K8IQY’s Optimized 38 Special
By Jim Kortge, K8IQY

First of all, I’d like to say that I have had many, many private conversations over the past few weeks with Ori Mizrahi-Shalom, AC6AN, and Glen Leinweber, VE3DNL about many of the changes that I have included in this article. Glen is the one responsible for all of the changes to the audio amplifier, and these were posted to the QRP-L gang via one of his posts. These discussions have been both enlightening and most useful in working out some of the details. I also have included many original ideas that we never got around to discussing, due to other priorities in our lives.

Here is a rundown of the changes in my version of an ‘optimized’ 38 Special. I’ll list them, and then take them one-by-one and discuss how to implement each, with details on building.

Modification Overview


The reason for this order is that it basically follows the layout of the schematic, and the radio. Hopefully, it will aid in following and understanding what the changes are, and why they were made. Each modification will list the original parts affected. Then I’ll describe what needs to be done to duplicate my 38S.

Optimized front end filtering:


Parts affected - T1 and C4. The reason for this change was to minimize the amount of feed through that I was hearing from stations on or near 12Mhz when band conditions were enhanced. Input transformer, T1, is wound with a 2 turn primary and a 17 turn secondary. Capacitor C4, changes value from a 0.01uF, to a 10pF unit. The combination of tuned transformer secondary and coupling capacitor match the input impedance of mixer U1, and keeps the bandwidth of the input filter as narrow as practical, consistent with adequate signal coupling.

Adjustable VXO setting capability:


Parts affected - L1. The reason for this change was two fold. I wanted to get as much range out of the VXO as was practical, and easily adjust the lower end to 10.1Mhz. This was accomplished by making L1 a 10uH choke connected in series with a 5-45pF trimmer capacitor. The choke was installed vertically, using the right PC board hole. The trimmer was installed vertically also, using the left PC board hold. The lead from the top of the choke was connected to the top of the trimmer, and the job was done. A 6.8uH choke would also work, but I didn’t have any of those. The lower end of the band can be easily set by adjusting the trimmer. With this arrangement, I get 23Khz of coverage. As a side note, my 38S does not have the RIT installed. I didn’t feel it was needed with the limited coverage of the main tuning.

2 pole, 500Hz bandwidth crystal filter:


Parts affected - C9, X2, R6, C10, C17, X3, C21, R9, and D6. None of these original parts are used except R9, which will be installed vertically. More on that later.

This is an ASCII rendition of the new filter. It uses 2 series mode crystals in a mostly conventional "ladder" arrangement. The only departure from a traditional ladder configuration is the output capacitor which is a parallel element, instead of the usual series element. This was done to properly match the input of mixer U3, which is 1500 ohms. 

                                12MHz      12MHz
   U1-4 -|---      |------||-----xtal-------xtal--------- U3-1

         P  |      S     270pF          |           |

     19t P  = 100  S 5t                 |           |
         P  | pF   S              270pF =     220pF =
   U1-5 -|---      |                    |           |
                   Gnd                  Gnd         Gnd
I used a transformer to get the 750 ohm push-pull output of U1 (pins 4 and 5) down to 50 ohms to drive the filter. The primary is 19 turns, wound as close together as you can get them, using #26 wire on a FT37-2 core. The secondary is 5 turns, wound in the gap left by the primary. I've got a 100pF cap across the primary to tune it to 10.1MHz. As you can see, the secondary drives a 270pF cap in series with the first 12MHz crystal, then another 270pF to ground, the second 12MHz crystal in series with the first, and a 220pF cap to ground on the output of the second crystal. So, a conventional ladder setup with crystals and caps, except for the final cap which is a parallel element instead of being in series. This is needed to match the 1500 ohm load at pin 1 of U3.

Traces to cut:


The transformer installs with the primary leads in the holes that Ori provided. They are the two holes, vertically oriented to the left of center of the T401 outline, when looking down at the component side of the PC board. The secondary uses the two lower holes Ori provided, within the outline, to the right of center. This will ground one lead, and attach the other to the 270pF capacitor on the input of the filter.

The 100pF capacitor which resonates the primary of T401 goes on the bottom on the PC board along with the two 270pF capacitors, as described above. The 220pF output capacitor goes in the C10 location. A wire is used at the C17 location.

This takes care of the filter itself, but to make it work properly, we have to make changes to the second mixer, U3, also. (Nothing is ever easy is it!)

A third crystal is used for X3. At the location of capacitor C21, we install a 4.7uH choke connected in series with a capacitor of between 82pF and 120pF. The capacitor value depends on where the crystals in the filter, and X3 are in frequency, relative to each other. I ended up using a value of 100pF, and the passband was just about perfect. Some experimenting will be required here, or just use a 100pF value and see how it sounds. The choke is installed vertically in the left hole provided for C21, the capacitor is installed vertically too, in the right hole. The tops are connected together. This point will also be connected to the switching transistor described below.

Finally, to make it work, we need a transistor to switch the capacitor out during transmit so that the offset is correct. Exactly the same approach Ori used. However, I found that diode D6 didn’t switch well enough, so a 2N3906, PNP transistor was used. Install R9 in it’s right most hole, and leave the vertical lead for a moment. Bend the emitter of the PNP transistor so that it is pointing 180 degrees from the base and collector leads. Install the transistor base into the hole marked for the cathode (bar symbol) of diode D6. The collector lead can go into one of the spare ground holes nearby. The emitter is connected to the top lead of resistor R9, and the the resistor lead is also brought around the transistor case and attached to the junction of the 4.7uH choke and the 100pF capacitor.

This completes the filter installation. For reference, the crystals I used are from Mouser, part number 520-HCA1200-S, and are a 12 MHz series unit made by ECS. I also bought 4 cyrstals made by Fox, but they are lower in Q than the ECS units, and are not recommended.

There is one artifact with this configuration. The transmit offset is shy of the required 900Hz by a few hundred hertz. What does this mean. If you call a station, he will probably use his RIT to tune you in optimally. If you call CQ, the returning station will be about 350Hz lower than you would like him to be. Don’t return him, cause you will find him chasing you all over the band. This small problem will eventually be rectified, but at the moment, its solution is waiting to be discovered!

900 Hz audio filter:


Parts affected - C15, R21, C35, C36, and R24. These changes should be made if the 2 pole crystal filter is used. Otherwise, the passband tuning established earlier will not be correct. Capacitor C15 uses a value of 0.1uF instead of the original 22uF. This reduces the low frequency coupling into the filter. Resistor R21 uses a value of 680 ohms. Capacitors C35 and C36 are 0.047 or 0.05uF, depending on what you have in your junkbox. Either will work fine. Resistor R24 is a wire to increase the output audio level for lower impedance headphones.

Optimized transmit preamp stages:


Parts affected - L2, C25 and C26. The number of turns on L2 is shown as 20 turns. My experiments have shown that this is one too many. L2 should be wound with 19 turns to get it to resonate correctly with the component values specified. To reduce the low frequency coupling from the 2N3904, pre-driver to the U4A stage, capacitor C25 is reduced from a value of 0.01uF to a 330pF unit. This change will prevent the transmit control waveform from being passed through the succeeding stages. I think this is one of the mechanisms responsible for the IRF510 heating problem posted to the QRP-L reflector. Capacitor C26 should also be a 330pF unit to properly load the output of U4A. A 270pF capacitor should be added to the output of the U4B stage. This capacitor can be installed either on the bottom of the PC board, or if you do your final amplifier using my approach, you can install this capacitor in the top holes provided for R101 and R103, since these parts are not used.

IRF510/MTP3055E 4 watt final amplifier:


Parts affected - C101, C102, R101, R103, and Ferrite Bead. Although most will probably use an IRF510 for the final, I used a Motorola MTP3055E MOSFET. I have several of these, partly because Motorola has a PSPICE model available for this device. It is not unlike the IRF510/511 series of devices.
First, cut the trace on the bottom of the PC board as indicated for the original 5 watt modification. In my version, C101 is not used, a wire jumper is used instead. A wire jumper is also installed for C102. This grounds the lower end of the 220 ohm resistor (R102) directly. Another one of the mechanisms responsible for the excessive IRF510 heating may well be the effect of using C102 and R103. These components integrate the 0 to 8 volt r.f. switched drive to the gate of the IRF510, causing the bias on the gate to run at nominally 4 volts. On many units, this bias level can cause excessive drain current. R101 is not used, nor is R103.
I also did not use the ferrite bead on the gate lead as shown in the original modification. My reading has lead me to believe that this causes instability, instead of reducing it.
This final amplifier implementation, along with the optimized low pass filter, renders a final that is clean, efficient, and free from parasitics. Mine runs 3 watts when on battery power, and 4 watts when on the bench supply at 13.8 volts. More importantly, I can run solid carrier for 30 seconds, and the MTP3055E is just slightly warm to the touch, using a small, finned, heatsink.

Optimized final low pass filter:


Parts affected - C505, L3, and L4. There have been numerous posts to the QRP-L reflector regarding whether to use a capacitor at C505, and if so, what its value should be. I ran my filter modeling program and designed a low pass filter, which preserved the original values for C28, a 820pF capacitor, and C29, a 560pF capacitor. My output filter uses another 560pF capacitor for C505. In addition, inductors L3 and L4 are wound with 13 turns, instead of the original 8 and 14 turns respectively.
One of the objectives that I wanted to achieve was to reduce the 12mhz component getting through the filter, since this is one of the unwanted byproducts coming out of the 2nd mixer during transmit. The 12Mhz component is not highly attenuated by the filter in the input of the 2N2904 pre-driver stage, so the output filter must take care of as much of what get through as possible. Based on testing with the signal generator, this filter is about a good as you can get for 5 poles. The 12Mhz component is attenuated about 15 dB, without substantially reducing the 10.1Mhz component.

TiCK keyer implementation:


Parts affected - D8, R15, R16, C31, R17, C32, and R18. This modification starts with cutting the PC trace between U4E and resistor R17. Just follow the instructions in the 38S manual. Another trace needs to be cut, which disconnects a side of capacitor C32, so that it no longer feeds point "A", as show on the schematic. The correct location involves the right most end of C32 (looking at the bottom of the PC board) from the trace that connects capacitor C15 to resistor R19. This disconnected point is then rerouted to the other side of R19, using a wire on the bottom of the PC board. Components D8, R15, R16, and C31 are not installed. Components R17, C32, and R18 will use different value from those shown on the schematic.
Install the TiCK keyer chip along with the components shown on the TiCK schematic. R17 will need to be a 10M resistor. C33 uses the original 0.01uF value. R18 becomes 4.7K, and C32 uses a 0.1uF capacitor. These values give a good sidetone telemetry level, but are still a bit too loud when transmitting. Somewhere, there seems to be another path that the audio is getting into the audio amplifier, other than the path we just provided.
That’s all the changes that I have installed to date. How all of it performs is a matter of opinion, but in the week that I have been running this configuration, I am delighted with the results. I’d still like to find the source of the ubiquitous "thump" that we all are hearing. And I’d like to track down the reason the sidetone is so loud during transmit. But these are minor annoyances.
Overall, the 38S is a really solid performer, even without the changes that I have made. I admire and appreciate the clever design Ori brought to the original configuration, and forethought to include room and allowances on the PC board for just the kind of changes I have described. With the changes, this ‘lil rig really performs. I’ve had numerous contacts with it, and repeatedly am astonished at the 579 to 599 reports I get, and the comments about being impressed with what 3 watts can do.
Have fun....see you on 30 meters with your 38S.

© Copyright - February 23, 1997
by Jim Kortge, K8IQY
PO Box 108
Fenton, MI 48430-0108

Return to KA7NOC's Home Page