3
\$\begingroup\$

I'm trying to use a Small Form-factor Pluggable (SFP) as the laser source for a homebuilt laser gas analyzer, which means I don't actually care about transmitting data. The plan is to send an arbitrary bitstream to an SFP transceiver, which will be a PWM signal with constant frequency.

Then, I'd set the bandwidth on the Rx channel to a reduced bandwidth, and use the duty cycle on my Tx signal to modulate Tx power.

I have a basic idea of how I'll set it up: send the signal to the Tx pins, attach all the other pins for serial communication and rate select, etc.

However, I'm trying to figure out the minimum frequency of the PWM signal that will work with my SFP. I'm planning to use this model, which has an advertised transmission rate of 155 Mbps.

The microcontroller I'm using maxes out at a 40 MHz PWM signal. So, if the Tx bandwidth on my SFP is set to full bandwidth, what will happen if I give it a 40 MHz signal? Will the SFP function fine and transmit a 40 MHz signal, will it function sub-optimally, or will it not work at all?

Additionally, how do I tell what the bandwidth is on the receiver if I set the Rx bandwidth to reduced bandwidth? I can see in the SFF-8079 document that there are options for selecting full or reduced bandwidth on the Tx and Rx channels, but I don't see documentation on my SFP (or any SFP) that clarifies what that reduced bandwidth is.

\$\endgroup\$
2
  • \$\begingroup\$ The reduced RX bandwidth is whatever it needs to be to improve the sensitivity for the lowest bit rate advertised by that particular SFP design. It might be no change at all if the specified sensitivity can be achieved without a change. If you want to find out what it is for a particular part, change your name to "Bob Cisco" or "Joe Juniper" and ask to speak to an apps engineer at the SFP manufacturer. \$\endgroup\$
    – The Photon
    Commented Jun 20 at 16:49
  • \$\begingroup\$ I see, thanks for the info. Looks like I'm Bob Cisco now. \$\endgroup\$
    – ewineteer
    Commented Jun 20 at 23:55

2 Answers 2

Reset to default
1
\$\begingroup\$

Too much for a comment, but not a full answer to your question either.

From the datasheet:

"Note:

  1. AC coupled"

Most SFPs for specific protocols are AC coupled and the data bitstream is tailored with this in mind. Unless there is heavy brains inside the module to decode and look at the data, it won't care that you're not sending Ethernet data, but it does block DC so you need to adjust what you are sending to have equally many 0 and 1 over some period of time in order to not be blocked by the AC coupling. "PWM signal with constant frequency" sounds not DC balanced (blocked/offset by the AC coupling), but perhaps you have something more elaborate in mind here.

I'll let someone with more experience in SFP modules answer for other issues you may run into.

If this is a one-off, you can probably open it and bypass the AC coupling, but it depends on if there are any more "smarts" going on.

\$\endgroup\$
4
  • 1
    \$\begingroup\$ I see, thanks for letting me know of this constraint. I think I can keep the signal AC coupled, but I'll have to rework how I modulate it. One question: what will happen if I send a 40kHz signal to a 155 Mbps SFP? \$\endgroup\$
    – ewineteer
    Commented Jun 20 at 21:05
  • \$\begingroup\$ @ewineteer Depends on the time constant of the AC coupling. 40 kHz is a lot less than 155 MHz. The SFP modules I have abused had 100 nF in series for AC coupling. If there are more “smarts” going on, it could be considered out of band. Try it and report back? Use some external means to PLL to 1024 or 2048 times your base frequency? Get a bare bones SFP? Short out the capacitor to make it DC coupled? \$\endgroup\$
    – winny
    Commented Jun 20 at 21:13
  • \$\begingroup\$ Oof, I'm sorry. That's a typo. I meant to type 40MHz. Think that'll work without issue? \$\endgroup\$
    – ewineteer
    Commented Jun 20 at 23:29
  • \$\begingroup\$ @ewineteer Probably. \$\endgroup\$
    – winny
    Commented Jun 21 at 7:04
1
\$\begingroup\$

I know of someone who distributed a 10 MHz reference (rectangular wave) clock in his lab using SFP modules. You'll be fine at 40 MHz.

If you want to avoid fiddling with the mechanicals of the SFP cage, and preclude errors with how you drive the SFP TX lines differentially (you need to drive them differentially!), you could buy the sysmocom SFP Experimenter Board v2; it's Free and Open Source hardware, so you're more than invited to instead make your own based on the schematics and hardware design files from https://osmocom.org/projects/misc-hardware/wiki/Sfp-experimenter .

\$\endgroup\$
4
  • \$\begingroup\$ A 10 MHz clock is DC balanced. A PWM signal is not (in general). \$\endgroup\$
    – The Photon
    Commented Jun 21 at 11:47
  • \$\begingroup\$ @ThePhoton usually, a clock is, indeed. In this particular case, I'm not sure it matters: there's an active part that will discharge the diode side of the coupling capacitors (the diode, be it the laser- or the photodiode), so I don't think that'll be a problem. But it's a good point, I hadn't thought about that. \$\endgroup\$
    – Marcus Müller
    Commented Jun 21 at 11:56
  • 1
    \$\begingroup\$ The diode doesn't come into it because there is a driver chip between the inputs and the optical device. But a high or low duty cycle signal passing through DC-blocking will drift down or up accordingly and might cause the output duty cycle to not accurately track the input, or even (in an extreme case) cause the logic to fail to switch. \$\endgroup\$
    – The Photon
    Commented Jun 23 at 7:17
  • \$\begingroup\$ Ah, so you'll need balancing logic. Hm. not hard per se, if one got programmable logic at hand, (in the reseiver, treat PWM as if it was a clock, divide by two; in the transmitter, always send the intended PWM pulse, followed by one with the opposite duty cycle) anyways, but probably exactly what ewineteer wanted to avoid. \$\endgroup\$
    – Marcus Müller
    Commented Jun 23 at 11:46

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