Design of Ham Vacuum Tube Transmitters
I started this article to document my steps in a transmitter design, but I realize two things:
Instead of building one rig, what about a general overview of how things work? Maybe that is better?
Common Vacuum Tube Transmitter Design Flaws
Old power supplies were often not that good. Most supplies used filter chokes, which can be a very good thing with high vacuum rectifiers, but the chokes were often the wrong size, and often used in the wrong way.
The primary advantage of a choke is to reduce peak currents that loop through the transformer, rectifier, and first filter capacitor. High peak currents, such as a capacitor input filter produces, don't just only aggravate power supply dc ripple. High peak currents cause abnormal heating and needless voltage drops in components. Capacitor input supplies can be particularly tough on vacuum tubes, and on old transformers using smaller copper sizes. Capacitor input supplies require copper windings to be sized for peak currents, to handle the apparent power factor (APF) of the heavy charging currents at the leading edge of each sine wave voltage peak. The capacitor input supply, in essence, runs only on line voltage peaks.
An example of this is the AL1200 (AL12 series) power supplies. I've measured about 40-50 amps peak current from the line sine wave voltage crest leading edge, while average current over the entire cycle is about 12 amps. This clearly calls for increased copper size, or designs that minimize series resistance, or regulation falls apart.
In comparison, a choke input supply (with properly sized choke) has a APF of nearly unity. Current is drawn in smooth step with voltage over the entire sine wave, even the desending slope of the waveform! Doing this allows smaller copper sizes, and requires far less attention to transformer and power line design. The load from power mains to the filter capacitor effectively looks like a resistance in a well-designed choke input supply.
The bottom line on this is the user has two choices:
As a closing thought, we often try to correct design issues through power wasting in bleeder resistors. A low bleeder resistance decreases the ratio between no load and full load current. If, for example, we waste as much power in the bleeder as transmitter power demand changes from zero load to full load, we now have a load change of 50%. We halve the load change as a percentage, so we might double regulation in a system by wasting power. A better solution might be to use a better initial circuit design, or the correct components, so we need minimal wasted power (heat).
By the way, a 50 watt bleeder dissipating 20 watts contributes just as much heat to the cabinet inside as a 25 watt resistor dissipating 20 watts. The BTU/hr is exactly the same. X watts of heat will be X watts of cabinet temperature rise.
General Advice on Supplies
When using a vacuum tube rectifier in a transmitter with widely varying load currents, a choke input filter with proper rating choke will be the superior system. If solid state parts are an option, a capacitor input supply is a much better choice for size, cost, and weight.
The required choke inductance is surprising, it will be pretty large. At 120 Hz (60 Hz and full wave rectifier), minimum bleeder current in milliamperes should be well over E/L. With 700 volts and 5 henries, the bleeder should draw well over 700/5 = 140 mA.
Another factor is voltage ringing or bounce caused by the choke inductance and filter capacitance. This "bounce" can result in wild but slow gyrations in supply voltage at a syllabic rate in a SSB transmitter of class B modulator, or at a code character rate using CW. This bounce is minimized using an excessively large value of filter capacitance and minimal inductance. Here is a screen shot of the bounce in my Viking Valiant when sending a string of dots. The scale is 100 volts per division, peak-to-peak bounce is around 200 volts:
Unless there is some particular reason to avoid solid state rectifiers, or unless APF is critical, we are generally much better off using solid state rectifiers followed by a capacitor input supply. If load current dramatically varies, it is usually best for a layman to completely avoid using a choke (even in a pi type supply filter). Contrary to old outdated opinions, transformers and components will live just fine in a capacitor input supply. While regulation will not be perfect, it will typically be far better than a poorly planned or crudely implemented choke supply. You can expect 10% regulation, or better, from a well-designed capacitor input supply. A choke supply is much more expensive and heavy, and will be about the same if you are careful. If the designer or builder is not careful, a choke input supply can easily have 40% voltage sag!
There are an abundance of bad power supply designs in Handbooks, Ham transmitters, and articles! Some are beyond bad, and should only be classified as "terrible".
I won't go into formulas, except to say a well-designed choke input supply with proper bleeder current outputs about .9 times the RMS voltage of a transformer, while a capacitor input supply with solid state rectifier outputs around 1.4 times. Keep this in mind while selecting a transformer. Also remember, most rectifier tubes have a pretty high internal impedance, and do not "like" capacitor input supplies. If you really want to run a tube type rectifier, you just might have to suffer with the performance problems and complexity of a choke input supply.
Dual Voltage Supply with Bias
A center tapped transformer can be used to supply stable PA HV and screen, as well as a lower voltage for early stages. This supply acts as a full wave CT for LV, and a FW bridge for HV. It allows equal voltage capacitors, and capacitances do not need to be the same. The low voltage supply can have more capacitance. better suiting the increased filter requirements of sensitive, low-level, stages.
This supply, with an appropriate size 500 Vct transformer, makes a pretty good 700V / 350V transmitter supply for 6146's, 807's, and similar tubes. It is simple, easy to build, and lacks complications of choke input supplies.
Bias voltages can be derived from a separate winding, a rectifier negative lead filter choke, or from a "backwards" transformer boosting tube filament voltage. A 120 to 24 volt transformer, for example, makes a cheap safe 6.3 to 30 volt AC (50 Vdc bias) supply. More on bias supplies later.
I thought, since I am designing a transmitter for myself, I would walk through all the steps here.
First, I decided to eventually run a pair of 6146 tubes. I want to measure the difference between 6146 tube types at HF, and document if they really are as different as claimed. These tubes are also a good choice for CW because they have high gain and are compact for the power level.
The 6146 data looks like:
Looking at tube characteristics we can pick a zero drive bias point based on limiting to 20 watts dissipation at 700 volts (20/700= .029 amps or 29 mA safe no-drive current). If we look below at the -50 volt curve (just below G) we would have safe plate current with just over -50 volts bias. This will eliminate any need for a screen clamp tube!
The rated output of a class-C 6146 at 700 volts is around 65 watts. This would be with about 600 volts of peak anode voltage swing, or about 400 volts RMS voltage (because the waveform has high harmonic content, RMS voltage swing is less than .707*peak we might expect).
At 130 watts for two tubes, the plate load impedance would be about 400^2 / 130 = 1230 ohms. If we compare this estimate to the approximation Ep/2Ip=R we have 700/.26*2 = 1346 ohms. I will design around my estimate of 1230 ohms optimum plate load resistance. It is probably closer, although this value is not critical.
Many articles suggest a network Q of 10-12, but in reality there is very little difference in harmonic performance between a Q of 5-6, and a Q of 10-12.
The calculated difference in 2nd harmonic voltage going from a Q of ~7.5 to a Q of 12 is only 1.4 dB. If harmonics are an issue, it is more effective to find other means to suppress them than to increase tank Q.
Good harmonic suppression is more a matter of construction and layout than it is tank Q, so long as a somewhat reasonable Q is selected.
Minimum Q for a Pi-network is generally accepted to be 1+(sqrt of Rp/Rl). This means if we plan for the highest impedance transformation we should be OK at all other settings. If we don't do that, the network might act "spongy" or tune backwards, with less loading capacitance decreasing plate current at dip.
If we plan on a 1.5:1 load SWR, we should use sqrt of (1230/33) +1= 7.1. Q would be 7.1. This also covers us at reduced power, because tank Q increases at a faster rate than the required increase in Q as the ratio becomes higher. We will never have a condition, at reduced power, where tank Q is too low.
We will plan for a Q of 7.1 on all bands into 33 ohms (not 50 ohms). This gives us headroom for SWR errors!!
Because of the large values required, 160-meter tank values determine the physical area required for the tank circuit.
For proper headroom, we need just over 500 pF at 1kV for the plate tuning capacitance. I decided to use a 208 pF capacitor with a 360 pF padding capacitor for the plate. This will provide a tuning range of 375-568 pF on 160 meters. The tuning capacitor will need padded on 160 and 80 meters.
For loading, I have an 1100 pF air variable. The loading capacitor will need padded only on 160.
Also see this link
I've settled on a power supply. The schematic below is a p-spice model of the supply. R5 and R6 actually represent internal resistance of the transformer. R1, R2, and R3 represent the load by external circuits, including voltage dividers and vacuum tubes:
This supply moves the choke into the negative rectifier lead so I can extract bias from the AC voltage appearing across the choke. I've done this in a Globe Scout and it works OK.
This is a full wave bridge that uses the transformer center tap to obtain half voltage for the low level stages and screen grids of the PA tube. R1 is the HV load, R2 the low voltage load, and R3 the bias system.
Diodes can be conventional 1N4007's.
Here is how the real-world supply actually tested in a load-pull as R1 was varied. It looks like I have a good 100 watt supply, and a marginal 150 watt supply. This should be suitable for a single 6146B tube, or perhaps a pair:
Supply ESR looks to be around 350 ohms or so. I tried several chokes and couldn't find anything better, and no matter how reasonable bleeder current or how much I increase inductance I can't seem to get less than 700 volts no load from my 700VCT transformer. I'm working on stabilizing the supply more if possible.