Here is a cheap and simple way to significantly improve the K2's CW filter(s).  With this mod, even strong signals will "fade to black" as they tune out of the filter pass band.  This mod also significantly reduces the "underwater" sound sometimes mentioned with regard to the K2's otherwise excellent CW filter.

I suspect that many hams have been conditioned by "big name" rigs of the past to accept poor ultimate rejection ( "that is how a HF rig is supposed to sound")  I liken the difference between good and mediocre ultimate rejection as similar to the difference between (good ) Dolby audio tape, and CDs. Yes, the tape gets the job done nicely, and in most cases is pleasant to listen to and "good enough"......ahh, but to a fussy ear, there is a huge difference.

Improvement you can expect from this mod:
Audible and measurable improvement in filter roll-off  performance for signals well outside the intended pass band. Reduced "underwater" character to audio, and slightly improved S/N ratio.  Here is the before & after data KT5X sent me.
K3IU's data indicate that the ultimate rejection of the filter is improved by about 10-20 dB. The improvement will be greatest for narrow filter settings and most useful during noisy and/or crowded band conditions. (still lots of time to do it before FD 2001  as I type this!) Read comments of other hams who tried this mod and liked it.
 

Problems: Due to slightly increased stray capacitance, and reduced stray inductance,  the filter pass bands will shift slightly. You will need to re-adjust the filter BFO settings to compensate. If you have spectrogram capability, no problem. If you set your filters by some other,  painful,  process, then you might think twice before making this mod.

This mod will not cure rigs that have abnormally poor ultimate rejection, and rigs that have especially good rejection will see minimal, if any improvement. Read comments from hams who tried the mod with no improvement.
 

NOTE: This mod has NO effect on the SSB (opt1) adapter filter. This is good, because we don't want to be messing with the SSB pass band on transmit. One ham wondered if a similar mod could improve the SSB filter, and here is my reply to that query.

Here is a sidebar all about ultimate rejection, and explaining why we want lots of it!

So how do we improve the ultimate rejection?

…easy!

Preliminary checklist:
Make sure you have followed all Elecraft's advice regarding keeping the varactor diodes, RP4 & 5 and crystal leads short (the packages need to be tight against the board) Use a pair of flush cutters to trim the crystal, varactor, and RP leads as short as possible. EXCEPT for RP4 pins 2&3 and RP5 pins 4&5…leaving these intact will make my mods below a little easier. If you have already done the pruning before reading this, don't worry, you can still work with flush cut leads...just not quite as easy.

Make sure the filter crystal cans (X7 - X10) are grounded with the shortest possible wire. Note: DO NOT solder to any place on the crystal can other than the top. Some cans are soldered closed at the bottom, and even welded cans can have their seals damaged by excessive heat applied near the base of the can.

On that subject, I have found a way of grounding crystal cans that gives me much better results than when I tried to follow the Elecraft instructions. Since we will be adding some extra ground wires later, you might want to have a look at that.

Parts list:

1. .01uF (103) capacitors, qty = 2. If you have some of those pepto-bizmo colored ones left over from building your K2, or installing option modules, those will work just fine, and are exactly what I used. If you have some .001, .033, etc. uF, then they should work just as well. Make sure they are ceramic, NP0 type is preferable...but nothing is critical. Physically smaller is better. Low voltage types are fine. just NOT Electrolytic, tantalum, etc. Some hams have asked about SMT caps, and I think they are an _excellent_ idea in this application. I don't "Do" SMT though, so can't offer specific advise on most appropriate package etc...whatever fits best, I guess.
2. A handful of discarded (you didn't really "discard" them did you?) component leads….as used for grounding crystals when you built the K2. Some fine (AWG 24 or so) bare bus wire will also work.

Procedure
 

1. Preliminaries:  Follow the Elecraft instructions, and remove the bottom cover (the heat sink stays on) and the top cover from your K2. If you have the SSB option, take it out so that you can access the RF board.
2. Add 5 additional ground wires to the crystal filter cans. This halves the inductance of the can grounding wire. Though slight, the inductance does allow the cans to float a little at RF, reducing isolation. The extra wires can be expected to give a 6 dB reduction in whatever leakage (again, I didn't measure) might exist.

Use that row of holes between the cans and the resistor packs to do this. I know, I know, the K2 manual says "don't use those holes." The reason the manual says this is that the wires will mechanically interfere with mounting the SSB board. We are going to install them so they don't cause any problems. This is done by running the wire slightly around the crystal, so it goes up the "face" (in some cases this means between two faces) of the can rather than the "edge" of the can. This photo shows my K2, with this mod,  with the SSB option installed. There is no reduction in clearance with the new wires routed this way.

Don't get carried away with this "wrapping" of the wire around the can. If it connects 1/8" from the round edge of the can, then the SSB board will have full clearance. The SSB board is well above the RF board, so it is OK that the wire starts out under the SSB board, as long as it is off the edge of the crystal by the time it is 1/4" or so above the RF board. You could angle the wire clear over to the other edge of the crystal, but that would mean you'd have a longer wire, with more inductance, and minimizing inductance is the whole point here.

The holes near X10 and X11 are a little off center, so route the wire around the nearer side. On X7, it is easier to go around the outside (opposite X8). On X8 & X9, the holes are centered, so go to whichever side suits your fancy.

DON'T cut these added ground wires off flush under the board. Leave them 1/4" long or so for now.

If you have the SSB adapter, you can re-install it now and check for/correct any interference with the new crystal grounding wires. Leave it installed, as we are done up-top.
 

3. Using the end pair (X7 & 11) of those ground wires which you just didn't cut off flush, install two .01 uF (103) capacitors. One from RP4-2 to the nearest ground wire. The other form RP5-5 to it's nearest ground wire. (pin one of the RP's is square) These give a stiffer RF ground to the varactor bias voltage bus….again by reducing inductance. Any RF voltage that survives the shunting effect of C166 gets a "free-ride" into the next filter section. These caps should reduce any such leakage by 10 dB or so.  Photo 3 shows the new capacitors, and also the wires that will be added in the steps below. Also visible in the photo, are the components for Elecraft's  Xfil2 modification. Use the minimum possible lead length.
4. Lay a discarded component lead on top of the grounded leads of the two caps just installed. This wire should touch all five of the added crystal grounding wires, and run about 1/16" above the board. (closer will add more stray capacitance to the filter). See Photo 3. Cut and try until it just fits, then tack it to all the ground wires and capacitor leads with solder. NOW you can cut off the excess crystal grounding wires. This cross wire is just to attract any E-fields that my try to bridge over the ground tracks that run between the crystal leads. It probably doesn't help a lot, but it sure won't hurt anything.
5. Bend a longish wire lead into sort of a "seagull" shape. (a narrow "V" with long "wings") Solder the center of this to the ungrounded lead of C166. See Photo 3. Be sure it doesn’t short to the grounded lead of C166. This wire should lay flat on the RF board. The point of the V goes toward the crystals, and the wings of the V extend near, but beyond the varactor pads. It may take a few tries to get it bent just right.
6. Add two cross wires from RP4-4 and RP5-3 to the "seagull" wire just added. These should run between D30 & 31 and D32 & 33, as shown in Photo 3 . Again, these wires should lie flat on the RF board. If you did not cut off these RP pins, then you can form an eye on the wire to encircle the pin, otherwise just tack them down. Also, butt them to the "seagull" wire, and add a blob of solder. Don't try to cross or wrap one wire around the other…it is more important to keep the wire against the board than to make a low resistance solder joint. These wires short any stray E fields coming off the varactor pads to ground. I left the seagull wings a little long, extending past the pads on D300 and D33. I am not sure if this matters or not, but the extra shielding seemed like a good idea. Note that these wires are NOT connected to ground. They connect to the varactor bias node, which connects to ground via C166, and the new caps we just added…this IS a pretty good RF ground.
7. Re-install  bottom cover. (must be done prior to filter alignment)
8. Re-Alignment: Align XF1-4 in CW and CW-reverse modes per manual or spectrograph instructions. Be sure to read KT5X's warning about BFO shift related to counter cable.
9. Reinstall top case  covers.
10. Most important step:  ENJOY your new & improved K2. Let me know how it works out for you.

MANY THANKS TO FRED MAAS, KT5X  for his hard work and enthusiasm in supplying objective test data, and encouraging me to make this a better article.

KOÆB's homepage