# Buffer for 5V, 500ps to 10 ns pulses

I am trying to make a buffer for short, 5V pulses. The pulses are negative (form 5 to 0 V) and last for 500 ps to 10 ns (Half maximum, full width). The load is capacitive, around 50pF, and as close as possible to the buffer. The input will probably be a differential pair. The pulse repetition rate is expected to be around 50MHz.

I considered several options:

1. Several fast logic gate, like SN74LVC2G04 in parallels. They could have adequate rise-time but the maximum frequency is low, around 150MHz.
2. Fast, high slew rate Op-amp like an AD8009. The bandwidth is better and this could work with pulses around 5ns, but probably not with shorter ones.
3. An RF amplifier, with a gain block or an RF FET. I don't know that much on those circuits beside basic class A amplifier and this will continuously consume power.

Thanks for any suggestions!

The pulses will be generated by some CML logic. The device driving the buffer will be similar to an ADCMP572. It's differential with a swing of 800 mV, so I need voltage gain, and enough current to get a good rise time.

• If you don't need power gain, you should consider using an appropriate pulse transformer. – Dave Tweed Dec 15 '15 at 12:41
• @DaveTweed The driving current of the previous component is in the order 20mA. I think I need power gain. – pserra Dec 16 '15 at 17:27
• You definitely need multiple drivers in parallel. The typical inductance of an IC pin is about 1nH, so if your load is 50pF, then you have a cutoff around 700 MHz (1/(2pi*sqrt(LC))), while the minimum frequency component for your shortest pulse is 1 GHz. So the limiting factor is the parasitic inductance more than just the speed of your device. – jpcgt Jan 13 '16 at 20:04
• If 20mA is way to low of power, consider looking at the first stage drive on a pico second laser? See how they do it. – MadHatter Feb 1 '16 at 17:38

Just to establish some ground rules here. What you want needs major amounts of current to get the job done.

Capacitance voltage charge from a constant current source is expressed as $$\ Ut = \frac{(I*t)}C$$

Given 5V and 50pF, that works out to 500mA of current just to get the pulse high within 500ps. And that's probably not too useful for you. So to get something a bit more reasonable like 50ps for 10% rise/fall time you're looking at 5 amperes of current to get the job done.

We're talking about a major high speed driver here. To look at it other way, your 500ps pulses are 2GHz if they were sinewave, which they're not. Some of the modern fast serial drivers can handle the speed but nowhere near the capacitance.

If you're really got your heart set at building this, you'd essentially need to design a basestation-level microwave amplifier. This is very much specialty equipment, but does not actually cost that much if you go to ebay. Google for microwave power amplifier. Note that you'd need to go up to 20GHz to get that 10% rise and fall time.

Alternatively positive emitter coupled logic might be coaxed to produce required amounts of oomph but designing such a monster is beyond the scope.

We built the circuit with FET drivers, like the UCC27517A, and extra logic to drive the FET driver. We relaxed the requirements on rise time, and used two circuits in parallel to get the pulse repetition rate we wanted, as the FET driver max frequency is to low.