I purchased and built an Elecraft XG-2 tri-band signal generator to
go with my K2 and my TS-850SAT. The manual for the XG-2
gives (roughly) this procedure for MDS testing:

1) Turn off your AGC
2) Select CW mode with a narrow filter, if you have one.
3) Crank the AF/RF gain up to full
4) Setup a VRMS DVM hooked to your ext. speaker jack

Measure the background noise on the DVM then hook up the
XG2, set to the proper band and set to 1 microvolt (-107dbm)
and re measure the voltage on the DVM. The first reading is N
and the 2nd reading is S+N. Do the following math:

Divide Signal (Vrms) "XG2 on" by rec noise floor (XG2 "off") = R
(Log(R))(20)+107=MDS

With this procedure on 40m my K2 scored -136.3dbm but my TS-850SAT
scored -137.2dbm. Puzzled by the results, I Googled and found those were
typical numbers, although the Kenwood seemed a bit on the hot side.

OK - Here's where it get's interesting - today I bought a Telonic Berkeley
rotary step attenuator, model 8107s. Now this device can add 0-59db in 1 db
steps. The manufacturer claims 0.2db accuracy and 0.4db insertion loss up
to 2ghz (worst case). To hook up to the XG-2 I needed a BNC male/male
adapter and a BNC female/SMA male adapter. To hook up to the radios I used a
SMA male/SO-239 adapter, 7' of RG-8x coax feedine into a MFJ 815c swr bridge
then through 2 coax jumpers (2' ea) with a SO-239 barrel connecting them.
Yes, I know - not an ideal low loss connection for MDS testing. I'll try
that later....

Anyways, I figure I've got 3db of losses (40m) through all those
connections/meter/coaxes. The XG-2 is supposed to be VERY accurate as far as
output levels go. Never the less I'm fudging the numbers based on W3AB's
recent tests of a XG-2 on an HP 8562A spectrum analyzer which showed a real
world output of -106dbm +/- 0.5db.

Let's add it all up: -106dbm - 0.4db - 3db = -110dbm.

If that is accurate, I figured I'd be able to crank in 26db of attenuation
before I reached -136dbm and lost the signals (I'm not measuring now - I'm
listening with headphones for the tone).

I went to 26db and the tone was EASY to hear! I eventually found that I had
to crank in **46 db** of attenuation before I lost the signal!

That's -156dbm for MDS, guys..... So what did I do wrong?

73's de Ken KG0WX - Kadiddlehopper #11808,
Flying Pigs #-1055, Grid EM17io, TS-850SAT,
Elecraft XG2, 4SQRP Tenna Dipper, Heath GD-1B