Tuesday, July 9, 2019

Arduino Keyer Analyzer



Every ham who programs MCUs and enjoys CW will probably do a keyer or two over time. My first was on the 6502 contained within my Atari 800 back in 1987.

We all know the basic ratios and WPM calculations, but there's a lot of other stuff involved in the logic of keyers. They aren't all the same - we have iambic, Ultimatic, Modes A & B and other variations.

Even if you have the above characteristics nailed down, there are still some subtleties hidden within the black box. For example, if you are sending a string of dits, how long do you have to release the paddle after the last dit before another one latches in?

I was working on programming a keyer and decided to put it on hold long enough to take on another programming project - to create a gizmo that would connect to a keyer's paddle input and keyed line output and make the timing measurements to answer my questions.

It determines dot, dash and space times in milliseconds and reports ratios and WPM. It determines logic type Ultimatic and iambic Mode A or B. It determines and reports the point at which another element latches in. It finds the delay between paddle closure and TX keyed line closure and measures the first dit to see if it is shortened. It will also send a string the user inputs on the tested keyer.

In reality, several of the attributes of a keyer can be determined just by playing with it for a bit, but others require a device such as this one.

To keep it as simple as possible, I/O is via a serial terminal screen such as the one integral with the Arduino IDE. Nothing more than an Arduino is required. The device is powered by the USB connection. I also incorporated an option for the device to report results in Morse (via your own keyer!), so the serial terminal would not be required.

An issue does arise when you want to test the keyer built in to your transceiver, since you don't have access to its keyed line out port. I dealt with this by making a little circuit to sample the RF from the transceiver while it transmits into a dummy load.

What about compatibility of the voltage level from the keyer's paddle jack to the Arduino? Well, I don't want to exceed 5 V input so I checked several keyers and they all passed the test. Here are some examples:


Keyer                                    Volts
K3                                         5.0 V
WA5BDU                             5.0 V   
Super CMOS II                    4.5 V
FT-991A                               3.3 V
Winkeyer 10                         5.0 V  
K3NG Arduino                    5.0 V


The "WA5BDU" is my Arduino keyer. I also tested the KX3 and the ATS-3 QRP transceiver and they were OK too.

Here's the menu which appears on the terminal screen:

WA5BDU Keyer Analyzer V1.0

Select:

   0 - Verify functionality
   1 - Timings, speed, ratios
   2 - Check Iambic & Ultimatic operation
   3 - Check Mode A or B
   4 - Check time for same element latch
   5 - Dot paddle open/close
   6 - Dash paddle open/close
   7 - Send PARIS for WPM
         * - Send serial port text

Here are some outputs from various requests from the menu:

Key Out idle state: HIGH
Dot length ms is 61
Dash length ms is 181
Space length ms is 61
1st dot length ms is 61
Speed: 19.7
Dot Ratio: 1.0
Dash Ratio: 3.0
Time to close TX line ms is 14
Time to get off paddle is 47
This is 77% of a space.
Speed using PARIS: 19.7


  The above are responses to menu items 1, 4 and 7.

Below is a table of results for various keyers I tested:

Keyer
Dot-R
Dash-R
Iambic
Mode
Latch wait
DLY to TX
WPM_20
WA5BDU
1.0
3.0
YES
B
53%
0 ms
19.7
WinKey 10
1.0
3.0
YES
A
98%
1 ms
**
Super CMOS II
1.0
3.0
YES
B
102%
9 ms
19.9
ATS-3
1.2
3.5
YES
B
98%
1 ms
19.3
FT-991A
1.0
3.0
YES
B
73%
17 ms
19.5
K3 (QSK)
1.0
3.0
YES
B
79%
121 ms
19.6
K3 (Semi)
1.1
3.1
YES
B
81%
13 ms
19.6
KX3 (Semi)
0.8
2.7
YES
B
76%
14 ms
20.8
K3NG Arduino
1.0
1.0
YES
B
102%
15 ms
19.9
K3NG U-Mode
1.0
3.0
NO
A
100%
15 ms
19.9
PK-Basic
1.0
2.9
YES
B
100%
6 ms
20

Below is my schematic for the keyer analyzer. There's not much to it until you add the circuit for sampling a transmitter's output.



So there it is. If you'd like to try it, an Arduino can be had for around $3 to $8 and my software is free to you. I've also written a manual for the thing with a great deal more detail. I've got the manual and the source code available in this shared Dropbox folder:

https://www.dropbox.com/sh/rsa4bb0ekdf69q0/AACfWsGcNrRTZgxEP4WOQyyAa?dl=0

I hope you enjoyed reading about it and possibly even trying it.

73,

Nick, WA5BDU

Friday, March 15, 2019

70 W eBay MOSFET HF Amplifier

I'd been working on my phasing receiver for about four years and decided it was time to team it with a TX section so I could make some QSOs and say "RIG HR IS HOMEBREW", a goal I'd had for some time. Not a QRP rig this time, but something in the 50 to 100 W range. I'd already taken a step in that direction by building a little Class E amplifier for 40 and 20 meters using three TO92 MOSFETs and supposed to produce 5 Watts out. (Lotsa luck on that ...) That could be my driver.

OK enough preliminaries. I read about the eBay amp on one of my ham mailing lists and I thought at the price, why not give it a shot?


I paid $18.57 ppd for it. It's said to be a class AB linear amplifier for SSB AM CW FM power amplifier for low-power radio power connection and suitable for FT-817 KX3 other small power stations.

It's basically a "bag-o-parts" and a circuit board. No schematic, no instructions. But you can find lots of stuff on the web, fortunately. Your fellow hams always helping out.



Here's what I got in the parcel from China. The board seems to have good quality. The parts look OK too - tubes for the one-turn transformer windings, ferrites, insulators for the TO220 parts, including shoulder washers.  And in my case, some SMT parts were pre-soldered. A nice bonus. The MOSFETs have had their part numbers removed. Ham community says IRF-530.

The board has some parts that you break out and use for the ends of the toroids, to solder the tubes that extend through the binocular balun holes to. This leaves two rectangular openings. The smaller ones are where the MOSFETs mount and bolt to the heat sink below. The larger one is where the big ferrite cores for the output transformer go. 

Finding some info ...

I found some very useful info on the web, including a YouTube video or two.

OE1CGS has some really good info, including a PDF he put together on the amplifier, some of which has been translated into English.

http://www.oe1cgs.at/50w-hf-verstaerker/

Finding a decent schematic can be difficult. You may find that some look good initially but lack sufficient resolution to read part numbers and values with confidence. I eventually did find a pretty good one.

Here's another good page on the amp, with measurements and good photos from DK9JC:

https://www.dk9jc.de/blog/equipment/142-diy-kits-70w-ssb-linear-hf-power-amplifier-ft-817-kx2-kx3

PD7MAA has more good info, Lowpass filter design info, a schematic and more good photos:

http://pa-11019.blogspot.com/2016/11/diy-kits-70w-ssb-linear-hf-power.html

Some discussion of the circuit ...

The circuit is fairly standard. It includes some frills like a regulator and pot to set the bias voltage / idling current. And a switch-around relay that feeds the antenna around the amp to the driver when 12 V is not applied, and through the amplifier when it is. There is a 2-pin header labeled PTT that when closed turns on a small PNP transistor which feeds the bias regulator and also feeds 12 V to the relay. The MOSFET drains are always connected to the 12 V line. I keep a shorting jumper across the PTT pins.

The MOSFETs have 150 ohm feedback resistors from drain to gate through blocking capacitors. Except mine were not 150 ohms as shown on the schematic. Was this a design decision, or did they just run out of 150 ohm resistors?  Who knows? Mine are 100 Ω. 

One thing that seemed to be something of a glaring omission was the lack of any kind of swamping resistors on the input to set a constant resistive load to the driver. I initially used my KX3 with ATU to try to get a good match but after reading comments and suggested resistor values from OE1CGS, I added the resistor (21 Ω). It's across the secondary of the input transformer. After that, my driver was looking at a much better SWR.

Some notes on building & testing ...

My building notes are exhaustive (tedious?) so I'll try to hit the high points. Of course, fit up the pieces and understand how they go together before soldering anything. The photos on various blogs I found to be helpful for that purpose. I copied some down to my own folder.

One thing to be careful of - the tabs of the two MOSFETs are head-to-head and there needs to be a gap between them or there will be a short. Keep that in mind. Maybe stick a little insulating shim like a piece of toothpick between them before soldering. 

And of course, the board is designed to sit on top of the heat sink and the MOSFETs bolt directly to it, with insulators. It's important to have full contact area and tight hardware. I think a few folks thought they'd test very briefly with no heat sink and blew their MOSFETs.

Since my board came with the voltage regulator components pre-mounted, I hooked up 12 V before putting anything else on the board and verified that it was working and that I could vary the voltage with the little pot.

After mounting all the parts, I wanted to set the bias. Even with no drive, I felt that I needed to hook up a load to the output in case of self-oscillation. Or maybe to prevent it. And 50 Ω across the input isn't a bad idea either. Another good safety practice might be a switch or hand key across the PTT pins in case the circuit runs away and you want to shut the bias off quickly.

The schematic shows 2.7 V bias. Most people find that too low. Adjusting it until the MOSFETs are carrying a little drain current is a good idea.

I tried various values of idling current, up to 120 mA or so, but eventually set on about 30 mA total current. The relay coil draws about 42 mA at 12 VDC, so subtract that off of the total reading from the supply. Or adjust bias to minimum, note the current, then increase until current increases by 30 mA. My bias voltage was 3.41 VDC at 30 mA total MOSFET current. Setting current higher might be OK but think about dissipation. 

For full power testing, a good dummy load is a good idea, since there's no SWR protection here. A way to monitor power - RF probe, power/SWR meter or oscilloscope is desirable.

A way to control the drive power is good too as gain varies on different bands. Maybe a 1 W output TX could be a good starting point. I think testing without an LPF into a dummy load is fine. Jumper across the LPF - using the supplied 0.1" pin headers.

I would not recommend going key down for longer than a few seconds, maybe 5 maximum until you know where you stand. I have a cheap Harbor Freight IR thermometer that I'd point at the transistor bodies and get a relative idea of how much they warmed up from a few seconds key down, and then to be sure they'd cooled down before doing another key down session. Feeling the heat sink is not a good way to see how things are going, in my opinion.


Some deviations in what was supplied and other info ...
  • MOSFET numbers are sanded off. Unknown P/N. I ordered 10 IRF530N from eBay.
  • Feedback power resistors are 100 Ω, not 150 Ω.
  • R1, R2 and R3 are 3900 Ω, not 10 kΩ.
  • R7 is 200 Ω, not 1 kΩ.
  • Pot VR1 (or VR3) not marked but measured and calculated as 785 Ω. Possibly a nominal 1 kΩ.
  • Some bloggers said adjust bias to 3.7 VDC, not 2.7 VDC. I used 3.41 VDC. It probably varies. With 3.7 VDC, my drain current "took off".

Results ...


With 1 W input:

BAND
SWR
WATTS
80
1.2
61.1
40
1.5
45
30
1.4
25.3
20
1.4
16.2


Of course I can get more power by increasing the drive on the higher bands. This is without a LPF, but I noted on 40 meters that power didn't decrease noticeably when I added the LPF.

On the air ...

So far I've just used it on 40. My Class E transmitter puts out too much power on 40 (4.2 W) and not enough (2.3 W) on 20!  So I built an attenuator for 40 meters to give 1.2 W drive. With that and a 13.0 VDC supply, I'm putting out a hair over 50 W. I'm a bit afraid to try for 70 W until I get more experience with it.

I lashed everything together with my homebrew phasing receiver, keyer and T/R switch and started making contacts. It's a lot of fun, as I expected. I've had some fairly long QSOs and some strings of contest QSOs with no performance issues. I put an antenna tuner in line and make sure the SWR is as flat as possible before transmitting.

BTW, the driver for the Class E section is my Si570 synthesizer, which also is used with the receiver. I had to add to the software to have it do the CW offset shift on key-down. I'll need other features (RIT, for example) shortly. But I'm getting off-topic ...

MOSFET failures ...

Everyone has them. I've had them twice. Once was cause by "human error" - I created a short with the lash-up of parallel resistors used for my 21 Ω swamping resistor. The second time it was simple over-heating, I think. Key-down too long and/or mounting hardware for the TO220 tabs not tight enough.



Here's a photo of it mounted on its over-sized heat sink, with LPF on the left and input attenuator below. The blue resistors are the feedback resistors. The gray ones in parallel with a smaller blue form the 21 Ω added swamping resistor. I hope to replace that with one properly sized resistor eventually.


Here's the schematic I downloaded from the PD7MAA site. It may not be hi-rez enough to build by as shown here, but I just want to give an overall view of the thing.

Some more about heat sinking and thermal issues ...

A large heat sink is a plus, but no matter how big yours is, a lot of heat has to flow through a small contact area to keep the MOSFETs from failing. Solid contact, good thermal conductivity through the insulator and adequate bolting force are necessary.

I was reading part 2 of WA2EBY’s popular MOSFET amplifier article in QST of April 1999. He noted that even with proper heat sinking techniques, his amplifier might exceed the maximum allowable temperature for the MOSFETs if it were held key-down for more than five seconds. However, he was able to send a string of continuous dits without overheating. So it's on the edge and if the duty cycle is kept withing typical ham CW/SSB values it will do OK.  I hope.

Final thoughts ...

Well, for my eighteen dollars and change, I think I got a decent little project and learning experience. I need to do more with it of course, and integrate it better into this theoretical transceiver I'm constructing.

Is it homebrew?  We worry about such things. Well - it's a nice board and bag of parts, but you get to find a schematic, fix a design flaw or two, build one or more LPFs, decide how to drive it, blow up some MOSFETs and learn from the experience.  So I guess it might qualify.


Tuesday, November 14, 2017

PIC Frequency Counter With Morse Output

It's a little strange to blog about a project that's mostly software, but even if I don't explore the software design in depth I can at least describe the resulting instrument.
For hams who program MCUs, a frequency counter seems to be one of several obligatory projects, with others being a keyer and a controller for a DDS or PLL type synthesizer. So I wanted to get in my frequency counter project.
Counters such as this one have been around for a while, often integrated into radios with analog VFOs. With the tap of a button, you hear the frequency accurately announced in Morse and thus the need for a calibrated dial is avoided.
I said "With Morse Output" but it's common to refer to the device function as AFA, for Audible Frequency Annunciation.
Before I ramble much more, let me list the features of my counter. Every programmer likes to add a wrinkle or two to previous implementations and I've added one or two:


Features
One button control to initiate count or menu actions
User programming of superhet IF frequency offset
Selectable 1 Hz or 100 Hz resolution
Option to suppress higher digits for faster readings
Can be built SMT or through hole
Small size - I used a RS proto board 1.75 x 2.75 inch
Setup saved on EEPROM and recalled on power-up
Three selectable Morse annunciation speeds
Selectable audible marker for each 1 kHz change while tuning

Specifications (approximate)
Accuracy:  Depending on the time base, can be as good as 1 Hz
Resolution: Selectable to 1 Hz or 100 Hz
Response time: 1 s for 1 Hz resolution and 10 ms for 100 Hz
Sensitivity: 35 mVpp to 180 mV through 6 meters
Idling current draw at 12 V input is 11 mA
My prototype SMT version will read to just over 140 MHz
Minimum frequency is about 50 kHz with input capacitor shown

Some hardware notes and schematic
I'd switched from PICs to AVR MCUs a few years ago but the PIC is unique in having a prescaler that can count really fast - independent of the chip's clock. So I went back to the 16F683 8-pin PIC for this project.
I looked at several other designs although I didn't have their source code to guide me. Some use Microchip's AN592 app note as a starting point, but its accuracy can degrade to 1 kHz or worse at higher frequencies. I liked the Stinger Singer counter sold by the Arizona SQRPions group and designed by Dan Tayloe. The use of a 74HC00 AND-gate package allowed precise gating of the input stream and also gating out and counting the remainder stuck in the prescaler, which can't be read directly.


Sorry it's fuzzy. I'll provide a link to a better schematic.

I built my DIP version on a Radio Shack 276-150 prototype board. Radio Shack is gone but I found some at B.G Micro. I also built a SMT version on a board I made using the toner transfer technique. You could omit the programming header if you have a pre-programmed chip.

Want to try it?
I've added a link to the HEX code. You can burn it to a PIC with a PICkit-2 or PICkit-3. Or I could program a PIC for you and mail it in the USA. Say $4 to buybye@suddenlink.net which is my PayPal address. Or four ones to my QRZ address. First make sure I'm still alive and capable of fulfilling the request by emailing me at kennnick@gmail.com.
I've also linked a manual below, with a lot more info on building and operating the counter.

Possible improvements
I like the feature I added to give an audible "tick" marker with every kHz of change. But it introduced an artifact. When this feature is ON, the counter must read the frequency continuously and the pin that drives the speaker is also used in counting logic. This results in a low "puttering" sound from the speaker that's almost but not quite inaudible. Of course you can turn it OFF if it's a bother. One fix I've considered is to use a separate crystal oscillator with a transistor for the timebase. That would free up one PIC pin to dedicate to the speaker. Another fix would be to use a higher pin count PIC. But a 14 or 18 pin PIC would make the counter so much larger. I could justify it by integrating a keyer function into the chip as others have done. That would make it worthwhile.

Files



73,
Nick, WA5BDU

Tuesday, March 7, 2017

PIN diode T/R switch

100 W T/R switch in box

I like to play with separate transmitters and receivers, either those I'm building or a few classic boat anchors I'm hanging on to.

Back in the 60s I participated in NTS (National Traffic System) nets where full QSK was part of your entry ticket if you wanted to be taken seriously. But even though I've grown used to the smooth "semi-break-in" of modern transceivers, the notion of "electronic T/R switching" has a mystique for me and I've always wanted to try it.

Description

First let me try a brief description of the switch if I'm capable of it (being brief), before I get into tedious detail over why I chose every minor component in the circuit.

The goal was for it to handle 100 watts on 80 through 10 meters and 6 if possible. Minimal SWR caused in the transmit path, less than 1.2 if possible. As much isolation as possible between the transmit and receive ports. I measured from 68 dB on 80 meters to 60 dB on 10 meters to 53 dB on 6 meters between the TX to the RX ports. That gives at most 0.1 mW to the receiver on 80 through 10. That's a loud signal but not destructive.

SWR to a 50 ohm dummy load is generally about 1.15, rising a bit a 3.5 MHz and 50 MHz.

The circuit uses back to back PIN diodes in the paths from antenna to transmitter and again from antenna to receiver. One pair is biased to conduct in PIN diode mode with forward current while the other is reverse biased with high voltage to the OFF or blocking condition. These biasing conditions are swapped to switch between transmit and receive.

The PIN diodes are actually common rectifier diodes which happen to have a PIN type structure. This is primarily in the interest of keeping costs low in the ham tradition.

Circuit




This is the main portion of the schematic, less the power supplies. The apparent complexity is reduced  when you see what's happening in the circuit.The two IRF830 MOSFETs control whether the diodes are in conduction mode or blocking. The '2222A inverter after the keying input causes the MOSFETs to be in opposite states, one on and one off. 

Consider the four current paths at the top from the 13.5 volt supply, when enabled these paths provide forward current of about 100 mA forward through each diode. Follow through the limiting resistor, two chokes, diode, then two more chokes and then the MOSFET to ground. One MOSFET will be ON and the other OFF, so either the TX or RX diodes will have forward bias current through them.

Next look at the 180 VDC supply coming from below and connecting to the drain of each MOSFET via 100 k resistors. For the MOSFET that is ON, that voltage is simply pulled to ground. But for the one that is OFF, the voltage appears between the cathodes of the diode pair, keeping them well reverse biased.

Besides antenna to TX and antenna to RS, there's one other somewhat optional path to consider, which is via the 1N4007 diode pointing down just left of the RX connector. This is a shunt path connection to provide an additional path to ground for any RF that makes it through the back-biased 1N4007s in transmit mode. In receive mode it is reverse biased with the high voltage supply and doesn't conduct.


Component choices


First, why two chokes in series at each location?  I copied from the design of WB9JPS (NA6O) on this. I'm not sure why he did it, but I have a fear of unknown self-resonance frequencies (SRF) in molded chokes when I'm trying to cover such a broad frequency range. I suspect the toroids with only ten turns will be fairly free of resonance which would address this concern. Probably just the toroids alone without the molded chokes would have been OK.

Next, the transmit path capacitors. For RF, ceramic capacitors such as type NP0 and dipped mica capacitors are usually specified. Capacitors described loosely as "film" types are said to have excellent characteristics for audio, but how do they do at HF and low VHF? Are they inductive?

So I got a little carried away with testing. My test consisted of putting the capacitor in series with the coax from my 100 W transmitter going to a dummy load. I checked SWR on 80 through 10 meters (and usually 6 as well). I held key down for 15 seconds and looked at temperature rise of the capacitor with my cheap IR temperature detector. Before hitting them with 100 W, I looked at SWR in a broad scan using my VIA (a sort of antenna analyzer).

I won't report all of the results here except to say that everything I tested did mostly "OK" - no huge SWRs or serious heating (one degree or so typical). I checked some large rectangular ceramic "disks", orange 400 V PETP 0.01 uF (three in parallel) film capacitors, large white 0.68 uF rectangular film capacitors also 400 V and four 0.01 uF SMT 500 VDC capacitors in parallel.

So I chose the big ceramics as  looking pretty robust and testing well. For other paths and bypassing I used SMT 0.1 uF, 500 V capacitors I have in my junkbox.

Now, the PIN diodes. Type 1N4007 are often used. Hayward used 1N4006s, which are also PIN diodes. For the transmit path, the 1N4006/4007 might be marginal for the 1.4 A rms required for 100 watts. Hayward and WB9JPS both used NTE5815. I think a 1N5408 is reasonably equivalent and I have them in my junk box. These are 3A diodes so I get some margin at the expense of greater capacitance. If I'd set my target at 50 W, the 1N4007s would have been OK in the transmit path. 

After these choices, there's not much that's critical.

Power supplies

My first idea was to use an external 12 to 14 VDC supply and provide the somewhat "high" voltage using one of those boost modules from eBay. But the little switcher was just too noisy, despite my efforts to filter it. In hindsight, I'm connecting it right to the receive antenna, more or less, so it would have to be super clean to be acceptable. So I went "linear" and power my switch from 120 VAC. 

For the HV, an obvious approach would be to use two back to back 120 to 12 VAC transformers to provide isolated 120 VAC and rectify that to 170 VDC. But I didn't have enough room so I stole a transformer from a Knight capacitor checker with magic eye to give about 130 VAC.  This one has to supply just about zero current so a simple capacitor filter following a bridge and it's done.

Oh, let me show the schematic before I continue:



Not much to explain in the low voltage section, but why the 78L12 regulator?  The unregulated 13.5 V output jumps to 25 V with no load. By design, the no load condition should never happen, but to protect the logic devices I have now and any I may add in the future I added the regulator. The 10 and 2.2 ohm resistors are my obsessive attempt to get to just the voltage I want for the bias current supplies. The low voltage supply uses a little plastic bridge while the HV uses four 1N4004 diodes. The 12.6 VAC transformer needs to be rated 400 mA or better.

Diode measurements and LTspice

My proposed diodes got a treatment similar to that I gave the capacitors. I wanted to be sure all this stuff really worked before committing to building. I made a jig to measure diode capacitance at 200 V reverse voltage and at or near 0 V with my AADE meter. 

1N4007:  2.1 pF @ 200V, 20.5 pF @ 0 V
1N5408:  9.2 pF @ 200 V, 76 pF @ 0 V

So the high voltage bias not only prevents the transmit waveform from forward biasing the diode but also keeps the series capacitance as low as possible.

I was trying to do a PIN diode T/R switch in LTspice using a downloaded 1N4007 model but the diode didn’t work like a PIN. I searched the LTspice Yahoo page and found a question by Don Huff W6JL with the same problem. He was told his model needed a Tt= statement in it for PIN behavior. I copied the text for his successful 1N4007 and now it shows PIN diode behavior.
Also, most models have Cjo specified at 0 V bias. So you don't see the lower Cjo you get with high reverse bias. I could change that to the value measured with high reverse bias, but of course it's only valid in that condition. But still to get a good prediction of performance it can be worth changing these values when switching between transmit and receive mode analysis.

PIN diode behavior and bias currents

RF switching with diodes can be done with "conventional" diodes, but the PIN diode is a different animal. Charge carriers that remain in the 'I' region of the diode allow current to flow in the reverse direction for a time after polarity switches to reverse during an RF waveform. So the diode is acting more like a low value resistor than a diode.

These charge carriers will be used up after a time, so there is a low frequency limit to this effect as well as a power (current) limit. This is why I'm not trying 160 meters. Note that the DC bias current serves to create the carriers, but its magnitude does not have to approach the maximum current of the RF waveform. I'm not certain how the required current magnitude would be calculated, but Hayward said, "I've found that a current of 100 to 200 mA is more than enough for 7-MHz operation at the 100-W level in a 50-ohm system." Note that at 100 W, peak RF current is 1.4 A.

Just as the PIN diode effects limit low frequency performance, the diode junction capacitance with reverse bias limits the amount of isolation obtained at the highest frequencies.

Board and enclosure

Recently I've been using the toner transfer method of board making for all but the simplest projects. I use SMT as much as possible since I don't like drilling holes. In this project I used a significant number of "leaded" components, but adapted them for surface mounting as can be seen in the photo.

The enclosure is a steel chassis I had on hand. I found that the two transformers generate a surprising amount of heat for their size and power ratings, so I'll probably drill some holes for ventilation when I put the cover plate on. 

Performance and enhancements

I made a number of measurements of isolation and SWR which looked good, but here I'm speaking of on the air performance. I used the T/R switch between my Knight T-60 transmitter and Drake 2B receiver during an on the air activity for several days on 40 (mostly) and also 80 and 15 meters. No failures or problems were experienced, just smooth T/R switching.

T/R switches such as this are often integrated into a "sequencing" system which keys various transmit stages, receiver muting, sidetone and so forth. I don't have much need for sequencing at the moment but might choose to design a keyer on a PIC or AVR chip so I can simply plug my paddle into the T/R box and have outputs to key the transmitter and/or VFO. A sidetone and muting are other options I might choose to incorporate.

Do it again, smaller

I allowed myself plenty of room for this project, which in the world of boatanchors is OK. But I think I could do a lot to make it smaller, especially if the target power level were 50 W or less. First, use a single choke in each bias leg, whether molded or toroid is found to be better. Then use more SMT devices, especially for the diodes and capacitors. And surely there's something smaller than IRF830s to handle 350 mA or so in saturation and block 200 V or more with margin.

The issue of generating the "high" voltage without RF noise will still require some thinking.

A problem and a fix

I was pretty pleased using the T/R switch with "novice" transmitters like the DX-60 and HT-40, with 30 to 45 watts out. But I got a surprise when I tried it with my HT-37, which puts out perhaps 100 W. 

On key down, the SWR zoomed. I eventually found that the T/R switch was switching between transmit and receive at a high rate - maybe 50,000 Hz or so. I tried many fixes and theories to no avail. At one point I found that a capacitor was "singing", but when I replaced it with a different type, it sang too. So it was responding to the problem, not causing it.

I found that I could start at low power and then run it up to 100 W with no issue, but keying at 100 W set off the oscillation.

I put it away a couple years and when I returned I started thinking about time constants. In NA6O's circuit, he used 0.022 uF capacitors from the drains of the MOSFETs to ground, and even sketched in the time constant. I didn't have that value in the voltage required in SMT so I substituted 0.1 uF. Realizing that this might be the problem, I changed those caps to two 0.01 uF units at each location and tried the switch with a 100 W transmitter. Success!

Potential builders please note: I have not changed my schematic as yet, so be aware of the change described above if you build a T/R switch based on mine.

Files

Schematics get a bit fuzzy and hard to read pasted into this blog. I'm going to put a schematic image and the ExpressSCH format file of the schematic into the Dropbox folder linked below. Also the ExpressPCB file for the board and anything else that seems useful.

References

"Electronic Antenna Switching" by Wes Hayward in QEX, May 1995. Also on EMRFD CD
"Experimental Transmitter" on the website of Gary Johnson WB9JPS (now NA6O):
http://www.wb9jps.com/Gary_Johnson/TR_Switch.html
QRZ.com page of Don Huff W6JL
Also see the Magic Box T/R switching system in 4SQRP.com, by K8IQY
I note that my schematic is very close to that of NA6O with component substitutions and power supply design variations noted.