Ok, now after spending time pulling hair out workng with the BIPOM MINI-MAX/ARM-E board and its very lousy documentation, I will have to say the Arduino is a dream to work with.

I do love the Frankenswitch, nice work there Mark!

Mmm.. the next version of this may truly be deserved of the title Frankenswitch!

I'm thinking of making up a new platform with two motors/switches, controlled from a single USB converter. The second switch would give us more flexibility before next February. That's when the antennas will need to be rearranged, as the analog PBS transmitter goes off-air permanently then.

There's a newer, digital HD, PBS station we can receive in place of that, but only with a different antenna configuration that conflicts somewhat with the existing setup of splitters and combiners.

Mmm.. I wonder if four switches might be even better.. ?

Cheers

You need to hook up a tiny little Jacob's Ladder to the thing. :-)

It's alive! ALIVE!

Instead of all these electromechannical switches, have you looked at some solid state SP2T or higher switches. Digikey has some switches that could be used. How about something like the NJG1519KC1-TE3CT-ND ?

About 0.5dB insertion loss up to 1GHz, 22dB isolation worst case. This switch is rated for 50 ohms, but the insertion loss is low enough to work in a 75 ohm system.

Easiest way to prototype this is to turn it upside down and bend the ground leads to touch a ground layer on a PCB. Then solder 1000pF SMT caps on the control lines to ground. Cut 5 coaxes as pigtails and use 1000pF SMT caps as DC blocks to connect the coax to the RF inputs/outputs. The performance should work well enough as a TV switch. If a PCB is needed then just follow the example in the datasheet.

Yeah, the thought has occurred to me, but apparently I'm not very good with my EMI/EMC stuff!

And this thing would definitely need to be very good to avoid degrading the excellent picture quality we enjoy here!

The one you linked to has only 24dB isolation, whereas the sealed mechanical switches I'm using here have 60dB+. There are probably better semiconductors to use, though, but I'm not yet advanced enough to lay them out and isolate things well enough.

However.. one of my next projects is to try building a low noise RF preamp for masthead applications. There are $2 /$4.50 chips from Digikey that quote < 0.5dB noise figures for this purpose, and with a bit of luck I might keep the total figure under 1.0dB.

Off-the-shelf pre-amps generally score noise of 3-4dB or higher, so this could be quite useful if I can pull it off. A commercial product with that kind of figure costs >$200 or so (I know, I have one of them here..).

Cheers

Mmmm.. still, it was rather nice of JRC to include a full PCB layout in that datasheet.. and it's not difficult to transfer that to some copper and etch it out.. Mmmm..

Cheers

Hi,

One way around the isolation challenge with semiconductor switches and PIN switches is to use them in a double pole configuration where you have a common "well decoupled from each other - as in series ferrite bead, series resistor, and parallel capacitor" on the RF switches. Each control line on the switch has this isolation filter. The output of one is fed to the input of the other arm of the switch, think of it as a DPST switch configuration from 2 SPST switches. Don't count on 2X isolation though, more like 35 dB + 35 dB = 55 to 60 dB maybe.


Some of the common design problems with semiconductor type of switches that result in much leakage and isolation degradation occurs due to:

1) Mis-managing the switch control decoupling (as FOLSOM mentioned)

2) Poor layout isolation (I've seen the RF input couple to the switch control line and lose 20 dB of isolation).

These switches need to be grounded VERY WELL. This includes the ground pad on the bottom of the IC package (if provided). If there is one, a via field (25 mil vias) to multiple ground planes works well. The device ground pins must go directly to a plane.

Place the decoupling capacitors within 10s of mils of the device. The RF traces take precedence though.

Try to make the traces that enter and exit the device are the same width as the pin to minimize VSWR losses.

If the frequency warrants it, usually above 700 MHz, use straight traces and curved bends (sometimes 45 degree and chamfered corners can be used depending on frequency).

Make sure you have the proper trace width to give 50 Ohm unless your target is 75 or 93 Ohms.

There is always discussion about FR-4, Rogers 4000 series products, G-Tek, Gore Speedboard, etc... Understand the material properties of the PWB, cost, CTE, Er, Loss Tangent, availability. FR-4 may or may not work for your application depending on your environment. The IPC and vendors have worked cooperatively to improve specifying & measuring critical parameters (see IPC-6018).

3) Whether the switch is reflective or absorbtive in operation and what works for your application. Be careful with layout of trans-impedance converters that accompany some RF switches and attenuators too.

4) Poor support from vendors and using their test circuits as "gospel". Many times these boards and stack-up are optimized and when the design is integrated onto a design with other circuit elements like 10MHz PLL Reference Clocks, mixers, attenuators, etc, "well what do you know, it doesn't operate as expected" becuase of other design considerations or circuitry partitioning. Breadboarding more of the circuit such as FOLSOM recommended is a very informed recommendation.



In summary, semiconductor switches can be used with proper attention to design & layout parameters.

Ross

One way to check if 20-25dB is enough is to combine them together and then put 20-25dB attenuation on the unwanted signal. If the picture quality looks ok with this configuration, then the switch should work OK. You would have to have some attenuators laying around though.

Yeah, that's clever! I don't have any attenuators around at the moment, but those can be simulated (I think) with a couple of 1:4 splitters, with terminators on the unused splitter outs. That stuff is laying around in my bins.

Cheers

Frankenswitch v2.

Wow, I'm so glad I get all of my terrestrial TV from one transmitter.

Now admittedly, it is a bit over the top.

We currently get reception from three different vectors: a big tower north of town (about 2/3 of the local stations), another south-east of town (the other 1/3 of the locals), and a very distant station far south-south-east of here (USA PBS station).

So this gizmo has a pair of 3:1 pushbutton switches (one for each analog tuner in the PVR), and six solenoid actuators to press the buttons. An FT232RL USB-serial breakout board provides control (in "bitbang" mode), and a ULN2003A darlington array drives the solenoid coils.

For added complexity, there's a small relay to cut DC power from the solenoid circuitry when idle (for paranoia). But it's kinda redundant now, since I later added a second, solid-state relay to power down the entire AC PSU (a brick which provides +27VDC@600mA) when idle (saves watts since the thing is only rarely active, and only then for brief instants).

It was a fun build!

Mark, maybe you and this guy should get together. Your minds seem to go in similar directions!

tanstaafl.

Yeah. I think I saw a commercial RFID cat door somewhere in Italy last month.. but without the mouse recognition system that that guy uses!

Cheers

I can't help thinking it is a good job Mark doesn't get his TV from a single transmitter. Just think how dangerously effective he'd be if he didn't have things like this to take up is "spare" time...

Yeah, but I swore years ago to only use those powers for Good..

Did you ever try the experiment to see if 25dB isolation was good enough? The solid state switch would be a lot cleaner solution.

Absolutely agreed -- it would be better, especially with a couple of these switches in series for higher isolation.

I have not tried it because of the difficulty (for me) of building / designing appropriate PCBs for 75ohm impedance with F-connectors (the one in the datasheet is for 50ohm with tiny SMA connectors).

I've figured out how to change the microstrip spacings for 75ohm, but am unsure about the other components and about actually manufacturing such a PCB with two (or more) of these circuits.

Thanks

While I can only admire any project that goes from USB at one end to MDF at the other, I have to say that if that had been my own TV watching then I'd have just bought however many extra USB TV tuners it took to solve the whole problem in software...

Peter

Tee hee!

And here I've been thinking that I can't even find enough TV worth watching on one channel, never mind enough channels to need all those switches.

Ahh.. but we don't watch TV, we watch MythTV recordings, with any commercials already stripped away from view.

Mostly from PBS, OntarioTV (like PBS), and CBC. Those are all full of "TV worth watching", as the catchphrase goes.

But also a bit of content from more commercial stations -- mostly movies recorded in the dead of night.

How do you find the commercial detection in Myth? I've always found it to be a little hit and miss. Certainly enough to never let it automatically flag and transcode without me checking first.

Sadly my Myth is installation is due to be usurped by a Humax Freesat+ box when somebody finally has some in stock. I want HD but my frontend just isn't up to doing it, and the cost of upgrading the box is as much as buying a Freesat+ box. I'm also partial to the fact I can power down a further two boxes at home and have something that 'claims' to draw less than a watt on standby.

I don't transcode, so the entire (unstripped) program is always there for me to view. But the commercial stripping in MythTV 0.20 (old version that I run) is VERY GOOD.

It generally only screws up at the end of every single HD program from CBC. CBC has a nasty habit of removing their onscreen logo for the final program segment before the end. This causes Myth to think that the last break continues through the final segment.

Other than that, it's pretty much flawless here.

Cheers

Are you aware that Myth trunk now supports the new (non-free) NVidia video API - VDPAU.

That means that an FE with a £30 fanless graphics card will do H264 HD with no CPU usage to speak of

And my BE uses BIOS wakeup and WoL to stay down and keep power usage down. (Nowhere near as little as consumer electronics but...)

We don't do commercial flagging - we just press 4<right> and we're done.

Oh well, I've found Freesat+ in stock at Argos. So I've got one arriving on Friday. I'd like to have kept on with Myth at home but I never got SWMBO to use it in preference to just watching TV so it was pretty much doomed to failure.

At least I've still got one at work, our journo's actually quite like using it.

That sounds cool, especially for you DVB-T folks.

Over here, MPEG2 rules the airwaves, and the fanless NVidia card I have doesn't seem to help with much at all -- most of the more useful decoding features fail to bind with the binary driver, so it ends up doing just about everything in software.

I guess the chipset must be too "primitive" or something.

So when I rebuild the software over the holidays (coming up), and replace it all with latest Myth SVN, I'm also ditching the NVidia card completely to save a few more watts.

I've got one of those nifty Intel DVI-D adapter cards (the mobo chipset has built-in Intel graphics, but no DVI output) that slips into the PCIe slot. Just a single chip and a few discretes on the card -- shouldn't generate any heat to speak of.

Cheers

Calculating the microstrip is really only the hard part. The rest of the values don't change.

Let me look at making a schematic that has a SPDT on each port to increase the isolation to about 45dB. I believe Digikey has all the necessary parts (SP4T, SPDT, and right angle F connectors) to make a solid state switch. This will need 4 GPIO and 3.3V to work.

The USB dongles I'm using have lots of GPIO, and onboard regulated +3.3V, so there are no issues there. Laying out and etching an SMT circuit board will be far more challenging for me.

There's a chap on another forum who's been suggesting PIN diode circuit solutions for this.

Here's the proposed circuit, and his explanation.