fets or bjts as voltage controlled resistors

[caress]

ok so i know this topic has come up before and i just read over these posts and articles:
http://www.diystompboxes.com/smfforum/index.php?topic=71037.0
http://graffiti.virgin.net/ljmayes.mal/comp/vcr.htm  --  fet as vc resistor
http://www.diystompboxes.com/smfforum/index.php?topic=24200.0 -- 4007 as vc resistor
http://www.diystompboxes.com/smfforum/index.php?topic=60662.0  --  the echobase uses a bc560 as a vc resistor. 

but...
i'm still not totally following how fets or bipolar transistors or whatever are used as vc resistors.

my questions are:
how can you calculate/determine the resistance in a setup such as the echobase?  fet?  lets say i wanted 10k?  100k?
what are some complications when using any of these setups?
achieving linearity appears to be the most difficult goal in using an fet/bjt as resistor(?), but what effect does that have when using an LFO or some other type of control voltage?  especially when using over a small range of only 10-20k?
any ideas about scaling the control voltage to be in the proper range for the fet, bjt, whatever?

anything else?

[R.G.]

how can you calculate/determine the resistance in a setup such as the echobase?  fet?  lets say i wanted 10k?  100k?
what are some complications when using any of these setups?

I think this is one of those places where the labelling (i.e. 'voltage controlled resistor') has tricked you into thinking you can just get what you want if you know the trick. It's not that easy. I think a better name might be 'voltage changed resistor' or (being a Texan) 'voltage herded resistor'. The 'controlled' makes you think that you can just set it to what you want. It's not that simple because the devices vary. A lot.

All readers of this forum should by now think "highly variable in everything" whenever they hear "JFET". 

You calculate/determine the resistance of such devices by either making them vary over the entire range of off (R=infinity) to on (R=really, really low) or by measuring what they are somehow. Getting a VCR (voltage controlled resistor) to be some exact value requires either measuring the exact device to know how that one piece of silicon responds, or uisng two matched devices, and measuring one while sending the same signal to both of them and counting on the matching. I guess I'm already describing the complications.

achieving linearity appears to be the most difficult goal in using an fet/bjt as resistor(?), but what effect does that have when using an LFO or some other type of control voltage?  especially when using over a small range of only 10-20k?

Linearity is difficult, because FETs have a small variable resistance range, and BJTs an even smaller one. The bottom line on that is that ANYTHING looks linear if you look at a small enough section of how it varies. Even diodes can be used as linear resistors if you restrict the signal level down to about 25mV.

However, using an LFO to control a VCR, especially over a small range, can be tricky. The classical example is the JFETs in a Phase 90. Those are being used as a voltage variable resistor. The JFETs are picked by type number to be a modestly good match to the LFO voltage, and then the LFO is tuned to the actual JFETs by a trimmer which sets the DC level of the LFO. And they're used over a wide range.

The best way to get a VCR to work well over a small range is to restrict the range with external resistors. A VCR usually goes from a few ohms to infinity (i.e. off). If you put a resistor in series with it and a resistor in parallel with it, then the series resistor keeps it from ever going lower than the series resistor. Same for the parallel resistor, keeps the resistance of the combination from ever going more than the parallel resistor. Now you can run the VCR over its normal huge range, and the variation for the whole setup will only vary between Rparallel and Rseries. Much simpler than trying to measure each VCR and tune it with only a voltage. Much more mistake proof.

any ideas about scaling the control voltage to be in the proper range for the fet, bjt, whatever?

Sure. Look in the data sheet. For a JFET, the Vgs is what controls the resistance. With Vgs=0, all JFETs are at minimum resistance of Rds, and that's specified there too. With Vgs = Vgsoff, the JFET resistance is infinite. The typically three to one range on Vgsoff is what makes this hard to deal with. The datasheet will also show you the maximum voltage you can get across your VCR without it stopping looking like a resistor and instead looking more like a current source. This is the region of the drain characteristic where the drain current/voltage lines slope and are not flat like they are at higher voltages. From this you can get an idea about what voltage is needed.

BJTs are simpler in some ways, worse in others. They have about the same resistance range, almost zero to infinity, but the thing that controls them can be thought of as either the base current or the base-emitter voltage. The range of collector-emitter voltages over which a BJT looks like a variable resistor is even smaller than the VCR range of a JFET, and you should not count on a BJT as a variable resistor for much over 25mV if you want low distortion. Using it as a switch is a different animal, and it can switch signals of a few volts on and off if you get through the variable region quickly.

For a BJT, the VCR range on the base-emitter voltage is quite small, usually only 0 to maybe 0.4 or 0.5V. That's nice, and the variation from device to device within that range is less than JFETs, although still not good enough to use a priori; you'll have to measure. That's why most BJT setups for VCRs use a large resistor to convert the control voltage to a control current into the base. The BJT is more linear in terms of conductance for current inputs than it is for voltage inputs. It's response to a voltage input is exponential with voltage.

Once again, if you must use a BJT as a small-range VCR, use resistors to "confine" its excursions.

In the case of using a BJT or JFET for a VCR on an oscillator like the Echobase, it's not really being used as a voltage controlled resistor. It's being used as a voltage controlled current source, which is a slightly different beast. In this case, the linearity with respect to the signal through it is not a concern, only how much current it lets through to run the oscillator at different speeds. That means that some of the trickinesses about signal size and linearity are less of a concern. You only have to figure out what the max and min currents are, and the LFO can then vary its conductance around that.

This is only the broadest overview of these issues. You've hit another one of those questions that are apparently simple, but have a lot of complications hiding under them.

[earthtonesaudio]

All it takes is one op-amp, 3 resistors, and one capacitor to make a voltage controlled PWM.  Pulse width modulation lets you switch the transistor fully on or fully off (which is what transistors are good at) for a variable amount of time.  If your switching frequency is fast enough, it sounds like it's "partly on" and you can't hear the switching.
Look up "square wave generator" using a single op-amp.

I like to use MOSFETs, capacitively coupled.

The only downside I can think of is the switching frequency is then coupled into your signal path.  So you have to deal with that somehow.

[snap]

The only downside I can think of is the switching frequency is then coupled into your signal path.  So you have to deal with that somehow.

the somehow part is the easiest to solve.

[R.G.]

The only downside I can think of is the switching frequency is then coupled into your signal path.  So you have to deal with that somehow.

the somehow part is the easiest to solve.

Right... 

[earthtonesaudio]

The only downside I can think of is the switching frequency is then coupled into your signal path.  So you have to deal with that somehow.

the somehow part is the easiest to solve.

Sarcasm?

Seriously though, to me the choices for voltage controlled resistors are either
a: go through a lot of trouble to get it to work within the linear reigon of the devices in question
or
b: turn it on and off really fast and then filter out a known (and fixed) frequency.

[JDoyle]

Seriously though, to me the choices for voltage controlled resistors are either
a: go through a lot of trouble to get it to work within the linear reigon of the devices in question
or
b: turn it on and off really fast and then filter out a known (and fixed) frequency.

b (part 2): ...and then go through a lot of trouble to get the damn click(s) out of the audio signal. If it is even possible.

Personally, I think the work is at worst the same, but with VCRs you aren't having to then REMOVE something from the audio signal and hope that the parts of that something removed that aren't in the audio range don't mess with everything else in the circuit.

I think it would be more work to filter out the clock signal AND stabilize the power supply than to put together a JFET/BJT test jig and measure parts.

But hey, to each their own...

Good luck!

Jay Doyle

[earthtonesaudio]

Well, yeah.  Jay is correct, if your oscillator is in the audio range you will hear a whine.  You will hear clicks if it's a LFO. 

But I was talking about making the clock frequency higher than audio.
You'd still want to filter it out of the audio path to be kind to tweeters and bats, but you'd never hear it.  With a high enough frequency, you can filter out the clock and not even touch the audio.

For what it's worth, this circuit here:
http://www.aronnelson.com/gallery/main.php/v/Schematics-etc/earthtonesgallery/Earth+and+Space+Wah+2.png.html
Uses the square wave generator on the same chip as the audio path (TL074 quad op-amp), with no bleed-through.  And this is one of the worst possible scenarios... variable frequency, variable amplitude, etc. 

The key to making it work well and avoiding headaches later is keeping the frequency high.


And to get back to the original post,

any ideas about scaling the control voltage to be in the proper range for the fet, bjt, whatever?

This is the nice part.  If your CV span equals the supply voltage applied to the PWM op-amp, you automatically get 0-100% PWM.  If the CV is less than rail to rail, you can either add gain to make it full scale, or decrease the supply rails for the PWM op-amp.


Anyway I'm sure this is way more than you wanted to hear about PWM, but it's just one of the viable options.  All have their pros and cons.

[brett]

Hi
when doing this with JFETs, I've found that devices with low Vgs-on seem to be more easy "tuned" (ie more negative.  Such as MFP102s and 2N5486s with Vgs-on of -5V, rather than J201s with Vgs-on of -1V).  Their Rds may be higher (more ohms).
cheers

[caress]

alright i've got a lot to digest here...

if it matters at all, i'll be using a noise source driving a sample and hold which i would like to have control a 5k pot.  guess i should check out the maestro s&h?

[frequencycentral]

alright i've got a lot to digest here...

if it matters at all, i'll be using a noise source driving a sample and hold which i would like to have control a 5k pot.  guess i should check out the maestro s&h?

The Maestro FSH is known to tick!

I like this S/H: http://fa.utfs.org/diy/roland100m/schematicsshow18.html

[caress]

The Maestro FSH is known to tick!

i know. i know...  i have a noise source and s&h circuit in mind that i'd like to try out already - i wanted to reference that schematic for the bjt vc resistor.  it's using a 2n3906 for the task.  is the trim that gets switched over as well centering the range?  not sure about that part...

I like this S/H: http://fa.utfs.org/diy/roland100m/schematicsshow18.html

hmmm.  that one's pretty nice with a low parts count.  i'll check it out...

i can't even remember which s&h i wanted to use.  it's in a schematic i drew up and uses an xr4212 and a bs170, but i can't figure out where it's from!