Ever since my previous adventures into TV antenna Frankenswitches, I've been looking for a simpler/cheaper way to accomplish something similar. A design for an antenna autoswitch that others could perhaps build/use for their own MythTV / SageTV setups.

DisEQc satellite multi-switches are widely and cheaply available. This past weekend, I picked up a simple 4-way DisEQc-2.0 switch for CDN$8 from Active Surplus.

Now, designing/incorporating a full DisEQc protocol front end for one of these would be optimal, but also complex and potentially expensive. So I'm thinking of something much simpler.

The disassembled DisEQc switch is shown below, and a larger version is here.

It appears to use pin diodes for the switching, with fairly straight and simple signal paths through those devices. One pin diode per input (four in total), plus two extras near the middle where they connect to the downfeed. Those two extras have me a little puzzled, but I think they're there just to provide additional isolation between the two pairs of inputs (left and right).

Each of the primary four pin diodes has a cascaded pair of transistors driving the bias, to switch it on or off. These four pairs of transistors are wired directly to GPIO pins on the unlabelled PIC controller chip on the board.

There is one extra pair of transistors, whose function I have not determined -- but that probably has something to do with powering the PIC. EDIT: they're for sending response packets from the PIC back to the receiver.

My plan, is to remove the PIC from the circuit board, and replace it with four TTL signals from the parallel (printer) port of my MythTV PVR box (or perhaps connect it through another one of those USB / GPIO gizmos).

Plus a ground wire, and a +12V supply line for the multiswitch, rather than trying to wire the +12V into the downfeed coax as is normally done with DisEQc.

Comments, suggestions, and wisdom would also be appreciated.
I'm a software/firmware geek, not an RF engineer!

Cheers

Mmm.. one thing it needs, is an added DC-blocking capacitor on each of the four inputs, to prevent the +12V switching voltage from running up to my antennas (and possibly shorting out there).

Any suggestions as to the appropriate value there for passing UHF/VHF television frequencies?

-ml

Is that the inside of a boxed commercial product? It looks to my untrained eye as if their sweatshop needs to get better orphans -- that's some pretty haphazard-looking soldering, especially considering this is RF.

Peter

Yes, that's the 35x52mm PCB from a 4:1 multiswitch. The RF paths look reasonably straight, for a CDN$8 device. Most of the crooked stuff is in the control circuitry.

Mmm.. The reference design I found for a 2GHz switch uses 56pF blocking caps.. so perhaps I'll try something close to that. But better suggestions are more than welcome!

All those coiled lines are feed inductors to provide DC with a high impedance to the RF traces. There is a shunt cap at the end of each one where the DC comes in. The crappy soldering of those caps shouldn't affect the signal path.



I would put in 1000pF to be safe. Even at 200MHz, the 56pF will be about 14 ohms. At VHF channel 2 the 1000pF would look like 3.5 ohms.

I think these switches normally pass 900-2000 MHz, so the feed inductors are probably tuned to about 1500MHz. You'll want to check your lowest channel to see if you have too much insertion loss. Also, these switches may not have the best isolation between ports (18-30 dB).

You may be able to remove the 12V supply if you want. You would remove the output transistor feeding the feed inductors and put your 5V control line there. You would want to scale down the 2.2Kohm by the common port to keep the same current through all the diodes as you would with a 12V supply (1kohm would probably work). Maybe put a 1000pF and 0.1uF on top of the already placed caps at the inductors to provide more filtering to the 5V supply.

If you REALLY want more isolation, you can hack this board to be the same as the schematic you posted earlier from the other message board. You would need +/- 5V though and another switch to get diodes:

• Make the 2.2kohm resistor 0ohm.
• At the feed side of the series diode on each leg, place another diode from the first diode pin to ground, anode on ground. There is plenty of ground by the diodes to scrape off some resist to make a close connection.
• In series with each control signal, place a 1kohm resistor.

The port you want enabled would be given a +5V signal, the others would be given -5V. The series diode would be turned on for the desired port, and for the unused ports the series diode would be turned off and the shunt diode turned on, providing more isolation.

Also, if you find the insertion loss too high at lower channels, you could cut the printed coils away and place discrete inductors. Digikey has some 0805 or 1206 muRata parts that should bridge the gap. This would be hard to tune without a network analyzer, but you could try some parts between 0.1uH and 1uH to see what gives the flattest response.

edit:
I wish I had thought of this; up until a week ago I had access to a network analyzer, and I have some switches laying around. I could have played with some values and given you insertion and isolation graphs.

Wow. Okay, lots of stuff to think about there.

Initially, though, I'm going for minimal modification -- just the DC blocking caps, and keeping the +12VDC PSU (since the PC has +12VDC readily available, no big deal there).

I fully expect the port isolation figures to be quite bad, especially compared to the 60dB+ that the mechanical Frankenswitches give. But it might be good enough for typical needs. We'll see.

But, alas, the one and only electronic component shop in town closes their doors at 5pm on Fridays, and doesn't reopen them until Monday mornings. So this project is (mostly) on hold for a few days now.

Cheers!

For what it's worth, here's the underside of the controller chip that I have now removed from the PCB. There were no markings at all on the top side, but here it appears to say ?5BQXCC on the underside.

Here's the hacked-up multiswitch. I've removed the chip, and the two transistors that dealt with the 2-way communications over the downfeed coax, and then added +12V, GND, and four control wires.

There is also now a 1000pF DC-blocking capacitor on each of the four coax inputs. I opted for non-surfacemount to begin with, because they were 1/3 the price.

The wires will all go to a USB/GPIO subassembly, for software control from MythTV. Later today, I hope.

Cheers

Okay, it works(!).

No idea about port-to-port isolation yet, as my various antenna arrays are all in use with the regular Saturday afternoon recordings, plus the Stanley Cup quarter finals.

Later.

The switch has now been tested with separate antennas feeding into each of the inputs, and the port-to-port leakage is minimal to non-existant. So the isolation is good enough.

If I connect an antenna to one port on the left-side of the switch, and leave the other left-side port unconnected, then the antenna's signal does leak over to the unconnected port. Much weaker, but sometimes viewable on screen. Ditto for the right-side pair.

But in practice, with antennas connected all around, there's no visible interference between ports. Woo-hoo!

Cheers

Mmm.. I think something like that might be quite useful. I wonder if a slightly less-intrusive modification might work: simply replace the +12V wire with the +5V from the USB cable, and change out the 2.2K resistor for a 1K value ?

I'll have to open it up again and try it.

Cheers

I dunno Mark, Frankenswitch v2 was my favorite.

Well, that works. As does a 470ohm value. Oddly enough, the current is still about 5mA with either the 1000ohm or the 470ohm in there. Perhaps the driver circuitry (two transistors, which I left in place) is regulating the current?

Anyway, this is really good now as-is. Completely bus-powered off of USB, so interfacing to the MythTV box just got simpler.


Mine, too.
V2 has less signal attenuation than the new hacked 4x1 diseqc switch does. All of our local channels are now strong enough for that not to matter (much) now. The distant weak analog PBS station is now off the air, and the antenna feed for the digital PBS replacement bypasses the switch anyway. Everything else is local and thus quite a bit stronger.

Another plus of v2 though, is that all of the coax connectors line up nicely along one side of the unit, for very easy connector access. Whereas the new 4x1 thingie has three on one side, and two on the opposite side. Very inconvenient, that!

Cheers

Can you measure that series capacitor on the common port? If it is a low value you can replace it with another 1000pF; that might reduce the insertion loss.

That's a good idea. I don't know how to measure the existing one, but it's simple enough to just replace it with 1000pF.

Thanks!

I've done this now, but really cannot tell if there's a difference.

Cheers

I finished the control board for this yesterday, and sorted out the multiswitches today -- three out of six were defective, and I ended up salvaging two of those using parts from the third.

The completed Frankenswitch v3 has four 4:1 switches tied to a single USB interface, with a pair of 74HCT139 demuxes providing fanout of the eight available GPIO lines, and a pair of 74HCT240 buffers acting as octal inverters to fix the inverted outputs of the '139 chips. Lots of wires/soldering to tie all of that together.

The switch is now installed in the Myth cabinet and working quite well. The remaining 4:1 switch (plus the dead one) is destined for a friend's Myth system.

Photos later, when the cabinet is next opened up to remove Frankenswitch v3.

Cheers

Dunno if you're still around, but I'm questioning that 1000pF value again. The very sensitive, ultra low noise, pre-amp I use for UHF/VHF television here, itself uses 68pF (with an inductor to ground), followed by 470pF on the input side.

I wonder if 1000pF is too high a value ?

Cheers

The inductor and 68pF looks like they are used for matching the input to 75 ohms, and the 470pF is used as a DC block. Are you having sensitivity problems with the switch?

At the high end of the UHF the 470pF or 1000pF will be past resonance, so it will look like an inductor. It should still be a low impedance so any loss should be low.

When I last tested it (still using Frankenswitch v2 in the setup), it looked good except for VHF-LO, channel 6 -- lots of noise (or reduced signal) on that one.

It was set aside for the summer, since the electromechanical v2 version works fine. But today I was repairing the pre-amp, and took note of the capacitors used there. Thus the query.

Thanks!

I wish I would have thought of using the switch before getting laid off. I had a spare laying around, and I could have given you insertion, isolation, and return loss graphs.

A few weeks ago, I picked up a different $7 4:1 DiSEqC switch, and today I modified it to work with the Frankenswitch v3 USB-based controller built previously.

The new switch has all of the coax jacks on the same edge of the casing, making it much friendlier to use.

Internally, each signal path has its very own dedicated pair of two pin diodes (eight in total), compared with the v3 DiSEqC switch's total of six diodes for the four inputs.

The PCB layout is also better than the earlier one, and this time I used SMT 470pF capacitors for the DC blocking caps (I already had some on hand) instead of the long leaded 1000pF caps used earlier.

The results are good -- +0.2dB improvement on the SNR reported by my best tuner, and visibly better on-screen analog TV, especially VHF-6.

I expect 1000pF would do every bit as well (or better), but I don't have any SMT ones on hand to try.

Cheers

I finally got round to working on this again, and now have four of the newer DISEqC switches modified and mounted behind the MythTV cabinet. This setup works very well.