Differences in audio characteristics between MOSFET and JFET devices

[fryingpan]

Unfortunately I'm not really familiar with JFETs (my analogue electronics course skipped them as they "are seldom used nowadays", in my professor's words), apparently they sit somewhere in between BJTs and MOSFETs but they're more similar to the latter as they have a high input impedance. As JFETs (at least "discrete" JFETs) are starting to become harder to find, at least in through hole format, I was wondering whether they could be easily replaced by MOSFETs, with only some minor component change. What do JFETs have over MOSFETs so that the FX building community favours them so much over other devices?

[Rob Strand]

apparently they sit somewhere in between BJTs and MOSFETs

JFETs aren't very close to BJTs at all.  BJTs work off minority carriers and the process of diffusion.

JFETs are quite similar to depletion-mode MOSFETs (which are slighly different to the more common enhancement mode MOSFETs.)  Both JFETs and depletion-mode MOSFETs require a negative gate to source voltage for biasing.

Some differences between the JFETs and MOSFETs are:
- For JFETs the gate is not isolated from the channel. It presents a reverse-biased diode. If you drive the gate positive you can forward bias it.
- For MOSFETs the gate is isolated by the oxide layer
- An additional thing about MOSFETs is they appear as a reverse diode across D & S.

On the whole the MOSFETs you come across will be enhancement types.  They tend to follow the structures of power MOSFETs and have high capacitances.   JFETs generally have quite low capacitance. On-chip MOSFETs can have much lower capacitances.  The higher capacitance tends to smooth off the sound.

The enhancement mode MOSFETs are closer to BJTs but only in the sense they require positive gate or base voltage to bias them.  In that sense the circuit shapes often look similar.

[fryingpan]

I was quoting Italian Wikipedia:

Il transistor ad effetto di campo a giunzione o JFET (acronimo dell'inglese "junction gate field-effect transistor", comunemente pronunciato "gèifet") è un tipo di transistor ad effetto di campo, da considerarsi un ibrido tra il transistor a giunzione bipolare e il transistor a metallo ossido semiconduttore a effetto di campo (MOSFET).

Anyway when I say that JFETs can be replaced by MOSFETs in audio circuits I'm not necessarily saying you should do so without any redesigning of the circuit (bias points are different of course) but with some minor redesign can MOSFETs work similarly to JFETs? That is, do they clip similarly for example? MOSFETs higher capacitance might not be of concern in guitar or especially bass applications anyway, or maybe clean amplification could be provided by a BJT in a previous stage as to reduce Miller capacitance. It's quite easy to bias transistors anyway, especially for "textbook" topologies which at the end of the day are audio amplification's bread and butter. Maybe it's a hassle or unworkable in preexisting traces but when making a board from scratch...

As for the difference between enhancement and depletion modes, apart from biasing differences what's the real difference in functionality, once you take into account each device's peculiarities? For example, don't MOSFETs have a triode region too where they work as a variable resistor? (To be honest BJTs appear to function similarly to FETs in their saturation region, although it would appear that they behave more like on-off switches because of the abrupt change in collector-emitter resistance).

[GibsonGM]

Hi Fryingpan,

Welcome to the forum. You're really jumped from that frying pan right into the fire, haven't you?      Great questions! 

What I'd do, to get a better understanding of the 'tonality' of these things is....actually breadboard a LPB-1, a JFET booster, and a MOSFET booster.   Learn how to mess around with them (emitter resistor, biasing, bypass caps...etc).    Use them to goose your amp, try it with little distortion and a Lot.   
The schematics are available all over the net, at General Guitar Gadgets, for example, and AMZ (see first page of forum, at top).   Read about how the basic gain stage of a BJT, a JFET, a MOSFET are put together...Youtube vids, runoffgroove.com, AMZ....

This will allow you to make your own conclusions.  I'd also suggest you make a clean booster (JFET, Mosfet perhaps), and use that to PUNCH the OTHER circuits you just built on breadboard (each is simply a gain stage, in reality.).      Then you can see what they do when they clip, and if you make sure the boost doing the clipping is pretty transparent, you will hear the difference.     You seem curious enough - this is the way to get your answer, not just by relying on what others tell you a "sound" is like, as esteemed and knowledgeable as so many here are! 

For ex., I think a BJT clipping at a moderate level is a little more harsh and shrill than a JFET..but that COULD be tamed a bit by playing with caps, etc.   I believe JFET and MOSFET sound a bit more "tube like", but that will land me in trouble with the Clipping Gods.

The triode regions of semiconductors are very small and clipping tends to happen more abruptly than in a real triode, BTW, just to answer that question for you.  VERY hard to tell the diff between them.    To me, nothing really 'beats' a tube, but that might also be due to some discrimination on my part     There is every chance in the world that a well-designed preamp based on any of the 3 devices mentioned here could fool a tube geek's ear!     

[Danich_ivanov]

I would say that the strongest point for jfets is that they work pretty much everywhere (at least pedal wise), and are easy to deal with, so having a bag of jfets is always a good idea. Mosfets and bjt's i found to be more specific, especially in the "dirt pedal" scenario, where you may end up using all types of transistors in the different parts to get desired sound, because each one sounds unique. But for buffers jfets are hard to beat, they simply do the job.

[fryingpan]

Actually I'm not necessarily referring to tube-like sound (by the way, I'm reading Teemu Kyttala's Solid State Amplifiers which is really an insightful read, although it's just a start, I guess). It's just that popular JFET devices are getting harder to come by (for example, you may still be able to find through-hole versions of popular JFETs such as J201s or whatever, but the prices are exorbitant for what they are) and common sense engineering practices call for using whatever is readily available because discrete components especially have no real "magic" properties (they are, in simple terms, chosen for their basic characteristics). I was just wondering whether the fact that few builds call for MOSFETs (or, conversely, that so many builds require JFETs) is due to a precise choice or it is just because most other popular projects utilise such components (and basically, do not reinvent the wheel).

[Rob Strand]

I was just wondering whether the fact that few builds call for MOSFETs (or, conversely, that so many builds require JFETs) is due to a precise choice or it is just because most other popular projects utilise such components (and basically, do not reinvent the wheel).

You can't theorize with this stuff too much.  Some people play around with different circuits/parts until they find something they like then release the circuit.   Others will just use MOSFET in a particular place in the circuit from the start- for no reason more than that.   Then they build they surrounding circuit and it sounds good and they have no reason to revisit the MOSFET section because nothing bad has come out of it.

If you do comparison tests between MOSFET vs JFET  vs BJT  you can get different results depending on what precedes it: ie. raw guitar, opamp, another transistor stage and also the size of the coupling cap.

[GibsonGM]

You COULD go on Ebay or Newark or (check other suppliers) and find some common JFETs, buy up like 50.  2N5457, etc.   Yes, J201s are getting slim, but they are still out there!    A few yrs ago I just bought a bunch of JFETs, MOSFETS.  BS170s.     Plenty to keep one going on boosts, dirt pedals - but obviously not cool if you need matched sets for a phaser or something.   That's what Smallbear is for! 

Yes, just build some things (gain stage-based), and play around as you read and learn about the device you're using.  It'll become clear pretty quickly! 

[R.G.]

We have repeatedly run into this in several places. In the vast majority of applications, characteristics other than a device being MOSFET, JFET germanium, whatever, THE CIRCUIT IS THE DETERMINER OF SOUND, NOT THE DEVICE.

Well designed circuits actually hide the device characteristics to make the circuit depend less on device characteristics. Not doing this usually means that each copy of a circuit has to be hand-tweaked to work right.

FETs of all kinds do "show through" because their very high gate impedance lets you design to it being "very high" instead of worrying about exactly what "very high" is, unlike the case of input iimpedance in bipolars.

It is not usually the case that a JFET is a JFET because of the wide range of both Vgs and Idss inside and across device types.

[Rob Strand]

FETs of all kinds do "show through" because their very high gate impedance lets you design to it being "very high" instead of worrying about exactly what "very high" is, unlike the case of input iimpedance in bipolars.

Friendlier to guitar signals/pickups.

It is not usually the case that a JFET is a JFET because of the wide range of both Vgs and Idss inside and across device types.

Less friendlier to circuit designers compared to BJTs, well unless you use pots.

[fryingpan]

We have repeatedly run into this in several places. In the vast majority of applications, characteristics other than a device being MOSFET, JFET germanium, whatever, THE CIRCUIT IS THE DETERMINER OF SOUND, NOT THE DEVICE.

Well designed circuits actually hide the device characteristics to make the circuit depend less on device characteristics. Not doing this usually means that each copy of a circuit has to be hand-tweaked to work right.

FETs of all kinds do "show through" because their very high gate impedance lets you design to it being "very high" instead of worrying about exactly what "very high" is, unlike the case of input iimpedance in bipolars.

It is not usually the case that a JFET is a JFET because of the wide range of both Vgs and Idss inside and across device types.

I know that well-designed circuits will minimise the influence of a given part on the sound, although sometimes they actually do. Simple boosters or distortion pedals with very few parts and no NFB or deliberate diode clipping anywhere, or maybe where chips are deliberately pushed beyond their linear operating range. Compressors, especially simple ones, which use optical elements as opposed to FETs or BJTs (and viceversa). I don't know, maybe sometimes JFETs are chosen because of their clipping characteristics (say, a softer knee, a large clipping asymmetry, etc.). Their high input impedance is certainly one fewer hassle but for example even BJTs (a large variety of them at least) in buffer configuration will have high input impedance and whatever comes after it doesn't necessarily need this feature. I gather that it is advantageous to use (certain) JFETs in onboard preamps because of their low current demands, for example, so I understand this choice. That was my main question basically, whether their popularity (instead of transistors or MOSFETs) was dictated by (real or perceived) advantages.

[Rob Strand]

I know that well-designed circuits will minimise the influence of a given part on the sound, although sometimes they actually do.

Here's some recent threads on the issue with JFETS:
http://www.diystompboxes.com/smfforum/index.php?topic=119698.0
http://www.diystompboxes.com/smfforum/index.php?topic=119707.0

You can only go so far.

[Eb7+9]

THE CIRCUIT IS THE DETERMINER OF SOUND, NOT THE DEVICE.

(*!*) ...



most audio hobbyists end up using jFET's as VCR (voltage controlled resistors) devices in compressors and phasors
or use them in gain stages, single ended or mu-stages mostly

as far as gain stages go, you'd have to look at things from a large-signal linearity point of view to really explain things

ie., HOW THE DEVICE AND THE CIRCUIT INTERACT TOGETHER to produce wave-bending effects
interpreted as a type of non-linearity of the dynamic range
(you can quote me here)

NL = that's one big thing that a musician might perceive here
a thing discrete circuitry tends to exhibit versus, say, large-gain NFB circuits (op-amps)

I'm pretty sure that's the MAIN reason why players dig Jack Orman's MOSFET booster circuit so much ...
looking at Jack's circuit, I can't see anything else going on there other than subtle-to-not-so-subtle dynamic manipulation of signal

http://www.muzique.com/schem/mosfet.htm

adapt this same circuit to run on a BJT and the magic is somehow lost ...
why, because in the same circuit they (circuit and BJT device) interact to produce
a way more linear (less musically interesting ?!) output

otoh, ...

the VCR application of standalone MOSFETs is more likely to draw fans at some point
as dynamic figures of 750mV (MOS) versus 100mV (junction) are suggested in analysis I've read

I have NOT done any MOS-VCR testing yet myself so I leave it to others here to check it out for themselves
but if true it would pave the way for marginally cleaner operating limiters and phasors that operate the FET way

not sure why we haven't seen more of this yet // or at least sign of peeps trying

the matching and characterizing of either types of FETs can be done easily with proper
understanding of circuit theory, and using logical testing methods

there's only one way of dealing with 2-D independent variance

conversely, I showed in the Paradigm Shifter how to deal with un-matched FETs in a single modulator setting
look it up // the same would apply here

...

btw, the super high gate impedances are only an advantage when dealing with piezo pickups
otherwise not a dominant feature in this comparison - at least when compared against modern high-Beta bipolars

[Rob Strand]

adapt this same circuit to run on a BJT and the magic is somehow lost ...
why, because in the same circuit they (circuit and BJT device) interact to produce
a way more linear (less musically interesting ?!) output

The capacitive loading on the pickup is a lot higher with the MOSFET; it gets Miller multiplied quite a bit.
Sometimes you can make things closer by emulating what is actually going on (ie. adding a CB cap to the BJT).

==================
I wasn't going to comment but:

the VCR application of standalone MOSFETs is more likely to draw fans at some point
as dynamic figures of 750mV (MOS) versus 100mV (junction) are suggested in analysis I've read

I have NOT done any MOS-VCR testing yet myself so I leave it to others here to check it out for themselves
but if true it would pave the way for marginally cleaner operating limiters and phasors that operate the FET way

The Electro Harmonix Bad Stone and the ETI phaser used CMOS gates as MOSFETs.  I suspect discrete MOSFETs can be use for modulation applications.  A trick to ensure better matching and tracking is to use a "dummy" MOSFET as reference/offset voltage for the others.  I've only toyed with them.  Another issue is the relatively low Rds_on which means you have to ride closer to the Vgs_off point, so matching becomes a bigger issue.  You can halve the problem with series MOSFETs.

Due to the high capacitance of 2n7000 family devices, using discrete devices is pretty marginal for compressors.  Basically things where the control voltage can change quickly it's going to be marginal.  For modulation devices the glitches can be minimized as the control voltage changes relatively slowly and you can always add low-pass filter to round the LFO off a bit.

[Rob Strand]

Here's two versions of the series MOSFETs:





Forgot to mention that Back-to-back series connection of MOSFETs can be used as an AC switch.
Also, the gate(s) only needs to be driven with one polarity.

[R.G.]

THE CIRCUIT IS THE DETERMINER OF SOUND, NOT THE DEVICE.

(*!*) ...
[...]
ie., HOW THE DEVICE AND THE CIRCUIT INTERACT TOGETHER to produce wave-bending effects
interpreted as a type of non-linearity of the dynamic range
(you can quote me here)

I just did.   
At the risk of letting you successfully start an argument, let me point out that you didn't quote that whole sentence, the first part of which goes:

In the vast majority of applications, characteristics other than a device being MOSFET, JFET germanium, whatever,

If there is a minority of circuits where the device does "show through", as the next couple of paragraphs discuss, what exactly is there to argue about here?

Especially If the infamous high feedback opamp can be designed into circuits which produce anywhere from subtle to downright kinky distortions?

I believe that I was pretty clear that the circuit can both show and hide certain issues of any device, so if the device doesn't change but the rest of the circuit can largely determine how much of it's kinkiness comes out, again, what exactly is there to argue about here?

[...]
conversely, I showed in the Paradigm Shifter how to deal with un-matched FETs in a single modulator setting
look it up // the same would apply here

Aaaaand we're still flogging that "revolution" I see. 

...
btw, the super high gate impedances are only an advantage when dealing with piezo pickups
otherwise not a ominant feature in this comparison - at least when compared against modern high-Beta bipolars

Only? Is that an opinion or an undisputed fact?