Connecting Multipule Radios via Cable 3

[daKlone]

Hiya,

This is really outside my area, so I'm hoping you guys can come up trumps for me.

I have a project where I need to connect a number of radios together via cables (for training purposes they are not allowed to actually transmit over air to each other).

Now, I'd be quite happy connecting two together using appropriately rated attenuators between them, but I'm a bit stuck once it gets to more than one. To make matters worse, we have to control the signal level between any two radios independantly.

Now, I've come up with what looks like an extremely complicated design, with fixed attenuators, variable attenuators and ampilfiers wired up in a right old mess. I can't help thinking that there must be an easy way...and you guys must know it!

Any advice appreciated.

 

[Comcokid]

I'm looking at a similar issue right now in connecting a small network of 802.15.4 devices together so problems can be forced on the network and results analyzed.

I'm going to put a signal splitter (i.e. power divider) on each unit. Perhaps a 4:1 resistive type. Unused outputs will get a termination. By connecting in different ways I can use either the 6 db splitting loss or the ~25 db port-to-port isolation.

I will connect between the units with cables and a combination of fixed or variable attenuators.

I'm planning on isolating each unit in a metal enclosure to prevent direct radiation around any cables/splitters/attenuators I may have.

Mine is still somewhat a complicated setup. Does anyone else know of some better tricks?

 

[VEBill]

In general, you decide how much end-to-end attenuation is required and then split it between the transmiter and the receiver. If (for example) you required 120dB pathloss in total, then 60dB at each end.

With multiple units, I'd consider simply joining everything together after the attenuators. With that much loss, you do not have to worry about matching (VSWR).

Pay attention to power ratings.

Stay well away from both the noise floor and the damage level and it should work. It's a very wide target.

Standby for further advice from others before you embark on this (just to be safe in case I'm missing something).

 

[BrianR]

I agree completely with VE1BLLs solution. Connect a power attenuator (or a dummy load with a sampling pickoff - same thing) to each radio and parallel all the attenuated outputs into a 50 ohm load to make sure the attenuators work properly.

It would probably also work just running each radio into a dummy load and putting all the dummy loads into a box together. There's bound to be plenty of leakage especially if you use single shielded coax to connect the load.

 

[pstuckey]

Since you have to control the attenuation of each path independently I would use an N-way (N=# of nodes) splitter as Comcokid suggests and put a variable attenuator in each of the paths. Terminate the one unused port on each splitter, which then be used as a convenient test point if you need to see the spectrum at an individual radio without interupting the link.

It's surprising to me that you would need any amplifiers in your setup. Even a large number of nodes (high splitter loss) the output power from a radio will probably have to be attenuated to simulate a link of more than a meter or two. PathLoss=10*n*log(4*pi*distance/lambda) where n=2 (ideal case).

Peter

 

[Higgler]

I'm curious how many radio's in the test scenerios and whether the test is over a wide frequency range (where you need good flatness of the coupling)?

I see two simple options if you don't need flat responses over a wide frequency range.

1) buy a small metal box, add bulkhead connectors that can take RF to the inside of the box. On the inside, add a wire to your connector (small antenna, expect poor VSWR). You can add attenuators on the outside to set levels and help VSWR problems once you do some S21 measurements. RF bounces around the box and coupling is very high due to the box even though you may think cutoff waveguide phenomenon would reduce coupling alot.

2) A microstrip/stripline alternative could be printed having a circular shape microstrip circuit. Connectors on the periphery (dozens) with 50 ohm printed lines from the connectors to the center section, print a circle shape where all the 50 ohm lines meet at the center. With this, you could add microstrip variable attenuators in the circuit and use external attenuators on the connectors. This is a poor microwave lens and will have some gain dips if you need very wideband coupling. You could do a similar setup with a group of cables having ends stripped and center conductors and grounds tied together. It's a big mismatch, but you will couple energy to all the other cables.

kch

 

[daKlone]

Thankyou for your suggestions, chaps - which to be honest only some of which I understand!

There are a number of test sets, with 10 to 32 radios per set. The radio frequencies are in the range of 1Mhz to 500Mhz, so a flat response across this range would be preferrable.

I wounder if I could ask VE1BLL and pstuckey to expand on their suggestions a little (as they were the one I think I understood!) in perhaps slightly simpler terms...I did say this was out of my area!

 

[daKlone]

I've drawn a quick sketch of what I now think I need to do, I'd appreciate it if someone could take a look and tell me where I'm going wrong!

It's one "node" of a 4-radio system.

It would be nice if there was a way of building up to larger systems using a modular approach, but I'm not sure that it is possible.

Anyway, have a peek at the drawing and tell me what you think....

http://photos.fotango.com/p/eba00381610f00000005.jpg

 

[VEBill]

Calculate the require pathloss. In other words, convert the power output and the desired input signal level to convenient units such as dBm. Then subtract one from the other to calculate dB. Do the same thing again for min and max to create a range. Triple check you math.

Without checking the math, it should be something like 120dB (but you need to do the math).

Divide by 2 because there will be two attenuators in any path.

You could build a metal box with lots of connectors all joined together. You can only use it WITH the appropriate attenuators.

Matching is NOT (!!!!) an issue when each radio has (for example) a 60dB pad on the antenna connector. That gives each radio a perfect match (120dB retun loss, in theory) no matter what happens past that point.

As mentioned by Brian, you might be able to make the link just using dummy loads at each end. But I saw one example where even 1kW HF wasn't heard a couple of meters away using dummy loads. Some equipment is better shielded than others.

 

[daKlone]

Thanks VE1BLL, but how do I get the dynamic variable element between any two points this way?

For instance, if I need (simultaneously) 120db between radio A and B, but 100db between A and C, and 150db between B and C, I don't see how I could do it using the method you suggest (but I could easily be missing something!).

 

[Comcokid]

If cost is a concern an inexpensive suggestion for a lot of shielded enclosures - go to the local hardware store and buy empty new paint cans.

 

[daKlone]

Right - empty paint cans!

Did you take a look at the diagram? What do you think?

 

[Comcokid]

My application involves small transceivers that transmit in burst and receive on the same frequency, so my approach was to have a single cable connection via attenuators/splitters to each radio. The power levels involved are low - 10 dBm max or 10 mW.

I am currently thinking out a similar setup and when I saw your post, I responded with my approach. But, I'm not 100% sure mine is the best either. VE1BLL about connecting everything togehter after the attenuators may work for me.

I gather from your diagram that your application may involve radios with higher power, or that transmit/received at the same time on different frequencies or may involve different antennas for transmit and receive. Hence the need for a circulator and different cable paths for transmit and receive. Is this correct?

 

[daKlone]

Yes, my application has high-powered radios all transmitting/recieving on different frequencies at the same time.

The reason I put a circulator in was to stop the "recieved" signal from being re-injected into the "transmit" path - is that the right thing to do?

 

[Higgler]

Sidenote on isolation,
"120db between radio A and B, but 100db between A and C, and 150db between B and C".
It may be difficult to get 150 dB isolation levels. EMI/RFI leakage considerations for the hardware may be a real challenge, especially at the higher frequencies. Connectors and circuits leak more at higher frequencies.

It may be interesting to do a sanity check with radio's next to each other having cables and loads on their outputs. They may talk to each other thru extraneous leakage and you may have to shield each one separately. That'll change the physical setup and the cost if they are leakers.

kch

 

[pstuckey]

A couple comments on your diagram, circulators at low frequencies (<500 MHz) are narrowband, typically ~10% bandwidth. The circulator might be OK to use for a narrow frequency range, but if Tx and Rx freqs are very different it won't work well. So I think trying to get rid of the circulator is a good idea. There should be something in the radio (possibly a diplexer) to separate Tx and Rx already. Some radios are designed to interface to a RF distribution unit, and if that is the case here and the radio requires some additional external separation between Tx and Rx, I would consider a diplexer.

Building up to larger is systems is going to be tough if all radios have to see all others. It can be done by adding more splitters, but the plumbing gets messy and you might be better off just using N-way splitters then. Sometimes in simulations like this it is more realistic that not all nodes see each other at the same time, then you could use extra ports on the splitters to bridge between groups (of say 4 nodes each).

In general, I think you are on the right track, make sure your first fixed attenuator can handle the Tx power. It would be helpful if you can provide more details, such as specific frequencies and power levels, and how many nodes will 'see' each other simultaneously.

Peter

 

[daKlone]

Thanks Peter!

There are between 10 and 32 radios per installation that have to "see" each other. Their maximum output power is 100W and various models cover the frequency range 1Mhz to 500Mhz.

Do you think I could do without the circulator altogether?

I just wasn't sure how to stop the transmit signal "feeding back" through the recieve path. I think I need distinct transmit and recieve paths (rather than just connecting everything together in a "lump") in order to be able to control the path loss between any two points independently and simultaneously.

 

[pstuckey]

I think it is likely you can do without the circulator, although ideally your simulation will mimic the real system as closely as possible. So if there are separate Tx and Rx antennas in the real system, then you may want to separate the paths (and use the same method of isolation as the real system). If the radio is (more likely) connected to one Tx/Rx antenna then you are probably ok without it.

100 W (50 dBm) Tx is alot of power. In general, to use inexpensive components you will want to keep the power well below 1 W. So you could use a high power, fixed 30 dB pad between the radio and the splitter. The signal would then go through a 30 dB pad, 32 way splitter, variable attenuator, 32 way splitter, and a 30 dB pad to get from one radio to the other. This would result in a MINIMUM attenuation of 90 dB (likely a few dB higher). And you could increase the attenuation from there with your variable attenuator. If we look at the path loss equation above (and assume for simplicity that the system antenna gains are 0 dBi - if not adjust the numbers accordingly) then 90 dB path loss corresponds to a range of ~3 km. If you want to simulate smaller path losses (shorter ranges) then you can use a smaller fixed pad, but then the splitter will get more expensive. Werlatone is one company I've dealt with that makes high power splitters. If you can use the larger pad, you can get less expensive splitters from Mini-Circuits.

Peter

 

[pstuckey]

BTW, the path loss example in my previous post is for 250 MHz.

Peter

 

[VEBill]

"dynamic variable element" - does the pathloss have to change over time? Or are you referring to the various losses over various combinational paths?

Most receivers have a huge range of acceptable input power levels (in volts, perhaps 0.5uV up to 10mV, square the v for P). That huge receive input range should FAR exceed the most-likely relatively small variation in power output from one transmitter to another.

In other words, if one of your transmitters is a bit higher in output power than the others, then simple put slightly more attenuation on that radio. The extra 3 or 10dB should still be well within the relatively huge acceptable range going into the various receivers.

You could make a matrix and solve for the optimums.

Or simply pick a resonable value for the power at the huge junction point (where everything connects, in the paint can :-] ), maybe 0dBm or something similar.

The key point is that the receivers will have a very wide range of acceptable input levels. Far more than required to level out anything you'll likely run across.

You'll need to do the math and check everything twice before you hit the PTT.