mike wrote...
I saw only one mention of the voltage to be measured. "Quite high"
is not an adequately descriptive term to determine safety. If the
voltage arcs across the resistor, you're gonna blow up something.
Indeed.
.. 450 ___________________
.. o--/\/--)__________________) scope, 50-ohms
.. gnd --' RG-174 coax
This technique is widely used to create low-cost in-place high-
frequency probes, but for lower voltages. For example, if R is
950 ohms you get a 20:1 divider. For the divider to work well
you need to avoid any capacitive coupling either to the resistor
or to its junction at the coax. For example, the resistor's self
capacitance may be about 0.2pF, which means the "probe" will be
flat in response up to 350MHz or so and then begin to increase in
response as the 0.2pF bypasses the 950 ohms shunting more unwanted
signal to the coax. So for those of us using 500MHz scopes this
technique is useful for 450-ohm 10:1 dividers and to a lesser extent
for 950-ohm 20:1 dividers, but no higher. In practice, we can make
multiple "probes" and solder them to the locations to be measured,
carefully positioning the small 1/4-watt axial-lead input resistor
to minimize capacitive pickup by the resistor's body.
Given a 20:1 attenuation, and a scope's 20 to 50V max direct-input,
this technique is limited to 400- to 1000-volt measurements maximum.
To measure to higher voltages, one must use higher-value resistors,
which must also be rated for higher voltages. But HV resistors are
larger and longer and therefore suffer from severe capacitive pickup
and loss problems, damaging the probe's believability at the higher
frequencies.
.. |<-- signal wire
.. |
.. | 0.2pF
.. | ,-----||-----, ______________
.. o--+-/\/\/\/\/\-+-)______________
.. | -+- -+-
.. |------' '--- gnd
.. |
For example, for a 50k resistor, as discussed above, 0.2pF of self
capacitance creates a 3dB error at only 16MHz. Yes, one could add
small trimmable caps, and with effort calibrate the setup in place,
but now you no longer have a trusty solder-and-measure method, but
a painful poor-performance unreliable technique. Consider, all HV
resistors have the resistive element spread throughout their length.
This will cause considerable trouble, as some capacitively-coupled
signal is lost to ground and other undesired pickup occurs from the
wiring carrying the measured signal. In the region of 5 to 25MHz,
the setup's frequency response will suffer both exaggeration *and*
attenuation, creating a real mess for anyone attempting to correct
the measurement. If you examine a commercial high-frequency high-
voltage probe, you'll discover multiple carefully-placed internal
shields designed to solve these problems. Even so, such probes
are generally limited to modest 50MHz maximum frequencies.
What's a hobbiest to do? I suggest that you consider capacitive
dividers. Remember the troublesome 0.2pF I've been talking about,
what if you had an intentional 0.25 to 1.0pF of "calibrated" input
capacitance, and no resistor at all?
.. | 1pF total = 175pF TVS
.. | \ ,--,____________________,-||-, ,--_/ ---0 scope 25pF
.. o---||--)____________________}----+-+-|_|-+- gnd 1.0M
.. | '--' 10' coax = 300pF '------------'
.. | \ Note the HF Kelvin
.. | surrounding shields wiring on the TVS
To make a 1:500 divider you'd need 500pF of downstream capacitance.
This comes from the scope's input capacitance, plus say ten feet of
small RG-174 coax, plus some additional adjustable capacitance that
you add in a little box at the scope. I know it'll be hard to make
a calibrated 1.0pF, so don't spend much time on that aspect. To
calibrate this beast, use a 10V square wave test signal and adjust
the box's capacitance to match whatever you come up with. You can
include some protection circuitry for the scope in the box as well.
A common silicon TVS will provide sub-nanosecond clamping, and its
rather high capacitance, often a problem when used to protect signal
pathways, is easily included in your 175pF downstream budget. To
be effective for ns-risetime pulses, use the Kelvin wiring shown.
One other thing, you'll use the scope's normal 1.0M input impedance.
Voila! Properly done, you have a true high-voltage divider capable
of accurate high-frequency transient measurements. Yes, it's an
ac-coupled probe, but its low-frequency cutoff (given by 1M and 500pF)
is a low 320Hz, which is quite useful when making measurements on our
common SMPS and PWM systems, with their 20kHz to 1MHz cycle rates.
The 500:1 ratio implies a 10kV capability, but unless you know what
you're doing and properly design and build the input capacitor, you
should not rely on it to handle 10kV without breakdown. But with
creativity and careful construction you may be able to trust it to
several thousand volts.
Here's an idea that occurs to me just now, which I have not tried...
Consider, 0.4" of 50-ohm coax is about 1pF -- this implies a scheme
like the cross-section drawing below will be compact, yet easy to
make, and may work well:
.. | 0.2" gap
.. | 0.6" shield /
.. o-------=========== ============== shield
.. | HV xxxxxxxxxxxxxxxxxxxxxxxx dielectric
.. | xx ===================== center wire
.. | xxxxxxxxxxxxxxxxxxxxxxxx dielectric
.. | =========== =+============= shield
.. | 0.4" overlap |
.. | /
.. | GND ---------------'
The outer 0.6" of shield, which is the input capacitor's HV terminal,
is held in place by an outer layer of heat-shrink tubing. Several
layers will provide good HV insulation to adjacent grounds. If this
probe adds too much radiative area for your HV signal, an additional
ground shield may be extended over the heat-shrink. When making the
HV conducting portion, and the ground shield near it, be sure to make
the edges as rounded as possible. Soldering a large bus wire around
the ends of both shield sections may help reduce the field gradients.
Note I've shown the coax's center wire terminating well inside the
dielectric. After you remove an outer 0.5" of sheild, which can be
used for the HV terminal, press the dielectric back, and snip the
protruding wire as far back as you can. Do this several times to get
it back about 0.2 inches. More is better. You'll be left with about
0.4" of shield-to-center-wire overlap. Seal the center hole with some
dielectric fashioned into an oversized cylindrical plug. Now add the
layers of heat shrink to hold everything in place. A short section of
insulated wire soldered to the outer-shield piece will stick out of
the assembly can be used to solder to your circuit's test point. A
similar short piece of wire soldered to the inner-shield will be your
ground connection, which you should always carefully make. This wire
carries the input capacitor's return current, so it should be short
and arranged with the HV wire to have a low inductance loop.
You can still kill yourself trying to hook it up...depending on your
definition of "quite high". Anything over 42.5V is generally
considered unsafe.
One other thing. It should be clear from my discussion, mentioning
soldering the "probe" leads to your system, etc., that all connections
are to be made with your system power OFF. Always stay well away from
high voltages. Keep one hand behind your back when it's powered up.
Well, this has been a long post, but we've covered some useful ground.
I'll save a copy in my computer for use later. It may come in handy.
