Re: Water probe: improvements

["Tesla list" <tesla@xxxxxxxxxx>]

Original poster: "Bob (R.A.) Jones" <a1accounting@xxxxxxxxxxxxx>

Hi Marco,

 More thoughts. Perhaps with that 100pf input C and output impedance are
your first corner frequency as a low pass.
You may have higher corner frequencies effects too but lets assume their
above your bandwidth requirements or we will deal with them later.
First reduce the C, with lower C diodes or connect the other side of the
diodes to a unity gain version of the input voltage note this can cause
instability.
Best of all dump them if you think you can.  Calculate the corner frequency
of what you have left and add that as a lead term to the amp.
The next corner frequency probably a lead will now be much closer.  If its
within range of your bandwidth requirement just add the hf lag at the same
frequency. You may need the hf lag
to ensure stability of the opamp anyway. It levels off the gain to give you
flat hf boost and can be selected experimentally as described in one of my
other posts.

Again it assumes quadratic terms caused by L are well out side your
bandwidth and or have a very low Q because of the high R. I am guessing but
they are probably very high frequency wave resonances of the tube probably
too damped to have much effect.

I hope this makes sense to you.

Bob

> ----- Original Message -----
> From: "Bob (R.A.) Jones" <a1accounting@xxxxxxxxxxxxx>
> To: "Tesla list" <tesla@xxxxxxxxxx>
> Sent: Monday, January 03, 2005 11:55 AM
> Subject: Re: Water probe: improvements
>
>
> > Hi Marco,
> >
> > I just reread your post and realized you had some compensation for input
C
> > of 100pf that's huge figure to compensate for and the amp may have run
out
> > hf gain hence the wrong edge height.
> >
> > I don't know your circuit but if the comp ration is not correct that
could
> > be the source of your problem too similar if not the same as above.
> >
> > Bob
> > ----- Original Message -----
> > From: "Bob (R.A.) Jones" <a1accounting@xxxxxxxxxxxxx>
> > To: "Tesla list" <tesla@xxxxxxxxxx>
> > Sent: Monday, January 03, 2005 11:42 AM
> > Subject: Re: Water probe: improvements
> >
> >
> > > Hi Marco,
> > >
> > > I checked it is correct,  Assuming that the problem is C and R ratio
> > > differences.
> > > No simple way to determine what the corner frequency of the lead lag
> terms
> > > are.
> > >
> > > You could do it experimentally and adjust the hf boost until the
initial
> > > rise height is correct then fiddle with the corner frequencies until
you
> > get
> > > the best flat top after the rise.  There will be quadratic terms
which
> > > inevitably you can not compensate perfectly  for with 1st order lead
> lags
> > > but they may be small or at a much higher frequency.
> > >
> > > Can you do a complex FFT on the step response to get the frequency
> > response?
> > > The signal does not look very clean for it.
> > >
> > > Bob
> > >
> > >
> > > ----- Original Message -----
> > > From: "Bob (R.A.) Jones" <a1accounting@xxxxxxxxxxxxx>
> > > To: "Tesla list" <tesla@xxxxxxxxxx>
> > > Sent: Monday, January 03, 2005 10:58 AM
> > > Subject: Re: Water probe: improvements
> > >
> > >
> > > > Hi Marco,
> > > >
> > > > What's the hash on the edge?  It looks like you don't have something
> > > > correctly terminated or you have current loops on the grounds.
> > > > Is your scope a diff input or grounded/earthed reference.
> > > >
> > > > Can you expand on what you mean by dielectric polarization?
> > > >
> > > > The last trace looks like the C division is lower than the  R
> division.
> > > i.e.
> > > > if you have a C divider and R divider in parallel and the C divider
> has
> > a
> > > > lower ration than R divider, then the rising edge will rise quickly
to
> a
> > > > particular level determine by C division ratio followed by a slower
> rise
> > > to
> > > > R division ratio. You will be able to compensate for this in the
video
> > amp
> > > > with the correct lead lag terms (hf boost) I think. I would need to
> > check
> > > > that analytically to be certain.
> > > >
> > > > Bob
> > > >
> > > >
> > > > ----- Original Message -----
> > > > From: "Tesla list" <tesla@xxxxxxxxxx>
> > > > To: <tesla@xxxxxxxxxx>
> > > > Sent: Monday, January 03, 2005 6:42 AM
> > > > Subject: Water probe: improvements
> > > >
> > > >
> > > > > Original poster: "Denicolai, Marco" <Marco.Denicolai@xxxxxxxxxxx>
> > > > >
> > > > > Hello all,
> > > > >
> > > > > During the past days I played with the smaller water probe
prototype
> > (a
> > > > > table top, 70 cm tall) I have at home. The step response time had
a
> > > > > decay as bad as the one of the bigger probe at the university
(about
> > 50%
> > > > > straight, then a 20 us decay).
> > > > >
> > > > > It was nice to see how sliding my hand along the water column I
> could
> > > > > either increase the decay time or reduce it, up to get some
negative
> > > > > peaking from overcompensation. But luckly I decided to first try
> > > > > something else than brutal force.
> > > > >
> > > > > 99% of the decay was actually due to the coax cable impedance. I
> built
> > a
> > > > > small video amplifier fed by two 9V batteries, including an
EL2244,
> > > > > protection schottky diodes, input AC coupling and compensation
> > > > > possibility. Output is balanced for a 50 ohm coax. Gain is now set
> to
> > 1.
> > > > >
> > > > > And voila'. The step response time went down to 20 ns
(nanoseconds).
> > > > > That's the fall time of the fastest 30V pulse I could generate at
> > home.
> > > > > This corresponds to a probe bandwidth of 0.36/20E-6 = 18 MHz. I
had
> to
> > > > > compensate a little bit with a trimmer to balance the capacitance
of
> > the
> > > > > schottky diodes at the input (about 100 pf).
> > > > >
> > > > > Yesterday I went to the university lab with my mighty videoamp,
full
> > of
> > > > > hope and...
> > > > >
> > > > > Well, the videoamp improved quite much the step response, but
> nothing
> > as
> > > > > radical as at home. I also tried sliding my hand along the water
> > column
> > > > > (now 2.7 m long, hanging from the roof at 6 m height) but I
couldn't
> > get
> > > > > the "overcompensation" with negative peaking.
> > > > >
> > > > > The benefit of the videoamp was also that now I could locate the
> > > > > oscilloscope on the floor, far, at the end of several meters of
coax
> > > > > without a noticeable signal degradation. And that's what I did. I
> > > > > noticed that removing the elevator from the TC vicinance bettered
> the
> > > > > response. And also rerouting the probe grounding horizontally
> straight
> > > > > to the wall helped. Then I moved the whole pulse generator box 4
> > meters
> > > > > far from the coil and, still, an improvement.
> > > > >
> > > > > To make it brief, I ended up with a response straight to 75% and
> with
> > a
> > > > > small "bump" lasting for about 10 us. With all the "stuff" near
the
> TC
> > > > > the bump whould be a decay from 60% lasting for 15 ns. So, once
the
> > > > > macroscopic problems from the coax are eliminated I bumped into
the
> > > > > isotropic capacitance problems. After the TC was freed from nearby
> > > > > objects I was left with...what?
> > > > >
> > > > > I think I figured it out this morning. It must be the dielectric
> > (water)
> > > > > polarization.
> > > > > As the step is generated from feeding 250V to the probe for 0.3 s
> and
> > > > > then shorting it to gnd, I think the DC component of that is
> > responsible
> > > > > for the little bump I wasn't able to counterbalance.
> > > > >
> > > > > Tomorrow we'll add a 1uF capacitor in serier to the pulse
generator
> > and
> > > > > I really hope to get rid of that bump too.
> > > > >
> > > > > I have scanned a printout of the probe step response. It's at the
> > bottom
> > > > > of the page at:
> > > > >
> > > > > http://www.iki.fi/dncmrc/work/hv_divider.htm
> > > > >
> > > > >
> > > > > Best Regards
> > > > >
> > > > >
> > > >
> > >
> >
>
>