Looking for ultra-low jitter 5 volt square pulse generator

[Agent]

I’m looking for something similar to the Texas Instruments LMK03318. Link:

http://www.ti.com/product/LMK03318

The problem is that I can’t tolerate anything lower than 4 volts and the maximum output is 3.3 volts. Maybe I could somehow amplify the signal without adding too much to the jitter?
I want something that will generate the exact same square wave every single time with precision to single picoseconds. To simplify, say I have a perfect square pulse with rise time = 0. I want this rising edge to occur at 1 ns consistently. Ideally this rising edge would only slightly deviate from 1 ns, 998 ps is okay and 1002 ps is okay. If Rubidium clocks have a jitter of +/- 5x10^-11 does that mean my perfect rising edge is guaranteed to occur no earlier than 950 ps and no later than 1050 ps?

I plan on splitting this output and feeding it into two 50 ohm loads effectively giving me 25 ohms.
Desired specs:
Amplitude: 4 volts into 25 ohms
Pulse repetition rate: 10 kHz
Pulse width: 20 ns
Rise time: ~1 ns
Fall time: Don’t care
Jitter: As low as possible

Sorry, if this was confusing. Ask for clarification if need be. Any insight, products, or designs are appreciated!

[David Hess]

The problem is that I can’t tolerate anything lower than 4 volts and the maximum output is 3.3 volts. Maybe I could somehow amplify the signal without adding too much to the jitter?

Fast comparators are good for that.  Their differential input gives them good power supply rejection.

I want something that will generate the exact same square wave every single time with precision to single picoseconds. To simplify, say I have a perfect square pulse with rise time = 0. I want this rising edge to occur at 1 ns consistently. Ideally this rising edge would only slightly deviate from 1 ns, 998 ps is okay and 1002 ps is okay. If Rubidium clocks have a jitter of +/- 5x10^-11 does that mean my perfect rising edge is guaranteed to occur no earlier than 950 ps and no later than 1050 ps?

I plan on splitting this output and feeding it into two 50 ohm loads effectively giving me 25 ohms.
Desired specs:
Amplitude: 4 volts into 25 ohms
Pulse repetition rate: 10 kHz
Pulse width: 20 ns
Rise time: ~1 ns
Fall time: Don’t care
Jitter: As low as possible

What is wrong with a low noise crystal oscillator followed by a synchronous counter to divide down to 10 kHz?  Additional jitter is limited to that added by the output stage of the counter which can even be external.  AC and LVC logic has about 1 nanosecond edges with light loading.

[Kleinstein]

The jitter form the rubidium source (and most other sources) is more complicated than just a single number can describe. For the short term it is much like a VCXO.  It is only on the longer term  (e.g. more than minutes)  that it is really good.

So the way to go is a crystal oscillator and synchronous divider (possibly divided in 2 chips: the last sync flip flop and the actual divider separate).
Not sure of 74AC74 or 74VHC74 are good driving 25 ohms.

[David Hess]

Not sure of 74AC74 or 74VHC74 are good driving 25 ohms.

They are not but a bunch of gates can be used in parallel or an emitter follower can be added to the output.

[tggzzz]

Not sure of 74AC74 or 74VHC74 are good driving 25 ohms.

See https://www.eevblog.com/forum/testgear/show-us-your-square-wave/msg1902941/#msg1902941 for an example of LVC driving 50ohms.

[Agent]

Not sure of 74AC74 or 74VHC74 are good driving 25 ohms.

See https://www.eevblog.com/forum/testgear/show-us-your-square-wave/msg1902941/#msg1902941 for an example of LVC driving 50ohms.

In that post you show measurements on two different scopes. I'm guessing that was done just to show what the signal looks like at 4 GHz vs 1 GHz? On the 1 GHz scope you show a rise time of 279.10 ps. A scope with 1 GHz of BW will have a rise time of ~400ps. How are you measuring a rise time under 400 ps?

[tggzzz]

Not sure of 74AC74 or 74VHC74 are good driving 25 ohms.

See https://www.eevblog.com/forum/testgear/show-us-your-square-wave/msg1902941/#msg1902941 for an example of LVC driving 50ohms.

In that post you show measurements on two different scopes. I'm guessing that was done just to show what the signal looks like at 4 GHz vs 1 GHz? On the 1 GHz scope you show a rise time of 279.10 ps. A scope with 1 GHz of BW will have a rise time of ~400ps. How are you measuring a rise time under 400 ps?

The two scopes were on stands at exhibitions; the reason I chose them was because they were available

As for the risetime, the tr=0.35/BW is only a rule of thumb. Plus any good manufacturer will quote a minimum bandwidth; the actual bandwidth will often be notably higher.

[klmbk]

You could possibly use MC10EP16 to translate LVPECL to PECL.

[cjheath]

Have you looked at TinyLogic, e.g. the NC7WZ14PS Schmitt inverter? Up to 6V, with ferocious output drivers into 50R, and can produce jitter in the sub-ps range. You might want to parallel two (non-Schmitt!) on the output for more drive.