Compensated Exp generators and VCO's

[Rick Jansen]

(This has become quite long, sorry :-)

About exponential generators
----------------------------
For the Moog filter I used an exp generator built with the 
LM3046 transistor array, and this works quite alright. For 
a VCO the specs of the 3046 are questionable. The matching 
between the transistors is ok-ish but not more than that. 
The current gain (hfe) is 50-100, not more. About the log 
conformity I have no data, nor on the bulk resistance of 
the transistors. (Anyone for this value?) 

The Formant VCO's and some other machines have used the 
Fairchild uA726 in the past, which was a precise dual 
transistor with built-in on-chip temperature 
stabilization. Alas, this has become an unobtainable part.

I wish there was a transistor array which did have great 
specs for the individual transistors, as substrate 
temperature stabilization is very easily done (one 
transistor a heater, one a temperature sensor, one a clamp 
and two the differential pair). Alas I'm not aware of such 
a part. Harris does have a CA3246, which looks like an 
improved 3046, but it is not available here in The 
Netherlands. It has a bit better specs for hfe, but again 
no data on bulk resistance or log conformity.

A new VCO
---------
A new project is a temperature stabilized VCO. The past 
week I've done some literature research on log and antilog 
(exp) generators, and I've come up with something I'd like 
to discuss. A first schematic is available in my WWW page:

  http://www.sara.nl/Rick.Jansen/Emusic/VCO/VCO.html

It features a precise and very log conformant dual 
transistor pair (National Semiconductor LM394), which is 
kept at a constant temperature with a heater circuit, as 
one of the biggest sources of error is the change in 
ambient temperature. This can be compensated in two ways:
- by using a tempco resistor, the infamous Q81's
- by keeping the temperature stable.

As for the Q81's: I cannot find them here in the regular 
shops, and thus I have abandonned this route and opt for a 
heated exp generator.

The heater consists of a piece of aluminium profile with a 
powertransistor and a temperature sensor attached to it, 
enclosed in a small plastic container with the dual 
transistor. The container is something like a small 
medicinal drugs package. Attach the cap to the circuit 
board and screw on the container. This way airflow has no 
influence. Of course further insulation with styrofoam or 
something is a good idea. 

As always there's also a downside to the heater solution: 
it draws quite some power, especially at startup time. So 
I will probably build a separate heater power supply.

Further errors in exp generators are introduced by the 
bulk resistance in the transistors. This can be 
compensated too,with the diode/4.3M resistor as shown in 
the schematic. The bulk resistance is about 0.5 ohm for 
the LM394.

By using a precise current source further log errors are 
prevented. 

The LM394 is a rather popular part in the world, and thus 
quite well available. 

Sawtooth generator
------------------
In the sawtooth generator the 2.2n capacitor is charged 
untill the LM311 voltage comparator's treshold is reached. 
Then a MOSFET switch discharges the capacitor. When the 
treshold voltage is reached the output of the comparator 
doesn't change back right away, it's held high for a short 
period, depending on the 22p capacitor and 10k resistor. 
This ensures the 2.2n is fully discharged through the 
MOSFET switch. 

The discharge time is kept very short here by using a 
HC4066 analog switch instead of a discrete (MOS)FET. FETs 
with a low Rds,on are hard to find, while the switches in 
the common HC4066 have an Rds,on of only 25 ohm! Normal 
MOSFETs have a Rds,on of ca 100 ohm. On the downside: a 5V 
supply is needed now. 

The discharge time of the capacitor introduces an error in 
the sawtooth. As the discharge time is always the same it 
gets increasingly important at higher frequencies. This 
can be compensated for by decreasing the treshold voltage 
at higher frequencies, so the charge time is a bit 
shorter, and thus linearity is ensured. The cost is that 
sawtooth amplitude gets a little bit lower at higher f.

Literature
---------- 
- Design with Operational Amplifiers and Analog Integrated Circuits 
by S. Franco 
- The Art of Electronics by Horowitz&Hill 
- Opamp Cookbook by W.G. Jung
- Elektuur nr 266, December 1985 pg. 38 'Kristal-oven' about a 
quartz-crystal oven.

In the schematic there is also a part devoted to triangle 
waves, but that part is in serious error. Comments and 
further discussions are of course very welcome!

Some questions bothering me right now:
- anyone know a value for the bulk resistance for LM3046 
transistors?
- anyone know another 4 or 5-transistor array with better 
specs than the LM3046?
- What did you do about temperature stabilization?
- Did you compensate for the discharge time in the 
sawtooth generator and bulk resistance in the exp 
generator?
- Do you convert the sawtooth waves to triangle or simply 
use separate triangle-wave generation? 

Rick Jansen
__
rick at sara.nl   http://www.sara.nl/Rick.Jansen
_____________________________________________
S&H's a module and s&h's looking good