For those early adopters who can't wait for AR magazine...
Nearest parallel resistor values for the 1dB/2dB/3dB/5dB/10dB/10dB/10dB Pi-Networks are:
869.548 ohms: use 6800 || 1000
5.769 ohms: use 6.8 || 39
436.212 ohms: use 1200 || 680
11.615 ohms: use 33 || 18
292.402 ohms: use 2700 || 330
17.615 ohms: use 820 || 18
178.489 ohms: use 22000 || 180
30.398 ohms: use 390 || 33
96.248 ohms: use 270 || 150
71.151 ohms: use 560 || 82
I recommend double checking the final parallel value for each pair before soldering them in, as it will be harder to find errors once they form a pi network, and I may have made a mistake transcribing the values.
Once each pi network is soldered, you can use the white silkscreen rectangle to write the particular pi-attenuator section's attenuation in dB.
The photos are detailed enough to give a rough idea of the 1% resistor colour codes.
The good VK5TR hath spake unto me that 20dB pi attenuators would verily invite excessive coupling, so yeah, values have not been calculated for 20dB pi networks.
Barry, VK5BW, very kindly did some tests looking at return loss, insertion loss, and VSWR.
Testing used the photographed step attenuator shown below, with the second BNC connector also soldered on.
1% metal film resistors were used for the pi-attenuator sections going from left to right: 1dB, 2dB, 3dB, 5dB, 10dB, 10dB, 10dB
In summary, for HF:
SWR < 1.1 across all bands up to and including 6m, and
SWR < 1.05 if > 5dB attenuation used, across all bands up to
and including the 6m band
Insertion loss of ~ 0.15dB at 150Mhz
At 150MHz, attenuation steps remain quite accurate for the 10 & 5dB switches
Bleed through with all attenuators switched in at 150Mhz estimated at < 1dB
SWR increases to about 2.0 by 150MHz.
Here are some fairly hi resolution photos. You can right click on the images and select "Open link in new tab" if you want to see them at full resolution and skip the default "fit to browser window" slide show format.
And now some close ups of the 10dB pi attenuator network, in case the above photos aren't crisp enough.
First, the top side of the board showing the switch, and the parallel and equal pi attenuator legs:
And then the underside of the board showing the series segment of the pi attenuator network:
Power handling will depend on the resistors used.
Wire wound resistors would add inductance and render the attenuator somewhat useless.
Diverging from the design (i.e. resistor placement WRT side of the board, adding shielding between or around sections, etc...) may improve performance, but without further testing you won't know.
The transmission line design assumes 1.6mm FR4 with 1oz copper, which is the standard PCB option from Hackvana.
Thanks again to Barry, VK5BW for running the tests, and the developers of the gEDA toolsuite under GNU/Linux, without which the barriers to learning PCB design would be much greater.