RE: Water probe: improvements II

["Tesla list" <tesla@xxxxxxxxxx>]

Original poster: "Denicolai, Marco" <Marco.Denicolai@xxxxxxxxxxx>

Hi Bob, all,

The videoamp (closed loop) gain is 2 and drops to 0 dB at about 56 MHz.
It doesn't rise 20dB/decade going down with frequency. The effect of the
100 pF added by the diodes is compensated by adding some resistance is
parallel with them. The step response "straightens" back as it was
without the diodes.

Some update about the probe.

1. I tried removing the protection diodes. No effect on noise or
response. Put them back.
2. I added a 390 ohm carbon resistor in series between the pulse
generator and the probe. That removed pretty well the hash (see updated
pictures on my web site). If I used a 600 ohm they would completely
disapper (not shown). So we don't need to bother about them anymore.
It's just the inductance of the driving wire resonating with the total
probe capacitance.
3. Now I employ the final configuration as for the TC running. That is,
voltage to probe comes from below, pulse generator grounded to the
floor, probe grounded to the wall, coax cable goes horizontally to the
wall (along with the grounding band), maximum clearance available.
Everything still looks good.
4. Signal got much cleaner by using thick Triax shielded cable. That is
going to be the one used in the final installation.

The last pictures on

http://www.iki.fi/dncmrc/work/hv_divider.htm

are shot with the best configuration as described above. The little 10
us bump is very hard to be removed. Pulse generator AC coupling didn't
help, forget about my "polarization nosense" :)
The bump simulated well as a 2.5 dB attenuation near 60 kHz. I simulated
a symmetrical gain increase of 2.5 dB for my videoamp. Tested it at home
it really reduced the bump in my little probe. Tested it with the big
probe: no effect.

Anybody any idea about what that bump is?

Anyway, you can see from the last pictures a fall time of about 50 ns.
That makes 0.36/50E-9 = 7.2 MHz bandwidth. As I was shooting for 1 MHz,
I am satisfied.

We are now trying to get the step response data to a CVS file. Possibly
today. The idea is to try to get the frequency response of the probe.
The next problem is to get more attenuation so that at full operating
voltage the videoamp will not get overdriven.

Best Regards

>
>  > Original poster: "Denicolai, Marco"
> <Marco.Denicolai@xxxxxxxxxxx>  >  > Hi Bob,  >  >  > 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.
>  >
>  > The original capacitance of the lower voltage arm of the
> divider is  > about 500 pF. The capacitance of the schottky
> diodes I added (one  > between input and V+, the other
> between input and V-) is 50 pF. So two  > of them makes 100 pF.
>
> Oh I did not realize the out put impedance was so high
> without the diodes.
> So removing them will only have a 20% change in your frquency
> response.
> 500pF does seem very high even given the high dielectric
> constant of water.
> The out put resistance is what?. At what frequency is the
> outputs C's impedance equal to the output resistance.
> This is the corner frequency of the hf drop off of your probe or lag.
> Lets say that's 0.056MHz so at 0.56MHz it will be 20dB down
> at 5.6MHz 40dB down
>
>  > I am really not confident the amplifier would last a
> second without them  > if the TC is turned on.
>
> They usually have internal protection diodes. If the output R
> of your probe is high you may not exceed the input current spec.
> More problematic is after such an event the opamp/probe
> output C my be very slow to recover so if your relying on
> that for clipping spikes during normal operation the spikes
> could get turned in to pulses with long tails.  An effect I
> belive in Terry's opto current probe.
>  >
>  > The opamp gain (with my simulations) should go to 0 dB at
> -100 degrees  > at 56 MHz which means its 20dB at 5.6MHz and
> 40dB at 0.56Mhz. assuming a first order roll off i.e. a lag
> so you would just have enough gain if your probe start to
> roll off at 0.056MHz .
>
> Bob
> .
>
>