http://www.1stpccorp.com/cb_power_cable.htm

Has saved the day for the 2nd time in a row!

Antenna tuners are kind of annoying. I'd like to achieve natural resonance by always having the right length of dipole flying. The following is a rough design for a unit that can ride along the length of a dipole wire antenna and coil up its ends as the frequency requires.

The power + comms line would be a two-conductor cable. With the device the farthest from feedpoint as possible, all the power + comms cable would be rolled up on the right spool. As it goes inwards, coiling up the antenna wire, it unrolls the power + comms cable. Due to layering on the spools the turns ratios of the two spools would not be equal, so a potentiometer at the swivel joint would control the error by varying the position of the power + comms spool to keep the angle of the joint constant.

A shaft encoder in the antenna spindle can communicate distance down to the control room. If integrated with some smart control electronics, quick tuning would be possible.

Two units like this would need to ride the dipole in sync with each other. I'm not yet clear on how to achieve a perfect electrical balance between the two dipole halves, but a balun at the feedpoint would help mitigate the impact of any small imbalance.

SPICE-Convergence-Problems.pdf
I wish I had found this years ago.

The metal box is where the amp project is being implemented. The black puck to the right is the AnTek transformer inside its magnetic shield. The two aluminum blocks behind the amp are the load resistors, each being 4 ohms / 200W.

I've been a Cadence user for the last 15 years. That kind of time frame has let me realize a pattern of usage with Cadence products. About 80% of the time spent in front of the product is dealing with Cadence bugs/quirks/workarounds, with the remaining 20% doing actual EE design work. Even though I've been aware of this pattern for quite some time, I accepted it since Cadence did really seem to have the best products on the market for the large part of those 15 years. Every release would fix old bugs which would give me hope. Hope which would ultimately be dashed by discovery of brand new bugs that were introduced in that new release. The time investment in discovering new workarounds with every release is not insignificant. But this is nothing compared to the poor basic quality of the 16.3 release. The schematic capture interface bugs in 16.3 completely kill the product, to such an extent that I've now deemed the software suite unusable.

The root cause looks to be common to processing keyboard shortcuts. In order to delete something, you highlight it and hit Del. Then witness nothing happening. You do this again, still nothing. Hammer on the keyboard 3 more times, and then maybe the component will disappear. Wanna place a wire? Press "w" and....nothing. Try it again, nothing. Hammer on it a few more times, move the mouse around a little, and maybe you'll luck out. It doesn't help that "w" is a toggle, so you better not bounce out of wire mode once you've been lucky enough to get into it! The same goes for other shortcuts as well.

I've put up with a lot from you, Cadence. But to screw up something so fundamental like this? You're off the island.

Enter: Altium Designer!

It seems the Protel boys have been quite busy over the last 15 years. In my earlier evaluations of Protel it didn't really hold a candle to Cadence. These days however, with Altium Designer 10, it seems to be quite a different story. Don't get me wrong though, Cadence is still the king of features and functionality (when it functions), but the delta between Cadence and Altium has shrunk to such an extent that Altium is an acceptable solution for my 90% case scenario. Individual results may vary, depending on your specific feature requirements.

I've been an Altium Designer 10 user for a grand total of 2 days and here are my initial notes on the product. First, the good:

1. Flawless keyboard shortcut processing! Yes, this is a weird thing to point out as #1, but I'm scarred you see.
2. No schematic rendering errors. No line segments hanging out in free space when you pan around.
3. When dragging a component, you don't end up dragging its pin, you reliably drag the whole thing. Wow, it seems I'm still venting about Cadence. 15 years is a long time, I tell you. Let's give this Altium-centric list a better shot starting with next point.
4. It embodies a well designed, component-reused, user interface. That is, they wrote a widget to do something once. Since they only wrote it once, they put a lot of energy into getting it nailed. That widget is then linked to from various locations whenever it needs doing. ONE UI for ONE type of task. Win!
5. There are no mysterious crashes or lock-ups. The software is robust so far.
6. The way they implemented keyboard shortcuts, as sequences of simple key strokes, is music to my ears. The sequences are easy to memorize because the choice of words is correct. After only 2 days, I'm typing frequent commands really quickly already. This is the equivalent of having a nice command line interface to a GUI. The best of all worlds. Absolute win on this one.
7. If you forget a full command sequence part way through, no worries! Menus pop up that quickly show you your next possible keystrokes. They disappear when you make a selection.
8. A wonderfully customizable interface. Choose the panels you want to see, dock them to various locations. Use tabs to flip between multiple panels docked to the same location. This is UI done right.

Now for some things that could use improving, but aren't quite bugs:

1. Don't make me annotate components, give them a number upon dropping into schematic.
2. Default to display value of more components, like voltage and current sources. Having to go into the Component Properties dialog is more complicated then clicking on a zero-valued visible property and just dealing with the Parameter Properties dialog. It also makes schematics more legible to have the value displayed all the time.
3. Make the MixedSim graph viewer colorblind-friendly. Most EEs are male, 8% of that majority will have issues using your product. Take a lesson from PSpice here, Cadence did it right in this case.
4. MixedSim looks to be multicore, although it doesn't take advantage of all the cores in the simple circuits I've tried so far. Perhaps it scales better with more complicated designs. PSpice is better in general, but MixedSim has been adequately capable to date.
5. Documentation on how SchLib/PcbLib/LibPkg/IntLib/Models all fit together could use a nice diagram. It wasn't easy to make new components with new models that end up in a nice IntLib package.
6. Provide WAY more components and footprints! Hire some contractors and have them bang out footprints for the whole SOT* and TO* series for example. Provide just a giant standard footprint library. Drawing these right is a PITA.
7. Multi-component select should comply to the Ctrl/Shift semantics standard in Windows. Don't redefine these old ideas. Use Ctrl.
8. Rectangular area select should include wires! Don't make me clean that crap up manually one by one. Ugh.
9. When using A/B cursors in MixedSim graph view show the difference of B-A in both dimensions, don't make me do math. There's a computer right here.
10. While the dockable locations are great, I'd like to see user-insertable dividers to provide more docking locations. For example on the left, it'd be handy to have left-top and left-bottom. Or even top-bottom-center docks. Given 1600 vertical pixels, the "Messages" and "Projects" panels could easily co-exist all on the left hand side.
11. Too complicated to make new component from cloning old. Unusual number of clicks required. Look at optimizing this user scenario. For example: I should be able to drag a component from the Libraries pane into my custom lib to use as a starting point for a new component. This would eliminate many clicks if Libraries is docked right, and SCH Library is docked left.

Now for things which I'd call quirks or bugs:

1. Can't figure out how to simulate just 1 schematic from a project. The "Active Sheet" option in the General MixedSim setup doesn't seem to do just the Active Sheet, but rather everything in the project. Had to move other schematics out of project to sim properly.
2. Fourier analysis fails with some current sources present. Gotta drop data collection for everything, and just collect for "active" nodes to get Fourier to work. It'd be better if Fourier would only ever be done on the "active" nodes, while retaining ability to collect data from all nodes. This is more in line with user expectations AND saves CPU time AND avoids whatever bug this is.
3. 4-way wire junctions have a bug in which the little ball joining them together magically disappears! You have to place an explicit junction on 4-way wire joints. This explicit junction looks pretty ugly too.
4. Print the maximally available precision in MixedSim cursor coordinates. The UI looks to be displaying 1/10th of step size, which doesn't give enough precision to make measurements without re-running at a needlessly small step size. The traces have the resolution, and so should the coordinates of the cursor.

That's it for 2 days of usage. The software is quite capable, and presents a much nicer learning curve than Cadence. I haven't touched the PCB functionality yet, so that may still be a huge disappointment, but I somehow doubt it.

The biggest plus: I haven't yelled at it yet.

Please get your act together and stop putting defective products on the shelves.

Sincerely,

The Consumer.

PS: I had to go through 5 digital scales of different makes and models before finding the 6th scale that actually displayed my weight properly. It is yet to be seen if it breaks in a week, or if the body fat % measurement works. Here is the winning product so far:

That's the Weight Watchers WW31X by ConAir. I like the small well-placed contact pads. These are the kinds of pads I can make complete and repeatable contact with to achieve consistent measurements. And now, here are the losers in chronological order of their disappointing me:

1) WeighMax W-BF440:

This scale worked for about 3 days before telling me I lost 50lbs overnight. At the return counter I just happened to be next to a guy that fixes scales for a living (imagine my luck!). His diagnosis was that one of the four pressure sensors died. This makes sense since I'm a little over 200lbs.

2) Weight Watchers WW67T by ConAir:

This one came pre-broken in that it showed my weight minus 50lbs. Back it went, and I set out to not leave the store until I saw a scale working in front of my eyes.

3) Homedics HealthStation Body Composition Scanner:

This one is the WORST. The two random samples I and a sales man pulled off the shelf failed to work in two different ways. The first failure mode was displaying 0.0lbs as the weight measure no matter what. The second failure mode was the display not coming up, except for brief flashes of blue. After the two defective products of this type, I switched models yet again.

4) American Weigh BioWeigh-IR BMI Fitness Scale:

This one did not fail to work per se, but after a good 5 minutes between a computer programmer (that's me) and the store sales rep trying to figure out how to get a weight reading, we both failed at operating this device. I have to say it's kind of a neat idea to have the display be portable and disconnected from the scale. Especially if you're extremely fat. But it's not a good idea to make a product so counter-intuitive that a technically inclined person can't figure out how to get a basic weight reading within 5 minutes. The manual was even consulted!

One point on quality of this product. The sales rep stepped on it with his heeled shoes, and weighing in at 170lbs, managed to deform the metal contacts on the face of the device. Do not leave empty spaces beneath thin metal surfaces when those surfaces are meant to take pressure.

Thinking $500 for a USB-GPIB interface is highway robbery I bought a cheap Chinese clone of the Agilent 82357B for $125. The Beiming Technologies S82357:

It worked for about 2 minutes (albeit poorly, wasn't able to get proper IDN strings) and then died. Luckily the Chinese seller was kind enough to issue a prompt refund, and I get to ship the thing back today. In the meantime I've decided to up the GPIB game and bought a National Instruments GPIB-ENET/100 instead of another USB dongle:

This looks to be an older version of the GPIB-ENET/100 product NI currently sells, but it should still be alright. Agilent also makes GPIB-Ethernet bridges, but they have the nasty habit of dropping support for them in their drivers. So screw that rat race, NI can get my money.

This LAN-based idea is much cooler than a USB or PCI adapter. It's not just Ethernet that it talks over, but it does full TCP/IP. It means you can route to this interface from anywhere; control your spectrum analyzer remotely when it's on a mountain, or whatever! Can even run the instruments over WiFi from a laptop. Hurray for not being tied down. None of this comes cheaply though, as the GPIB-Ethernet bridge is $425. Still a fraction of the $1,150 MSRP though.

The last advantage should be better compatibility with LabView. I've read several reports now of Agilent interfaces being quirky when used in NI land.