1
\$\begingroup\$

I'm experiencing long fall time issue for a 0.5ns pulse duration and 3MHz optical pulse source. The manufacturer test report provides the following pulse shape for the optical pulse. I don't know how they measured it nicely, but I tried to obtain the similar result using a fast photodetector, a 20GS/s 4GHz oscilloscope in 50 Ohm input impedance mode and a 1m coax cable(RG-58). In my case the scope shows the following:

I have this fast photodiode operating in photoconductive mode. At page 4 they provide a model and according to given parameters like Cj. So I modeled this photodiode and my instrumentation(coax cable and scope) in LTspice as follows:

enter image description here

In simulation, the pulse at the scope end degrades a lot especially the falling edges stretches a lot. But I'm not sure if that is the root cause in my case.

Can the 1m coax be the reason for long fall times at scope? What else can be the reason? How to remedy this and obtain shorter fall time?

\$\endgroup\$
11
  • \$\begingroup\$ Are you sure that you have 50 ohms at the scope? Many modern scopes have a 50 ohms internal option, but you have to select it, it is not the default. \$\endgroup\$
    – Mattman944
    Commented Oct 7, 2021 at 18:59
  • \$\begingroup\$ I meant that. I use the scope's internal 50 Ohm. \$\endgroup\$
    – floppy380
    Commented Oct 7, 2021 at 19:03
  • \$\begingroup\$ Your bandwidth is at least 1 GHz so your coax is not even close to the correct model. Use a t-line model; I believe they are available in LTSpice. Then try bypassing the coax in your sim and comparing results. Mess around with things. \$\endgroup\$
    – Andy aka
    Commented Oct 7, 2021 at 19:14
  • \$\begingroup\$ @Andyaka How did you quantify 1GHz? And is coax capacitance dependent on freq? \$\endgroup\$
    – floppy380
    Commented Oct 7, 2021 at 19:40
  • \$\begingroup\$ At these frequencies that coax is a transmission line, not a lumped capacitance, so you can't just model it as a capacitor. The extremely slow fall time (~10ns) seems unreasonable given the speced diode capacitance (< 1pF) and the 50 ohm termination, so I suspect something is wrong. You should be on the order of 100ps. The battery is charged right? You're certain your scope is configured correctly (no post filters running)? \$\endgroup\$ Commented Oct 7, 2021 at 19:46

2 Answers 2

2
\$\begingroup\$

A fast RC time constant calculation by using C = 82 pF and R = 50 ohms gives time constant = 4.1 ns. It's well in accordance with your slow pulse falling. The diode + the DC bias source charge the capacitance faster when there's light, but the capacitance is discharged only by the 50 Ohm load. The diode sinks nothing.

You should have a proper preamp just after your diode (max distance say a couple of centimeters). That preamp should be able both to source and sink the needed current. Proper matching at the distant end of the cable gives some tolerance to the impedance matching at the preamp end, but the current source and sink capability must exist.

\$\endgroup\$
8
  • \$\begingroup\$ Also see manufacturer i.stack.imgur.com/0gCoo.png It shows "20mVΩ" What does that mean? \$\endgroup\$
    – floppy380
    Commented Oct 8, 2021 at 2:29
  • \$\begingroup\$ Some comments say coax capacitance is not realistic at these frequencies. is my coax model correct? \$\endgroup\$
    – floppy380
    Commented Oct 8, 2021 at 2:32
  • 1
    \$\begingroup\$ @floppy380 That millivoltohm is not a physical dimension. 20mVΩ shows 2 separate things: 1) You have sensitivity 20mV/div in use and 2)You have the termination resistor switched ON. The capacitance 82pF/meter is ok. Omitting cable inductance and the fact that capacitance and inductance are distributed along the wire (=Heaviside transmission line model) will cause wrong simulation results with fast pulses that you want to have. You fortunately have good impedance match at the distant end which reduces the error. (continues) \$\endgroup\$
    – user136077
    Commented Oct 8, 2021 at 6:03
  • \$\begingroup\$ (continued) But you really should use proper transmission line model. Many simulators like Micro Cap have it with no trickery. Alternatively you should use so short cable that its length is well below a quarter of the wavelength at the highest meaningful frequency component (calculated with the signal speed in the cable). \$\endgroup\$
    – user136077
    Commented Oct 8, 2021 at 6:05
  • \$\begingroup\$ So the problem here appears to be an issue of the COAX model than it is of the actual diode? With your suggestion of a preamp, what if you sink it with a resistor to measure the photodiode current as a voltage? I worked with similar items before, so this question interests me. The recommended circuit for my item recommended a resistor from the diode's output to ground as one readout method. \$\endgroup\$ Commented Oct 8, 2021 at 14:30
0
\$\begingroup\$

Sanity Check: For simple RC circuits, you can use the old oscilloscope bandwidth - rise time equation (valid for single-pole filters) to estimate the rise and fall time of your circuit. In this case, you have an RC circuit of 50 ohms and 82pF, thus the bandwidth is about 39 MHz \$ \left (BW = {1 \over {2 \pi R C}} \right )\$.
Using the oscilloscope rise time versus bandwidth formula: \$ rise\_time = {0.35 \over BW} = {0.35 \over 39MHz} = 9 ns \$. This is pretty close to your fall time of approximately 15 ns. Do you have an additional 50pF of wiring capacitance floating about (layout + scope input capacitance) or the bandwidth limit enabled on the scope? Of course, this completely ignores that you have a transmission line setup.

Normally, you have a high speed buffer in close proximity to the photo diode to minimize capacitance across the diode and provide a robust drive for the cable.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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