Upgraded Spec. of a 40 year old Tek SG 504

[Mosaic]

Hi all:
As an exercise in RF detection I refined the TEK SG 504 leveled sig gen leveling head to give the instrument around +/- 0.05dB flatness across the Ghz range at the 4Vpp (16dBm) level.
http://www.nscainc.com/uploads_2014/product_guides/TEK_SG504.pdf

The new head used new detectors and a custom PCB  on 0.8mm FR4 with BNC connectors (optional SMA).

The measurement was done (after warmup) via a broadband 10dBm attenuator into a Boonton 4210-4b micro watt meter to mitigate any VSWR errors.

Thus actual delta p-p measurement levels (less adapter losses) =>  5.88 dBm - 5.78dBm (0.1dB pp) = .005Vrms, or +/- .0025Vrms for a +/- 0.05dBm swing.
As 5.78dBm = 0.425 Vrms => the % precision deviation is  +/- .0025/0.425 x 100 = +/- 0.588% versus 4% OEM.


This makes it more useful as a reference tool for measuring 'scope bandwidth  or determining RF amp/mixer IP3 etc.

Most of the improvement was obtained from PCB size reduction, improved ground plane coverage, a modified BNC (m) launcher, and using parasitic stub trace 'tuning' to improve return loss leveling across the band.

I hope to do a Hackaday page for DIY and offer a kit for convenience. I put up a few excess calibrated units from my development on Ebay as well.

Another persons 'substitute' head I  tested had a 0.5dBm dip @ 800Mhz or so on the same test jig. Basically  5 x 0.588% = 2.94%.  Thus that is within the 4% level OEM spec, but I needed a better result.

Here are a few pics:
https://www.dropbox.com/sh/xmf4hgaxvpjyc2j/AAAuzDz2vfhjWrACqvMcXT8Ea?dl=0

[AF6LJ]

Nicely Done.   

[Earendil]

Hmm. Looks interesting.
And what about the harmonic content of the signal generator? Have you factored those into your calculation too?
Also the Tektronix claims 4% on the full temperature range (-15degC to 55degC). Have you tried varying the temperature of your circuit and see if your measurement results change?

[Mosaic]

There's internal temp compensation via a matched detector diode which is also built into the OEM Tek unit. Temp shifting is not an issue. Also the housing is fairly isothermal with a very small air volume and insulating material compared to the OEM product.
What makes the new unit better is it contains stub TX line tuning to knock out resonances in the band.
Typical flatness is +/- .05dB as I noted. I have built units down to +/- .025dB!

As to the harmonics or new out of band resonances....no noticeable difference using my calibrated Spectrum Analyzer and  S12 testing with a 3GHZ VNA.