Design delay circuit

Where I work we have several rooms (one in QA, at least one in engineering) wired for testing our protocols over very long cable runs, to flush out timing issues like this. The larger of these rooms has I think 10,000' of cable with taps every 150 feet. It looks very impressive because it lives in a huge cable run on the ceiling - it looks like the data feed to some massive supercomputer on the other side of the wall. Only this computer is always the OTHER side of the wall, no matter which side you're on - because this cable goes nowhere and does nothing :)

The smaller room has only 5,000' of cable. It's wrapped in a neat spiral around the walls of the room (which is small and square).

Can you just roll it up on a cylindrical form, or are there coupling issues to watch for?

...Jim Thompson

Hi all i have a problem. In my circuit i have intoduced a delay time with a coaxial cable, the delay time is approximately 30 ns. My question is: Is there a solution without a coaxial cable?? Please help me I'm deprived of hope!! Thanks very much! Bye Bye Pazzeo

You can get passive LC delay lines in 8 pin DIL packages.

Bongiorno Pazzeo,

Passive delay lines are one solution but they tend to be a bit costly. Another solution is a simple LC delay. It depends what you want to delay. If it's digital then an LC followed by a Schmitt trigger would be what I would do.

If it is high frequency analog stuff then you would have to crack out the Williams Filter Design Handbook to design it with discrete parts. Or, as Andrew suggested, use delay lines. Keep in mind that delay lines need to be carefully matched for their characteristic impedance. Otherwise their frequency response looks like a pretzel. I usually drive them with a current source and terminate the end of the line with whatever the data sheet says it wants.

Just don't expect a 30nsec delay line to give you hundreds of MHz of bandwidth. If you need large bandwidth the coax is the ticket. You could use a small one and curl it up if space is a problem.

Regards, Joerg

Hi Jim,

Yes, but I have seen longer delay lines than that. If you take the good stuff (mil) it can be a tiny coax and then the rolled up package isn't so bad. Some lowpass effect can be compensated so even lossier stuff can suffice if $$$ is an issue.

Once I did that the poor man's way. I cracked out the old Motorola MECL book to calculate the Z and meandered the line between VCC and GND planes. Worked great and cost next to nothing. It just looked odd, but only on the plotted layout.

Regards, Joerg

Hi Jim,

I meandered it and used ground fingers plus vias in between. That scheme should be doable as a coil as well.

Regards, Joerg

I thank you very much for your help. I need large bandwith but i don't work with coax because the space is small and i search a method to reduce the line of delay in microstrip . Another problem is that I want a variable delay time and with coax i don't do it because i don't do cut and paste the cable every time. The delay time must be around 30ns! Bye Pazzeo

"Joerg" ha scritto nel messaggio news:%u_Ed.9576$ snipped-for-privacy@newssvr13.news.prodigy.com...

Hello Pazzeo,

Let us know how much bandwidth you need. Also, is it digital or analog? That makes for a huge difference in available solutions (with digital being pretty easy). If digital, do you have Schmitt trigger inputs available? What is your group delay flatness requirement over the bandwidth if analog?

You could do it on the circuit board if the board is large enough. But you'd have to pretty much dedicate a whole layer that is protected between ground/supply planes since 30nsec is quite a lot of trace length. As usual with RF layouts, no sharp turns and corners. You could have meander taps that are connected to a multiplexer to select delays. Again, if it was digital there are many other solution such as controlled LC networks with FETs, PIN diodes and other stuff. Works for analog, too, but it depends on the required group delay flatness over your bandwidth.

There are also schemes that transfer the signal into 'sound' in a ceramic and then back to electrical. It is used for TV applications but this can get pretty esoteric unless it is a very large volume production run.

Ciao, Joerg

Hi , the bandwith is from 400Mhz to 800Mhz and it's analog... "Joerg" ha scritto nel messaggio news:vSdFd.1086$ snipped-for-privacy@newssvr14.news.prodigy.com...

Hello Pazzeo,

That is next to impossible with discrete components and 30nsec delay. Not that it is totally impossible but the tuning for such a structure to obtain a flat group delay could become quite laborious.

I think if you don't want coax the scheme with a long trace on a circuit board sandwiched between solid planes is worth considering. But it must be calculated carefully so you keep a very defined trace impedance. If you have ground traces side by side with the signal trace and lots of vias in the ground traces you could get away with a pretty skinny pattern. Of course, you would have to adjust for frequency response but that is pretty easy.

If you opt for a small pattern you may have to use a spiraling trace or at least partial spirals, in order to avoid sharp turns that could ruin the impedance and cause reflection. I have done a few of these but not this long. Most of mine were up to 30cm or so.

Regards, Joerg

Thank you for you help. I will use spirals, but now i have a problem. I simulate the circuit with ansoft designer. In this software i 'don't know how insert spirals. Do you know about this? Have you a help for Ansoft??

Bye Pazzeo