meg ohm input impedance. You previously used 5 meg ( 2 x 10 in AC parallel) which would be 400hz. That might explain part of the sound you got.

So working it though, to get 40meg you'd practically use 2 x 100meg. Each resistor would produce 180uV of noise the two would be 253uV. Guessing at 100mVRMS for the disk output the signal to noise ratio would be -52dB at low frequencies. Now big input resistors' noise will be in parallel with the disk so the noise will roll off about 40Hz so I don't think that will be a problem unless your amplifer has an unusually good low frequency response. A TL072 is 18nV/rt-hz so that's 2.5uV over a 20Khz bandwith. Small compared to the resistors. Assuming a x1 gain the output voltage offset would be 0.45V over the full temperatue range (very conservative). A fresh 9V battery will give you an output range of 3Vpp i.e 2.55v with a 0.45v offset. So if the output is AC coupled that should not be a problem, but rather depends on how much output you get from the disk and that has not been specified. The TL072 will roll off at 3MHz so that gives some measue of EMI protection but best to screen as much as you can. You could add 1k resistor and a 3n3 cap low pas filter on the output.

All in all it will work but the bootstrapped FET buffer is a much better approach, especially for noise.

[Helmholtz]

You previously used 5 meg ( 2 x 10 in AC parallel)

I understood that the 2 biasing resistors were AC bypassed, with a single 10M or 22M resistor between the virtual (4.5V) ground and the opamp's input?

[Mick Bailey]

I understood that the 2 biasing resistors were AC bypassed, with a single 10M or 22M resistor between the virtual (4.5V) ground and the opamp's input?

Yes - that;s right.

What I'm not understanding is the parallels with a condenser mic input where there can be quite a small source capacitance (as little as 6pf) and large (220M to 5G Ohm) input impedance, set mainly by large value resistors. Why isn't noise a factor with those circuits?

[nickb]

Yes - that;s right.

What I'm not understanding is the parallels with a condenser mic input where there can be quite a small source capacitance (as little as 6pf) and large (220M to 5G Ohm) input impedance, set mainly by large value resistors. Why isn't noise a factor with those circuits?

I think it's just the same. Say you have a 0.5G bias resistor and 10pf capacitance then you have a low pass filter with a pole at 30Hz. I expect suitable high pass filtering say about 50Hz in the preamp reduces it further. As far as specs are concerned, A-weighting helps again (-40db at 30Hz) reflecting the insensitivity of the ear.

My imagined circuit arrangement of the op amp buffer was two large resistors connected to the non-inverting input and with the inverting input connected to the output. This saves one resistor and one capacitor.

[Helmholtz]

I think it makes more sense to look the actual S/N ratio and not at equivalent input noise alone. A more efficient source/higher signal voltage will allow for more circuit noise.

[nickb]

I think it makes more sense to look the actual S/N ratio and not at equivalent input noise alone. A more efficient source/higher signal voltage will allow for more circuit noise.

Absolutely. The problem was the only spec given was 80pf so you have to work with the cards you're dealt

[Helmholtz]

Absolutely. The problem was the only spec given was 80pf so you have to work with the cards you're dealt

Yes, I figured and appreciate your considerations.

What I meant was, when comparing established preamp circuits for different sources, typical signal voltages also should be taken into account. A stronger source will allow for higher input resistance without getting objectionable output noise because it requires less total gain.