TDR Logic, mental lapse

The approximations could be breaking down, or I could be misremembering. I could have a squint in my reference books, but they're at the lab and I'm at home.

(I'm in the middle of writing an expert report on rotating lidars for vehicles, and could use a break.)

Cheers

Phil Hobbs

Resonances can also get you much higher or lower Z, of course. I'd look it up in Kraus or Balanis, but as I say they're at the lab.

Cheers

Phil Hobbs

The image method is trickier in the electrodynamic case, but for spacings

Yes, but with separate spacers. It's commonly known as "ladder line".

Cheers

Phil Hobbs

Do vehicle lidars measure the time delay of one echo per shot? They could maybe digitize the entire return signal and form a more 3D picture of what's out there. Lotta DSP.

Is there any height information?

If you connect both wires of a twisted pair, you get the even mode impedances, which are different from the differential-signal odd mode impedances. In one case, the wires help each other and in the other case they fight.

I have a Spice model of a shielded twisted pair, and it's messy. It demonstrated the hazards of bootstrapping a photodiode over a longish twisted pair.

I've never done that. I'd be concerned that the prop delay would be different top and bottom, so I'd get ugly differential pulses at the end. Eventually the bottom one has to via up to the parts on top.

As the two conductors of a differential pair get farther apart, they become two independent single-ended lines. The odd and even mode impedances become equal.

We try to keep diff pairs equal delay for fast stuff. It's a little tricky to turn corners, if picoseconds matter.

There is a rarely-used form of stripline

gnd =========================

sig+ -----

sig- -----

gnd =========================

which Saturn knows about.

and

gnd =========================

sig+ ----- ----- sig-

gnd =========================

which it does too, somewhere else.

The open wire telegraph and telephone line impedance might have been close to 600 ohms (thin wire, large spacing).

Wires behave like well behaving transmission lines only when the wires are at least a significant fraction of wavelength long.

In baseband telephony with maximum audio frequency of 3.4 kHz, the free space wavelength is in the order of 100 km. Typical subscriber lines are only a few kilometers long, so only a small fraction of the wavelength.

If you try to measure the line impedance at such small fractional wavelength, it is different from calculated impedance. The impedance remains constant as calculated when the line is several wavelengths long (i,e, at higher frequencies).

To clarify that, if a transmission line has ohmic loss and you measure the sine-wave impedance at one end of an infinite run, it will increase as the test frequency decreases.

An ohmmeter or TDR will show an ever-increasing curve vs time.

I think that's right.

Yup. The longer-ranged ones make 5-10 ns pulses by discharging a few-nanofarad 0402 cap through one of those triple-stacked Osram lasers, usually using a chip-scale GaN FET as you'd expect.

My measurements and simulations indicate that the total inductance of that loop is less than 400 pH, about evenly split between the source and drain circuits. You still need 40-50V to get the dI/dt fast enough!

The most common kind (Velodyne, Waymo, Uber, Hesai, etc.) uses an array of TX/RX pairs arranged at various elevation angles. The separation is usually 0.3 degrees or so near the horizontal, and a bit more further up or down. The Velodyne one's elevation range is about 27 degrees. Since you know the range, azimuth, and elevation for each shot, you get a full

They're mounted inside a housing that rotates at around 300-700 RPM. The measurement rate is about 1 Mpel/s, so that's about 100,000 points per cloud.

Cheers

Phil Hobbs

Yeah OK.. I've never tried to drive twisted pair fast. I'm not sure I want to think about it inside a gnd shield. Jeesh, I'm mostly clueless with coax. :^)

Oh boy... how long and how fast? I did a driven shield thing on cable down ~2 feet of probe and ~1MHz max freq. It seemed to work fine.

Sure... it's only a silly idea. (you could try some added length on top to balance the via.. oh and then you've got to put the receiver on the bottom :^) Oh, the other bad thing about the idea is it radiates all over. (none of the field lines go to ground. ideally.)

Huh.. is that because they get closer to other grounds than to each other?

I had a transmission lines course in EE undergrad. (Henry Neubauer) But all I remember is ladder networks. It would have been great (for me) if there was some lab. I could have a lot of fun with a spectrum analyzer, tracking gen and coax stubs/ loads.

SA and TG are ~$3k (from rigol ~1.5 GHz) So if you bang on stub tuners, can you see 'em in the reflection? Surely, how well, what's lost?

Thanks. I've never done stripline... at some point you need the right tools/kit.

I'm still just thrilled with the idea of differential signals and the surface of zero potential they create.

(It gives clarity to these balun (balanced- unbalanced) transformers that I've used with some confusion. The TV antenna 300 ohm twin-line to 75 ohm coax for instance.)

George H.

Say Phil (if you're allowed) can you say anything about the detectors? are they APD's or SPAD's? Who's waiting for the consumer app to bring the price of spads down.

George h.

What is the "surge impedance" as opposed to any other impedance?

ut the

I'm not clear what you mean by "split the wire in the middle". Are you say ing you cut the wire in half, attached one end to the signal and the other end to the ground? Are the two new wires in proximity?

Varies. Normal APDs have better dynamic range than SPADs, especially if you want amplitude measurement. With a SPAD, you measure only the toe of a strong return signal.

Well, you can get those nice On Semi MPPCs for $40, which isn't so bad. They're dramatically better than SPADs for most things IMO, unless you've got some super swoopy quench circuit.

The On Semi ones with the 10-um pixels have a reset time of about 5 ns, which is pretty stellar. OTOH their fill factors aren't that great.

Cheers

Phil Hobbs

This is crazy:

Under $2 in quantity.

Oh I guess my dream is to do entanglement cheap. ~404 nm laser to 808.. or something like that. photon counting in the NIR.

George H.

Looking for precise time correlations with very small, slowly-varying intensities is a good application for SPADs.

A few feet of shielded pair between the pd and the amp, wanting a few ns rise time. Too much delay. The customer wanted bootstrapping and it made no sense. They'd never heard of bootstrapping until I did it for them on another system.

Windowless EUV photodiode.

Basically the e-field coupling gets too weak. One wire sees the entire universe as ground and doesn't notice the opposite-polarity voltage on its distant mate. Imagine the virtual ground plane getting very far away from the wire, many diameters. That approaches free space.

TDR is great fun. Run your finger down a trace and watch the bump move on the screen. Locate the vias by touch. Electronics can be very tactile.

We have a giant 50 pound-ish spectrum analyzer klunker. I'm thinking of getting a small Rigol or something that would fit on my bench.

It's just another name for the characteristic impedance of a transmission line: