[sdiy] Question about my high freq compensation

[Ian Fritz]

Hi Ray --

Well, Oren has already helped you find a better way to hook up the HF 
compensation, but here are just a couple of extra thoughts.

First, the reason that it is correct to connect R13 to the servo 
output:  The high-frequency error comes from the change in Vbe caused by 
parasitic series resistance multiplied by the output current.  Since the 
servo output follows the output current, it is the proper source to 
generate the correction voltage to the input Vbe.  The output of the 
summing amp (where you originally connected) follows the log of the output 
current and so has the wrong behavior for proper compensation, even though 
it moves in the correct direction.

Second, everyone should be reminded of the clever improvement on this 
compensation circuit developed by Rene.  Rene's circuit has a 10k trim pot 
connected to the servo output, followed by a diode to ground.  The voltage 
from the trimmer's slider is fed back to the input base through an 
appropriate resistor (depending on the amount of parasitic 
resistance).  This structure more nearly mimics the output side of the 
converter.

Finally, I would agree with Oren that many of the advantages of using the 
expensive, high- performance transistor pair are obviated by the relatively 
slow switching circuit and also by op-amp offsets and leakage.  If fact, my 
most successful designs have used the exact  opposite approach of yours, 
namely medium performance converter transistors (CA3083 array) and high 
performance op amps and switches, along with separate Rbe and reset-time 
compensation.  At some point you might want to try using a less expensive 
pair and looking at whether the performance is really much worse.

Best regards,

   Ian



At 12:09 AM 12/9/2003, Ray Wilson wrote:
>Hi List
>
>I would greatly appreciate it if some of you really smart people would look
>at my oscillator here:
>
>http://www.musicfromouterspace.com/analogsynth/OctVCO_LM394_FixedSquare.html
>
>and let me know if I have goofed on the high frequency compensation.
>
>My explanation (which may be incorrect is this) is this:
>
>Here is what I think is happening. I know the circuit looks wierd but  you
>must remember that as the voltage gets higher on the inputs of the inverting
>summer made up of IC1-B and associated resistors that the voltage on the
>output of IC1-B gets lower and lower. I think the high freq compensation
>works by adding a secondary path for current to pull the base of the top
>transistor a little lower as the input voltage gets higher and subsequently
>the frequency gets higher and higher so that the discharge time of the
>integrating capacitor (which becomes more significant at higher frequencies)
>is compensated for by causing the oscillator to oscillate a bit faster than
>it would without the secodary current path.
>
>Please let me know if I have goofed so I can fix it or if I got it right in
>which case I can have a beer and celebrate.
>
>Thanks and Cheers
>
>Ray