Designing with VCA's

[Cliff Schecht]

I've never designed with a VCA before, but today I had the scope out and was trying to get a VCA to give me some output, but couldn't get it to do much of anything. Has anyone used or designed with the LM13600 OTA (with the linearizing diodes and buffering)? I was really just bs'ing the values using the datasheet and was trying to use a 555 timer for the variable gain and a schmitt trigger oscillator for the input signal, but I'm sure I was doing quite a few things wrong, as I just kinda threw it together on the breadboard. I had dual rail voltage in the form of two 9 volt batteries doing plus and minus 9V.

Basically, I'm trying to use a (an?) LFO to control the variable gain stage of a VCA, to try to get a choppy type trem effect. Well actually I was thinking that you could make some sort of switchable or blendable type control that allowed one to blend between, say a square wave and a sine wave. This is really just for a personal learning experience, I'm not expecting the next big tremolo to come out of this, but any help in getting me closer to something that works would be appreciated greatly Smiley.

Edit: Sorry about the topic reposting three times for anyone who saw that, my browser decided to freeze up for a second.

[Paul Perry (Frostwave)]

The LM13600 data sheet VCA does work. BUT......
1. it is VERY easy to destroy a LM13600 by running excess current thru the control pins. Be sure there is a sufficiently large series resistor.
2. You might want to build just the VCA, and make an adjustable control voltage supply from a 10K pot across the supply. That will give you an idea of whether the VCA part works, and over what range, and then you can go on to the 555 part.

I find it is always a good idea to build the separate parts of a circuit & test each one as I go. Because, otherwise I have a big circuit & dont know which part(s) don't work

[slacker]

check out this http://www.geofex.com/Article_Folders/VCA%20Applications.pdf from R.G.

[gez]

Don't have the data sheet to hand (link always helps) but most circuits for the LM13700 etc have low input impedance, so a buffer on the input is always a good idea. 

I take it you have a resistor from the output of the OTA (not the output of the buffer) to ground/Vref (if single supply), to convert the current into a voltage? 

Try it without the linearising diodes and, as Paul suggested, with a variable resistor (and stop resistor) to the Iabc pins instead of the output from a 555/whatever, just to check that it's working as an amp. 

PS  Unless you shape the square into a trapezoid, your tremolo is going to click like hell.

[R.G.]

The LM13600/13700 are really not VCAs, they're Current Controlled Amplifiers. They are OTAs, and have a transconductance of about 19.2 Iabc. That means for each differential volt into the two inputs, they put out 19.2*Iabc on the output. The output is also a current, and must be changed into a voltage by running it through a resistor.

The simplest way to set these things up is to make a bias voltage of half the power supply, well bypassed to ground, and then run a single load resistor to that bias voltage. The voltage gain of the amplifier then becomes Vo/Vin = 19.2*Iabc*Rload.

If you want voltage control, you can almost get it. The Iabc pin is tied to the V- supply by two diode drops. So if you put a resistor between your control voltage and Iabc, the control current Iabc becomes Iabc = (Vcontrol-2*Vd)/Rb, and that approaches Vcontrol/Rb as Vcontrol gets much larger than 2*Vd. When Vcontrol is less than 2*Vd, nothing happens. Vcontrol can be anything as long as Rb scales it down to make Iabc be less than 1ma for the 3080, 2ma for the 13600/13700. If you exceed those currents, the chip turns into a Darkness Emitting Diode.

There is another way to get real voltage control. The current into a diode is an exponential function of the voltage across it. If you feed the Iabc pin with a low impedance voltage which ranges from 0V to about 1.4V and is current limited to less than the destruction value, the gain is an exponential function of the applied voltage, which is exactly what you want for linear-sounding volume control. It takes some messing about and scaling with opamps and such to get a reasonable control voltage (like maybe 0-5V) set up this way, but it's doable.

Then there's the input voltage. The input is a raw bipolar differential amp, one base being the + input and the other being the - input. This input has a linear range of less than 25mV. Get much over 20mV and you get distortion. You may or may not like this, some people do. But it's not good for things you don't want distorted. That's why all of those circuits have under-1K resistors hanging from the inputs - they are part of a voltage divider to divide the signal down to under 25mV max. That makes for a low input impedance and a low input voltage, so you have to get enough gain back to amplify the signal back up to audible. In general, dividing down a signal like this and reamplifying it makes for a noisy setup. This is why OTAs have a reputation for being noisy. They're used under conditions which make them that way.

The LM13600/13700 and NE3280 have linearizing diodes on the inputs. These pre-distort the input signal in a way that the OTA undistorts, so you can pump in higher signals  before getting lots of distortion. Works OK, but lowers the amount of gain you can get out of them. That may or may not matter in your application.

Outside of voltages which are destructive and/or simply make the inputs not work, there is no fixed relationship between the input bias voltage and the output DC level. The output is a differential current. When the + input is higher than the - input, the output sources current; when the - input is higher than the + input, the output pin sinks current. If the output is shorted to any voltage between V+ and V-, that's fine, the current just runs into the voltage. If there is a resistor between the output pin and any voltage between V+ and V-, a signal appears across the resistor. The DC level is whatever level the non-signal side of the resistor is tied to. It the resistor is not in the middle of the power supply, then the signal will bump into the power supply early on the shortest side so you still need to bias in the middle of the power supply for biggest undistorted signal swing.

But the inputs do not need to be at the same bias voltage as the output . They can be at some other DC level, and they can't control the DC level of the output.

There is some bleed through of the control voltage movements. This is because of mismatches on the input devices primarily. For lowest feedthrough, you need to have some means of slightly DC imbalancing the two inputs to make up for this mismatch and cancel out the control voltage feedthrough. There are lots of ways to do this. The dynacomp uses a scheme of unbalancing the two by unbalancing the DC bias voltages on the two with that 2K pot, which also serves as the input signal divider.

So bias the inputs, bias the outputs; divide your signal down to under 25mv peak; load the output with a resistor to half the power supply, or two equal resistors to V+ and V-, which is the same thing; and put currents under 1ma/2ma into Iabc.

[Cliff Schecht]

Thanks for the great replies guys! I'm looking forward to getting something up and running, I've been running this idea through my head for a while and finally got a scope to play with.

[Harry Palms]

http://www.wiseguysynth.com/larry/schematics/dbx_vca/2150data.pdf

http://thatcorp.com/datashts/2181data.pdf

http://www.thatcorp.com/datashts/dn137.pdf

To see this type of chip in action; real time use in an effect, scrounge up a used DOD milkbox compressor and reverse it. It'll set you back maybe $25 (I got one for $13.50 shipped) but it's a good lesson and you'll end up with a decent pedal too.

[toneman]

PAiA's 9710   VCA/BalancedModulator uses LM13600s or 13700s.

PAiA provides the schemo for FREE here--

http://www.paia.com/prodimages//9710scha.gif

now U know   

[Cliff Schecht]

Speaking of PAIA, I'm getting to meet the companies owner tomorrow, apparently he works with my dad .

[Cliff Schecht]

Also, for future reference, around what frequency should I be running my LFO at?

[Cliff Schecht]

I've figured it to be around 1/4Hz to about 3 or 4 Hz, does that sound about right?

[gez]

Upper limit would more likely be 12 Hz or so.  More even...

[Cliff Schecht]

Upper limit would more likely be 12 Hz or so.  More even...

One of my engineering buds was saying that if you go over about 5hz, you'll get some sort of weird type of effect that would occur, something like a ring modulator. I'm not sure what frequency it occurs at exactly, but I'm guessing the two frequencies mix into each other like a ringmod and create a nasty dissonant type effect.

Maybe I should ask a better question. What time ranges am I looking at for typical tremolo settings?

[gez]

Your friend is correct, but that happens at frequencies higher than 12Hz.