This area will contain some modification notes for the SMiTe, our little surface-mount 80-meter transceiver.

This is a note from Gary, N3GO, one of the KnightLites' ENGINEER EXTRAORDINAIRE, describing a mod to the SMiTe that should REALLY increase the usefulness of our little rig.

Greetings Gang;

I got a query about whether one could mod the KnightSMiTe to transmit in the lower sideband. (i.e. below the receive frequency. This is desirable if one wishes to check into a sideband net (as an emergency appliance).

The answer was yes, and the solution should appeal to us all. I sure wish I'd figured this out before I decided to release it for production :-)

The mod fix tunes the transmitter on the KL Net frequency... The TX no longer moves with the tuning of the RIT control capacitor. The RIT function remains unchanged and the result is a RX/ TX offset that's user selectable. On two units tested, I was able to achieve nearly a KHz offset. This means that you'll be able to call CQ and hear stations that zero your carrier when responding.

To achieve this result do the following:

Remove diode D2 and connect an N-channel J-FET (J309 or equivalent) as follows:

Source - Connect to pad where the cathode of D2 was connected. (This is also connected to the pad where the rig is keyed)
Gate - Connect to any convenient ground reference. (I choose the bottom of C3 and R2)
Drain - Connect to the junction of Y1 and C1. (I found the top of C1 most convenient for me.

How it works:

The gate of the N-channel J-FET is negative with respect to the source during receive and thus the channel (drain to source path) is pinched off which presents an extremely high resistance (and considerably less capacitance than a 1N914 diode) and is effectively an open circuit. As such it has no effect on the oscillator tuning network (C1 and L1 in series) to which it is connected and C1 permits tuning of the oscillator frequency during receive.

When the rig is keyed, the source is grounded placing the gate and source at the same potential ( zero volts). The drain to source resistance of the FET is minimum under this bias condition effectively connecting the bottom of the crystal Y1 to ground by short circuiting (and thus disabling) the oscillator tuning network (C1 and L1).

The result allows the crystal to operate at (or very near) its marked series resonant frequency during transmit and the tuning network serves as a true RIT.

For best performance, adjust the receive frequency as high as possible for maximum TX/RX offset which will result in a receive tone of several hundred Hertz.

72

Gary, N3GO
Raleigh, NC

Prof Jay and Gang;
I did the legwork to resolve the question of FET substitution for D2 on the KnightSMiTe. So far, I haven't found an N-channel J-FET that doesn't work in this application, so if you have any N-channel J-FET not included here, give it a try.
For the record, I tested the following 4 devices and all produce identical results... (i.e. fixed TX with independent RIT):

     MPF-102       (Radio Shack P/N 276-2062 @ $0.99 each)
     2N3819          (Radio Shack P/N 276-2035 @ $0.99 each)
     J-309              (don't recall where I got these)
      NTE-312      (Same here... Perhaps Dan's Small Parts???)

     MPF-102   &   J-309      -     Drain         Source    Gate

     2N3819                   -     Source        Gate      Drain

     NTE-312                  -     Gate          Source    Drain

Gang;

As a test exercise on my recently procured copy of the ARRL Radio Designer circuit simulator, I attempted to optimize the KnightSMiTe for power output while maintaining the harmonics below the FCC limits of -30 dBc (30 dB below the (output) carrier). I incorporated the result on my own KnightSMiTe and am happy to report the outcome was as predicted by the simulator.

To keep this brief, the following modifications will improve the KnightSMiTe transmit performance without degrading receiver performance (other than susceptibility to 40 meter broadcast interference... not an issue for most CONUS users).

Change the coupling capacitor C4 (between the emitter of Q1 and the base of Q2) from 82 pfd to .02 ufd.

Change C7 and C8 (the output filter capacitors) from 820 pfd to 470 pfd.

Change C9 (output 40 meter trap) from 220 pfd to 180 pfd.

The power output will increase to nearly half a watt (the bad news is that the rig now consumes 100 ma of current at 9 VDC). Transmit efficiency appears to be just over 50 percent with this mod, but the rig will now put out quite a healthy signal. I was able to extract a full watt at 12.6 Volts, but eventually burned out Q2. I was using a small package (SOT-23) 2N3904 for my test.

Those with the hefty 2N2222 may get more output power... (it has higher gain than a 2N3904) and it will handle the higher power dissipation.

GL and 72

Gary, N3GO

You can simply change C7 and C8 and get nearly the same result. C4 and C9 only improve output power slightly. In addition, you may in fact wish to leave C9 alone (i.e. 220 pfd) to maximize second harmonic attenuation and enhance rejection of 40 meter broadcast band interference.

Here is a message from Gary, N3GO, sent to a SMiTe owner who was having some trouble. This message can serve as a nice summary of troubleshooting the KnightSMiTe!

The KnightSMiTe is quite easy to troubleshoot, so don't despair. First plan of action is to determine what's not working... and the important assumption here is that it was working for some period of time implying that it was assembled correctly. If that assumption isn't correct, it would serve you well to verify the diodes and tantalum capacitors are installed in the correct orientation. Inspect the transistors to verify they are not installed upside down :-) These critters are so small it's actually possible to install them that way... experience speaking here :-( Given that you're certain it's properly assembled and an actual failure has occurred, finding the fault should be straight forward.

With 9 Volts power applied, proceed as follows:

1). Connect a short wire (or test lead) to the emitter of Q2 (in receive mode). You should hear a 60 Hz buzz at the output of the audio amplifier. This confirms that all is well with the receive detector on :-)

2). Next, tune an external receiver to around 3687 KHz. (+/- 700 Hz)... still in receive mode, and with the short wire still connected. Your signal if it exists will appear in this region of the spectrum and confirm that the oscillator is working.

3). Finally Key the rig (dummy load desired, but not required with 9 Volt supply), and verify that the frequency of the oscillator shifts to 3684.4 (+/- 100 Hz or so) assuming you have replaced D2 with an N-channel J-FET. The frequency will only shift slightly (100 to 400 Hz or so depending on the trim capacitor setting) if your KnightSMiTe doesn't have this modification. This test verifies the keying circuit operation...

Completing these steps will identify where the problem has its roots :-).

At this stage, you go to the ABC's of troubleshooting :-)

A). Verify power is applied to the circuit and is of the proper voltage (in both receive and transmit modes). A weak battery may work in receive, but drop out when switching to transmit due to the increased power supply load.

B) Verify the voltages around each device (particularly the one that appears to be inoperative) is reasonable (i.e. zero if it's supposed to be at ground or isolated via a coupling capacitor such as the output filter, or the same as the battery supply if directly wired that way, and somewhere in between if isolated from supply and ground via resistors or transistors.) Note that inductors look like direct connections (short circuits) for this test. For example the base of Q2 (the output transistor) is connected to ground via the inductor at the base during transmit but should be about 1 to 1.2 V lower than the supply voltage in receive mode, while the collector is connected directly to the supply via the inductor on the collector.

C) Verify that voltages don't appear where they shouldn't appear. For example, a common problem reported with the KnightSMiTe is that with no antenna connected, a small DC voltage appears at the antenna terminal which should be isolated from DC via the coupling capacitor C6. The problem manifests itself as motorboating in the receiver and reduced or no transmit output power. The cause (in every reported case thusfar) is a solder bridge across C6 during assembly.

A failure of any of the above tests should point you to within a component or two of the problem. Since you've already changed the two transistors, I'll assume step 1 passes Armed with these two pieces of information, I predict that the failure is due to a solder bridge or open, so inspect carefully... with lots of magnification. Check for shorts first with your ohmmeter in the low ohms position (if you have a continuity "beep" function, so much the better... The test will go fast in that mode.

In any case, the KnightSMiTe is quite robust and you're not likely going to be able to damage anything catastrophically... including connecting the battery backwards :-) At high voltages (12.6 Volts and above however, you run the risk of burning out Q2 due to overdissipation if not well matched on the output.

Gary, N3GO

This is a note from Pierre, KA2QPG:

I have a Pixie II, which I built from the K5FO schematic found on the Web. It is a fine minimalist rig, and I have made several QSO's with it, but the selectivity of the receive section leaves a lot to be desired. Particularly annoying are the SW broadcast stations which fade in and out with music and voice, sometimes wiping out the CW. SMiTes and other Pixie variants also have this problem.

Several have suggested audio filtering, which is possible with additional circuitry. But I wanted to keep the rig in its minimalist beauty, while improving its performance. So I concentrated on the RF end, and after some experiments I came up with a fix that eliminates almost all the BCI with no active components: a quarter-wave shorted stub.

The idea is nothing new. I found it mentioned in a 1947 ARRL Antenna Book! All you need is an electrical ¼ wavelength of 50 ohm coaxial cable (44 feet of polyethylene RG-58 or 53 feet of foam RG-58, for the KL frequency) and a "T" adapter to attach it to the antenna jack of your rig. I used an Ethernet BNC "T" from my computer job. One end of the coax has a BNC plug, and the other end is soldered short and taped up.

When I plug this stub into the "T" while listening to the Pixie, the broadcast interference just disappears, leaving only 80m CW signals audible. It's really an impressive improvement. Also, the high, steady carriers I usually hear (probably from the crystal's second harmonic beating with 7 MHz broadcast carriers) disappear.

I would recommend this filter for anyone who wants to use a Pixie or SMiTe for actual communication. It would work on 40 meters with the appropriate scaling, but would probably not be so effective because many broadcasters are right near the frequency of operation.

72 de
Pierre Thomson KA2QPG

Note: Pierre also is looking into an LC bandpass filter and suggests the following reference:

For plans of an LC filter suitable for multiband field work, see June 1994 QST, page 32, for an article by N1AL.

He also has another tip:

Another simple fix that helps the CW copy... add a largish inductor (a potted 100 mH in my case) in series with the 10 uF gain cap at the LM386 chip. That gives some preference to low audio freqs and reduces the "ultrasonic CW" to comfortable levels. At this time of the year, it also cuts the general QRN crash level by 10 dB or so.

Sam, AE4GX, offers these notes:

I am MODing the PIXIE 2 to try to get as much out of it as possible. I am working on two problems.

BC Breakthrough: I have eliminated it with a tuned LC toroid circuit tapped up from the grounded end(about 25%) to join the output of the LP filter. There is a 5 turn link on the grounded end of LC circuit that connects to the antenna and ground. (This basically looks like a tiny antenna coupler). I was thinking about replacing the LP filter with this same tuned circuit and link arrangement. I think I have seen this tuned output circuit configuration in one of the old SPRATS. I can see I still may need the LP filter to reduce the harmonics but was thinking that the tuned circuit directly at the collector of the PA Q2 might help the selectively. Has anyone tried this?

Sensitivity: I can hear CW signals but the volume is fairly low. Has anyone played with the bias resistors to the PA Q2 to see if the RX mixer action could be improved without affecting the power output during TX. It looks like the bias resistors are effectively out of the picture during TX. I wanted to try this before adding any AF Filter/preamp. I see a fairly BIG OSC signal at the BASE of Q2 and was wondering if it could be OVERDRIVING the DIODE detector action during RX?

Trying to increased my QRP knowledge. Any comments would be appreciated.

Sam AE4GX

(and more from Sam:)

I continue to try to improve my PIXIE2. The latest change is to put a tuned LC toroid/cap circuit tapped down(about 30%) and connected to output of the LP network with the end of the new LC grounded. A link coupled winding over the cold end of the LC circuit connects the antenna and ground. This acts like a miniature antenna tuner and eliminates the BC feedthrough even using my 125 endfed longwire. Before this change the PIXIE was completed swamped with BC interference.

After making a QSO with Lou in Knoxville, TN (I live in Atlanta, GA) on 7.040 (2-way QRP at 5w both about 569) I asked could he listen for me about a 1KHz lower on the PIXIE. Changed gear(same antenna my inverted L longwire at 35 ft) and called. He came back with a 519 and he got a 559. Another PIXIE 2-way QRP QSO.

Right after the call I checked the PIXIE into a dummy load with my scope and calculated 18milliwatt (my battery under load 6.6volts at 60ma..???? not very efficient). The input into the LC added circuit was about 50 milliwatt.

Again I chalk up my personal records with this amazing $10 rig and I'm still not done yet. Stay tuned!

Sam AE4GX

(and yet more from Sam:)

When I ended the last segment of this story I had a miniture antenna tuner coupled to the antenna connection of my PIXIE. I saw I putting out a very low 18mw and knew something was wrong since I was pumping in about 400mw. Taking the new tuner addition out and checking it found two interesting things: First, the LC was tuned to 5.9MHz so I had to take off some turns and retweek the cap. Second, using my MFJ SWR analyzer and a dummy load I found the SWR very high : >>3:1. So I put two four turn links (one end of each link grounded) on the cold end of the LC and rechecked the SWR. Now its <2:1. One link goes to the antenna and the other link goes to the rig's output point.Putting it back on the rig and checking power got about 180 mw output across dummy load. BIG IMPROVEMENT.

Now Local AM broadcasts breakthrough is GONE but occasionally I hear foreign broadcast. Found out that it was station on 7.535 kHz coming in at S9+60 db on my ICOM756 with the endfed 135 ft longwire. Plan to put a series LC circuit from input to antenna tuner to ground. Should eliminate this guy or at least greatly reduce him.

Now for the BIG change: VXO operation. I hated the fact of the two fixed frequencies. I was moving the xtal frequency using the two different caps (switched) to ground. This originally gave me about 800Hz shift. Good for another PIXIE (I worked AE4NY's PIXIE at two feet. HiHi.)

So I wound a FT50-61 ferrite toroid with 31turns and put it and a 25pF var. cap in series with the Xtal. Now get 6KHz swing (7038 to 7044KHz). BETTER. Took the original offset switch and switched in or out a very low cap to provide tx/rx shift. Works great middle of the range but too much or too little at the ends. BTW the 0 - 50% of 25pf shift covers 2/3's of the range. So the fix offset would be different between the ends. Hum ....Plan to put two var. caps in rig and switch them manually between rx/tx. This should give me flexibility between tx/rx operations over the 6 kHz range. Watch out guys her I come. VXO(6kHz), 180mw and a wire antenna. How's that for PURE QRPing.

Sam Billingsley AE4GX
e-mail: sbillingsley@usaninc.com

Mike, DF2OK, shares some ideas and experience on filtering for BC interference on his 80m SMiTe (and is working on 20m versions of the SMiTe!):

I've noticed, that the audio-amplifier is overloaded in the evening because of the strong signals on 7MHz BC band. My mods are:

Protection against wrong polarity with two diodes, one in the 9V bus and the other in the ext. (13V) bus. The voltage lost and the higher inner resistance of the diodes in this circuit are the reason that I connect an elko of 330 uF from +UB to ground behind the diodes, 9V in of the TRX.

As I wrote some time ago, I reduced some hiss by soldering a 2,2 nF Cap from Pin 1 to Pin 5 of U1.

The higher external power of about 11 Volt and up produces a buzzing noise out of the audio amplifier. Trying around I get this terrible noise off by connedting a 4.7 nF Condensor from Pin 8 to nearby ground of U1.

The next step was to increase the value of C10 up to 100 uF. After this mods where done, I was able to use this nice rig in the late evening and night here in DL, too.

I found some solutions of reducing BC interference by searching in the Internet of 'PIXIE2' as a keyword in searching machines.

I found for me, that a low-pass LC circuit helps a lot.

Between C15 and Pin 2 of U1 I put a choke of 1 MilliHenry in series and a capacitor of 0.1uF from Pin 2 to ground. All these components are not SMD, but what shall I do if you have none, hi :-)

Changing the band on which the SMiTe operates involves changing the crystal and the output L and C. Here is a response from Gary, N3GO, to Mike's, DF2OK, query about getting the rig on 20m:

    Try the following:

            68pF  for C7   
           15 pF  for C8
           15 pF  for C9
           1.5 uHy for L4

I did an optimize of the circuit using ARRL's Radio Designer. I haven't tried these out on the bench though so you are on your own. I think the values will work well though. The Pi values should give you best power output into 50 ohms with 2N2222 and C9 resonates with L4 to trap the second harmonic (28 MHz) to keep you within FCC limits on harmonics (< -30 dBc). 10 or 12 pF may be a good compromise on C9 if the 3rd harmonic comes up too high. It will suppress the third at some sacrifice to rejection of the second.

L3 should be OK as is (i.e. 22 uHy). It serves as an RF choke and should still work fine at 14 MHz. Reducing it to 5 UHy doesn't degrade performance much though so do as you wish with it.

Ditto with L2. Same story here. Reducing the value may improve efficiency here just a bit, but I doubt you'll notice much difference. It will increase the drive into the Q2 but probably won't increase your output power much.

>I know, that the circuit C9 L4 is a resonant circuit for >about 7.2 MHz. So how to modify this?

Yep... It's a harmonic trap. Adjust C9 to null the second and third harmonics equally. This will minimize transmit harmonics as well as suppress strong signals in those bands from creating interference in your RX. I guess you don't need to worry much about FCC requirements :-)

You didn't mention L1... You may need to reduce it's value to 30 to 50 uHy. Leave it alone until you determine if it needs changed though. The "problem" that it may introduce is it may cause your oscillator to stop when you attempt to push it too far. If it becomes annoying to you, reduce the value to one that gives you the largest tuning range whilst maintaining oscillation.

For 14 MHz, you may want to see if you can get a higher frequency NPN for Q2. This will improve your output power, and transmitter efficiency.

Here are some very interesting notes from Todd, KB0HQU:

Some of the folks wonder about modding the SMiTe. Well, I have a suggestion for how to do it! It requires a little mechanical finesse. In other words, don't drink a pot of MY coffee, the way I like to make it, and then try to do this.

Gang, we have a whole back of the board to put things on. If you have access to copper tape, you can put that down for pads. If not, you can nibble out little squares of pcb and glue them down for pads (old idea, new area :-). If you don't have that, then you can run little runs of bus wire around, and just solder the parts directly to the wire. Optionally, you can super glue the parts down before routing the wire. Here is the really tricky part: get a Dremel or some other way to turn really small drill bits. Then get a really small drill bit. A pcb bit is what I have in mind here; the 1/16" bit is too big. Something on the order of 0.040" or 0.030" is fine. Drill from the backside into the copper pads where you need to make the interconnect. Then reflow the solder on the top around a piece of wire you stick through to make a via. If you did that first, you could then bend and route the wire flush to the board on the back, and solder the added parts to it.

Now, this puts parts on the back of the board, and you can't mount it flat down on a surface. But if the parts are all on one side, you can still use a piece of double-sided foam tape Radio Shack has, or used to have, some). If the parts are all over the back (meaning you are very mod-some :-), then just cut little pieces of this stuff, and put it where parts aren't, and stick it down in your tin or whatever. After you do your mods this way, folks won't be able to tell at a casual glance that mods were even made. Heck, at a casual glance, most non-hams can't even tell that it's a RADIO :-)
Todd

Here is a hint from Gary, N3GO:

replacing D2 with a J-FET is quite worthwhile and I humbly apologize for not solving that issue prior to releasing the final layout :-( I was focused on getting the rig out before Christmas and agreement was unanimous that most folks on this list might consider room for improvement a design "feature" :-)

If you are interested in the Knightlites and Amateur Radio contact Paul at

AA4XX@bellsouth.net