Generate 1ns Pulse w/Frequency 680MHz | Prathap Chandra

[Chandra214]

Please suggest a method to generate a 1ns pulse. with 680MHz frequency. i.e ton: 1ns.
the pin should also be able to generate pulses with pulse width anything from 1ns to 4s with 1ns step, i.e the pulse can take any value b/n 1 ns and 4s with 1ns step.
This pulse is used to trigger a laser.
Thanks,
Prathap Chandra

Edit1: The pulse should be well shaped square wave.

 

[vk6kro]

This might come close:

http://www.home.agilent.com/agilent/product.jspx?pageMode=SC&nid=-536902257.0.00&cc=AU&lc=eng

They have models that go to 500 MHz.
Even their 80 MHz model is about $6000 US. Very nice though.
One 500 MHz model is about US$12146.

 

[Chandra214]

we are not looking for a instrument solution.
This should be implemented on a PCB. Right now we are considering MGT based FPGA for the solution.

 

[f95toli]

What kind of rise/fall time do you need? 1ns is not difficult; but in order to get a "well shaped" pulse you need a very short risetime.
Most commercial low-frequency(<1 GHz) generators will have risetime of 2-3 ns which is far too long in this case. It is doable, but the equipment is quite expensive and I seriously doubt a DIY solutin would work.

Also, you do realize that it is not possible to transmit a "proper" 680 MHz wave in an 1ns wide pulse envelope, right? The width of the pulse is too short and you will get lots of high frequency components.

 

[Chandra214]

What kind of rise/fall time do you need?

We would need the rise time and fall time to be within 50ps.

Also, you do realize that it is not possible to transmit a "proper" 680 MHz wave in an 1ns wide pulse envelope, right? The width of the pulse is too short and you will get lots of high frequency components.

680Mhz is the frequency at which 1ns pulses need to be sent, i.e 1ns "high" time and 0.57ns "low" time. so the total time duration would be 1.57ns leading to 680MHz frequency.

 

[negitron]

A 1.57 ns cycle time is 639 MHz, not 680. And I'm afraid that if you have to ask, a 50 ps rise/fall time is far beyond your design capabilities. You cannot use ordinary components and PCB construction for this sort of thing; stray capacitances and inductances will swamp your signal, turning it into mush.

 

[f95toli]

I have a SRS DG535 in my lab which is pretty good, but even with the fast fall/risetime modules the times are about 100ps; 50 ps is VERY difficult.
I have used a pulse generator that could manage something like 30 ps; but that was an Anritsu generator with a price tag of about $300 000...

Also, the "rule of thumb" is that you need a bandwidth of cables etc of about 1/(risetime/3) in order not to distort the shape of the pulse; meaning you are actually oishung the limit of what can be done using standard SMA connectors etc with 50 ps.

Again, it is possible but it is not easy and very expensive.

Would it be possible to generate the pulse using a "passive" method? I might be wrong, but I thought pulses for fast lasers were usually generated using delay lines etc; not with active components.

 

[Chandra214]

A 1.57 ns cycle time is 639 MHz, not 680. And I'm afraid that if you have to ask, a 50 ps rise/fall time is far beyond your design capabilities. You cannot use ordinary components and PCB construction for this sort of thing; stray capacitances and inductances will swamp your signal, turning it into mush.

Sorry.. the cycle time is 1.47ns

I have a SRS DG535 in my lab which is pretty good, but even with the fast fall/risetime modules the times are about 100ps; 50 ps is VERY difficult.
I have used a pulse generator that could manage something like 30 ps; but that was an Anritsu generator with a price tag of about $300 000...

Also, the "rule of thumb" is that you need a bandwidth of cables etc of about 1/(risetime/3) in order not to distort the shape of the pulse; meaning you are actually oishung the limit of what can be done using standard SMA connectors etc with 50 ps.

Again, it is possible but it is not easy and very expensive.

Would it be possible to generate the pulse using a "passive" method? I might be wrong, but I thought pulses for fast lasers were usually generated using delay lines etc; not with active components.

The application needs a lot of buffer memory, so we can make use of a FPGA for two operations, one is for FIFO implementation on four 12 bit high speed ADC's and for generating 1ns pulses. ( virtex 5 LX30). Frankly speaking, generating this kind of a pulse is giving me a lot of nervous times
Still have to explore possibilities with passive circuits, like using a SRD and likes.
http://www.furaxa.com/Documents/FWP-v1.2.pdf

Thanks,
PC

 

[negitron]

Again, it is possible but it is not easy and very expensive.

Agreed. I've been thinking this problem over all morning and I came to the conclusion that there really needs to be a compelling reason why an off-the-shelf solution cannot be used. In the long run, it would really be much less expensive in terms of both time and money. If you really must have a custom-engineered solution, you're probably better off going to a third party rather than try to re-invent the wheel by yourself.

 

[Chandra214]

http://www.micrel.com/_PDF/HBW/sy88922v.pdf
This can do the trick, Hopefully!
Do you foresee a problem with this?

 

[vk6kro]

You wanted this:
"the pin should also be able to generate pulses with pulse width anything from 1ns to 4s with 1ns step, i.e the pulse can take any value b/n 1 ns and 4s with 1ns step".

That reads like you want period in 4000 million steps. ie 1 nS to 4 seconds in 1 nS steps.
That would take a very serious piece of equipment to generate such accuracy and display the claimed result.

The link you sent does not seem to do this. It didn't seem too clear what the link was doing except that it did mention lasers at the top, so maybe that will be enough for you.

 

[Chandra214]

With four transmitting channels, 680MHz can be distributed to 170MHz each.
So each channel(pin) has to send out a 1ns pulse with 4.88ns delay between each ( worst case scenario).
If I can find a circuit to generate a 1ns pulse like below:
http://www.edn.com/article/CA476913.html

can i shape the pulse better? or can the Micrel Laser driver receive this and transpire a better shaped pulse?

If I am to multiplex all four channels to drive the Laser, How can that be approached?
May be a very high speed MUX?


- Prathap

 

[trambolin]

Just a quick question have you ever considered how much energy needed to transmit this signal properly without getting lost along the way. I agree with negitron on that. And do you think they will withstand this load, without getting damage after some number of pulses?

I am just asking out of curiosity

 

[f95toli]

With four transmitting channels, 680MHz can be distributed to 170MHz each.
So each channel(pin) has to send out a 1ns pulse with 4.88ns delay between each ( worst case scenario).
If I can find a circuit to generate a 1ns pulse like below:
http://www.edn.com/article/CA476913.html

can i shape the pulse better?

The circuit you are linking to has a rise-time of 520ps +-100 ps according to the text; this is ten time what you asked for earlier.

I don't know how much experience you have with microwave circuitry, but I hope you realize that the difficulty and complexity (and price!) increases by many orders of magnitude when you go from operating at around 1-2 GHz (which can be done using "ordinary" high-speed electronics as in the circuit you linked to) to 20-30 GHz which is what is needed for a rise-time of 50 ps.

 

[Chandra214]

Just a quick question have you ever considered how much energy needed to transmit this signal properly without getting lost along the way. I agree with negitron on that. And do you think they will withstand this load, without getting damage after some number of pulses?

I am just asking out of curiosity

No we haven't considered it yet. But the assumption is:distance between the pulse driver and the transmitter is as less as possible.

The circuit you are linking to has a rise-time of 520ps +-100 ps according to the text; this is ten time what you asked for earlier.

I don't know how much experience you have with microwave circuitry, but I hope you realize that the difficulty and complexity (and price!) increases by many orders of magnitude when you go from operating at around 1-2 GHz (which can be done using "ordinary" high-speed electronics as in the circuit you linked to) to 20-30 GHz which is what is needed for a rise-time of 50 ps.

it seems, we might have to compromise with rise and fall times, may be upto 200ps.