1G resistor

You should see that rise when you have 1200v on the part.

Cheers

It's a linear network, so the compensation is independent of voltage. So I need to add capacitance across the 1G resistor, or I have to make a higher ratio divider, like 5000:1 or something, to get the lower compensation cap up to a realistic value.

I guess if the resistor is soldered to pads on the PCB, the solder will lift it up a bit and the equivalent shunt C will increase. It's a very sensitive function of the resistor height above the board. I could cut out some ground plane, too.

I understand, but C= q/V and what was your test voltage? 1200v?

Cheers

I don't know how well behaved and reproducible 1G precision resistors are these days but the 10G and 100G we used in Faraday amplifiers were highly batch dependent. The most interesting feature was that in addition to capacitance they also had a networked battery storage like characteristic that cannot be modelled as a single time constant.

The most complex part of the whole thing was compensating the resistor so that when a 10v step signal was applied the settling time to full ppm accuracy was under a second. You will see similar things at 1G but on a shorter timescale. If all you want is some indication it isn't a problem but if you want the waveform about right it could be.

Basically drive it hard with a voltage step function for different lengths of time and watch how the decay curve goes on a scope.

Does it have to be so high? 0.1G and these quirks become ignorable.

--
Regards, 
Martin Brown

My experiment was done with a 10 volt pulse. The real thing will be pulses from near zero to +1200 or so.

I'm assuming that the 1G thickfilm resistor is linear, which I expect it to be. It will probably have a small voltage coefficient and mediocre tempco, not a problem here since this is just a monitor pickoff.

Next I measured shunt capacitance, glued down to a PCB with a layer-2

Interesting. I'd expect the middle part of the resistor to look like an RC transmission line. You can buy 1 Tohm 5% and 10T 10% radial lead resistors from Mouser, plus I have some in my drawer so I'll have to investigate. A good excuse to use my Keithley 405 micro-microammeter, which has a 300 fA full scale range. (Of course it takes a couple of hours warmup to get down there, but it's pretty good for 1965.)

The PCB fringing is pretty well bound to dominate the TC unless there's a slot. It's not that small, for a start, and it has a gigantic tempco.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC 
Optics, Electro-optics, Photonics, Analog Electronics 

160 North State Road #203 
Briarcliff Manor NY 10510 

hobbs at electrooptical dot net 
http://electrooptical.net

2% or so accuracy would be plenty for my application. We want the customer to be able to check that we are indeed making output pulses. The monitor output isn't even a customer requirement, it's a freebie.

PPM measurement at these voltages would be a whole nother problem.

I've got samples of 0805 resistors up to 1T ohms, and they seem to work. They get hard to measure up there.

My pulser has a DC-restore sort of output circuit to add a 0 to 250 volt baseline bias to the 0-1200 volt pulse. The optics guys want that. The ultimate load is a small capacitance, a PLZT gadget. The baseline injector droops if it sees much resistance to ground, hence the 1G pickoff resistor.

I just want to get the pulse response flat. I guess I'll have to add a couple pF of real capacitance across the 1G and not rely on the resistor or the PCB capacitance. A modestly lower value pickoff resistor will still need AC compensation.

I was shocked at the low value of shunt capacitance the 2512 shows, and how radically it depends on proximity to the PCB ground plane. Lifted a quarter inch off the deck, it increases from 0.006 to over

0.05 pF.

When I play this game, I remove all ground plane from under the part. And you can make little HV capacitors from intentional PCB copper in the blank areas.

--
 Thanks, 
    - Win

I thought about using FR4 as the capacitance. That would add vias, which is OK but not great at 1200 volts. But the capacitance would be poorly predictable and would probably vary between board batches. I'll probably parallel the resistor with a string (three or four?) of series low-pF 1206 caps. Nuisance.

At least I did the experiment and didn't assume that the free-air shunt capacitance would be enough.

So would I, but the gross eyeballed waveforms sure look like a 1G resistor with a small shunt C. A small jump and then the exponential charge.

Some other notes:

Glued down, I see about 0.006 pF.

Lifted up a couple tenths of an inch, 0.032 pF

Flat down again, with a grounded shield (screwdriver) over the resistor, about 0.003 pF. So a lot of the C is in the air hopping over the resistor. You can buy omega-shaped metal things, surface mount test points, that might be used to shield a resistor on top.

Rotated 90 degrees, resistor up on edge, I see 0.018 pF. But we'd have to hand-solder it that way; our pick-and-place won't do that.

You can buy 1 Tohm 5% and 10T 10% radial lead

An unplated slot would probably increase the capacitance. The ground plane shorts out the PCB and resistor bulk capacitance.

--

John Larkin         Highland Technology, Inc 
picosecond timing   precision measurement  

jlarkin att highlandtechnology dott com 
http://www.highlandtechnology.com

Probably so. The question is whether the 1000 ppm/K of the remaining FR4 contribution is a problem or not. If not, terrific. 2512 is probably big enough that it's OK--with ground on L2, the aspect ratio is pretty large.

But if you don't care about loading the 1200V, why does the resistor have to be so big?

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC 
Optics, Electro-optics, Photonics, Analog Electronics 

160 North State Road #203 
Briarcliff Manor NY 10510 

hobbs at electrooptical dot net 
http://electrooptical.net

I do care. The baseline offset voltage is added to the grounded pulse generator by a second-order AC-coupled baseline-restore sort of circuit that droops if it's loaded.

The customer load is about 50 pF, pure capacitance. The major droop contributor is the pickoff resistor.

--

John Larkin         Highland Technology, Inc 
picosecond timing   precision measurement  

jlarkin att highlandtechnology dott com 
http://www.highlandtechnology.com

Ah, so capacitance to ground is OK, I see. In that case, a bit of Cu-clad Kapton over top would knock it down further, and you could keep your creepage distance.

However, using a slot would get you a nice stable 0.05 pF, which you could trim out with a 2k pot and a 22 pF NPO hung on the output amp.

If the amp is inverting, it'll cause fewer stability worries, but it should be OK even noninverting since you're tapping it down to a gain of less than 1.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC 
Optics, Electro-optics, Photonics, Analog Electronics 

160 North State Road #203 
Briarcliff Manor NY 10510 

hobbs at electrooptical dot net 
http://electrooptical.net

I already use a bunch of ACPL227 optoisolators on the board. It's rated for 3KV RMS. The capacitance, from 1-2-3-4 to 5-6-7-8, is 1.1 pF.

So it looks like a good capacitor to slap across my 1G resistor. More C than I'd prefer, but workable.

I was thinking about using the FR4 as the cap. Cut a window in the ground plane under the resistor highside pad, and run a copper pour on layer 3 or 4 from the lowside, to peek through the window in the ground. But how to estimate that capacitance? Dremel?

0.062 thick FR4 runs about 15 pF/sq inch side to side. But this is more complex.

--

John Larkin         Highland Technology, Inc 
picosecond timing   precision measurement  

jlarkin att highlandtechnology dott com 
http://www.highlandtechnology.com

beautiful "electrometer"!

What was the composition and construction of such resistors?

Yup, fabulous instrument. They show up on ebay.

The manual says to replace the high-range resistors every 6 months. I bet mine have never been replaced. It seems pretty accurate.

On the lower current/higher ohms ranges, I can shuffle my feet on the carpet from 10 feet away and wiggle the meter.

--

John Larkin         Highland Technology, Inc 
picosecond timing   precision measurement  

jlarkin att highlandtechnology dott com 
http://www.highlandtechnology.com

It says check high range ones every 6 months- not replace. No doubt you're are done aging and have settled in.

Page 34 has a fun list of original parts supplies. Some are still around.

ISTM John's groundplane *adds* a distributed capacitance to ground.

.--Cs-.--Cs-.--Cs-.--Cs-.--Cs-. | | | | | | | | | | | | Cg Cg Cg Cg Cg Cg | | | | | | === === === === === ===

If the to-ground shunt capacitances Cg dominated, the result would be lengthy R-C delays.

The total capacitance Cg would be given as a 2510 flat plate over a thickness of Al2O3, then a groundplane. If the groundplane is under FR-4, we have...

6.5mm .-.----------.-. / / / /| / / / / / 2.5mm / / / / / .-'----------'-. / / '-'----------'-' /

----------------------' / /////// FR-4 /////////| /

-----------------------' =====================