rev AM Nov 14/09
HF Amplifier Stability at VHF
A great deal of empirical or voodoo engineering surrounds stabilizing amplifiers. While other areas of troubleshooting and engineering follow logical steps, few builders follow logical planned steps to test for, define, and correct stability problems. Much of the problem is simply not understanding the cause of instability. It is difficult to find detailed information describing why we need suppressors and how the suppressors actually function. The absence of readily available accurate information causes builders to rely on cookbook recipes, and sometimes regretfully pathological science.
Cause of VHF Oscillations
Just as in most planned oscillator systems, the most sensitive control element in the tube generally has the largest influence in determining oscillation frequency or if the system will oscillate. This is why we almost always put the primary frequency control elements of a VFO or crystal oscillator in the control grid circuit. We also need enough gain in the feedback system to overcome losses. Positive or regenerative feedback has to exceed system losses or the system simply won't oscillate.
The control grid has the most control or influence on anode current. This a case where the name, CONTROL grid, describes the function in almost every circuit. We certainly don't have a control cathode or a control anode! The grid is the dominant control system for dynamic operation of the tube.
The anode system normally has the highest RF voltage swing because it normally has the highest impedance at radio frequencies; most radio frequency current flows through the anode path. The grid to anode path is by far the most likely feedback system. The normal mode of VHF oscillation in HF PA's is at a frequency where the tube becomes a tuned-plate tuned grid oscillator. The control grid to anode path generally has the highest possible gain in the amplifier system, and that is why this part of the system is (by far) the most problematic area of the amplifier system.
The control grid system behaves like it is connected through a parallel-tuned circuit. The stray capacitance is primarily between the grid element inside the tube and ground, generally via the filament and other connections. The inductance is via the grid leads inside the tube through the socket to the actual chassis connection. At some frequency, the grid capacitance will parallel-resonate the total inductance of the grid-to-ground path.
L1 and C2 is the primary stability problem, since the parallel resonant combination it floats the grid off the chassis. Unfortunately we cannot greatly affect L1/C2 because they are mostly inside the tube. We want L1/C2 to be resonant as far above the operating frequency as possible, and to have the lowest possible impedance below the operating frequency.
L3/C4 (including C3,C6 path and C5 path) allows the anode to change voltage with current changes in the tube. We want L3/C4 to be resonant below the grid resonant frequency, and to have the lowest possible impedance. Ideally we would want it to be a zero impedance at the frequency where the grid is resonant, or to appear as a low-to-modest value shunt resistance. For maximum efficiency we want L3 to have zero loss resistance at the operating frequency.
In most amplifiers the grid and anode systems (in particular the grid) dominates system stability.
The feedback path is through C1.
The anode is actually the second most problematic area, but it is an area we can most easily alter and modify. The anode has stray capacitance to ground. The path from the anode to chassis has series inductance. Stray capacitance at the anode parallel tunes the path to ground and forms a parallel resonant circuit. This resonance greatly increases anode impedance at some very high frequency.
The grid has significant capacitance to the anode, and this capacitance forms a feedback path.
With all of this, the circuit has everything needed to become a tuned-plate tuned-grid oscillator.
If feedback loss (attenuation) from anode-to-grid is less than tube gain at some frequency, the tube may oscillate. The final requirement is the phase of unwanted feedback must be a value that causes regenerative or positive feedback. These requirements are the same in any oscillator.
Once again, the conditions required for instability are:
- Gain must exceed attenuation in the feedback path
- The grid must have a sufficiently high impedance for the amount of available feedback to cause a stability problem,
- The anode or other element involved in the oscillation process must have a sufficiently high impedance at the same frequency as the grid to cause a stability problem
- Feedback phase must be within the correct range to obtain positive feedback
If any one of these four requirements are not met, the tube will not oscillate! This is true no matter how high Q is in any individual path, or if the tube has suppressors or not.
Claims have been made that tubes will remain stable for years, and a "sudden event" will make the tube break into an uncontrolled oscillation. That absolutely can not happen unless one or more of the four important system parameters above significantly change. If one or more of the above parameters change in a way that allows oscillation, the tube will oscillate continuously until operating voltages are removed. The notion a healthy system can go along for hours, weeks, or years and suddenly break into an uncontrolled oscillation that damages components is highly unlikely unless a major component significantly changes characteristics.
Quite often, in fact most of the time, oscillations are not damaging. Consider the oscillator in a transmitter. The oscillator rapidly comes up to a state of equilibrium and stops increasing in amplitude. We never find an oscillator that can output more power than the same amplifier tube can provide operated as an amplifier, so the idea that a tube that saturates at a few amperes of cathode current can provide 50 or more amperes of "big-bang" current from an oscillation is clearly a very ridiculous idea. The cathode can't magically produce more current as an oscillator than the saturated emission would permit in any other service. Such big-bang claims might make good fictional theater, but they aren't factual.
The most common effect of VHF oscillations are creation of spurious signals; not bangs, pops, or arced bandswitches. Bangs and pops are caused by gassy tubes or other problems, while arced bandswitches (if caused by an oscillation) are generally caused by oscillations at or near the desired operating frequency!
Location of Suppressor
Suppressors are normally found in anode systems, even though other locations might also work to suppress oscillations. A VHF suppressor must be located between the tube element and a low-impedance path to ground at VHF. This is because the suppressor must be able to load down one or more portions of the unwanted oscillator circuit. The actual working circuit causing a VHF oscillation is almost always entirely different than what appears on the actual component-based schematic. The cathode, an element commonly involved in low-frequency instability is rarely involved in VHF oscillations, other than supplying electrons and stray capacitance to ground.
A VHF oscillation, if it happens to occur in an HF PA, is almost always rooted in the system behaving like a "tuned-plate/tuned-grid" oscillator.
Most of our modern PA's are grounded grid (cathode driven). Cathode driven operation requires one or more grids be directly grounded to the chassis (at least for RF) with the lowest impedance possible. This is necessary to shield the output from the input, and assure operating frequency stability and purity of emissions.
Using the anode for suppression generally works best because the grid or grids can remain well-grounded for RF, provided the best operating frequency performance.
The Most Unstable Tubes
The most problematic tubes for VHF oscillation have relatively large elements and long thin leads. Tubes of this type have low gain or are unusable at VHF because elements in the tube (shunt internal capacitance combined with series lead inductances) are actually resonant at VHF.
Internal connecting leads size and length are often the major problem. The longer and thinner the internal (and external) leads, the less stable a tube becomes. Long thin leads move the self-resonance lower in frequency while increasing element impedances. This allows even tiny amounts of anode-to-grid feedback capacitance to cause unwanted self-oscillation.
- A few examples of common troublesome tubes are 811A's, 572B's, 833's, 4-1000A's, 3CX1200A7's, and 3CX1200D7's.
- A few tubes of moderate instability are 3-500Z, 3-1000Z, and 4-400A's.
- Examples of tubes having virtually unconditional VHF stability are the 3CX800A7, 3CX1200Z7, 3CX1500A7/8877, 3CX3000F7, and 3CX5000/3CPX5000/YU-156 series.
Looking at the above tubes, it is the tubes with the thinnest and longest leads that are most troublesome. These are also the tubes with poorest VHF performance when used in amplifiers intended to operate at VHF.
The most troublesome tubes above tend to oscillate in the lower-VHF range, between 30 and 100 MHz. The typical instability frequency of an 811A or 572B is around 80-100 MHz, assuming grid leads are short and direct to the chassis.
The plot on left is the feedthrough loss of an 811A tube in a special shielded fixture. The tube can oscillate on any frequency where loss is around -25 dB or less.
Moderately stable tubes tend to oscillate at 100-200MHz. 3-500Z's, for example, generally are most unstable from 180-200 MHz.
A very important thing to remember is the closer the tube's instability frequency is to the operating frequency, the more likely it is to have damaging oscillations. This is because the tube might actually oscillate on or very near the tank circuit's resonant frequency. It also is much more difficult to stabilize a tube with a low grid frequency resonance without severely impacting
desired operating frequency efficiency and gain.
Anode Circuit Layout
Anode circuit layout can contribute to VHF instability. Long thin leads from the tube anode connector to the chassis at VHF are a problem. Problems can occur when thin (and long) plate blocking capacitor leads, thin and/or long wiring, and poor mounting of the plate tuning capacitor are used. Remember, this is a VHF path also, even if the amplifier only intentionally operates on HF.
To maximize stability:
- Use wide anode circuit leads from the tube to the tuning capacitor
- Mount the tuning capacitor directly on the chassis, or on a large metallic groundplane area that is thoroughly bonded to the chassis at many points
- Use a low-inductance plate blocking capacitor
- Keep all leads as short as possible, even if it is at the expense of having wiring "look pretty" with all perfectly aligned 90-degree angles
- Use the chassis as a groundplane and as an input to output shield. Keep the tank RF current path common to the grid grounding point in a grounded grid amplifier
- Don't ground tank capacitors exclusively or primarily to a front panel.
Grid Circuit Layout
The grid circuit layout is probably the single most important area for insuring a stable design. Long thin leads from the tube grid connector to chassis are a problem at VHF. Problems often occur from physically thin and long bodied grid capacitors or thin and/or needlessly long grid wires or wiring. The best idea is to ground grids directly to the chassis through ground lugs mounted directly on the chassis immediately adjacent to grid pins. Always think "zero length grid leads"!
To maximize stability:
- Use wide low-inductance grid leads from the tube socket directly to the chassis, connecting grid grounding leads to the closest possible point. Ideally use ground lugs right at the grid pins (rather than using socket mounting screws) for grounding.
- Use low-pass pi-network or parallel tuned networks as input matching circuits.
- Mount any swamping or grid load resistors right at or on the tube socket so leads are very short.
- Mount the low-pass or bandpass input matching system near the tube, or use exceptionally low-impedance transmission lines to reach the input matching system.
- Keep all grid connections as short as possible, even if it is at the expense of having wiring "look pretty" with all perfectly aligned 90-degree angles.
A Common "Bad Grid Idea"
One of the very worse things in modern grounded-grid triode PA's is the idea floating grids on capacitors adds useful negative feedback. This is similar to what Collins did in their 811A amplifier, and Japanese manufacturers copied the bad idea into their power amplifiers. Heathkit was also a victim of this engineering gaff. I think I know where and how it all started.
When I was designing PA's in the late 70's and early 80's, an employee of Eimac (who was also an author of many articles and a popular Radio Handbook) put considerable pressure on me to float the grids of 3-500Z PA's through small mica capacitors. He called the circuit a "super-cathode driven" amplifier. He wrote letters and called frequently, asking why I would not float the grids through small mica capacitors.
I believe this quite likable fellow creatively "borrowed" this idea from the Collins 30S1, which was actually a proper application for this type of system. This system works in the 30S1 because the 30S1 is a cathode-driven class AB1 tetrode. The 30S1, unlike the copy-cats, has zero control grid current. The grid does not shunt the upper capacitor divider with a drive-varying grid resistance. Essentially R1 (see the circuit below) is infinite in the Collins 30S1. The 30S1 also, unlike the copy cats, has a directly grounded screen that always shields the RF input from the RF output.
His theory was pretty simple on the surface. Floating control grids through small mica capacitors would form a capacitive voltage divider, with the small grid-to-ground capacitors forming the grounded half of the voltage divider. The even smaller internal cathode-to-grid capacitance would form the upper leg of a voltage divider. The driving power would be increased by this negative feedback (the grid would partially follow the cathode voltage, reducing effective grid/cathode voltage and effective drive power), and the amplifier would be "cleaner" and with reduced gain be a closer match to exciter power.
After some thought and a few questions, I learned no one actually measured performance or calculated feedback over a wide range of operating frequencies and grid currents. If this is a standard capacitive divider, the sampled feedback voltage would be constant in both amplitude and phase regardless of frequency, power levels, and tuning. To be a capacitive divider, the capacitor's reactances would have to totally dominate the system impedances.....and there is where the wheels fell off his idea.
The basic circuit he promoted, and that Heath and others used, was similar to this one:
The grid connects at the junction of C1 and C2, while the cathode connects to the top of C2.
C2 is the internal stray G-K capacitance of the tube
R1 is the time-varying grid impedance
R2 is only added to allow us to see the input impedance change of the divider on a probe model.
Sweeping the system from 100KHz to 30MHz shows us the following:
We find a huge spike in grid-to-ground impedance at 2MHz, and very uneven response above that range. By manipulating the value of L1 (the grid chokes) we can move the spike around, but we are ALWAYS left with some frequency where the grid just won't be grounded at all! Heath for example had the spike below the 160-meter band.
This is a very serious violation of good engineering practices in any grounded-grid PA, and is actually at the root of stability problems in some popular PA's. Collins, for example, had a series of field modifications to the 30L1 grid system. The best idea for the 30L1 Collins would have been to abandon the silly notion this system adds stable controlled negative feedback, and change the amplifier back to a true grounded grid. If Collins wanted negative feedback in the 30L1, the PROPER method would have been the addition of a resistor in series with the cathode feed point near the tubes.
There are obviously several major flaws with the super-cathode drive concept, when it uses a capacitor divider.
Grid current causes grid-to-cathode impedance to constantly vary with drive level. When grid current is absent, the grid-to-cathode impedance is nearly an open circuit. Grid-to-cathode capacitance dominates the upper half of the divider, and everything appears to work as planned. Unfortunately, a problem appears whenever the grid draws current. Even the tiniest amount of grid current causes grid-to-cathode impedance to decreases rapidly. With only a few dozen milliamperes of grid current, grid impedance drops to a few hundred ohms or less. As grid current is drawn, the decreasing grid impedance dominates the upper leg of the voltage division circuit!
There are also new potentially destabilizing resonances added in the grid path.
This system causes three major problems:
- Grid drive is effectively reduced as operating frequency is increased. This is the opposite of what we need! We need more drive to offset system inefficiencies on higher frequencies.
- Feedback starts to show significant phase-lag with increased drive, especially on lower bands.
- Grid-to-chassis impedance at VHF and LF is increased, making the amplifier much less stable. An SB-220 heath amplifier for example required nearly twice the parasitic choke inductance when the "super cathode" circuit was used. Still, because of pressure from this person, the circuit was added!
When I tested several amplifiers with this alleged "super-cathode" system added, IMD performance decreased significantly under some operating conditions. Stability also significantly decreased. Several amplifiers I tested using 572B, 3-1000Z, and 3-500Z tubes all had higher intermodulation distortion and required larger parasitic chokes when this super-cathode system was added!
Unless you have a class AB1 tetrode or pentode, ground the control grids directly with short heavy leads or use low-inductance high-value capacitors with very short leads in your cathode-driven PA! This system really does not belong in any grounded grid triode amplifier. Get rid of it.
What Does the Parasitic Suppressor Do?
The parasitic suppressor normally has two components in parallel, a resistor and an inductor. At low frequencies, the path through the inductor dominates the system. At very high frequencies, the resistor dominates the system (assuming it is a low-inductance resistor).
One common problem is people assume brown carbon resistors are non-inductive. That isn't the case. For an example, look at the following resistors:
All of the spiral-conductor resistors above have significant inductance at VHF, and make very ineffective suppressors unless the reactance is cancelled. Only the true carbon composition resistors are useful in non-resonant standard suppressors.
This is a typical suppressor system, including inductance of the anode lead:
In this case V1 represents the tube. The following is a simulation of currents in the suppressor:
Starting at 30MHz, the ratio of current in the inductor to current in the resistor is:
Frequency -I(L1)
-I(R1)
30MHz
0.0047
0.0015
60
0.0041
0.0026
90
0.0034
0.0034
120
0.0029
0.0037
160
0.0024
0.0041
190
0.0021
0.0042
220
0.0018
0.0043
This tells us something very important. The INDUCTOR dominates only at low frequencies. At 30MHz, current in the inductor is three times current in the resistor.
At 190MHz, in the range of the instability frequency of a 3-500Z, the resistor has twice the current as the inductor.
This tells us any changes in INDUCTOR design or inductor Q (such as use of nichrome wire) mainly lowers low frequency Q. It would have virtually no effect on very high frequency Q of the system.
- The dominant factor in controlling VHF Q is the resistor value, and any reactance in the resistor path
- The dominate factor in determining HF Q and performance is the inductor value, and any changes in inductor Q
This has been my point all along with the Measure's nichrome suppressor. Measures claims, incorrectly, his suppressors provide lower VHF Q while, in fact, they do exactly the opposite! A typical Measures hairpin suppressor actually produced significantly higher system Q in the anode of a 3-500Z (nearly twice the VHF Q), because the equivalent Rp of the suppressor in series with the anode lead was lower!
The reasons HF PA's arc are explained at other pages of this site, and include incorrect relay sequencing, load faults, as well as improper tuning and exciter transients.
Reducing VHF Q
If we want a lower VHF Q, while maintaining high LF Q and efficiency, the system must shift current into the resistor faster as frequency increases. The suppressor must also have higher Rp, so it dominates the anode path inductance that is in series with the suppressor.
While Measures openly touts his "low-Rp suppressor", the fact is a low Rp suppressor results in higher anode system Q!
A Truly Improved Parasitic Suppressor
In order to reduce VHF Q, we must have a resistance dominate the anode system. This means, in a frequency sweep simulation, the ratio of currents in the resistance to current in the inductance must be as high as possible. Let's call that slope the rate of transfer.
The rate of transfer can be increased by adding a small value of capacitance in series with the resistor:
The old suppressor was:
Frequency -I(L1)
-I(R1)
Ratio
30MHz
0.0047
0.0015
3
60
0.0041
0.0026
1.6
90
0.0034
0.0034
1
120
0.0029
0.0037
.78
160
0.0024
0.0041
.58
190
0.0021
0.0042
.5
220
0.0018
0.0043
.42
The new one:
Frequency -I(L1)
-I(R1)
Ratio
30MHz
0.0069
0.0026
2.6
60
0.0050
0.0055
.9
90
0.0027
0.0052
.52
120
0.0019
0.0050
.38
160
0.0013
0.0048
.27
190
0.0011
0.0047
.23
220
0.0009
0.0047
.19
Graphically we see the currents are:
The green curve is current through the inductor, the red curve shows current through the resistor. Notice how flat current is in the resistor, and how sharp roll off of current in the inductor becomes.
This means we will have very low anode SYSTEM Q starting at a low VHF frequency of 50-60MHz, and continuing up to UHF. Dissipation in the resistor is still reasonable at HF, efficiency and tank Q at the operating frequency remain high, yet VHF suppression is greatly improved.
Selecting Component Values
Optimum resistor value can actually be determined by measurement, or determine empirically.
If the anode path is long and thin, the impedance will be high. A high anode path impedance (thin or long leads) requires higher values of resistance, because we want the resistor to dominate the anode system impedance. The best value for a resistor is generally one that is approximately equal to, or slightly higher than, the anode path reactance at the frequency of instability.
That impedance can be measured on an impedance test set, or other ways by creative engineers or technicians, but as a general rule long, thin anode leads like 811A's require 100-150 ohms of resistance while shorter thicker anode leads like those in 3-500Z tubes require 50-100 ohms of resistance. Stable tubes with external anodes often can just use anode lead resistance, using brass or other materials, to adequately dampen anode path reactance.
The inductance has to present a significantly higher reactance than the suppression resistor value at the frequency of instability. This causes the majority of current to flow through the resistance at the very high frequency, and not the inductor.
If you look at amplifier designs, you will find tubes like 811A's generally have higher value resistors and many turns of wire in the suppressor. Tubes like 3-500Z's have significantly fewer turns, especially when grid leads are kept very short and direct to the chassis, and lower value resistors.
The more unstable the amplifier tube, the larger the inductor and resistor must be.
One way to view this is to consider the frequency response of a Hi-fi amplifier. Larger values of plate load resistors in amplifier stages reduce higher-frequency gain. The same is true in HF PA's.
Lower frequencies of instability require larger inductors, so the RF path is shifted over to the resistor at a lower frequency.
Uses For Improved Suppressors
Series-resonant suppressors are used with slightly inductive resistor paths, and larger-than- normal shunt inductors. A small capacitor is placed in series with the inductive resistor path, and this capacitor series-tunes the resistor path. This results in a very rapid shift of current into the resistor as frequency is increased. This works well with amplifiers operating at 1/3 to 1/2 the instability frequency, minimizing resistor heat while providing perfect stability.
Typical applications are 3CX1200A7 and D7 tubes, 572B tubes, and 811A tubes.
Shunt suppressors with series-resonant tuning are also sometimes used, the normal application is very high power stages with substantial anode-to-tank currents. These suppressors consist of a series R/L/C system, where the C is normally just stray capacitance to the tube anode. Sometimes these suppressors take the form of a ferrite block placed between the anode and chassis. The inductance of the block series-tunes stray capacitance, and the losses act like a damping resistance in series with that path. I've stabilized 50-100kW VHF transmitter designs using shunt suppression.
Other Instability
Some PA systems are prone to oscillation at low frequencies. Yaesu and Dentron amplifiers using 572B's, and the Collins amplifier using 811A's are good examples of production amplifiers with stability problems.
These amplifiers tend to oscillate NEAR the operating frequency.
All of these amplifiers, except the Yaesu, use tubes with high anode-to-grid feed-through capacitance and no neutralization. Worse, the Collins floats the grids for RF, reducing the already poor isolation of anode-to-cathode feedback path in the 811A.
Yaesu uses one of the poorest engineered feedback systems of all, with a capacitor from the output of the pi section back to the cathode! Phase shift in that path would vary wildly with tank circuit tuning and load impedance on the PA, as would the amount of feedback!
The Yaesu amplifier is a particular problem with Chinese 572B tubes, because grid mu is lower. Negative grid bias has LESS of an effect on cathode current, so the Chinese (and Russian) tubes draw extra quiescent current when the antenna relay is open. This additional current allows the tube to amplify while the amp is in standby. Since the antenna and input source are removed in standby, and the improperly designed feedback path to the tank output remains in place, the PA oscillates near the operating frequency with no load! Voltage in the tank builds up to many thousands of volts, because no energy is extracted to a load. The fact the oscillation is at a low frequency allows the bandswitch to see the full voltage, and it fails.
Amplifiers can create extremely large voltages when RF is applied and a load is not present!
All of the amplifiers discussed above would be greatly improved by:
- Adding a proper bridge neutralization circuit like Heathkit, Ameritron, and Gonset used in 811 amplifiers.
- Grounding the grids either directly or through low reactance very-short-lead capacitors, directly between the socket's grid pin and chassis.
- Using the improved suppressor outlined above to de-Q the amp at lower VHF.
Conclusion
I hope this information is useful, and helps people understand what really goes on in a parasitic suppression system. As time permits, I add more articles about curing unique problems in amplifiers, and diagnosing amplifier failures. I hope these pages are a good start.
Please pass this web address along to others.