The Case for a Grounded B+ & other Innovations in the Design of a Vacuum Tube RF Power Amplifier

Why ground the B+??? “Because I can” is a good enough reason for me. But there's more: A huge design issue in a multiband RF amp with a pi-network output, is feeding DC voltage to the plate. An RF choke with high enough impedance for the lowest band will have a series resonance at, or near the higher frequency bands. One solution would be to locate the RF choke at the low impedance end of the pi network, and feed high voltage DC directly to the plate through the tank coil, with a blocking capacitor feeding the output. But this doubles the peak voltage on the input capacitor. Electrons don't care which part of a circuit is grounded, so why not ground the B+, and feed negative voltage to the cathode (or filament)?

Another advantage of a grounded B+: the plate directly connects to the input tuning capacitor eliminating the inductance of a blocking capacitor in the VHF parasitic path (through tuning cap to ground).

As an added bonus, a grounded B+ with B-minus below deck eliminates the need for lethal voltages topside, making it safer for “cut and try” experimenting on the tank circuit.

Bias supply and T/R switching
Of course, the filament transformer primary-to-secondary insulation needs to withstand full DC voltage, and the negative bias voltage needs to be referenced to B-minus. In my amp, the bias rectifiers are fed by the primary of a cannibalized wall-wart transformer, its secondary connected across the filament transformer secondary. The plug prongs are mounted directly on the bias board, which is supported on the terminals of--and insulated by--the plastic grid current meter, mounted below deck. A high-voltage blocking capacitor (which must withstand stand B-minus and bias voltage) is in series with the input. The grid(s) need to be bypassed to ground with capacitors of similar voltage rating.

T/R bias switching bias is controlled optically by an LED. Voltage for the 120 v DC T/R relays is developed by a 10 watt Zener diode which conducts bleeder current to ground. This insures that, should there be a power supply failure, for grid protection, excitation cannot be applied to the tube(s). For 24 VDC relays, the Zener wattage might need to be higher. With other relay arrangements, deriving LED (or neon) optical bias control voltage from the same source that activates the relays would also provide protection.

Bandswitching
With the input tuned circuitry below deck, instead of attempting to mechanically link its rotary bandswitch to the output switch, LEDs are arranged around the output switch to indicate the position of the input switch. I like lights, so all glow red except for the band of the input switch, whose one gang selects the corresponding LED, turning it yellow.

Convection cooling
The 811-A tube sockets are mounted on a thick sheet of plastic near the bottom of the chassis. Above them, a wide opening to allow convection air from below to cool the tubes. This arrangement also shortens the plate-to-ground VHF parasitic path.

Full disclosure: The idea of grounding the B+ came from an AM broadcast transmitter I installed at KUOA back around the mid-1980s.

The Bandit​

73,
Jim

EARTH: LOVE IT OR LEAVE IT​

 

Look, Ma--no plate blocking capacitor!​

The parasitic choke coils connect VIA alligator clips to facilitate checking the resistors. The sheet copper plate bus is from scraps found on the copper roof of Maybee Hall, on the Tulsa University Campus, where the studios of KWGS were located during my tenure as CE. The enclosure is of an old vacuum tube regulated DC power supply--a gift from a faculty instructor at John Brown University, where I was CE for O&O stations KUOA, and KMCK.
73,
Jim

EARTH: LOVE IT OR LEAVE IT​

 

Interesting....

 

W9JEF:
I distilled your concept to one that I can understand and note a few items and questions:

1) Your diagram does not show a plate output capacitor, however your written explanation states that there is one. Am I missing something...clarify please?

2) Since 811's are zero bias tubes, besides wanting to measure the grid current, why not ground the grids directly ....this would also reduce 'parasitic' resonances ?

3) Your schematic does not show plate "parasitic" parallel coil and resistor 'VHF' chokes, however, your actual build appears to show them. Clarification please?

4) I am currently building a parallel 811/572 amp and have discovered that there is no added advantage to feeding the cathode through a 'balanced' 2 capacitor network. In my amp a single cap works just fine. Wondering if you have experimented with this aspect and if so, share your results?

5) It seems that this configuration A): Reduces/precludes the plate high voltage personal safety risk (assuming you want to experiment with the anode circuit) AND B): Reduces 'VHF' parasitics and these are the 2 advantages of this configuration. Is this correct?

6) R3 and R4 resistances are not specified. Values ? Also, R3 and R4 appear to be redundant....my similar testing showed no significant performance advantage in balancing the currents as per your configuration. Ditto for any 'safety' risks. Have you performed any testing with only one of these connected to the filament circuit? Results?

I look forward to your responses.

20+ WPM 1970's Extra

 

The only plate capacitor is the variable that tunes the input of the pi-network. The conventional plate output/DC blocking capacitor was referred to as added inductance in the VHF parasitic path (through the plate tuning capacitor). Since plate electron flow is through the tank inductance and to ground through the RF choke across the low-impedance output, there's no blocking capacitor.

Grounding the grids directly would put them at B+ potential. Hence, the capacitors to put them at RF ground, but maintain grid voltage in respect to the cathode, which is at high B-minus potential.

Since the schematic is simplified, there seemed to be no need to show the parasitic chokes.

It was considered to be “good engineering practice” to balance the excitation to both sides of the filament.

I think the main advantage of placing the plate RF choke across the low impedance end of the pi-network is that it greatly reduces the effect of series resonance. An RF choke with satisfactorily high impedance at 1.8 MHz will likely have a series resonance at, or near the higher frequency bands. I recall a RF power amp project featured in QST having a cooling fan for the RF choke directly feeding the plate. Another advantage is the need for one fewer component; since for safety, there's already an RF choke to ground at the low-impedance output.

R3 & R4 provide a measure of current limiting in case of an arc. In my amp, 27 ohm, 10 watts. Since feeding the DC to just one side of the 811-A filament would produce an effective almost 18 volt (peak-to-peak) AC bias at the other end; again, “good engineering practice” (YMMV). The main objective being to test the efficacy of grounding the B+, taking time for additional experimentation seemed to be of dubious value.

It's pleasing to converse with a fellow home brewer. Do (or did) you have a callsign?

73,
Jim

EARTH: LOVE IT OR LEAVE IT​

 

Ooops ..Re: # 2 .....Should have known that....that's what happens when I haven't finished my morning coffee.

Re: # 4....I totally and intuitively agree with having balanced feed to the cathode is 'Good Engineering Practice' ....my Eimac handbook shows it also, but that's about it, no additional explanation, which is troublesome since I like to know 'why' a design aspect is being used preferably with actual test results....In my comparison tests I measured just the RF power output at the plate side between balanced feed and the single cap feed.

I had not considered the series resonance of the pi output aspect....good point!

Your post has given me quite a bit to think about....need to process some more and wedge in some time on my "Plywood PA" to see how it works on my layout.

Additional question I would ask is regarding the filament choke and any experience you might share....I generated a post last week about this but didn't get any responses....don't understand why....it's a reasonable question regarding common mode currents and filament choke design.

My question is in the 'Homebrew' section and is titled something like: " "Best" Filament Choke..."

I would value your inputs...

Thanks!

73

20+ WPM 1970's Extra

 

To accurately determine the difference a balanced feed would make, you should use an oscilloscope to observe any difference in AC ripple on the envelope. Especially comparing B-minus feed to a single filament lead v. a balanced feed (such as using a filament transformer center tap, or in the case of my amp, balance using a pair of resistors). That nearly 18 volt AC peak-to-peak on the unfed side may show up.

Thanks. Rather than jump back and forth between threads, I thought it best to reply here:

Was there a direct connection to the field strength meter, and measured at what point ? Are the cold ends of the chokes bypassed with capacitors on the order of .01 mf? If not, I highly recommend it.

I dug into my files, and here's the best view of my filament choke (seen at the bottom of the shot):


It's mounted topside of the quarter-inch plastic upon which the 811-A sockets are mounted. Bifilar wound on what I remember (from 2008) as an Amidon ferrite core. Looks like about 20 turns each. It's #12 solid enameled, which allows more turns-per-inch than your #10 stranded/insulated house wiring, and the slightly greater voltage drop would be insignificant.

73,
Jim

EARTH: LOVE IT OR LEAVE IT​

 

No direct FSM connection...small vertical antenna measuring 18 inches away at the outboard filament xfmr....no cold end caps. Yes, I know that good engineering practice is to have them there BUT if we assume that there are CM currents and balanced equivalent currents then the blowby should be minimal given the 10uH value at 40m...Spice analysis shows better than 50-60 dB of attenuation. My professional EE opinion is adding those bypass caps or ferrite 'Prayer Beads' is an expedient 'band aid' given the simple circuit involved. Moreover, my past shield room experience shows that adding 'Prayer Beads' will typically shift the emission frequency lower but the CM currents are still there. Anyway, this is a very in-depth subject that cannot be explained adequately in this forum method...

My foregoing FSM observations prompted my subject/request regarding my post about rod vs toroid filament choke only because I have already resigned myself to winding a bifilar choke, but wanted to know which, toroid or rod, was a better.

Additional research and inquiries show that this test data is not available.....no worries.

Thanks for sharing...I appreciate it.

73

20+ WPM 1970's Extra

 

Call them BandAids, but bypass caps are imperative when it comes to RF filtering. Unless the entire RF generating unit is enclosed in a Faraday cage, and leads to the filament transformer effectively filtered, your field strength meter is going to see stray radiation beyond what proper filtering would result in.

73,
Jim

EARTH: LOVE IT OR LEAVE IT​

 

CORRECTION: It was Kendall Hall on the TU campus, where the studios of KWGS were--and still are--located.


Regretful to say, I never got around to installing a vertical ground plane on that copper roof.

73,
Jim

EARTH: LOVE IT OR LEAVE IT​

 

Does the Bandit use the four 811A's?

If so, I'd guess the Kilovolts DC meter on the right side of the panel isn't really indicating kilovolts DC, as 3500V would blow the tubes right out of their sockets.

I love those big meters, however. Unless one is lucky enough to find meters like that on the surplus market from a seller who hasn't checked market value lately, meters like that "today" can run $100 each.

 

The note on the meter is intended to instruct the op to multiply the reading by 0.3, so the voltage on the 4 is a conservative 1080.

The Bandit moniker is the combination of Band and dit (as in Morse). And also, the two top meters are reminiscent of a bandit's mask. The “smiling” grid current meter at the bottom is upside-down for easier reading.

A couple decades ago, I got lucky at a second-hand store, and picked up several of those meters for cheap (don't remember the actual price). I suspect that they came from the estate sale of an area ham, or perhaps from a U of A physics instructor.

73,
Jim

EARTH: LOVE IT OR LEAVE IT​

 

The grounded B+ is nothing new in the history of RF power amplifiers. The Harris 5kW AM broadcast transmitter installed in the late 1980s at KUOA, where I was CE, featured a grounded B+.

However, the 2016 ARRL Handbook (the latest edition in my library) shows only these 3 ways to feed B+ to an RF power amplifier:


(A) shows the conventional circuit, but it's noted that since the choke does not present a sufficiently high impedance on the wide range of frequency bands, the heat could burn it out.

(B) shows the choke at the low impedance end of the tank coil through which plate current flows, but this doubles the peak voltage on both the Tune and Load capacitors.


(C) has the blocking cap taking the voltage off the Load capacitor, but not the Tune capacitor.

(D) is the omitted way, grounding the B+, eliminating the problems of (A), (B), and (C).

Does anyone have a later edition of the Handbook; and if so, does it include option (D)?

73,
Jim

EARTH: LOVE IT OR LEAVE IT​

 

I have a 2018 and it doesn't show D, either.

IMHO, at amateur power levels, grounding the B+ creates more problems than it solves. But that's just me.

 

There are the tradeoffs. Could we discuss the problems of grounding the B+ at amateur power levels?

73,
Jim

EARTH: LOVE IT OR LEAVE IT​