6
\$\begingroup\$

I'm making my own PCB connectors using the "multicomp" range of crimp connectors available from farnell (ex. http://uk.farnell.com/multicomp/2226a-04/crimp-housing-1-row-4-way/dp/1593508 with terminals at http://uk.farnell.com/jsp/search/productdetail.jsp?SKU=1593529&MER=baynote-1593529-pr), but while I'm able to get a good DC connection and solid mechanical connection (the terminals don't pull off easily), the AC performance of the resulting cables leaves a lot to be desired; I don't have a lot of test equipment here but with a simple home-made test (wiring them up to an FPGA development board and looping back to the same board and seeing how fast a signal I can send through them and receive it reliably), it appears I can only get up to about 500KHz before I lose the signal. I'd have expected to be able to do better than that, as I've seen this type of connector used for signals I assume must be higher frequency than that (e.g. USB2).

Any ideas what I can do to improve the performance beyond this? What kind of frequency should I be able to usefully achieve with this type of connector?

EDIT: I'm putting them onto CAT6 cable. Is this a suitable cable or do I need something heavier?

\$\endgroup\$
  • \$\begingroup\$ What are the source and input impedances in the test rig? The higher the impedance, the more the cable capacitance rolls off the high frequency response. \$\endgroup\$ – Kaz May 2 '13 at 18:08
  • \$\begingroup\$ Using Cat6 cable, are you sending balanced (differential) signals? How many pins are you allocating for ground? \$\endgroup\$ – The Photon May 2 '13 at 18:20
  • \$\begingroup\$ I'm unfortunately not certain what the impedances are; the device is a Cyclone IV, whose data sheet specifies a leakage current of 10uA. The signals are not differential, no. There is only a single ground. \$\endgroup\$ – Jules May 2 '13 at 18:25
  • \$\begingroup\$ You can put your own shunt resistor at the input to configure the input impedance. Try the test with various low-ish impedances at the input. \$\endgroup\$ – Kaz May 2 '13 at 18:33
  • \$\begingroup\$ A few feet of Cat6 will probably not even pass audio without noticeable HF roll-off, if facing some FET input stage with gobs of input impedance. High impedance also invites noise. \$\endgroup\$ – Kaz May 2 '13 at 18:40
2
\$\begingroup\$

Here's a comparison and some numbers:

I run single crimped wires of length between 3" and 24", sometimes with multiple hops, and at 2 megabit data rate. The signal is 5V UART TTL, and the driver is an Atmega microcontroller (25 mA nominal pin current) with a 70 ohm current limiting resistor and a dozen or so 40 picoFarad MOSFET inputs on the other end.

This all works fine; oscilloscope shows the signal is decent, and communication works.

When I added a TVS diode with 2 nF capacitance across this bus, the signal degraded enough that I could not keep 2 Mbit data rate. The 3 dB filter frequency of a 70 ohm, 2 nF low-pass filter is about 600 kHz, IIRC, which would explain the signal degradation.

So, by comparison, you have one of three problems (or a combination):

1) The driver that is emitting the signal is not very strong. Some microcontrollers can only drive a few milliamps on their pins, for example, which translates into a high-impedance source. 2) The load you are driving is high capacitance somehow. 3) The wires you use add significant inductance.

The fix in 2) and 3) is to remove the cause. The fix in 1) is to re-drive the signal with a buffer or line driver (or perhaps a MOSFET gate driver, which can drive several amps!)

| improve this answer | |
\$\endgroup\$
  • 1
    \$\begingroup\$ It sounds like the 8mA current produced by my FPGA is probably too small for the job then. Adding a buffer will be tricky because it's on a development board, unfortunately, but maybe I can find a way of building an external driver board that can fit directly onto the header pins. I'm assuming the capacitance introduced by such a connection wouldn't be too high... \$\endgroup\$ – Jules May 3 '13 at 8:04
  • 1
    \$\begingroup\$ Before you do that, get an oscilloscope and look at the signal to try to figure out the exact problem you're having! \$\endgroup\$ – Jon Watte May 4 '13 at 6:45
1
\$\begingroup\$

I wouldn't run CAT6 cable for HF analog signals... too much interference, too much induction.

When it comes to high freq signals (typically greater that 500kHz) the quick and dirty solution is RF specific equipment. As described before, the interior capacitance of DC wiring and cabling terminals can cause a lot degradation and reflections. Coaxial cable is great solution. Expensive though. Take a look at SMB and SMC cables and terminals. SMC is small enough to fit on PCBs.

| improve this answer | |
\$\endgroup\$
  • \$\begingroup\$ For clarification, my signals are digital, but are being degraded too much for the receiver to understand them. \$\endgroup\$ – Jules May 2 '13 at 18:32
  • 5
    \$\begingroup\$ @Jules - There are no digital signals. Every signal is really analog, "digital" is just a simplification that makes certain things easier to understand. \$\endgroup\$ – Connor Wolf May 3 '13 at 3:51

Not the answer you're looking for? Browse other questions tagged or ask your own question.