dermel'ing

You need to try a modern dental air turbine drill. 500,000+ RPM makes engraving more like sweeping away talc with a tiny paintbrush.

Ah, left-hand solder!

Copper isn't very strong.

I guess I could get some smaller cutters, but I can do 0603s and SOT23s now. And I'll never be able to do MSOP10 or US8 type packages... I use adapters for those.

and

Yes, I was thinking copper plumbing would serve well instead of using steel pipe fittings.

I have a fixture I made from 1/2 inch copper pipe many years ago. It has survived serious overload without damage and is extremely strong and rigid.

Making the frame out of copper pipe would be simple. You can cut it with ordinary pipe cutters instead of having to thread steel pipe which is difficult and expensive. You can assemble it for a test to see if everything fits before soldering it together.

Copper plumbing comes in various sizes, but the cost increases drastically above 1 inch:

After my experience with 1/2 inch pipe, I think 1 inch pipe would probably be plenty good enough for the frame. After it is assembled and cleaned, a coating of clear Krylon spray would keep it looking nice:

This would also serve to keep PCB prototypes looking good instead of having to coat them with gold.

For some of us, good enough is good enough:)

with

I should mention that making the base a complete rectangle allows you to use a three-point support. This means you can sit the machine on any reasonably flat surface and have it work without wobbling.

A four-point support will have one leg that doesn't quite hit the table, and the machine will wobble back and forth as the carriages slew.

Rigid and stable are the watchwords.

Here's how I used to do it in the stone age, before I could get reasonably priced 2-5 day PCB deliveries or was spoiled by an in-house PCB shop.

I just hacked this out: The "cut" is about 0.010" wide. The way I made it was to put TWO blades in a Stanley utility knife like these:

Insert TWO blades into the knife, with a piece of paper or tape between them as a spacing adjustment. Adjust the blade spacing for the desired span. The paper or tape works fine for 0.020" spacing but no more because support by the aluminum handles is lost. I milled my own handpiece for wider spacing.

The way it's used it to scribe the copper that I want to remove using the dual blade knife, preferably in one pass. I could go over the cut grooves more than once, but then I risk going astray, as I did trying to make something worthy of a photograph. Don't do that and try it my way first. The initial dual blade cut is only a cutting guide. Also, I usually use a steel rule to keep the knife going in a stright line. Curves are more difficult.

Ideally, I would cut all the way through the copper and into the fiberglass. That's not necessary. Instead, grab a second Stanley utility knife, of any type, which has only a single blade. I use it to follow the grooves cut in the first pass with the dual blade, to deepen both knife gooves until it looks like I'm through the copper.

Next, find a #1 slotted jewelers screwdriver or sharp pointed knife and pry the copper off the PCB starting at one end of the cut. This takes practice and a sharp screwdriver, but can be done. Don't use an Xacto blade because they're rather weak and don't like being used as a pry bar.

Once I have a small amount of copper peeled upwards, I grab it with a pair of pliers and SLOWLY pull the copper away from the PCB. If I've cut all the way through the copper, it will peel away smoothly. If I missed, it will tear and I get to start prying again.

At one time, I had a 2 blade contraption made specifically to lay

Good luck, be careful, and don't cut yourself.

Fixed by using adjustable feet. Machine tools do that.

Cheers, James Arthur

Way too complicated. Get an X-Acto or Ace hobby blade:

Grind the cutting edge flat so you don't cut yourself later, and grind the end to a 120 degree angle. Grind the straight edge to a bevel so you can set the knife to the proper angle to cut the copper.

Use the blade upside down to cut a thin slot in the copper. Sharpen as needed.

I don't think you are going to have much luck cutting 50 ohm traces on 30 or 60 mil copperclad. You need much thinner material and a different approach.

Now we need a cheap CNC router with 0.1 mil repeatability. The cheap Chinese units won't do the job. The best they can do is around 1.98 mil. Commercial US units are way too expensive. A DIY using copper plumbing frame might do the job.

The next problem is FR4 is going to be too lossy at microwave frequencies. Rogers is extremely expensive. We need to find a thin double-sided copperclad material that works into the tens or hundreds of GHz. Ceramic is interesting but might be difficult to use.

Any ideas?

I find adjustable feet don't work. You have to put the machine in exactly the same spot every time. If you move it a slight amount, the feet are no longer touching and the machine wobbles. If it moves itself due to slew rates, you are in trouble.

The alternative is to bolt or clamp the machine to the table. That is a possible solution, and I use it with my grinder that vibrates horribly as it hits resonance. But it is a bit of a pain when you have limited space and need to move machines to and from the work area.

Three point solves the problem, except in the worst case of massive resonance. I hope a PCB CNC won't hit that.

Neither is hard freehand with a Dremel. Draw lines with a fine-point Sharpie first.

An inch or two of FR4 is OK at 10 or 20 GHz.

I was talking about the pcb thickness.

I need to go much higher.

I've used FR-4 at 1.9GHz and somewhat regretted it. The problem is the dielectric constant Er which varies from 4.3 to 4.8. There are also variation is material and plating thickness. Er and coef of thermal expansion are also sensitive to the direction of the fiberglass weave (anisotropic) in all 3 dimensions.

Eventually, I got things to work, but ended up with only one PCB vendor willing to make them to the tighter Er tolerances required, for an outrageous price, of course. If it's only one short run, I suggest you consider using semi-rigid coax and connectors instead running RF traces on or inside FR-4.

This board makes clean 100 ps or so pulses with 45 ps rise/fall times. That's roughly a 7 GHz bandwidth.

It's a conventional FR4 board. The tricky part was fanning out to the SMA connectors from the tiny pads on the chip-scale output driver.

We did a bunch of ATLC simulations to get the edge-launch SMA footprint and stackup right.

The Si-List has long discussions on the problems of FR4 at microwave frequencies, including most of the things you mentioned:

On the other extreme, the TI AWR1243 76- to 81-GHz FMCW Transceiver is meant for inexpensive automotive applications such as emergency braking and highway driving. The pcb stackup shown on page 39 uses Rogers 4835 on the top signal layer, and FR4 for the rest.

I suppose if you are going to make tens of millions, you might be able to get a reasonable price for the pcb. But if you only need 10, forget it!

So I need to find a different way to make microwave pcbs for development work. I'm thinking of copperclad polyimide glued to a plain FR4 substrate. Then I need to make the traces on the top layer, and connect to the ground plane on the top layer of the FR4. Making the traces and drilling the holes is not so bad, but electroless copper plating is a real bear.

I keep looking...

Very nice looking PCB, but you don't have any one or two inch RF traces between the drive and the connector. More like 1/8" trace. If you look at 12GHz satellite TV front ends, most use some kind of PTFE laminated PCB. I don't know what to suggest for a better PCB as I'm not up to date on the available technology. Here's an article by Rogers that explains some of the problems: See Fig 4 for a loss vs freq graph for various PCB materials.

Looks like ATLC2 not the original ATLC I didn't know it had been ported. I'll probably ruin my evening playing with it. Thanks.

Am 11.07.2017 um 04:15 schrieb Jeff Liebermann:

I got quite good results even with filters on FR-4.

The filter on the right side was simulated with Empower Electromagnetics some 10 years ago and home-etched on 0.8mm Bungard FR-4.

It hits quite closely, and is good enough to separate the harmonics in a frequency multiplier. Adjusting the coupling and making it compact took quite a time with the simulator.

It is probably not wise to switch manufacturers based on the price of the day. Ask your manufacturer what their favourite prepreg is and prescribe it for the future. Always using Isola-xxx in their standard process should return quite consistent results.

Put test microstrips on unrelated boards and you soon see what to expect. Avoid unusual processes. We had delamination with Polyimide prepregs, it needs hotter baking and has short shelf live. The manufacturer may be tempted not to throw it away when the time has come.

The inductive part of a resonator is less impressed by varying Er than the high impedance end. Rhea describes in his filter book how to make Hi-Q resonators on FR-4. Make them very wide and too short and replace the hi-impedance end with paralleled capacitors.

I'm not sure if it is exactly that book, mine is light blue and

cheers, Gerhard

Easy; bevel or birds-mouth the pipe and weld. When the weld cools, it's a fixture.

Yes, you have to cut the pipe. My idea was to cut one of the legs in half, assemble, then weld the pieces back together.

Then I remembered an old fixture I made from 1/2 inch copper pipe, and how strong and rigid it was. I looked up the prices, and found a 1 inch copper pipe with the needed plumbing fixtures would be economical and probably more than strong and rigid enough for the frame.

Well, maybe a half inch. Conventional FR4 boards get ugly with long traces; like the doctor said, don't do that.

Part of the problem is dielectric loss, but the bigger problem is usually the skin loss in the copper. Standard boards have a black oxide treatment on the copper, to make it stick to the epoxy better, and that's horrible lossy crud. Just peel up a trace and see.

Shiny-bottom copper over teflon laminate is great for high frequencies, but adhesion is terrible. You can cut fancy patterns easily with an x-acto and then peel the copper off easily.

We recently did a tester board that wound up having about a foot of trace end-to-end, with four switching relays in the path. The relays are fine, but trace loss is pretty bad. A fast step comes out as a

If

I used ATLC and Rob used ATLC2, and they agreed pretty well.

ATLC2 is even more fun than ATLC.

We have Sonnet Lite too... one of my kids learned how to drive it, and it's occasionally useful.