1G resistor

On Sun, 22 Nov 2015 03:02:53 -0800 (PST), Phil Hobbs Gave us:

For a control loop I can understand feed forward. For a simple monitor, however?

I will go look up my old 15kV supply and see what we did. It was a medical supply so should be a pretty accurate monitor circuit.

I wish you people would learn how to use a news client intelligent enough to conform to simple Usenet conventions for line length. Or configure the ones you are using.

News: It is NOT "my old client's fault". It is the NON-conformity of web based news access and basic laziness that no one ever held the idiot developer's feet to the fire for ignoring it.

A slot under the 1G is free, and nearly eliminates the to-board capacitance. The main to-board capacitance left is the 1G's pad on the cold end, which is fine, absorbed into the compensation cap.

Cheers, James

Yup.

Not that I can see.

Cheers, James

Don't unplated slots cost extra? It's another PCB process step.

nce.

h is

don't see unplated costing extra, if the outline is milled you can do the slot at that step

a plated slot would have to be milled before plating so that might be an extra step

-Lasse

Have you been drinking? That dead short on OUT1 through the ACPL is going to be tough on your pulser. And you have some pretty large time constants there. What's the duty cycle profile of your pulses?

On Sun, 22 Nov 2015 07:33:53 -0800, John Larkin Gave us:

Our Au over Ni PCBs are the same price no matter what. They quote on via count and pad count and square inch of PCB. Odd shield layer shapes do not matter as it is simple photo etch. Pads do as the mask changes too. We were even allowed to have a slot on the top layer that was not on any subsequent layer. So a gap slot under an SMD was even possible

Was the slot plated or unplated?

--sp

This uses a low-CTR optocoupler to improve the waveform.

So, you haven't followed the the thread, Fred? Makes you look like the fool you are. As usual.

The downside is unpredictability. I don't know what the step response would look like; the resistor still has distributed capacitance to space.

I don't have good enough numbers to get a believable simulation of this, so I might breadboard. Cutting out the ground plane under the resistor may help. The solder may lift the resistor up enough to help. But a discrete parallel capacitor has to help swamp ugly distributed RC effects.

It's hard to buy HV caps. It looks like the universal (stocked, affordable) part is 10 pF at 1 KV, in 1206. Two or three of those in series gives me 5 or 3.3 pF, which is starting to waste a little energy but should make a pretty flat attenuator.

I can probably do this without a trimpot. Everybody is ragging me about trimpots. The Thevenin impedance of the wiper becomes a problem when the upper capacitor is several pF.

Version 4 SHEET 1 2036 680 WIRE 800 -160 -128 -160 WIRE 1552 -160 864 -160 WIRE -128 80 -128 -160 WIRE -48 80 -128 80 WIRE 112 80 32 80 WIRE 160 80 112 80 WIRE 272 80 240 80 WIRE 320 80 272 80 WIRE 432 80 400 80 WIRE 480 80 432 80 WIRE 592 80 560 80 WIRE 640 80 592 80 WIRE 752 80 720 80 WIRE 800 80 752 80 WIRE 912 80 880 80 WIRE 960 80 912 80 WIRE 1072 80 1040 80 WIRE 1120 80 1072 80 WIRE 1232 80 1200 80 WIRE 1280 80 1232 80 WIRE 1392 80 1360 80 WIRE 1440 80 1392 80 WIRE 1552 80 1552 -160 WIRE 1552 80 1520 80 WIRE 1712 80 1552 80 WIRE 1824 80 1712 80 WIRE -128 176 -128 80 WIRE -32 176 -128 176 WIRE 112 176 112 80 WIRE 112 176 32 176 WIRE 160 176 112 176 WIRE 272 176 272 80 WIRE 272 176 224 176 WIRE 320 176 272 176 WIRE 432 176 432 80 WIRE 432 176 384 176 WIRE 480 176 432 176 WIRE 592 176 592 80 WIRE 592 176 544 176 WIRE 640 176 592 176 WIRE 752 176 752 80 WIRE 752 176 704 176 WIRE 800 176 752 176 WIRE 912 176 912 80 WIRE 912 176 864 176 WIRE 960 176 912 176 WIRE 1072 176 1072 80 WIRE 1072 176 1024 176 WIRE 1120 176 1072 176 WIRE 1232 176 1232 80 WIRE 1232 176 1184 176 WIRE 1280 176 1232 176 WIRE 1392 176 1392 80 WIRE 1392 176 1344 176 WIRE 1440 176 1392 176 WIRE 1552 176 1552 80 WIRE 1552 176 1504 176 WIRE 1712 176 1712 80 WIRE 1824 176 1824 80 WIRE -128 224 -128 176 WIRE 112 224 112 176 WIRE 272 224 272 176 WIRE 432 224 432 176 WIRE 592 224 592 176 WIRE 752 224 752 176 WIRE 912 224 912 176 WIRE 1072 224 1072 176 WIRE 1232 224 1232 176 WIRE 1392 224 1392 176 WIRE 1552 224 1552 176 WIRE -128 336 -128 304 WIRE 112 336 112 288 WIRE 272 336 272 288 WIRE 432 336 432 288 WIRE 592 336 592 288 WIRE 752 336 752 288 WIRE 912 336 912 288 WIRE 1072 336 1072 288 WIRE 1232 336 1232 288 WIRE 1392 336 1392 288 WIRE 1552 336 1552 288 WIRE 1712 336 1712 256 WIRE 1824 336 1824 240 FLAG 112 336 0 FLAG 272 336 0 FLAG 432 336 0 FLAG 592 336 0 FLAG 752 336 0 FLAG 912 336 0 FLAG 1072 336 0 FLAG 1232 336 0 FLAG 1392 336 0 FLAG 1552 336 0 FLAG 1712 336 0 FLAG -128 336 0 FLAG 1824 336 0 SYMBOL cap 96 224 R0 WINDOW 0 -46 23 Left 2 WINDOW 3 -56 54 Left 2 SYMATTR InstName C1 SYMATTR Value {CP} SYMBOL res 48 64 R90 WINDOW 0 -42 56 VBottom 2 WINDOW 3 -37 54 VTop 2 SYMATTR InstName R1 SYMATTR Value 100Meg SYMBOL cap 256 224 R0 WINDOW 0 -46 23 Left 2 WINDOW 3 -56 54 Left 2 SYMATTR InstName C2 SYMATTR Value {CP} SYMBOL res 256 64 R90 WINDOW 0 -42 56 VBottom 2 WINDOW 3 -37 54 VTop 2 SYMATTR InstName R2 SYMATTR Value 100Meg SYMBOL cap 416 224 R0 WINDOW 0 -46 23 Left 2 WINDOW 3 -56 54 Left 2 SYMATTR InstName C3 SYMATTR Value {CP} SYMBOL res 416 64 R90 WINDOW 0 -42 56 VBottom 2 WINDOW 3 -37 54 VTop 2 SYMATTR InstName R3 SYMATTR Value 100Meg SYMBOL cap 576 224 R0 WINDOW 0 -46 23 Left 2 WINDOW 3 -56 54 Left 2 SYMATTR InstName C4 SYMATTR Value {CP} SYMBOL res 576 64 R90 WINDOW 0 -42 56 VBottom 2 WINDOW 3 -37 54 VTop 2 SYMATTR InstName R4 SYMATTR Value 100Meg SYMBOL cap 736 224 R0 WINDOW 0 -46 23 Left 2 WINDOW 3 -56 54 Left 2 SYMATTR InstName C5 SYMATTR Value {CP} SYMBOL res 736 64 R90 WINDOW 0 -42 56 VBottom 2 WINDOW 3 -37 54 VTop 2 SYMATTR InstName R5 SYMATTR Value 100Meg SYMBOL cap 896 224 R0 WINDOW 0 -46 23 Left 2 WINDOW 3 -56 54 Left 2 SYMATTR InstName C6 SYMATTR Value {CP} SYMBOL res 896 64 R90 WINDOW 0 -42 56 VBottom 2 WINDOW 3 -37 54 VTop 2 SYMATTR InstName R6 SYMATTR Value 100Meg SYMBOL cap 1056 224 R0 WINDOW 0 -46 23 Left 2 WINDOW 3 -56 54 Left 2 SYMATTR InstName C7 SYMATTR Value {CP} SYMBOL res 1056 64 R90 WINDOW 0 -42 56 VBottom 2 WINDOW 3 -37 54 VTop 2 SYMATTR InstName R7 SYMATTR Value 100Meg SYMBOL cap 1216 224 R0 WINDOW 0 -46 23 Left 2 WINDOW 3 -56 54 Left 2 SYMATTR InstName C8 SYMATTR Value {CP} SYMBOL res 1216 64 R90 WINDOW 0 -42 56 VBottom 2 WINDOW 3 -37 54 VTop 2 SYMATTR InstName R8 SYMATTR Value 100Meg SYMBOL cap 1376 224 R0 WINDOW 0 -46 23 Left 2 WINDOW 3 -56 54 Left 2 SYMATTR InstName C9 SYMATTR Value {CP} SYMBOL res 1376 64 R90 WINDOW 0 -42 56 VBottom 2 WINDOW 3 -37 54 VTop 2 SYMATTR InstName R9 SYMATTR Value 100Meg SYMBOL cap 1536 224 R0 WINDOW 0 -46 23 Left 2 WINDOW 3 -56 54 Left 2 SYMATTR InstName C10 SYMATTR Value {CP} SYMBOL res 1536 64 R90 WINDOW 0 -42 56 VBottom 2 WINDOW 3 -37 54 VTop 2 SYMATTR InstName R10 SYMATTR Value 100Meg SYMBOL res 1696 160 R0 WINDOW 0 -71 38 Left 2 WINDOW 3 -80 74 Left 2 SYMATTR InstName R11 SYMATTR Value 2Meg SYMBOL voltage -128 208 R0 WINDOW 0 40 112 Left 2 WINDOW 3 -39 181 Left 2 WINDOW 123 0 0 Left 2 WINDOW 39 0 0 Left 2 SYMATTR InstName V1 SYMATTR Value PULSE(0 1000 1u 1u 1u 5m) SYMBOL cap 32 160 R90 WINDOW 0 0 32 VBottom 2 WINDOW 3 32 32 VTop 2 SYMATTR InstName C11 SYMATTR Value {CS} SYMBOL cap 224 160 R90 WINDOW 0 0 32 VBottom 2 WINDOW 3 32 32 VTop 2 SYMATTR InstName C12 SYMATTR Value {CS} SYMBOL cap 384 160 R90 WINDOW 0 0 32 VBottom 2 WINDOW 3 32 32 VTop 2 SYMATTR InstName C13 SYMATTR Value {CS} SYMBOL cap 544 160 R90 WINDOW 0 0 32 VBottom 2 WINDOW 3 32 32 VTop 2 SYMATTR InstName C14 SYMATTR Value {CS} SYMBOL cap 704 160 R90 WINDOW 0 0 32 VBottom 2 WINDOW 3 32 32 VTop 2 SYMATTR InstName C15 SYMATTR Value {CS} SYMBOL cap 864 160 R90 WINDOW 0 0 32 VBottom 2 WINDOW 3 32 32 VTop 2 SYMATTR InstName C16 SYMATTR Value {CS} SYMBOL cap 1024 160 R90 WINDOW 0 0 32 VBottom 2 WINDOW 3 32 32 VTop 2 SYMATTR InstName C17 SYMATTR Value {CS} SYMBOL cap 1184 160 R90 WINDOW 0 0 32 VBottom 2 WINDOW 3 32 32 VTop 2 SYMATTR InstName C18 SYMATTR Value {CS} SYMBOL cap 1344 160 R90 WINDOW 0 0 32 VBottom 2 WINDOW 3 32 32 VTop 2 SYMATTR InstName C19 SYMATTR Value {CS} SYMBOL cap 1504 160 R90 WINDOW 0 0 32 VBottom 2 WINDOW 3 32 32 VTop 2 SYMATTR InstName C20 SYMATTR Value {CS} SYMBOL cap 864 -176 R90 WINDOW 0 -23 62 VBottom 2 WINDOW 3 -51 8 VTop 2 SYMATTR InstName C21 SYMATTR Value 5p SYMBOL cap 1808 176 R0 WINDOW 0 -59 29 Left 2 WINDOW 3 -58 64 Left 2 SYMATTR InstName C22 SYMATTR Value 2.5n TEXT 184 -64 Left 2 !.tran 10m TEXT 664 -40 Left 2 ;2512 1G Resistor with parasitics TEXT 768 -112 Left 2 ;2 x 10p 1KV TEXT 40 -112 Left 2 ;T840 500:1 Output Pickoff JL Nov 22, 2015 TEXT 368 504 Left 2 ;2512 element capacitance to ground plane is

Coffee and milk this morning.

There's no short. The ACLP is a cheap 1 pF capacitor rated for 3KV RMS. I included the capacitance note as a hint to newbies.

The output can go up to 90% duty cycle from a few Hz up to tens of KHz. Rise/fall times in the 5 us range, all of which I want to pick off with reasonable fidelity.

I don't know specifically, but don't see why it's much different than an unplated mounting hole, process-wise.

Cheers, James

"Not made here" syndrome? :-(

I mentioned the proper way to build a divider like this, at the beginning of the thread. As did many others. You all are still trying to convince John, but he's stuck with the "ooh shiny" he's got in his head.

He'll build a prototype, put some pads around it, do more breadboarding/prototyping, and release the thing with whatever components work, maybe with a few bodge wires. Is that listening to the people he's asked for ideas? No, not really. Is that proper design? Meh.

Tim

Well, it's not as though lives are at stake. A proto run is probably 5 bills and a week's time.

And asking for input doesn't mean we don't have to make our own call and live with the results, good or bad.

Seems like a few minutes with a Dremel and a couple of hours with a scope would clear up most of the uncertainty.

Cheers

Phil Hobbs

Dividers are easy. Wideband, precise dividers aren't easy. The distributed RCRC... inside the big 1G resistor makes the step response ugly. Just cutting a hole under the resistor won't magically result in clean DC-to-1MHz step response. I don't want to bet a PCB spin that it will.

I'm not stuck on anything. I've changed my design several times, and Spiced a number of situations.

Kluge wires are exactly what I don't want. Given that femtofarads matter here, it's unlikely that kluge wires could fix a problem.

I admit that I intend to release it with whatever components work.

Is that listening to the people he's

I've listened to several good points, and several bad ones. Drill a hole and pray is not my idea of proper design. Engineering is about numbers.

The safe thing looks like paralleling the 1G resistor with some real capacitance. 5 pF (two 10P 1KV 1206 in series) looks OK. A Spice sim of that still shows a couple per cent of non-flatness on a step, from the distributed RCRC in the resistor; a couple per cent is tolerable. Cutting a hole in the board, or removing a bit of the ground plane, would probably improve that some.

These are 0.5% and COG at 2pF, seems like way better chance for a one-time trim :

I suspect those tolerances are +-0.5 pF, not +-0.5%. But the board will be all surface mount.

The only commonly stocked and affordable caps in this range seem to be

The 1 pF optocoupler is cute, but the waveforms simulate better with 5 pF. Darn, I really wanted that opto on the schematic that way.

Your in-air measurement suggested that lifting the cap off the board gives something very close to the expected value.

To confirm ahead of time you could repeat your on-board measurement as before, but after Dremelizing underneath. Or you can make the real PCB with a slot under the 1G AND a parallel cap, and use whatever works. Or you could mount the 1G on its edge (and get free coal from the assembly crew for Christmas:-).

Cheers, James Arthur

Lifting it up about a quarter inch gives a "prompt" capacitance of about 50 fF, but the following RC charging curve seems to vary shape with height, so there is likely some RCRC going on, as well as the pure shunt C. I don't assume that a hole under the part is the equivalent of it being a quarter inch off the board.

I'd like to design it now and have it work without futzing, or needing a board spin.