JFET input opamps, why not MOSFET input?

hrs

Joined Jun 13, 2014
355
Hi,

Opamps like the TL07x have JFET inputs. Why not MOSFET or CMOS? Very often guitar effects have either such a JFET opamp on the input or a JFET transistor configured as a buffer. Why not some type of MOSFET transistor configuration?

OBW0549

Joined Mar 2, 2015
3,566
Opamps like the TL07x have JFET inputs. Why not MOSFET or CMOS? Very often guitar effects have either such a JFET opamp on the input or a JFET transistor configured as a buffer. Why not some type of MOSFET transistor configuration?
Going by my memory (which I admit may be a bit dim) of the time in the late 1970's when these were introduced, I believe the semiconductor fabrication processes in use at the time permitted fabrication of p-channel JFETs alongside the BJTs used in the rest of the op amp, allowing for much-reduced input bias currents (the main point of having FET inputs). As I recall, MOSFETs were not compatible with those fab processes.

Someone may come along with a better explanation, but I believe that's more or less the way it was.

The attached document provides some history.

Attachments

• 207.9 KB Views: 5
Last edited:

hrs

Joined Jun 13, 2014
355
Going by my memory (which I admit may be a bit dim) of the time in the late 1970's when these were introduced, I believe the semiconductor fabrication processes in use at the time permitted fabrication of p-channel JFETs alongside the BJTs used in the rest of the op amp, allowing for much-reduced input bias currents (the main point of having FET inputs). As I recall, MOSFETs were not compatible with those fab processes.

Someone may come along with a better explanation, but I believe that's more or less the way it was.

The attached document provides some history.
I was thinking along the lines of linearity, noise, speed etc. Fab processes never occurred to me, thanks! And thank you for the document, it looks interesting.
MOSFET input op amps exist. I thought there was one way back. See: https://prom-electric.ru/media/CA3140E.pdf
This modern one has a MOSFET input for extremely low bias current:
Wow, that ADA4530 is $25 at digikey! Must be an expensive process. Analog Ground Joined Apr 24, 2019 456 I was thinking along the lines of linearity, noise, speed etc. Fab processes never occurred to me, thanks! And thank you for the document, it looks interesting. Wow, that ADA4530 is$25 at digikey! Must be an expensive process.
20 femtoamps is a mind-boggling low current. The PCB design and assembly is a lot of fun to insure that level of performance by eliminating leakage currents. Leakage paths have to be greater than 50,000,000,000,000 (5e13) ohms to 1 volt. They are pretty much only used as preamps. Special connectors are used to attach to sensors.

crutschow

Joined Mar 14, 2008
28,580
One significant difference is that JFETs generally have lower noise than MOSFETs.
There are some MOSFET op amps, but you will notice that their noise level is generally higher than the JFET op amps (and JFET amps often have lower than noise than BJT input op amps).
Some low-noise audio op amps have JFET inputs.

dl324

Joined Mar 30, 2015
13,857
As I recall, MOSFETs were not compatible with those fab processes.
That changed in the 80's. Intel had a process they called BiCMOS; their process designater was P651. I don't think there were every any real technical hurdles that prevented anyone from doing it. In the early 80's designs for memories and, later, microprocessors, were transitioning from bipolar to NMOS and then to CMOS (though some companies had a PMOS only process).

Around 2000 they started using germanium again; with silicon.

Audioguru again

Joined Oct 21, 2019
4,129
Guitar effects used vacuum tubes and an input resistance of 2M ohms and some still do 60 years later.
Many effects are used today with antique germanium transistors.
Old stuff? Yes. Sir Mick Jaggar is 76 years old today and maybe also his amplifier and effects.

MisterBill2

Joined Jan 23, 2018
10,602
20 femtoamps is a mind-boggling low current. The PCB design and assembly is a lot of fun to insure that level of performance by eliminating leakage currents. Leakage paths have to be greater than 50,000,000,000,000 (5e13) ohms to 1 volt. They are pretty much only used as preamps. Special connectors are used to attach to sensors.
Bob Pease wrote quite a bit about working with the very high input impedance devices and many of his insights are quite valuable. The summary is that it takes a lot more effort to be successful in dealing with those systems.
And what specific property would you want in a mosfet that is not found in the present devices?

Always, there must be a reason for a change, UNLESS you are a software company that sells operating systems. Then, there does not need to be any benefit at all for the users from whatever changes are inflicted.

jpanhalt

Joined Jan 18, 2008
11,088
I was thinking along the lines of linearity, noise, speed etc. Fab processes never occurred to me, thanks! And thank you for the document, it looks interesting.

Wow, that ADA4530 is $25 at digikey! Must be an expensive process. This note by TI compares the three input types, BJT, Mosfet, and JFET: http://www.ti.com/lit/ml/slyt701/slyt701.pdf Comparison of input impedance and bias currents begins about page 11. Also of note, input bias current of CMOS inputs is affected by protection circuity that may be included. As for prices, for the 3 op-amp inputs discussed: OPA211 (BJT):$7.23
OPA320 (CMOS): $2.08 OPA140 (JFET):$3.61
All prices are Digikey, cut-tape, quantity =1.