4
\$\begingroup\$

My goal:

I am trying to phase lock a 74HC-based digital circuit from a clock reference at 10~15MHz pulse repetition rate. The pulse width is around 400 picosecond. The peak voltage is sometimes below 700mV, depending on the coax cable length.

I don't actually mind the shape of the clock; even a conversion to sine wave like waveform is accepted, as long as I can drive my 47HC based electronics from this weak pulsed trigger. Should I look for designs to pre-condition the clock reference? I generally aim at a low-cost solution, else it would be an over-kill for my digital board (~$50).

Clock reference at 10~15MHz pulse repetition rate

Circuit diagram of the clock reference

The clock source is generated by a photodetector. The following shows the internal circuit layout provided by vendor. I cannot modify the circuit inside the photodetector as it will void the warranty terms.

Clock reference circuit

Things I tried:

1. 74HCU04 amplifier

The following layout from this blog removes the pulses instead of amplifying it:

schematic

simulate this circuit – Schematic created using CircuitLab

2. NPN class-C amplifier

I observe ringing at around 300mVpp but I cannot see the pulses:

schematic

simulate this circuit


[EDIT] Things achieved so far based on answers

George Herold reminded me that my current-sourcing signal needs to be converted to a voltage before amplification. So I now connect the 50Ohm termination resistor in series with my amplifier gates.

As I only have HCU04 at hand, I tinkered with the capacitor-resistor combination to amplify the signal at multiple stage:

schematic

simulate this circuit

In stage 2 and 3, I have to AC-couple the signal so that it can be biased at VCC/2 at which amplification effect is the largest. Otherwise the baseline of the pulse signal will be drifting away to either VCC or GND where there will be no amplification. The 4th stage works simply like an inverter to give me a rising edge.

Amplified signals

Is there any design rule for this? I am afraid the result may not be reproducible on PCB layout.

\$\endgroup\$
  • 1
    \$\begingroup\$ Just thinking out loud- if the pulse width is ~400ps, any amplifier you're going to use is going to need to have frequency response out to the GHz range, wouldn't it? That might be why neither of those circuits worked quite right for you. \$\endgroup\$ – TheNoonMoose Nov 11 '14 at 17:55
  • \$\begingroup\$ So this is at the end of a piece of coax? Can you gain it up before then? Something like your transistor circuit. But 1 M ohm is going to be much too big. Maybe 1k or less. (make R2 smaller) There's the old pulse stretcher trick a diode into a cap with the a longer reset resistor to ground. But I'm not sure if a normal diode (1n4148) will be fast enough. (oh looking at a st9011 data sheet, that will not be fast enough.) \$\endgroup\$ – George Herold Nov 11 '14 at 18:00
  • \$\begingroup\$ Don't know if this helps at all, but you can go much faster than 10 to 17ns propagation delay of the HCU family; a 74ABT04 has a propagation delay of 3.4ns @5v. \$\endgroup\$ – tcrosley Nov 11 '14 at 18:23
  • \$\begingroup\$ Is a GHz amplifier the only solution to my problem? I mean I don't even need gain linearity -- as long as I don't miss the pulse count I am ok with that. \$\endgroup\$ – Antony Nov 12 '14 at 15:02
  • \$\begingroup\$ As for amplifier solutions, should I look for designs that amplifies charge instead of voltage? I have this in mind because only a tiny energy is stored in that 400ps pulse -- any circuits that detect voltage change will inevitably absorb the finite charge influx and flattens the input signal. \$\endgroup\$ – Antony Nov 12 '14 at 15:12
2
\$\begingroup\$

It's all about capacitances.

74HC-whatever is too slow for that. Input capacitance for each "block" of your circuit is 3.5pF. This is flattening your 0.4ns pulse.

ST9011 transistor is too slow too. Datasheet says: 370MHz bandwidth and 1.5pF collector-base capacitance. You probably also have some extra parasitic capacitances there.

400ps is quite short time, maybe not enough to charge your whole circuit capacitances. Wiring and PCB are very important when you deal with such short pulses or high frequencies.

You need ~2.5GHz transistor and correct wiring/PCB to eliminate parasitic capacitances.

2.5GHz because 1/0.0000000004s = 2500000000 Hz

High frequency transistors are not expensive, I just found random 6GHz transistor (BFR93A) and it costs less than 0.30$ where I live. You may look for something with higher gain or just try this.

| improve this answer | |
\$\endgroup\$
2
\$\begingroup\$

OK with the added circuit diagram things are much clearer.
The photodiode is providing a current. So you'll turn this into a voltage with the load resistor. You'll want the load R and any amps right next to the BNC. (Capacitance is not your friend.) Most likely you are using a 50 ohm resistor. But there is no reason you can't try something bigger. Try several hundred ohms and see if that helps things. If you do end up needing an amp (transistor) Phil Hobbs recommends the BFG25AX (now obsolete and fast disappearing.) If you need more than just a bigger R, then ask and we can find a substitute.

| improve this answer | |
\$\endgroup\$
  • \$\begingroup\$ Could you suggest any schematics for transistor-based amplifier? Any technical papers to refer to so that I can calculate the appropriate capacitor/resistor values? \$\endgroup\$ – Antony Nov 17 '14 at 10:11
  • \$\begingroup\$ @Antony, Uck! Those 74HC04's as amps stink. (ok I've never used them, but unless your are stuck on a desert island with zero budget, get the right part for the job.) First conatact maker of the part and ask them. Second hang a much bigger resistor to ground, right at the output of the bnc. Then stick a comparator on that. (A nice fast comaprator..LT part number tomorrow.) \$\endgroup\$ – George Herold Nov 17 '14 at 22:27

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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