Tuesday, June 1, 2010

AM Transmitter, 813 Style, Part 1 (PA Deck)

(This is the first part of a three-part series. Parts 2 and 3 can be found here and here.)

[Note: I've changed the circuit slightly from my original publication in this post. Refer to the 19 August 2010 Addendum below.] 

Some time ago I toured the shack of a friend, W7MS (Mike), in Reno, Nevada. I was very impressed by his collection of boatanchor equipment, but I was especially wowed by his RCA BTA-250M Broadcast Transmitter that he'd converted to 75 meter operation. 

RCA's BTA-250M was designed to generate 250 watts carrier output using a pair of 813s modulated by another pair of 813s. Mike had done a great job of restoring his radio, and the four 813s, lit up side-by-side, were beautiful. 

After I left I began thinking...I had a box full of 813s up in the attic. I wonder if... 

Well, skipping ahead...about half a year later I finished constructing my 75 meter AM transmitter. Like the BTA-250M that inspired it, it too uses four 813 tubes: two in the PA and two in the modulator. It's designed to be driven by an external audio and RF source, and I use a Johnson Ranger to drive mine (identical to how John Staples, W6BM, drives his 813 transmitter, as described in Electric Radio, issue 15). 

 My transmitter generates output carrier RF power in the range of 200-350 watts. Here's the schematic of the PA Deck :

 
(Click on image to enlarge)

Notes on the schematic: 

1. A large part of the design is based upon the 80-meter 813 amplifier described in "One-band Kilowatt Amplifiers," which can be found in the 1961 - 1968 editions of the ARRL Handbook. I designed a different pi-network using the equation in the Wingfield equations (reference recent ARRL Handbooks). 

2. Per the original "One-band Kilowatt Amplifiers" article, the amplifier doesn't require any neutralization on 80 meters, so none was added. 

3. There's a two-pole, three-throw rotary switch that's used to select screen-grid current monitoring (either the left tube, the right tube, or both tubes together). Monitoring screen current independently allows (allegedly) for tube matching. 

4. Originally, I didn't have parasitic suppressors in the plate circuits of the 813s, but, when I first started testing the deck, I was seeing a lot of high-frequency stuff on the 'scope I'd connected to the "RF Sample" Output BNC, and I thought that this might be parasitics, so I added the two plate suppressors. They changed nothing, and I later discovered (using my spectrum analyzer) that the high frequency crud was all harmonically related to the fundamental -- that is, it's the natural byproduct of a Class-C amplifier, and that there were no parasitic oscillations. I decided to leave the plate suppressors in (out of laziness), rather than remove them, but, per the "One-band Kilowatt Amplifiers" article, they shouldn't be needed on 80 meters. 

5. Given the high-frequency harmonic components that I was seeing on my spectrum analyzer (up to and beyond 200 MHz), I built a metal cage around the entire PA deck to minimize unwanted EMI radiation. 

6. The input network is the same as the one described in "One-band Kilowatt Amplifiers." C38 was changed from 0.001uF to 0.01uf to give a bit stiffer connection of the input network to ground (because there's no neutralization required, this capacitor doesn't need to remain 0.001uF that was used in the original article). 

7. The pi-network's inductor is a three-inch long piece of air-inductor stock that I had in my junkbox (2.5" diameter, 6.7 tpi, 12 gauge wire). This length gives a max inductance of about 15 uH, but I tap it at around 10.8 uH. 

8. To design the Pi-Network I first calculated the load that I needed to present to the plate using the equations for Class C RF Power Amplifiers found in the RF Vacuum Tube Amplifiers section of older editions of the "Radio Handbook," published by Editors and Engineers. (For my calculation I used 350 watts out (carrier) at a B+ level of 1650 VDC. This gave me a Plate Load (RL) of about 2600 ohms.) 

Then, given this load and the desired Q (Q should be in the range of 10 - 20; I chose 12), I used the Wingfield equations from the ARRL Handbook to calculate Pi-Network components. I put all of these equations into an Excel spreadsheet to allow easy manipulation and experimentation "on paper." 

(Note: equation nomenclature changed in later editions of the Editors and Engineers "Radio Handbook" from that used in earlier editions, and I believe an error crept into the text. The best way to determine if an edition is in error is to compare the variable being solved-for in the description of the Class-C calculation steps (particularly steps 6 and 7) against the variable being solved for in the same steps of the "Sample Calculation" that follows this description. For example, the 18th edition of the book, the terms ebmin and epmin are swapped between their use in the description of the equations and their use in the "Sample Calculation" which follows the description. If you're putting your equations into something like Excel, watch out, or you'll have a problem! I fixed this by assuming the terms ebmin and epmin were correctly used in the "Sample Calculation," and so I swapped them instead when they were first mentioned in the prior description of the calculations.) 

19 August 2010 ADDENDUM: 

I noticed that, while transmitting, the RF power output would slowly increase from 250 watts (my initial setting) to 300 watts over a period of about 3 minutes of continuous transmitting. And this effect would reoccur after I let the transmitter idle for awhile (i.e. cooling down) and then began transmitting again. 

In other words, it acted suspiciously as though the Pi-network's "loading" setting was changing with heat (its capacitance decreasing). So I pulled the RF Deck out of the rack for some bench testing. 

If heating were an issue, as a quick test I transmitted for a few minutes, then powered-down (letting the HV decay to 0 volts!) and felt various components in the RF deck. Most felt OK, temperature-wise, but one capacitor, a 500 pf, 20 KV cap that I had placed in parallel with the "LOADING" variable-capacitor, was suspiciously warm. Hmmm...could this be the culprit? 

I was using two sections (out of three) of the loading variable-cap (both sections connected in parallel). I wired in the third section of this cap (giving me 1800 pf max instead of 1200 pf) and removed the fixed 500 pf HV cap that was in parallel with the loading cap. 

Powered back up, tuned the transmitter for 250 watts carrier output power, and after three minutes...it was still 250 watts! Problem fixed! Apparently the cap was lossy and, with the tank-circuit currents, it was heating-up and changing its capacitance. 

I was a bit concerned that, with the three sections of the variable-cap wired in parallel, adjusting the LOADING control for a desired power might be a bit touchy because the capacitance might change too quickly as I turned the knob, but it's actually quite acceptable (admittedly, I have BIG KNOBS on my controls, which help when making fine adjustments). In this new circuit configuration, LOADING adjusts power from a min of about 180 watts to a max of about 370 watts RF output (carrier only, as measured using a Bird 50 ohm dummy load). I typically run the power at 300 watts carrier output. 

The PA Deck schematic page (above) is now labeled "Rev. 2", to differentiate it from the original Rev. 1. The changes incorporated into Rev. 2, are:
  1. Delete C81 (500pf, 20KV fixed cap).
  2. Change C30 from a 1200 pf max variable cap to an 1800 pf max variable cap.
(Note: Other schematic pages are still at Rev. 1).
   
The culprit!

Fixed cap removed and 3 sections of variable-cap wired in parallel.

And here's the finished transmitter, up and running!

Some additional photographs showing construction of the PA Deck...
   


Building an RF "cage" around the PA using scrap sheet metal I purchased and had cut-to-size at a local metals recycling place.

 
I grounded the metal base of each 813 in the PA section at two different spots for each tube using flexible "fingerstock." (The fingerstock flexes out of the way whenever a tube is inserted or removed). I don't know if this is necessary, but I recall reading about it somewhere (can't recall where, though, at the moment).
    
The angled piece of black material between the tubes and the front panel is actually a rectangle of PCB material that I painted black and stuck into the PA Deck to deflect the fan's air up and out through the screen material on top of the case. 

 The finished PA Deck:
 
Other Notes: 

1. The plate voltage when not transmitting is about 2300 volts DC, but it sags down to around 1800 at 250 watts out. This sag is probably due to the transformer itself coupled with the capacitor-input filter (rather than choke input filter) that I'd decided to go with (hey, I already had the caps in the junkbox). Modifying the supply to a choke-input filter may give me more output power, but honestly, it's more work than I think it's worth, so I'm leaving it as it is. (Note to self, though: next time, do a load test on the transformer and filter before installing everything!) 

2. Screen voltage is about 350 volts idle and drops to 300 volts when transmitting. 

3. At 290 watts out, HV reads about 1775 VDC, Plate current is 230 mA, Screen-grid current is 51 mA, and grid current about 20 mA. (Therefore efficiency is about 71 percent). 

4. When running at a Pout of about 290 watts, plate voltage of 1775 VDC, ebmin of about 300 volts (assumed), and efficiency of 70% (plus other assumptions per the "Radio Handbook" equations) these numbers work out to a plate load of about 4000 ohms. For an inductance of 10.8 uH in the pi-network, Q (given a 50 ohm load) calculates to be about 17, so we're in the ballpark of a Q between 10 and 20. 

5. Note the following pi-network Q relationships (using the Wingfield equations):
  • As output power increases (by changing loading capacitance), for a given value of pi-network inductance, pi-network Q will decrease.
  • As frequency decreases, for a given value of pi-network inductance, pi-network Q will increase.
How does it sound on the air? 

You can listen to a clip of the 813 Transmitter on W6THW's website here. It's the track labeled "K6JCA 813 RIG (AM)". (The rig was putting out about 300 watts, carrier power. Mic is a Heil PR-40 run through a Beringer 802 Mixer/EQ box, which feeds the Johnson Ranger's microphone input.)

References: 

 Articles:
  • "A Modern One Kilowatt AM Transmitter," W6BM, Electric Radio, #15, July, 1990
  • "813 Triodes as Modulators," W6BM, Electric Radio, #57, January, 1994
  • "An AM Kilowatt Using 813s 1989 Style," WA4KCY. Electric Radio, #5, September, 1989
  • "One-band Kilowatt Amplifiers," ARRL Handbook, 1961 - 1968 Editions, ARRL
  • "Class-C Amplifier Calculations," Radio Handbook, Editors and Engineers, 18th Edition (1970) [See note in text above re: error in equations.] Or one could use an earlier edition of this book, such at the 15th edition (pages 153-156) which doesn't have this error.
  • "Tank Output Circuits,"ARRL Handbook, 1997 edition, ARRL, pages 13.5 - 13.9 (Describes the Wingfield pi-network equations.)
Websites, Transmitters Websites, RCA BTA-250M Manual Websites, 813 Data Sequencer Designs:

Caveats!
 

Standard warnings apply: First, I may have made mistakes when writing this post or in my design. I cannot guarantee everything is correct. Second (and most importantly), this design uses high voltages that can kill you. Be cautious and BEWARE!