Accurate & fast signal generation ???

[manitooo]

Hi all.


I would like to generate a digital signal (with only a rising edge then a falling edge), but I would to have an accurate control on the time betwenn to 2 edges.

For the moment, I think I can generate a signal with a 6.25ns resolution

Does anyone know how to have a better generation, or if it exists a "raedy to plug" solution"

plz help ?
thx a lot in advance

 

[MonkeyBusiness]

If you need to generate a single pulse, or even a train of pulses, you can hook up a counter and control the count value.
Fast logic families for 2.5Gbps and even 10Gbps exist. Look at https://onsemi.com high speed logic solutions (ECL and similar families).

 

[Kripton2035]

for really precise pulse (even sine waves) you can try the direct digital synthesis chips (DDS) from analog device. the fastest one is up to 1GHz clocking, giving output sine waves up to 500MHz, and milli-hertz resolution.

 

[flatulent]

transmission line pulser

If you only need a few values of small pulse widths, a transmission line pulser can be used with different lines for each width. You can also use an XOR gate with one input from a slower generator and the other input from the same source via a transmission line. The pulse width is equal to the delay time through the transmission line. This method will produce another pulse when the main pulser goes to the other polarity. You can use fast logic, like eclipse lite from Motorola or the 10 GHz more expensive types. You can also try some RF grade transmission line switches to cut in or bypass a binary weighted set of line lengths (delays).

 

[manitooo]

quite fuzzy but I think U seem to mainly use delay propagation parameters to generate such signal, isn't it ?

But is it possible to have a programable signal in such way, instead of having one line for each width ?

 

[flatulent]

pulse width is line delay

The pulse width is the line delay. This is the only military approved way to generate pulses. They disallow one shots. The usual way is to use a lumped element LC delay line, but for your short pulses a regular line is better.

The RF switches are to switch in different lengths of line for different pulse widths. Some logic families have programmable delays in ps range increments.

A detailed view of how the XOR-delay line single shot works is as follows. In steady state, the line to one input of the XOR and the output of the delay line to the other XOR input have the same logic level which makes the XOR output low.

When the input line changes state, one input to the XOR changes state immediately while the change in logic level at the input of the delay line starts traveling down the line. Before it reaches the end and the other input of the XOR, the inputs to the XOR are different and the output goes high. After the chagnge of logic level gets to the output of the delay line and the other input of the XOR both inputs to the XOR are at the same logic level and its output returns to zero. Therefore, the XOR output pulse width is equal to th delay line time delay.

If you want only high transition triggering, you supress the low going triggering by putting a AND gate at the output of the XOR and feed the input line to it as well. That way a low going input line will block the XOR output.

The first oscillator I described was only to trigger the one shot if you wanted a periodic pulse stream like from a lab instrument pulse generator.

The pulse width can be varied by changing the delay through the delay line. The line can be a series of transmission lines that are switched in and out and can also include the silicon variable delay IC for very fine tuning. The transmission lines are in a binary sequence of delays so that they can be switched in tandem or bypassed with a DPDT switch to form about (2^N) different equally incremented delays from N lines.