[SOLVED] Analog Delay Line 50Mhz

[Basil1402]

Hello.

I need to build an analog delay line with the following criteria :
- process 50MHz sinusoidal signal (dynamic and DC level not important)
- delay = 20 ns (ie one period)

The point is : I really need a delay and not a phase shift. For instance, if the amplitude of my input signal is changed, I need the output to change accordingly 20 ns later. I am not sure if I am crystal clear here :|

Do you know any structure (passive or active) to do that ? I tried LC ladder, with no real success.

Thank you for every hint you might give me.

 

[BradtheRad]

I once heard color tv's have to delay an internal signal (chroma or luma). The time required is a microsecond. So they wind several hundred feet of wire around a tube. It works.

 

[KlausST]

Hi,

what about 4m of RG-58 cable?

Klaus

 

[Basil1402]

I forgot to mention that it's for an embedded system. So long cable won't be ok. But thanks anyway.

 

[jiripolivka]

Try Murata, they make a selection of miniature ceramic delay lines like you can use.

 

[Basil1402]

As a matter of fact, it looks promising. But I have the feeling that they are designed for digital signals, don't they ?

 

[zorro]

Hi Basil,

Please try all-pass filters.
If the bandwidth of your signal is narrow or some (linear) distortion can be tolerated, a very simple one can be enough. Otherwise, a higher order could be required.
Regards

Z

 

[tggzzz]

You used to be able to do that kind of thing with SAW delay lines. I don't know if they are still made or if they have the right specs.

 

[Dan Mills]

SAW acoustic lines were what was usually used in analogue PAL telly for time aligning the luma and croma components.
Sort of old school these days.

Regards, Dan.

 

[FvM]

Low-pass filters with smooth characteristic (e.g. Bessel) have almost constant group delay in the pass band and are better suited as delay line than all-pass filters, I believe.

I don't see why LC ladders don't work for you. They are often used to implement small delays. You should calculate something like 10th order for one period delay with flat group delay up to the intended bandwidth.

 

[tggzzz]

SAW acoustic lines were what was usually used in analogue PAL telly for time aligning the luma and croma components.
Sort of old school these days.

That was one use; they had many others with a variety of characteristics, e.g. in radar.

 

[FvM]

PAL delay lines have to store several hundred signal periods, this would be in fact hard to accomplish with LC delay lines. But a single period is quite easy.

 

[D.A.(Tony)Stewart]

All analog scopes have video delay lines for the signal to be displayed so the can trigger the sweep before it hits the phosphor

A cable is a distributed network of L, C with a characteristic impedance with a source load R to match to get a flat response with no echoes or high return loss.

**broken link removed**
**broken link removed**

The Vbat ought to be a video source with same source impedance as Load.

It can be constructed with solid magnet wire around a form but should be at right angles or planar with ground plane.

Interwinding capacitance should be low by gapping each turn by >3d (wire) so they do not couple to each other.

Consider ground plane or tin/brass shield around it solder to ground plane with tabs.

Then choose your impedance eg. 50 or 100 Ohm and calculate L in nH then use a calc. for turns and diameter, D with wire thickness d.

Did you forget to match impedances ? L orientation or capacitive coupling?

More info http://www.rhombus-ind.com/dlcat/app1_pas.pdf

- - - Updated - - -

However your requirement must define bandpass ripple, phase shift and band reject levels to be testable .

Note that time or group delay in a bandpass is the derivative of phase shift over the band, so you can't have zero phase shift.

I really need a delay and not a phase shift

 

[Basil1402]

Thank you all for your answers. I have new paths to investigate now.

My main focus now :
- Murata Ceramic Delay lines
- Coax wire (even though quite bulky)
- LC ladder

Thanks again, I consider my problem solved:thumbsup: