[Neuroscience] Re: Fast drug application 'liquid filament' help!
(by jonesmat from physiology.wisc.edu)
Tue Jul 31 00:09:44 EST 2007
On Jul 25, 8:57 am, gtswan... from gmail.com wrote:
> Hi Bill,
> We use Flavor B almost exactly as described by Matt. It is a very
> cheap and workable solution for fast application, if you can manage to
> get the bimorphs in the first place. I have just a few points to add
> to Matt's very nice description. We always solder a resistor and
> potentiometer in line with the leads to the bimorph, which reduces
> ringing and provides additional control over the speed of the
> translation. In our lab, we typically lift whole cells into the
> laminar stream instead of patches. With small cells, like transfected
> HEKs, the 10-90% rise times of AMPA receptor currents are typically ~1
> I've been using my first bimorph for nearly 10 years now. They are
> quite durable. Sadly, the expensive manipulators and all the other
> associated equipment are not nearly as reliable.
Good suggestions. We also usually put something between the power
supply and the bimorph. For the last seven years we've just been using
a capacitor in parallel with the bimorph (i.e., directly between the +
and - poles of the power supply). We've tried numerous configurations
of capacitors and pots, in series & parallel. All have worked to
greater or lesser degrees, and the single cap is the one that ended up
working well enough and being simplest. It may matter whether you're
using a constant voltage or constant current supply (we've used
constant current for the last seven years, so maybe the single cap is
best for that configuration?).
To expound a bit on this issue, the whole point is to put some sort of
an RC filter in line with the source, so as NOT to give a perfectly
square voltage/current step. The problem is that the piezo and pipes
are physical semi-rigid objects, so they have an intrinsic resonant
frequency. Applying energy at that frequency will make them vibrate
strongly at THAT freq, regardless of the desired movement step. An
ideal square step contains ALL frequencies (i.e., its Fourier
transform is white), including the resonant freq of the bimorph/pipes,
and therefore a true square step will almost certainly cause the
"ringing" that I mentioned earlier.
So smoothing the step is a good way to avoid much of the ringing (but
requires some fiddling, so as not to smooth it so much that it slows
down the exchange time). Another excellent way of doing this is to
filter the command pulse you give from your acquisition software. This
is actually a lot more flexible, because you can shape it however you
want, rather than as a single RC exponential (which also has an abrupt
starting edge, so isn't perfect). However, fiddling with the waveforms
in software turns out to be just as much of a pain as fiddling with
resistors and caps in hardware, so we typically just use the cap.
Another thing that we've found makes a HUGE difference is to "etch"
the inside of the flowpipes with Hydrofluoric Acid, so that the glass
septa separating the pipes is as thin as humanly possible, without
actually dissolving completely. This has two major benefits. First, it
makes the solution interface very thin, so that fast exchanges are
easier with smaller movements. Second, it also helps reduce "ringing",
for reasons which are still a mystery to me. Maybe because there's
less of a turbulent zone between barrels, I don't know.
Anyway, the "etching" thing is a whole saga unto itself. I have
detailed notes about our method, if anyone's interested. It is a PAIN
IN THE ASS, but is often worth it.
More information about the Neur-sci