multi pin connnector high frequency compatibility

[sdouble]

I do develop charge- and current- sensitive preamplifiers.
Depending on the needs they could be single channel devices with coax input/outputs or multi-channel devices.
In the latter configuration, I typically have a motherboard hosting several daughter boards.
A typical daughterboard is shown below. The interface connector is a 13-pin 2.54mm pitch E-TEC connector for which 2 (square)  pins are manually removed to increase the high voltage clearance.
The inputs/outputs are high voltage , detector input, positive and negative low voltages, several gnd pins and 1 output pin.
Now the issue : when dealing with low frequency design, I have no issue with the set up but when I try to translate my high frequency design into this system, everything collapse. I typically face massive oscillation issues. I do blame the connector since the very same circuitry operates flawlessly in the single channel designs.
That being said, do you have any clue on how to improve the situation ? It drives me crazy

[tooki]

I worked on the PCB and mechanical design for several silicon detector amplifier boards (first a 4-channel version using carrier boards for the detectors, and a later 16-channel version using carrier boards). For the 4-channel carrier board version, I used very short pogo pins. For the 16-channel version, I used spring-loaded mezzanine connectors from Samtec which are rated to ≥5GHz. The amp itself is specified to 4GHz. A separate pogo pin was used for the high voltage supply.

Anyway, while I didn't design the original circuit (though my supervisor and I suggested some small modifications, which we then implented), proper layout proved to be crucial: the user's original PCB layouts oscillated, while my layouts were stable, and he was able to finish his PhD using them. I do not know how the 16-channel version's performance compares to the 4-channel version, since I switched departments, but the amplifier blocks themselves use a 100% identical layout on both boards, with the only differences being the different connectors to the carrier boards, and the longer traces from the amp blocks to the output coax connectors, so I'd expect the carrier board connector to be the most significant source of differences.


Some key approaches were very clean grounding with carefully placed via shielding, loop areas kept as small as possible, adequate space between channels, and of course proper impedance-controlled traces.

I'd have to check the design files, but I think the Samtec connectors used were FSI-110-03-G-S-M-AD. (Or possibly the 6mm high variant.) That series is rated to a minimum of 5GHz, with the 3mm high version rated to 8GHz for single-ended signals. I used alternating pins for signal and ground.

I hope this helps!

[sdouble]

thanks a lot for all those detailed info. Very helpful.
I looked at the SAMTEC reference. Please, forgive my ignorance, but I don't get how this connector mate to the mating connector.
In my current design, I have a header soldered to my motherboard and the daughterboard show in my previous post plugs into the mating female header.
The one you are referring to obviously operates very differently.

[tooki]

Oh, sorry, I wasn't clear enough in describing the connector! On that connector, each contact is a leaf spring. Each contact simply presses against a pad on the other PCB. There is no mating connector.

(You just need to use a gold-plated PCB, which you're probably doing anyway.)

I chose a model that has alignment pins on both sides (top and bottom), so that the bottom alignment pins ensure the connector aligns properly with its solder pads, and the top alignment pins ensure the other PCB aligns properly with the connector. The boards need to be squeezed together fully, whether by some other aspect of your mechanical design, or by screws right in the connector (like the one I used).

But of course, you don't have to use this style of connector; depending on your situation, a traditional one with male and female connectors may be advantageous. (In the boards I worked on, a primary design goal was to create a large number of "disposable" carrier boards, so avoiding a connector on the carriers was a huge savings: make one expensive amp board and dozens of cheap carriers. The earliest design, which I assembled dozens of pieces of but which was not designed by me, had the detectors wire-bonded directly to the amp board, and after a few uses, the bonding pads wouldn't accept bond wires anymore, and the entire amp board had to be discarded. Hence the goal of separating the bonding pads from the amp circuitry.)

[sdouble]

I eventually understood how this connector works.
Based on your suggestion, i do think that I 'll go for the MEC2-08-01-L-DV-WT.
The pitch seems ok and there is a separated "area" that I'll use for the high voltage pins.
Not sur about the leakage current at 800V-1kV.. we'll see.