Hello!

I need to switch 10x 50-ohm lines on 6 different paths for each line. Up to now, I've solved this case by using a lot of analog switches where each of the 50ohm line had 6 different analog switches connected to them. The signal on the lines is 0-5V up to 160MHz. The issue that I am having is that the circuit does not perform well at high frequencies, presumably because of all the parasitic capacitance of the switches combined. I can achieve correct data only up to cca 85MHz. In order to reduce that, I am thinking that I'll utilize analog multiplexer with enough bandwidth for my application. I am wondering, to properly terminate the lines, can I just use mux internal ON resistance into account (because it is nearly 50ohm)? Or should I still use 50ohm resistors?

Also is there any other way to achieve that frequency passthrough for my design? Perhaps any special ICs exist for my use case?
Maybe I am looking at the wrong way - please let me know.

Thanks in advance.

Best regards

 

Analog video switching might be what your looking for .
Many many chips.
May be the adv3226 type is what your after.

 

There are switching arrays made for that purpose, packaged and connectorized and ready to play. Pick your connector typeand control scheme. They use reed switches for switching .Pickering has been making switching systems for years.. Very common for automated testing systems, but not cheap

 

to properly terminate the lines, can I just use mux internal ON resistance into account (because it is nearly 50ohm)? Or should I still use 50ohm resistors?

Using the internal ON resistance is problematic since there is typically a large variation in that value (look at the spec sheet).
You need a switch with an ON resistance low compared to 50 ohms and then use a 50 ohm resistor termination.

 

Hi,

• What is the aim of the project?
• Is each signal only routed to only one path or multiple paths simultaneously?
• What design specification are you aiming to achieve? For example - max insertion loss for each channel, maximum crosstalk between channels etc
• You mention "I can achieve correct data only up to cca 85MHz", are these signals digital (TTL) in nature?
• How do you know the data is correct? What measurements on the data have you taken e.g. Oscilloscope traces?

You need to work from measurements to understand what is happening.

However here are some other points to consider:

• Have you used 50 ohm traces on the PCB?
• Have you considered signal crosstalk in the PCB design?
• Have you used a single analogue switch per package or multiple switches per package?

Best regards
James

 

Contact the PICKERING company directly! They are well qualified to design and create switching system packages. It is what they do. My best advice is to take their advice.

 

How much on/off isolation do you need? That will have a bearing on what type of switch / layout will work for you.

 

Hello!

@drjohsmith Crosspoint switches unfortunately will not be suitable for my application, because I need to switch a single input bus to the one-of-the-six output buses. Also each of the line can act as IO. So the switches should conduct in both ways. Following drawing shows the situation:


@Kiwiboy
1. The aim of the project is to control the IO group selection programatically.
2. Each of signal in a group is wired to the corresponding signal in a group selected with a ''mux'' drawn.
3. The design specification is that a data is correctly propagated further through some kind of a switch. Crosstalk and insertion loss should be minimal, so that signal is not corrupted, however this should not be a problem I presume.
4. Signal is TTL, however I would need to sample analog signal on the other side because we use calibration mechanism where we set the threshold voltage of a signal received to compensate different voltage levels on a signals etc..
5. Each of the trace represent a data segment which are then captured on the other side and translated into something meaningful

6. 50ohm traces are indeed used on PCB, also crosstalk is minimized with careful PCB layout + ground plane below
7. Each of the analogue signal is in their own package. We are thinking that parasitic capacitance can be the problem here. I was also looking other possible candidate such as DG2034E, but first, I want to make sure that this is indeed the problem in the first place.

@HasBeen I am not enitrely sure what you mean by that.. Can you provide more information? Thank you!

Best regards

 

Given the list of difficult but reasonable requirements I would repeat my suggestion made in post #6. The one unanswered/unasked question is how many of the switching packages will ultimately be required? That will certainly affect the price.
The big challenge is that this project is quite far outside of the common design realm.

 

Your saying a 50 ohm bus of data.
With a square wave on them , but needs to be bidirectional.
This is sounding a contradiction.
Is this digital or analog signal?
What do you mean by bus ?
Whats the signal that's no directional ?
If analog , what are the requirements for bandwidth , thd, snr ?
If digital , what rise fall time do you need, what pulse compression / expansion do you want , what propergation delay.
Tell us a lot more about the inputs and output characteristics and we can help , if you keep us being feed little crumbs were liable to miss lead you.

 

Thought I posted this but don't see it now.
This article may be of interest.

 

Thought I posted this but don't see it now.
This article may be of interest.

good link,
the OP now wants bi directional signals though !
all very strange

 

Data communications do sometimes travel both ways. Consider USB as an example, and the new 2-wire Ethernet scheme. So bidirectional makes perfect sense to me. The smart pressure transducers communicate in both directions, although not at the same time.

 

Data communications do sometimes travel both ways. Consider USB as an example, and the new 2-wire Ethernet scheme. So bidirectional makes perfect sense to me. The smart pressure transducers communicate in both directions, although not at the same time.

Totaly agree @MisterBill2 , some communication is bi directional, I2C is another example ,
as is a good old fashioned telephone,
point I was highlighting is the specification the OP wants is coming out in dribs and drabs,
wonder what the next constraint is going to be ?
OK, if the OP is refining ideas as they go,but wonder if there are any more requirement that are needed,
or are they not needed ?

 

I suggested a company that has been producing a wide variety of switching systems for at least 30 years. That is Pickering company. And, by the way, most of their stuff meets MIL specifications as well. It might even be on their shelf, ready to ship.