[Elecraft] K3 Filter Characterization using Spectrum lab

[Mike Scott]

I have noticed significant and unpleasant audio filter spectrum shape
differences by ear when I change filter bandwidths as well as change from
LSB to USB or CW to CW reverse.
This is not an issue at CW filter bandwidths but can be annoying at phone
bandwidths using the 6 KHz filter. When LSB is equalized for a flat pass
band (minor tweaks), USB audio is highly skewed at high frequencies. Level
at 2.7 KHz is 14dB above the level at 500 Hz in USB.

I am wondering if anyone else has notice the effects I am reporting on here
and whether there is something I can do about it.

Unless I have a bad filter, I see a possible need to save Rx equalization by
selected filter and operating mode. Either that or something is wrong on my
end. My guess is that the Tx audio has the same issue but I have done
nothing to test that yet. The changes in sound quality are not trivial.
Filters are properly enabled and set up with proper offset and gain. There
were no signals in the pass band when the data below was collected.

I collected data to characterize the issue using Spectrum Laboratory V2.7,
using K3 line level out.
The two filters characterized here are the stock 2.7 KHz five-pole filter
and the 6 KHz 8-pole filter.
I illustrate measurements at BW = 2.7 KHz and BW = 2.8 KHz below with USB
and LSB. I first set up the Rx audio equalization to obtain a maximally flat
pass band on LSB and the 2.7 KHz filter at 2.7 KHz BW.

For LSB
The 2.7 KHz filter at 2.7 KHz BW is flat (zero slope) from 250 Hz to 2750 Hz
with an absolute level of -62 dB at 1KHz, an arbitrary reference point to
compare overall gain changes.
When I shift BW to 2.8 KHz (now using 6 KHz roofing filter) the audio
spectrum has a slope of -1.7 dB per KHz and the 1 KHz level went down 2 dB.
The LSB data isn't too bad. The problem is USB on the 6 KHz filter.

For USB
The 2.7 KHz filter at 2.7 KHz BW has a spectrum slope of +1.7 dB per KHz and
the 1 KHz level is -64 dB.
When I shift BW to 2.8 KHz (now using 6 KHz roofing filter) the audio
spectrum has a slope of +5.3 dB per KHz. This is very noticeable to the ear.
There is also a droop or saddle in the filter response down 6 dB in the
center of the pass band and the overall level drops 8 dB (S-meter drops 2-S
units on background noise). The overall effect is unpleasant with the
accentuated highs and it requires an audio gain adjustment and I lose
S-meter calibration.

Detailed data...
			FL2		FL3 (Rx Eq. adjusted for flat
response)
	LSB		2.8Khz	2.7KHz
Level at 1 KHz		-64 dB	-62 dB
	Slope		-1.7 dB/Khz	0 dB/KHz
	Saddle	-4 dB		0 dB
			
			
			FL2		FL3
	USB 		2.8 KHz	2.7 KHz
Level at 1 KHz		-72 dB	-64 dB
	Slope		5.3 dB/KHz	1.7 dB/KHz
	Saddle	-6 dB		-3 dB

Saddle is defined as middle of spectrum maximum drop in level from the high
points somewhere in the spectrum. The data usually "looked" like a saddle
except for when it was flat.

Mike Scott - AE6WA
Tarzana, CA (DM04 / near LA)
K3-100 #508/ KX1  #1311