From: Sar Saloth (sarsaloth_at_yahoo.com)
Date: 2002-03-08 20:26:53
I have tested that circuit but with a different OpAmp. Usually I do
something like connect a capacitor or resistor to a function generator and
use that as variable portion of the load.
I do believe that you might have had it stable in your configuration, but
then we had better use the identical components. I do have the equipment
to do such a test if you want me to test a particular combination. I don't
mind doing that and sending you any waveforms.
As I said before, the same circuit (although with Opamps that really DO
make it unstable) has been in production for 15 years with very people
noticing a problem. I can't tell you what product it was, but many have
been made. Nevertheless, I can see some low level noise that shouldn't be
there, and when I shock it it does indeed oscillate until powered down. I
guess my point is that just because people don't notice a problem doesn't
mean that everything is OK. I apologize, this is rather nitpicking, and
with your explanation below I understand better. It is rather dumb to
compare the TLC27L4 with the LT1114.
A similar circuit is working quite well in the comEEG prototype.
>However I must admit, that the testing level is quite low
>(on the oscilloscope I could not see oscillations).
>To be sure a more thorough test should be done with a function
>generator that injects a square wave into the control loop.
>
>(any volunteers here with the right equipment ?)
>
>IIRR the important trick in the virtual ground circuit was the
>ESR of the tanatalum C at the output of the op-amp. An ESR of about
>500mOhms provided a negative phase shift of about 90 degrees at the
>0dB point in the bode plot of the loop gain, resulting in a total
>phase margin of about 90 degrees (what is very stable)
I didn't realize the ESR was so huge! That could make a big difference,
especially for an OpAmp that is designed for some level of capacitive
load. Relying on such ESR is definitely component (mostly OpAmp) and
capacitor specific, and possibly also supply voltage and slightly dependant
on the actual output voltage, so it would be necessary to verify the
stability with the final configuration.
I just suggested the additions of the three passives because it is much
easier to take them out than to put them in after the board is built.
Also, even if you are operating within the Cload (linear-tech's terms)
OpAmps you will still have significant overshoot, which would increase any
noise on the virtual ground reference. Those curves are done for
capacitances without the ESR though, I suspect. It was interesting that
the LT1114 data sheet shows up to about 60% to 80% overshoot with 10nF
load. Too bad they don't provide more data for loading the outputs.
About your analysis - sorry for being obtuse, what did you use to get the
90 degrees from at the 0dB crossing? Does that include the op amp output
impedance? This is getting me thinking, if the OpAmp output has a part in
it, that might explain why the same circuit with an LTC27L4 OpAmp seemed
stable until there was a significant perturbation - enough to cause the
gain to drop due to loading the amplifier, which is when it went into
oscillation. Another mystery solved ( to MY satisfaction anyway) !! I
love this group!
Anyway, I recommend putting the parts on the board just in case. If you
want me to test anything, I can pay for shipping.
In the next two weeks I have to spend time sourcing parts, so I will look
at these OpAmps more. Which OpAmp was used in the original design in the
COmEEG?
Thanks, this is getting embarrassing.
>Regards,
>
>Joerg
_________________________________________________________
Do You Yahoo!?
Get your free @yahoo.com address at http://mail.yahoo.com
This archive was generated by hypermail 2.1.4 : 2002-07-27 12:28:40 BST