AC coupling is easy

Hmm air wire a resistor string... ~10 1/4W through hole?

George H.

Production would murder me in my bed.

--

John Larkin         Highland Technology, Inc 

lunatic fringe electronics

You could blame it on me. :^) googling, there are low capacitance R's.. mostly low values for RF stuff. (one was 10k ohm) What's the C (end to end) for a resistor string in a SIP? You could cut off the intermediate leads.

George h.

You have to relieve the ground plane by several times the dielectric thickness all around, or the fringing will dominate. A 50-mil slot on two sides of the pads should give you almost another factor of 2.

I often use through-hole parts to reduce the capacitance to ground. Air is nice that way.

The nice thing about that is that it would tend to compensate for dielectric constant variations. FR4 is not famous for repeatable epsilon. Maybe put some individual pads nearby, connected to the output, tweak it up with a Dremel, and drill out the pads you don't need. That way you can still ship the Rev A boards, just with a little rework, and you can drill a stack of several boards. Then if you need to roll the rev for other reasons, just get the board house to drill the holes.

Normal 1206es are 200V, with some up to 400V. With that ridiculous waveform, you could conceivably get hot spots in the resistive element due to capacitive coupling within the package.

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

Is it not possible to bias the pulse generator for zero average?

|| .........||.......... -------- -----

+1

Or if customer cannot cope with -42V back bias then give them C-coupled only and let them provide their own bleed resistors :)

piglet

Not easily. Given the available board area, not at all. I don't have any negative supplies.

The DC would only be about +30 average, which I don't think would bother the customer's gadget, but he wants zero. The negative side of the customer's gadget is real ground, so I can't bias that up to fix the offset.

--

John Larkin         Highland Technology, Inc 
picosecond timing   precision measurement  

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

A joke solution: Put 2 sm Rs upright & a 3rd across the top. Add sm Cs on 2 of those. Hand solder :)

NT

There's too much I don't know. The fringing thing, for one.

Maybe the alumina substrate makes the e-field sort of linear across the chip. If the resistor element is high value, it's embedded in that nice uniform field.

How do narrow pulses affect breakdown voltage and clearances?

Maybe I'll drop in a single 2512 and get on with life. Something like this would work:

The bottom side creepage would be around 200 mils with that part. Only

6 volts/mil.

--

John Larkin         Highland Technology, Inc 
picosecond timing   precision measurement  

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

I have a pulse generator that makes sorta gaussian pulses, 1200 volts peak, around 7 ns width, maybe 5 MHz. My customer wants zero DC component into his gadget, so I'll add optional AC coupling for picky people like him.

A 1206 resistor should be OK at 600 volts peak, so I can do this:

Looks easy. It's not.

The PCB pads will have about 0.8 pF of capacitance from point X to ground. If R1 is big, all the voltage appears across it. If it's small, it has giant current spikes on both pulse edges. I can't find values that make the voltage across R1 reasonable without dissipating about a watt in R1. Using three resistors doesn't help.

Cutting away some layer 2 PCB ground plane reduces Cx, to wild guess maybe 0.5 pF. Not good enough.

I could add a cap across R1, about 0.8p, but that takes room and I'd have to add a new part to stock.

I could use PCB traces and pours to make the 0.8p, or to partly guard the x node, but I'm not sure exactly how.

I could order some special high-resistance high-voltage resistors. Some serpentine pattern, not a regular resistor with laser trim slits.

I could use a single stock 1M 1206 resistor and hope it survives. ====================================================

Use an inductor instead of a resistor?

--
Regards, 
Carl Ijames

I thought of that, but it would have to stand the 1200 volt pulse.

--

John Larkin         Highland Technology, Inc 
picosecond timing   precision measurement  

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

I thought of that, but it would have to stand the 1200 volt pulse. =====================================================

I pretty much assumed that you had thought of it but since you didn't say I took a chance :-). I was hoping that with the wide range of possible values and the low current you could find a coil with a good enough voltage rating. Can you use two coils or one coil and one resistor? Or is that right back to the problems with two resistors?

--
Regards, 
Carl Ijames

Can you use an inductor that has better insulation? Wind your own if it is simple enough - some magnet wire can withstand several KV. Here is an article on the subject and they are talking about 1K2 pulses:

Insulation info:

John

Right. The PCB pad capacitance makes any series string bad. If I lower the resistances enough to swamp the pad capacitance, they get too hot!

--

John Larkin         Highland Technology, Inc 
picosecond timing   precision measurement  

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

C coupled, resistor string to ground

So, instead of ground plane under the intermediate pad(s), put a guard electrode, driven by the raw pulse.

If I make the Vx AC potential the same as the generator voltage, I fry the second resistor. The trick would be to have the pulse voltage equally distributed between the two resistors, to equalize power dissipation and peak voltage. To do that, Vx must pulse to half of Vgen. To do that, Vx has to be halfway guarded. I don't have the 3D EM simulation tools, or the time, to do that.

A HV capacitor in parallel with R1 would work. Of course, its pads increase the capacitance at point X.

--

John Larkin         Highland Technology, Inc 
picosecond timing   precision measurement  

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

You didn't like the dial-a-pad idea?

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC / Hobbs ElectroOptics 
Optics, Electro-optics, Photonics, Analog Electronics 
Briarcliff Manor NY 10510 

http://electrooptical.net 
http://hobbs-eo.com

Kinda labor intensive. I'm working on rev C of the pulser now, which is embarassing enough. Rev D would be mortifying.

The expensive 3KV resistor is probably the way to go.

Revs A and B both had pulse monitor output connectors, which were

1000:1 or later 100:1 scope pickoffs, to monitor the pulse. Both were terrible. Same sort of problem as the AC coupling. 100:1 scope probes overheat when I look at the HV pulse, so it's not easy for them either.

My brilliant solution for rev C is to delete the monitor output.

--

John Larkin         Highland Technology, Inc 
picosecond timing   precision measurement  

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

More, or just as, importantly: it needs to be physically small, so as not to have complex high order responses (an inductor is a spool of wire, aka some combination of transmission lines) that dirty up the pulse waveform; and it needs to be air cored, or relatively large so as to avoid saturation of the core material.

Tim

--
Seven Transistor Labs, LLC 
Electrical Engineering Consultation and Design 
Website: https://www.seventransistorlabs.com/