Recommended way to make a permanent one-off circuit

[rhb]

I use those boards all the time and they work just fine.

For somethings, yes. For others, no.

You need to match the tools to the problem. No surprises there.

Amen!

If you *really*  want to quantify electrical performance of a particular construction style, connect an SMA-F connector to a 50 ohm non-inductive load using that method  and subject it to the tender mercies of a Tek 11801 sampling scope and a 20GHz SD-24 TDR head.  3 mm  copper foil on single sided FRG4 is good up to several GHz.  A tip I learned from Thomas H. Lee of Stanford in his excellent "Planar Microwave Engineering".  Best radio book ever.

I have the kit to do this.  So long as the requested setup is simply a 50 ohm SMD load connected to the SMA-F,  I'm willing to run a few tests of the performance.

Every impedance  discontinuity will be obvious with fraction of a mm resolution.

One approach would be to use copper foil from the SMA-F and then put a segment of each method leading to the load.  I  like to make little boards to demonstrate stuff using the 11801.  So having gotten curious about the difference between expectation and reality I'm going to put together a few test cases.  Suggestions welcome.  Because it's all time domain you can see each segment separately with several segments in series.

Have Fun!
Reg

[Infraviolet]

I can recommend stripboard (also called veroboard) for if you're in a hurry. But if you aren't in a rush then PCB fabrication is so cheap thesedays that it is worth it for even little things like this, and an actual PCB gets you solder mask to prevent shorting by contact with metallic objects and silkscreen so you can put labels and notes to self about pin functions.

[tggzzz]

I use those boards all the time and they work just fine.

For somethings, yes. For others, no.

You need to match the tools to the problem. No surprises there.

Amen!

If you *really*  want to quantify electrical performance of a particular construction style, connect an SMA-F connector to a 50 ohm non-inductive load using that method  and subject it to the tender mercies of a Tek 11801 sampling scope and a 20GHz SD-24 TDR head.  3 mm  copper foil on single sided FRG4 is good up to several GHz.  A tip I learned from Thomas H. Lee of Stanford in his excellent "Planar Microwave Engineering".  Best radio book ever.

I have the kit to do this.  So long as the requested setup is simply a 50 ohm SMD load connected to the SMA-F,  I'm willing to run a few tests of the performance.

Every impedance  discontinuity will be obvious with fraction of a mm resolution.

One approach would be to use copper foil from the SMA-F and then put a segment of each method leading to the load.  I  like to make little boards to demonstrate stuff using the 11801.  So having gotten curious about the difference between expectation and reality I'm going to put together a few test cases.  Suggestions welcome.  Because it's all time domain you can see each segment separately with several segments in series.

Please stop making me jealous. The best I can do is a 140ps/4GHz Tek 1502.

One trick I sued to sell a surplus one at a harvest was to have a 20cm long wire[1] and run two fingers along it. The two fingers were clearly visible on the display

Before anyone thinks that's ridiculously fast, I've measured jellybean logic with a ~250ps transition time (74LVC1G family). Yes jellybean logic hits microwave frequencies; deal with it

[1] actually a microstrip, but that doesn't change the point.

[dietert1]

Yes, i remember using twisted pairs of that transformer wire for fast signals, with stubs less than 2 or 3 mm. Result is similar to those LVDS double traces on FPGA proto boards. Anyway in air 1 nsec is 30 cm wavelength, so even a single wire on the small protoboard should be good enough for that. One thing to avoid is ground bounce, so one needs a redundant Gnd network.
One reason why i am rarely drawing boards for prototypes is i don't need to and i prefer not to work out designs to send them to a chinese company where i don't know anybody.

Regards, Dieter

[rhb]

Please stop making me jealous. The best I can do is a 140ps/4GHz Tek 1502.

One trick I sued to sell a surplus one at a harvest was to have a 20cm long wire[1] and run two fingers along it. The two fingers were clearly visible on the display

Before anyone thinks that's ridiculously fast, I've measured jellybean logic with a ~250ps transition time (74LVC1G family). Yes jellybean logic hits microwave frequencies; deal with it

[1] actually a microstrip, but that doesn't change the point.

I managed to pick up a 1502 and really like it.  It's much more portable.  The 11801 is a beast.  But oh so cool.  I bought one after I bought one of Leo Bodnar's pulsers and saw the supplied  test printout using a CSA803A and SD-30. I had to have one and managed to get one for ~$150. It required reseating all the connectors several times to get it to boot and run the diagnostics which revealed the NVRAM had failed.  Once I replaced that it's been fine since.

Sadly, there are no schematics available.  The Tek Museum has all the records from a member of the design team.  But the lawyers don't want to vet the material for release.  There is a bunch of 3rd party stuff in them.

Happily, I later picked up a 2nd unit, so I've got a parts mule if I run into problems.  Even better, I now have 4x SD-22, 1x SD-24, 4x SD-26 and an SD-32 in an SD-30 enclosure I bought on ebay from Tek.  It has the 2.8 connector and the scope reports it as a 32.  My guess is it was a prototype to test whether changing the connector would increase the BW.  I have a 2.8 to 3.5 adapter on it and that reduces the BW to the SD-30 spec.  I've not done a detailed comparison and even if the parts are the same, the 32 may have required selecting  the best parts.

[amyk]

If it's simple enough I'd not even bother with a PCB and just go for point-to-point wiring.

[donlisms]

That was good enough for Jim Williams. 

[n4u]

When it works then its ok - dont bother in opinions. For low frequency cirsuit without care about noises u may even solder it all with wires without any other connections. As long as it meet the requirements its ok.

[rhb]

That was good enough for Jim Williams.

It did indeed.  But there are not many in the Jim Williams or Bob Pease category.  I have used the Williams-Pease style many times, but it's generally difficult to make significant design changes.   Those usually require another board.  But hey, if you can design the first transistor op amp in the breakroom at Fairchild over lunch on a visit from Philbrick where Bob worked at the time,  neatness *doesn't* count ;-)

BTW It was  not an IC.  It was discrete dies wirebonded together inside the can.

Have Fun!
Reg

[ccktek]

Another enthusiastic vote for Manhattan.  You can quickly make pads of various sizes from PCB material with one of those inexpensive metal hole punch kits.  The pads are glued down with superglue, ready for soldering in seconds.  They can be removed with a sharp sideways tap with a screwdriver. 

https://www.amazon.com/ABN-Power-Punch-Heavy-Duty/dp/B07D7HJQ5Z/ref=psdc_2225057011_t1_B00OM4GVX0

[edpalmer42]

As a Time-Nut, I needed a circuit that would square up an input signal and divide the frequency down to reduce the number of phase wraps.  I used a piece of copper-clad board and dead bug components.  It looks absolutely horrible!

Then I tried to measure the Allen Deviation noise floor ...... okay, not so horrible after all!

Ed