Meander line

Hi,

Depending on where you have space, it might be best to wind up some micro-coax and solder center lead and ground down on both ends. Should at least minimize crosstalk.

Thanks. I agree that would work fine, but this is a high-volume, very low cost system, so I can't afford it.

Cheers

Phil Hobbs

Den torsdag den 27. oktober 2016 kl. 23.48.50 UTC+2 skrev Phil Hobbs:

how much delay you need?

Sorry, but I missed the post about that project.

Assuming it's a fast sampler in the hundreds of psec and delay doesn't mean several seconds I'd consider adding a couple of layers next to your ground plane (which should usually be layer 2). So the next new layer #3 would be your meander trace and #4 another gound plane. If in dire need to save every penny you can also move the power plane to #4 and then sandwich the meander trace between GND and PWR. A well bypassed power plane is like ground from an RF perspective. Make sure vias are too close by because they can cause echoes on really fast stuff.

This is 430 ps delay one-way, theoretically 50 ohms. Risetime is 95 ps.

It occupies about 0.18 square inches. It could be tighter, but you'd have to worry about coupling between the loops, which would cause pre-shoot.

How much delay do you need? Delay trashes risetime.

Can you put the delay before the pulse generator? That could be done with LCs, lots of delay in a small area. How about running through a single/dual/triple TinyLogic gate to get a few ns?

Sonnet Lite is pretty good to sim this sort of stuff. It's free.

Two layers is interesting.

You could run GND lines into the channels and via them to GND every 0.100".

Not necessarily, depends on the board material.

It doesn't really work if you must pulse-echo targets that are physically really close to the board. You'd either need a delay between pulser and sampler or somewhere between sampler and target (like an extra piece of coax). Sampler and target is often better because if the pulser is a unipolar diode solution it'll issue a big fat pulse all by itself.

I had a hard time talking a client out of two-layer for a 5GHz pulser application. But once we did that resulted in a major performance boost and the cost increase for multilayer is not substantial these days. It was like as if someone had turned on the high-beams.

Phil, you could talk to these guys but chance are that going to 4-layer is cheaper:

On a a two layer board, you are restricted to microstrip, which is dispersi ve (and delay does mess up rise-time, as John Larkin says).

If you can got to a four layer board you can bury a length of strip-line, w hich is symmetrical and non-dispersive, like good quality coax.

You may need a better substrate than FR4-epoxy bonded glass-fibre, dependin g on the frequencies involved. When the wave-length gets down to the glass- fibre cloth mesh spacing things start going wrong. Rogers have substrates d esigned for microwave frequencies, but they aren't cheap.

It's the SPICE model thread from yesterday.

Cheers

Phil Hobbs

Found it. I did one of those pulse-echo deals early last year. That was real fun. I probably could have done it without SPICE but it would have taken a lot longer.

The only problem was that my fastest scope only goes to 1GHz and I had to use the pulse-echo circuit itself together with hard targets so it could kind of measure its own performance. That's how we always did it in ultrasound.

"Some"?

The ns/mm^2 of two layer board is pitiful, even with a lot of crosstalk / dispersion being tolerable.

You're better off using something with wire: twisted pair, a proper delay line component (hah, and blow your entire budget on one part), or a common mode choke or something.

"Data line" CMCs, of modest values, look suspiciously like old-timey wound delay lines. Some of them have plots of diff mode vs. freq. that look like they have suspiciously good TL properties (i.e., the impedance curve has several harmonic peaks and troughs).

Tim

-- Seven Transistor Labs, LLC Electrical Engineering Consultation and Contract Design Website:

... as opposed to the 'lumped constant' approximate delay lines that come in expensive little packages. A couple of RJ-45 jacks and a meter of cat-6 cable would make a fine 60 ohm (two pairs in parallel) delay line; the extra pairs could either double your effective length, or extend the delay.

It's the frequency response of the dielectric that is bothersome at UHF, where the plastic in cat-6 will disappoint. PCB materials like polyimide (Kapton) flexible circuit material should work well, though, and will be less expensive than teflon. Thick-ish polyimide meander circuits should be well-behaved to the few-GHZ range:

If you have a narrowband signal, of course, any transmission line is a good inductor for Z mismatches where the line is higher Z than the drive, and a good capacitor if the line is lower Z.

Would sending the signal around the perimeter of the pcb, (with some ground guards) be a bad idea?

George H.