OK I've acquired some parts and been able to complete an experiment. I suppose the Tek and HP parts aren't in high demand and that's why I was able to acquire them in short order.
Tek P6150 x10 tip
HP 54006A x20 tip
PRL DIY 953r (x20)
If you recall from above, I have a Tek P6150 kit with poor quality x10 tips. The x1 tip is quite good (looks new AFAIK) presumably because it never got used.
I acquired an 54006A x20 probe body and resistor (not a whole kit). I used it with the SMA cable from the Tek kit, I suppose a high quality cable with good connectors and good construction.
The PRL trace is for a Pulse Research Labs "PiNet" universal attenuator kit. It's a PCB between 2 BNC connectors. Would be easy to make the whole thing yourself, but the cost would be close and you wouldn't have the nice enclosure. I used an 0805 0.1% thin film 953r resistor -- note that this is actually closer to the nominal 950r than the 54006a which is only within 1% (I only have 1 resistor, not a full kit, and measured it as 956r). Not that that matters for this experiment. I wired it just as a straight through resistor, not as a Pi pad.
So, you can see that the P6150 and 54006A traces are extremely clean, with the 54006A having a 6% better rise time. I guess that implies something about the capacitive load of the probe, although the P6150 is supposed to have only 0.15pF vs the 0.25pF of the 54006A. However as I said my P6150 x10 tips are is sorry shape, smushed into the barrel a bit so this is damage which possibly affects the response. "New" tips are over $125 on ebay so that's not very attractive, esp. since it would be impossible to evaluate the actual condition without having it in front of you to inspect, and clearly they do get damaged so even if you got a new one it won't last.
The PRL trace is a bit slower than the others. I guess that represents higher capacitance? For the cost, and the 3% difference from the P6150 I think that's great. It is 10% off of the 54006A though. Still, it's ok. The bigger problem with the PRL trace is the overshoot. Where does that likely come from?
The PRL trace looks exactly the same whether I use an SMA connector with 6" RG316 pigtail coax ($6 preassembled from amazon), or the good x1 probe and gold plated ground tip from my P6150 kit.
Which is very good news. It means my theory to put the resistor at the other end of the cable might be valid. It means the overshoot and the extra capacitance comes from the construction of the "attenuator", and not so much from the bare wire "probe". If I can address that with better PCB design, better connectors, or other fix, then the probe end can just be a consumable wire pigtail, with numerous DUT attachment possibilities.
I've also attached a pic of the PRL board. Note that it's 2 sided, the other side is in parallel to the first side. I suppose the layout is not ideal for series resistance with lowest possible capacitance, but would I be able to do significantly better with a purpose built PCB, without the extra ground "stubs" etc? And specifically would I likely be able to address the overshoot? Full pdf of the kit is
here.
When I get a pulser going, I'll redo this so as to get a better appreciation of the performance. I only have a 200MHz scope, but that should be good for 1.75ns rise time so that is still significantly better than my current test circuit.