Q: Biquad Antennas

Do you have an URL for a floating reflector biquad?

While both are at ground potential, I consider the surface to be a ground plane when the spacing from the element to the plane doesn't matter, while a reflector has a critical spacing between the element and the grounded surface.

A ground plane would be under a vertical antenna.

I'm of the school of grounding the reflector, but there are cases where reflectors float, such as in yagis.

snipped-for-privacy@cox.net (Arthur Shapiro) hath wroth:

You rang? I've been wanting to rant on the topic of biquad construction. There are far too many erronious biquad articles.

The antenna will work with either a grounded or floating reflector. It doesn't really matter. I've done it both ways with no obvious difference in performance.

Note that the 2 quad elements are literally floating above the ground plane in these models.

You might be looking at this construction article mounts the N connector directly on the reflector without a coax extension section to the quad elements. Scroll down to the biquad section. This is the WRONG way to do it and will not work at all. Even the author claims horrible performance in the article. Little wonder why.

This is more of the same, with the antenna wrongly constructed: |

pair of wires shown, between the connector and the biquad elements should be a 50 ohm coax section, not an open wire pair.

Even more biquads built totally wrong:

the same problem. Amazing.

This one attempts to solve the problem by moving the connector ground to the antenna elements:

the ground and exposed center ping lead length are still excessively long. Also, don't try to drill circuit board material with a large wood drill bit while holding onto the board. Great way to rip up your hands.

So much for the wrong way to build a biquad. The following are some proper biquad constructions articles. Note that the coax cable extenders *THROUGH* the reflector in all examples. The center conductor is only exposed when it hits the quad elements and then as little as possible.

(dead?) (For USB radio)

A really marginal construction idea (mine):

mistake I made was to support the ends with nylon spacers. The nylon barely works at 2.4Ghz and connects to the antenna at the worst possible high impedance point. Removing the nylon spacers increased the gain about 1dB. However, I've only measured 8dBi of gain but it should be 2dB higher, so something else is still screwed up.

Note that there is really only one critical dimension, the cut length of each quad element. Whether it makes a perfect square is totally irrelevant. The length of the "loop" should be 1 wavelength or

12.5cm. The problem is from where to where does one measure. The end that hits the center conductor is easy enough, but where the ground end hits the coax shield is a problem. If the two quad elements don't hit the shield in the same exact place, half the length BETWEEN the elements need to be included in the loop length. It's tricky and varies with the method of construction. In all cases, the loop length is never longer than one wavelength (12.5cm) and always slightly less. It's probably safe to cut the loops a bit short.

However, the cut length is seriously critical. At 2.4GHz, the difference in length per MHz is: 3*10^10cm/sec / 2400.0 MHz = 12.5 cm 3*10^10cm/sec / 2483.5 MHz = 12.1 cm (12.5cm - 12.1cm) / 83.5Mhz = 0.005cm per Mhz With 5MHz channels, you have to be with 0.025cm to hit any channel.

The only reason that it works at all is that the biquad antenna is fairly broadband and will work reasonably well even if miscut and mistuned. The VSWR and gain curves versus frequency are fairly flat over a wide range of frequencies.

The height of the antenna only affects the VSWR (how close to 50 ohms the antenna appears) and does not affect the center tuning frequency much.

Without a directional coupler and signal source, or network analyzer, your chances of getting it perfect are minimal. So, build one:

was a site showing how to make a directional coupler out of two pieces of semi-rigid coax cable, but I can't find it.

You can use Netstumbler to test antenna gains:

everyone seems to make the same mistake. They mount the antenna on a camera tripod and try to connect to a distant access point that's also at approximate ground level. The problem is that the Fresnel Zone extends to the ground an makes a mess of the numbers. Assuming a typical camera tripod 1.5 meters off the ground at both ends, the maximum range that can be used for testing is 130 meters (for 0.8 times Fresnel zone radius). without the Fresnel Zone issues, reflections from the hard street will cause problems. Maybe do the test over an RF absorbent grass lawn or park will give better results. It's easy to tell if you've got reflection or Fresnel Zone problems. Just raise or lower the antenna. If the gain moves, you've got reflections.

Here's one; hope it is clickable as the URL is split weirdly as I construct the message:

Art

So after digesting your pleasantly voluminous message (thank you!), let's see if I understand. I was really hoping to mount the N jack on the ground plane, and it sounds like that is OK, even if not universally done in your references.

But you assert that the critical factor, whether or not the jack is electrically connected to the ground plane, is that short 15-18mm gap between the jack and the actual antenna. It HAS to be a piece of coax - the alternative of two soldered pieces of copper wire or a notched cylinder of half inch copper pipe will severely detract from performance. Correct? I assume that's an SWR issue.

And doubtlessly the scraps of cable TV coax I have about the house are the wrong impedence.

Art

I pieced together the URL. Pretty funny seeing the dish on the car roof.

I can't really tell if the ground wire connects to the reflector or not. Since they didn't use PCB but rather copper sheeting, it is clear they didn't make a connection from the back.

I saw Jeff's reply and if floating is good enough with him, I would float it only because soldering to a ground plane is work. Generally you want to tin the area where the connector will be soldered, then try to solder the connector without melting the inside. I would much rather just epoxy the connector to the reflector, which means the only wire you have to solder is the one that brings ground to the biquad element.

You will undoubtably ruin one or two until you get the hang of soldering to them, so have a few on hand. If you don't ruin one, then build another antenna! If you have a ham flea market in your area, these connectors are a quarter. Buy a handfull.

If it is really necessary to use a coax fed, then I would solder the copper pipe to the ground plane. If the RG-8 insulator isn't inserted, then you can blast the pipe and copper with a torch without fear of melting. So scratch the epoxy I mentioned in the other post.

I built mine just using copper wire to reach the biquad. However, I don't have any gear to measure SWR at wifi frequencies, so my construction may not be optimal.

snipped-for-privacy@cox.net (Arthur Shapiro) hath wroth:

Ummm.... you missed my point. The coax cable needs to extend all the way to the antenna. If you mount the coax connector on the reflector, you will still need to attach a short piece of coax cable or simulated coax cable made out of tubing, from the connector to the antenna. That's not easy. It's MUCH easier to run the coax through the reflector and deal with only one messy soldering job, than to put the connector on the reflector and end up with two messy soldering jobs.

No, I said the critical factor was the length of the pieces of wire used to make the quad antenna elements. The height of the quads above the reflector is not particularly critical.

What I said about that gap was that it MUST be coax, not two pieces of wire.

Severely is an understatement. As several of those that built it like that mentioned, it doesn't work.

No. It's NOT a VSWR issue. The exposed extra wires in place of the coax cable become part of the radiating elements and totally mangle the radiation pattern. There will also be a significant change in VSWR, but it's not as damaging as having the pattern and gain all screwed up.

Actually, I use RG-6/u and RG-11/u (75ohm) coax for LOTS of antenna projects. (RG-59/u is garbage). If I'm trying to squeeze every bit of gain out of the antenna or match the tests to the model, I use LMR-240 or LMR-400. However, if I'm trying to throw something together that's not particularly critical, I'll use 75 ohm coax. The VSWR and resultant loss is negligible. See:

VSWR is only 1.5:1 which is about 0.18dB loss. Not much. The

75 ohm coax actually has less loss than the equivalent length of 50 ohm coax, so the loss is even less.

The big problem is connectors. I use CATV F connectors and quad shielded RG-6/u. Going from F to RP-TNC or RP-SMA is a bit of a problem. I make my own adapters or use 2 adapters, but I suspect I'm losing something in the connection.

Anyway, follow the examples I listed that are known to work. If you wanna be creative, at least build one that works so you can compare performance.

snipped-for-privacy@cox.net (Arthur Shapiro) hath wroth:

That's one of the examples of how *NOT* to build a biquad. Same problem I mentioned in my rant. The coaxial cable needs to extend all the way from the connector to the quad elements. I don't see any easy way to salvage the design and recommend you consider one of the others that do it correctly.

Incidentally, you might want to look into a sector antenna.

(very slow loading) gain, wider coverage area, and rather easy to build.

the coax feed.

They have a variant I haven't seen before, which using a director with the biquad:

Refering page:

Ahah...I think you're starting to beat some sense into me- thanks!

In my case, I've already purchased the appropriate connector to N pigtail for my particular card, so I'd really like to use the bulkhead-mount N jack that I purchased (25 cents? Hell, it was five bucks, but at least is silverplated.) So mounting on the reflector is logical, and then apparently I'll use a little stub of some coax soldered to that jack to support the biquad appropriately off the plane of the reflector. If I've still missed the boat, hit me again. And to just dwell on one of your points: half-inch copper pipe with a wire suspended inside is NOT a coaxial equivalent - correct? That's a shame, as it fits nicely around the business end of the N jack.

I don't seem to have any copper-clad pc boards in the scrap box right now, so it's time to find either that or some sort of craft store that might carry copper sheeting.

Art

snipped-for-privacy@cox.net (Arthur Shapiro) hath wroth:

We try harder.

Sigh. Well, Mr Miso(?) found a good construction article that will make good use of your N bulkhead connector:

don't duplicate the sloppy screw mounting of the N connector flanges. I think that's just to illustrate how it goes together. Otherwise, it's an excellent construction method.

If you must do it with a piece of coax, then make sure that the center conductor can slide up and down. Extend the center conductor from the shield. Solder the coax center conductor to the connector center conductor as close to the insulation on the coax as possible. Then slide the shield over the N connector and solder it to the connector or reflector all the way around the connector. Unless you are VERY good at soldering, you're going to have problems doing this. (At least I did when I tried it and I'm really good at slobbering, err... soldering).

Personally, methinks assembling the coax section out of copper tubing and wire makes more sense. See the URL above.

Naw, you're close but are trying to make things difficult. Methinks the problem is that you don't seem to realize how critical cable connections can be at 2.4GHz. Even the smallest amount of exposed center conductor will cause an impedance mismatch. Fortunately, it's not too horribly critical in the connector extension (as evident from my use of 75 ohm coax) as long as it doesn't radiate. That's where I think you missed my previous remarks. Using two wires in place of the coax feed are a problem because the wires will radiate. If I have time, I'll do an NEC2 model of this wrong way to build a biquad and demonstrate the problem.

It most certainly is a coax cable. If you look at the biquad URL's I listed as being properly constructed, some use coax cable simulation out of copper pipe.

The trick is to get the impedance close to 50 ohms. For air dielectric. Z = 138 log b/a Z = impedance in ohms b = inside diamter of outer tube a = outside diameter of inner conductor For 50 ohms, b/a is about 2.3.

Trick. If you need support for the center conductor, I've used styrofoam peanuts from packing material held in place with hot melt glue.

Try copper roofing flashing from the local hardware store. The stuff on rolls is fairly cheap.

Sheet aluminum also works, but you might need to do some glue and goo instead of soldering to the reflector. A sloppy way is to take a large copper or brass washer and drill 4 holes that align with the 4 holes in the N connector. Stuff the coax braid between the reflector and the washer and tighten the screws. Instant compression joint.

I'd going to make another biquad as Jeff really believes the coax is a must. However, I would go for the copper pipe as it will make a better base to held to the biquad driven element. I use a pipe cutter, but to make the tilt on the end, I suppose a hacksaw will be required.

I don't know if this is a good idea, but where you solder the N connector to the copper, you might drill some hole to act as thermal relief, i.e. make the copper a less efficient heatsink. I would make them small as all you are trying to do is break the thermal path. Again, tin the copper first.

I also want to protect the design from getting banged around. I'm assuming a fiberglass radome won't be a problem.

Jeff Liebermann hath wroth:

Sigh. Well, the above is wrong. I hate it when that happens.

I built a model of the screwed up biquad antenna construction method with 4NEC2. I expected to find lousy gain, a weird pattern, and sky high VSWR. What I got instead is normal gain of about 10.5dBi, a normal antenna pattern, but a sky high VWSR (6.7:1). See:

you want to play with the NEC2 model, see:

It's still a lousy way too build a biquad due to the very high VSWR, but it's not as lousy as I originally predicted.

Grumble, grumble...

No ham would put up with more than maybe 3 on VSWR, so I'm going to go with the copper pipe. It is also better mechanically.Those wifi cards get pretty toasty as it is, so the lower the VSWR the better.

Is the director element worth the effort? Those guys are artists when it comes to bending and solder wire. My results are not nearly as clean. I can't see how they soldered the director so cleanly without any mechanical support to hold the wire in place. The usual practice when you solder is to twist the wires first but clearly that won't do in this kind of antenna construction.

snipped-for-privacy@sushi.com hath wroth:

Thank you for not suggesting an antenna tuner. As for tolerating VSWR, well my junk HF antenna is really awful should be replaced. I tolerate high VSWR, but just barely. Unfortunately, my ancient HF rig doesn't.

If you could make the two wires to be 50 ohms instead of whatever they turn out with 0.200 air spacing, the antenna could be build without a coax section. It would be tempting to try a piece of G10/FR4 circuit board, with a 0.1" wide trace for 0.062 board, with everything else grounded and copper tape over the board edge. It's stiff enough to support the biquad. Hmmmm....

The radios made to tolerate infinite VSWR at the antenna if they have an external RF connector. I'm not so sure about PCMCIA or USB cards that have had a connector or pigtail added. 50mw isn't going to blow up or overheat anything. However, the higher power boards are another story. Dunno. I've never blown up a board and seen no reported radio failures resulting from antenna experiments. My guess is that some home made antennas have fairly awful VWSR. There would a flood of "it blew up my wireless router" postings if there was any possibility of failure.

I don't know. I'll have to model it. I don't think it will work with a wire support. I think it should be insulated. However, after my last bad guess, methinks it best to wait for me to run a model. the author YU1AW obviously ran NEC2 models through 4NEC2 to get the plots on his web pages, but apparently didn't do the same for this antenna. Methinks it might be an experiment that needs work. Dunno.

Did you see my calculations as to how critical the cut lenths are? At 2.4GHz, the difference in length per MHz is: 3*10^10cm/sec / 2400.0 MHz = 12.5 cm 3*10^10cm/sec / 2483.5 MHz = 12.1 cm (12.5cm - 12.1cm) / 83.5Mhz = 0.05mm per Mhz If you're 0.05mm off in your cut lengths, you're 1MHz off frequency. Miss by +/- 2mm and you're outside the 2.4Ghz band. Either gets some decent test equipment so you can do cut-n-try, or be VERY precise in your cutting and soldering. However, that's only if you want perfection. The gain and vswr curves for a properly built biquad are very flat, which means you can be fairly sloppy and still have something that works.

From carpentry: Measure twice. Cut once.

Incidentally, I'm borrowing an old Wiltron network analyzer, which should allow me to post pretty pictures and accurate construction measurement. One catch. It's broken and I have to fix it first.

Two things come to mind regarding the measurement accuracy. One, since it is the length of the loop that matters, maybe we shouldn't be so concerned about getting the wire to fit the pattern as much as just measuring the right length of wire, then just bending it.

The other thing that comes to mind is maybe a helix is the way to go. At least it is very forgiving in accuracy, i.e. it is a wide bandwidth design. The gain isn't all that high on a volumetric basis, but if you can't build the biquad on target, you don't achieve the theoretical gain anyway.

I've noticed the satcom guys use a 4 turn helix and a dish. Now of course in satcom, you need circular polarization.

Exactly. It's the end to end length that's important, not how well you simulate a square (quad) element. I sometime make diquads using a circular loop. Works fine.

Well, you lose -3dB from the mismatch between circular and linear polarization. That's a small price to pay for insensitivity to polarization differences. For a point to point link, two helicals are insensitive to single bounce reflections. They tend to be very broadband (if properly matched).

However, all is not wonderful in helixland. They're difficult to build. They're difficult to match to 50 ohms. They tend to become huge for high gains. (You should see the monster pair that I have). If wound on cheap PVC, the PVC intruduces some loss.

Some of the SETI and ham satellite nerds also use circular polarized patch antennas and dual mode cans:

What they're trying to accomplish with a helical dish feed is match feeds pattern to the f/D ratio of the dish. See: the section on dish feeds in section 6. Sorry, there's no sub-section for helical feeds yet. The basic idea is to have the -3dB beamwidth of the antenna hit exactly the edges of the dish. Half the power stays in the dish, the other half goes off the edge. (That's part of the 50% typical dish efficiency). The helical is nice because it's really easy to adjust the beamwidth by simply adding or subtracting turns.

I may give that a shot. BTW, the biquad doesn't seem to be in the ARRL antenna book, or at least in my 16th edition.

I've been making fiberglass tubes for lightweight containers if fiberglass is any better than pvc. It is certainly lighter. Not cheap though. If anyone is curious, I can write it up as there are quite a few tricks to construct composites in your garage.

Does the dish reverse the polarization?

snipped-for-privacy@sushi.com hath wroth:

There are some that would consider that to be a compliment. The way a biquad works is quite easy. You start with a single full wave loop antenna. End to end impendance is 100 to 120 ohms. Put two loops in parallel, and you get about 50 ohms. The quads are fed symmetrically (in phase) with a half wave seperation, so there's no cancellation. Add a reflector, and you get some gain.

I used to buy the fiberglass scrap from the local sailboard and sailing shops. Fiberglass is way better than PVC. The acid test is to put the material in the microwave oven and see if it gets hot (or melts). There are many different types of PVC with different characteristics. The white stuff stays fairly cool. The grey stuff gets too warm. The black stuff melts.

Incidentally, an easy test is to take your access point and slide a piece of PVC pipe over the antennas. If the material is truely RF transparent, there should be no change.

Yes. Incidentally, the correct term is that it reverses "sense" as in RHCP and LHCP. Each surface reflection changes the polarization sense.