Load Google Translate Introduction: I hate connectors. They're big, unreliable, messy, expensive, and introduce crosstalk. If I could put my whole system on one board I would. But when the system is spread across hundreds of meters, or even just involves different circuit board technologies or mix-and-match board sets, I need connectors after all.
The specific need I have at the moment is for a board-to-board connector. The two boards will both be clamped flat to the same aluminum heat sink, so the usual vertical mate board-to-board connectors won't work. I have about 20 low-speed digital lines, some power supplies, and about 12 x 50 Ohm analog RF signals in the 20 to 200 MHz band, where I need low (-80 dB) crosstalk. A set of 10 SMA connectors would sort-of work for the analog RF signals; I want to improve on SMA in terms of cost, density, and assembly time, without giving up much in the way of crosstalk. I hope the low-speed digital and power will come "for free".
I looked at using a single high density connector (Samtec QMS/QFS) but wasn't happy with the amount of board space needed to fan in and fan out. Two or three smaller connectors would work, but if they are high density, I risk reliability due to assembly tolerances putting stress on the contact area. Low density connectors have larger antenna areas, which is bad for crosstalk.
Many connectors I have looked at come close. Maybe I can leverage the current industry excitement about high-speed serial interconnect. I sense that I need a creative way to relax one key spec, but still get the job done.
- Larry
Hi Roland I have a question. Not so long ago I decided to create a circuit with floating cores. Idea is that main single core can be split into small ones (with shorter bit range) and joined back into big. For example one 128 core splits into 8 16 bit cores or 2 64 bits. Idea is that program code which is usually much smaller than 32 bits can be processed better. It is a kind of optimization of used area and number gates involved in processing. (I heard about one company which created circuit with split cores in 2006 but later there were no any info on that project.)
But now I got serious problem with instruction decoding. Each combination of cores requires special connection to the instruction set. So I got to many signal lines which create numerous time delays and circuit becomes very complicated for simulation. My question is - Is it possible to create high speed interconnect between instruction decoders and core units (arithmetic units, memory) so that each core will prefetch instruction independently and total time delay will be enough small?(not like in reconfigurable systems).
Roland,
I am researching possible board to board connections for a design change that I am working on. Currently we utilize a multilayer mixed signal board that has some issues with noise from the transmitter section being transferred to the ground plane. I would like to isolate the transmit section with an air mote around the entire section and move it from a four layer to a two layer. My connection questions come from the most efficient way to transfer the 3v transmit initiation pulses along with the monitoring signals to and from the floating daughter board. Also the transfer of the positive and negative 24V power supply voltages across the gap. The maximum frequency of the pulses is 40kHz, so it is very low in the realm of skin effect. Most of the systems will be operating at 10kHz down to 350HZ. The transmit section power requirements would be less that 400mA for each supply. Would a board to board riser connection be a better option or a copper ribbon be a better design with respect to the power and signal range?
V/R
Mike Burr
R&D Engineer
© 2009 Premier Farnell plc. All Rights Reserved
Premier Farnell plc, registered in England and Wales (no 00876412), registered office: Farnell House, Forge Lane, Leeds LS12 2NE