VCA Circuit

In my quest for VCA circuits I ran across this one:

http://electronicdesign.com/analog/op-amp-and-two-jfets-form-voltage-controlled-amplifier

Not sure how well it works in real life, but it spices really nice!  Some CV bleedthrough (?) but none of the massive attenuation followed by massive gain (to avoid large signal distortion) nonsense that's going on in most VCAs.

Sim works great with all resistors set to 10k, negative to zero CV, MPF102 JFETs, rail-to-rail I/O opamp.

LTSpice here: http://www.mediafire.com/download/tnfrm7netlp7knb/VCA_2014-04-13.asc

Inherently large signal, 80dB dynamic range - so: what's wrong with this circuit and why have I never seen it before?  It looks like a dream come true, but I'm a noob at this stuff and so am likely missing some key flaw.

"so: what's wrong with this circuit and why have I never seen it before?" - Dewster

It looks beautiful, elegant, clever, simple... Too good almost to not have been seen before - But I havent seen it before!

EDN Design ideas are often remarkably clever - People are often drawn to publishing their simple clever ideas there - they pay $300 a pop, and give you the chance of winning something bigger..

This one looks like a cracker - if its as good as it looks and doesnt have some issue we havent spotted... Many thanks for bringing this to our notice!

Fred.

"Too good almost to not have been seen before - But I havent seen it before!"  - FredM

There's gotta be something fatally wrong with it!  It would likely work best with a dual JFET where temperature and characteristics would be closely matched.

Even Rod Eliott doesn't list it in his otherwise fairly comprehensive VCA techniques web page:

http://sound.westhost.com/articles/vca-techniques.html

"EDN Design ideas are often remarkably clever..."

Some of the best circuits show up there.  One was a pitot-tube air-speed gauge for hang gliding - a linearizing, temperature cancelling, multiple feedback loop, logarithmic/exponential monster constructed from a quad opamp and a BJT array.  It blew my mind at the time (1997) and remains beyond my current analog skills level.

Ah, here it is: http://electronicdesign.com/test-amp-measurement/linear-pitot-tube-air-speed-indicator

Dewster,

I have been playing with that VCA simulation, and I cannot fault it - Yes, matched transistors thermally linked, or ideally on a common substrate, would really lock it tight - but its good even in its basic form..

The biggest problem with FETs in shunt (or pass) mode is non-linearity, which causes distortion - This circuit reduces the distortion greatly .. Its not as good as an expensive THATS Blackmer VCA, but for its cost and simplicity it looks good enough for most musical instrument applications IMO (VCAs in musical instruments dont often need to be nearly as "good" as those used in mixing consoles or Pro-Audio processing gear).

The other use I can see for this VCA (with some modifications) I can see is as a theremin mixer - Add some rectification and filtering 'round the opamp and have a direct audio difference out thats easily biased to provide whatever adjustable distortion one wants, so you can go from "clean" to as "dirty" as one wants.

Oh, one other thing - Rod Eliott misses quite a large number of commonly used VCA's - particularly ones commonly used for musical instruments which arent suitable for studio equipment... This is something even those designing electronic musical instruments often forget - We can have a different set of requirements - we dont always need signal to pass through without "modification" of parameters the "modifier" isnt designed to modify .. For example, VCA's that distort when over-driven may be unnaceptable in a mixing console, but desirable in a synth, and VCF's where the Q can be turned up to cause bursts of oscillation when overdriven are totally unnaceptable for mixers but highly prized in analogue synths!

Fred.

Hi Dewster,

Have you prototyped this yet?

I transferred the schematic over to Proteus, and it simulates fine - except the JFET models fpr the same parts behave hugely differently - I needed to alter V4 to -0.5 0.5 100 to avoid distortion. My MPF102 model is a simple generic model - but I substituted a full SPICE model of a similar JFET and got similar disparity between the two programs.

I am inclined to trust LT-Spice more, and basic calculations indicate its more likely to be telling the truth - but I cannot be sure - and this disparity is a real worry (not because of this circuit, but because of other simulations I am now worried about).

I need to examine my Proteus setup to see if ive done anything stupid to the simulation parameters (it has a global setup area, and if you mess anything up here, all simulations can get fuckex up - this allows one to tweak simulations and get them to run with low accuracy and get 'round convergence errors etc - but its dangerous if one accidentally does something silly and dont change it back) but just rhought id mention it in case LT-Spice was the problem.

Fred.

PS - I notice you are using the LM6132A .. Thats one of my favorites! ;-)

"Have you prototyped this yet?"  - FredM

Not yet but soon.  It's tax time which means the SO and I are spending days tallying up what the IRS and the state of NJ already know (and will indeed inform us if/when we make a mistake).

"I notice you are using the LM6132A .. Thats one of my favorites! ;-)"

Yes, I had some left over from an old project (speaker test microphone preamp) that someone sent me, never used one until a couple of days ago (in a capacitor ESR meter I'm spending way too much time designing).  Super nice opamp, rail to rail I/O, low power, low voltage, and the simulations match reality pretty well.  Works well as a comparator too (have one in an oscillator that develops ~3 microsecond pulses across 2.5 ohms at 0.25V).

Fred: try this model for J175 and make the CV positive:

.MODEL J174 PJF VTO=-7.1507 BETA=1.24461E-3 LAMBDA=2.01968E-2 RD=3.95403 RS=3.95403 IS=9.7275E-16 CGS=1.58E-11 CGD=1.88E-11 PB=3.7207E-1 FC=5E-1

The author seems to have picked a PJFET with high VTO and high IDSS.

Hi Dewster,

I will try those - but if they are "true" then they are miles off any JFET I have measured (or derived values from in real life from circuits).. VTO at -7.15 in particular.. And miles off from some of my models (Alas, generic models in Proteus dont show their values, so I must derive these by building simulated test circuits)

But I think it comes down to VGS(off) in my Proteus J174 model being much lower - I have been running test circuits and it seems this model is giving something like a VGS(off) of about -2.

I am not worried anymore though - everything behaves as expected when I use models I create myself, and when I use the same model in both LT-Spice and Proteus I get the same results..

I have been playing with this VCA and derivations thereof and cannot find a flaw - thermally link the transistors and its a core circuit with wide application and almost ideally suited AFAICS to many theremin related things - As a VCA, as you pointed out, its clean and easy - no need to mess with the signal path (attenuation/amplification) and as maximum attenuation corresponds to 0V on the gate, setting up the CV is one simple trim using 2 resistors, all one needs is to set the incoming gate voltage so that the fet acts in its linear zone, and this is easy because Vds is is so low..

I think CV trimming will be required in any practical application due to the wide variation in VGS(off) from part to part (-5V to -10V for the J174 ?)

So far, in simulations (using a few added components for AC coupling and biasing / adjustment ) I have "made" mixers, frequency doublers, audio ring modulators, wonderful distortion circuits, squarer, and of course VCA ;-) - A whole load of circuits I have been using H11F1's to achieve can be easily replaced.

I think the selected parts with their high VGS(off) are probably best - but I want to play with more common JFets having lower VGS(off) - simulations indicate these will work ok but perhaps not be quite as linear / low distortion..

Fred

Set the VCA circuit up on my bench today.  MPF102 NJFETs, LM6132A opamp with +/-9V supplies, all fixed resistors 10k, a 1k 10-turn trimpot feeding the input Rs with a 0.47uF input cap on the wiper connected to a function generator, a 22k pot from ground to -9V for CV.

It works!  It looks as nice as the sims!  Sweet!

Feeding it with a 6V p-p 1kHz sine and trimmed for minimum output with CV=0, I'm seeing ~3mV p-p output with CV=0; 5.6V p-p output with CV=-2.9V.  This is a dynamic range of 65dB.  It could be better as this is the limit my scope with maximum averaging can resolve, and includes some noise.  With CV=0 and pinching one of the FETs for a bit (to increase temp) the output rises to ~50mV p-p; pinching the other FET makes the imbalance go through zero and reverse phase, so thermally coupling the FETs should help ameliorate this.

Output amplitude vs CV looks pretty darn linear:

Works above 100kHz even with my rat's nest breadboarding.  Output has an offset of 163 mV with CV=0.  As you say Fred, the maximum CV would need gain trimming, but minimum gain conveniently corresponds to CV=0 (with a bit of trimming of the input pot).

This thing almost makes me want to build an analog Theremin!  (Quick, someone shoot me!)

===========

[EDIT] Not surprisingly, I can get a stronger post-trimming null if the JFETs are better matched.  Two of my MPF102s have VT of -2.65/-2.85V and IDSS of 10.3/11.1mA respectively, and these work better together than with the third MPF102 that has VT=-2.05, IDSS=6.91mA.  Using this third FET with one of the others gives a trimmed minimum signal of around 7mV, or 58dB dynamic range.  Still not all that bad I suppose.

The more I play with this VCA, the more I love it!

Found a source of new dual JFETS (these are still being made in through-hole packages)

http://www.linearsystems.com/products.php?pr=jfet-amplifiers--duals

These are selling at £3 upwards in the UK (one-off, TO-71, minimum order £25 ..) I bought a batch to play with (3 weeks delivery) and will let you know how it goes - Should have some spare to sell in one-off if anyone was interested - I got a special price so will have 20 (inclusive of tax and postage they cost me just under £5 each), but will probably need to pay full price on any subsequent order - bordering on too expensive (same sort of price as a Blackmer Log VCA).

I suspect I will likely use this VCA for non-critical stages where a pair of cheap matched fets glued together will do the job.

Fred.

This is the VCA I have been using - at £3 its great (or theres a pre-trimmed version at £6) but can be a bit awkward to interface.. http://www.mouser.com/ds/2/407/2181data-7040.pdf