HF RF SSB CW Amplifier Dynamic Electronic Bias

[ Home ] [ Up ] 

EBS-1 or HF Amplifier Electronic Bias System theory

The idea behind automatic electronic bias is a reduction of quiescent current and heat in high power radio-frequency amplifiers. The basic idea is when RF drops to a very low level, quiescent current is reduced. This reduces heat and saves a small amount of energy.

In order to function properly a few conditions have to be met:

• The bias system can never go deep class C or force the cutoff bias to needlessly high levels. Resting bias has to hover at or around tube cut-off voltage.
• The bias system must respond rapidly even with very little radio frequency input voltage. Linear operating bias must be switched on faster than syllabic envelope rise, and there must be adequate hang or turn-off delay at the falling edge of exciter RF output. This requires a fast attack system with long hang time.
• The bias system has to switch on with drive levels far below levels where high PA current is required. This is to minimize turn-on clipping.
• The bias voltage, when switched on by RF, has to be stable. This will minimize intermodulation products.

Automatic or dynamic electronic bias systems work very well so long as conditions above are met. Designs can improve when we look at and correct past mistakes.

Poor Bias Systems

The major problem with the old Alpha 77 EBS system was the designer forced the 8877 tubes hard into excessive cutoff voltage. 20 volts of bias would have been adequate, but the system forced the bias well above that level. Under no-signal conditions the excessive bias pushed the tube(s) far up into class C instead of being at the edge of class B. The Alpha 77 bias system had to pull the bias from an unnecessarily high cut-off voltage down to a few volts. When the bias starts out much higher than necessary, the electronic bias takes significantly more time to turn-on. This causes chopping and distortion at the start of each RF syllable and insures very low level signals are not amplified, at least for a short period of time. The small time delay in reaching linear operating bias squares the envelope rising edge, and causes splatter. It is the SSB equivalent of key clicks, except the spurious products repeat at a syllabic rate.

The ideal electronic bias system or automatic amplifier bias system would have the tube hang at a very slight conduction current so it is still into AB class, but AB class with minimal quiescent current. This will make resting dissipation virtually the same as full cutoff. It's a good idea to have a self-bias resistance shunting the electronic switch in any EBS system. A shunt self-bias resistance will hold the PA system slightly in conduction, but with a minimal amount of quiescent current. A 2-5k ohm resistance should be good enough for most high-mu triodes. You might want to read about cathode bias at this link.

Typical Electronic Bias Circuit Operation

In the circuit below, C2 serves only to couple the 1.8-30 MHz transmitter energy into a standard diode voltage doubler. In order to properly double, the system needs a dc path from the R6 side of C2 to ground. R1 limits diode and Q1 base current. C5 is an RF smoothing filter. R2 loads the base of switch Q1. Q1 has a 1.4 volt threshold, although a standard bipolar would switch lower. Q2 is a switch that turns on when the base goes below resting bias voltage. It follows bias voltage down until D3 eventually clamps at around 9 volts.

D3 will set the "on" bias voltage. R8 limits diode current in tube arcs. D4 clamps the open circuit to a reasonable voltage, and can normally be omitted if triodes or tubes are actually used in place of R99 (see what happens in the cathode at the cathode bias link up above).

C6 adds a little hang time. R5 limits discharge current of C6 while C6 recharges to an "off" state through R4. C3 is just an RF and transient bypass. There is a great deal of additional external capacitance not shown.

There actually is no part of this circuit with "audio" unless Q1 is not fully on. When Q1 is fully on, the collector is simply grounded. It is only when RF drive goes below two volts or so that Q1 passes any "audio". A 2 volt level typically corresponds to a power level of about 40 milliwatts into 50 ohms. At times background blower or room noise will have that level. (40 mW is 34 dB below 100 watts PEP.)

The Signal Detected

There seems to be a little confusion about what comes out of the ANTENNA connector in an HF radio. The exact words used were:

From the above response to an e-mail from me, it appears some people think audio of "50 Hz" appears at radio's antenna connector. I've even heard people describe how a push-pull RF amplifier has less intermodulation or splatter, which is also just as silly! It's easy to understand this confusion, and it's equally easy to straighten it all out (at least with most technically-minded people).

If we modulate an HF SSB transmitter with a single tone, like a 50 Hz tone, the transmitter output actually contain no 50 Hz energy at all! When the suppressed "carrier" is set at 3.800 MHz and we modulate with a steady 50 Hz tone, the RF output frequency is a pure sine wave of 3,799,995 kHz. This is why the 50 Hz tone passes through antenna tuners, coax, and radiates from a 3.8 MHz dipole! At the receiver the "TONE"  goes through a series of bandpass filters, none of which need to respond to 50 Hz! Why? Because the tone is converted to radio frequencies by the transmitter!

In the receiver the RF tone is converted to new frequencies, eventually reaching a final "mixer" or product detector that reinserts a local carrier oscillator creating a new "difference frequency" on the proper sideband, and passes the newly generated difference frequency out to the audio stages. It is at that point, and that point only, the 50 Hz finally reappears. As a matter of fact if you mistune the receiver slightly you never get the original 50 Hz in this example, you get some other frequency. 

If we look at that same 50 Hz modulated 75-meter SSB single tone on a spectrum analyzer we would see this spectrum:

The single tone RF output, regardless of audio tone input frequency, is a single narrow RF spike that actually is a steady unmodulated RF carrier! Years ago Collins and other companies generated CW this way, by using a 1000 Hz tone injected into the audio of a SSB exciter or transceiver. The output frequency of any SSB transmitter is indeed a single RF frequency when modulated by a tone. This radio frequency tone, or plurality of radio frequency tones (for speech or two-tone tests) far above the baseband audio frequency range, is actually what the electronic bias system detects.

At any instant of time during modulation of a SSB transmitter, watching with an oscilloscope, we see the waveform similar to the waveform below on the transmitter's RF output port:

Note there is no energy in the modulation at 50 Hz!! It is all on 75 meters (in this example) as it should be.

How would we model this single-tone SSB system in spice? Like this....

Note voltage source V1 is a 8-volt peak-to-peak RF (2.828 volts RMS) generator. It feeds a 50 ohm load. The power is .16 watts, or 160 milliwatts.

For a two-tone source, typical of speech, we would need to simulate the actual RF signal. Say we had a tone of 50 Hz and 2500 Hz. The 3.800 MHz LSB transmitter would now output two frequencies, 3.799,995 for the 50 Hz tone, and 3.797,500 for the 2500 Hz LSB tone.

The resulting drive waveform looks amazing like a two-tone test instead of a carrier, because it is a two-tone signal!

The waveforms and circuit shows none of the problems caused by using the incorrect simulation of audio frequencies to represent a radio frequency signal.

When the bias signal switches on, which actually occurs in about  .17 milliseconds on the initial leading edge, bias stabilizes and sits at about 9 volts. This is significantly faster than the rise time of a CW signal (normally 5 ms) and so causes very little distortion to the applied waveform. Only the very first cycle in a series of cycles is modified. This model says the EBS-1 (very similar circuit, not exact) should "turn on" at about 2 volts peak. Let's see what happens below with a "real one".

EBS-1 Test

Enough of spice models, let's look at how an EBS-1 really works.

.

Source voltage is a 35-volt power supply.

Dropping resistor is 150 ohms, 7 watts.

This is the audio test where audio frequencies were applied.

For RF, the coaxial line with BNC connector is attached. Audio lines go into a low power 3.8 MHz SSB exciter.

Below: Threshold of the EBS-1 is 10 volts peak-to-peak at 50 Hz. There are 3 volts of 50 Hz ripple on the bias line! This would be no good!

The scope has 10x probes.

In this case with audio improperly applied to the RF port of the EBS system, we see a saw-tooth waveform with about 7 volts p-p ripple in the bias line.

The applied audio level is 5.8 volts peak, or 11.6 volts peak-to-peak, and the EBS has not turned fully on. This is why the EBS, with stock component values, cannot be connected to the microphone or audio lines of a transceiver. It must be connected to the RF input of the amplifier, and it must be tested and evaluated with the RF signals it was designed to operates with.

The problem with the test above is every HF radio we use does not apply audio to the EBS-1 detector. The radio applies RF!

Let's apply 3.8 MHz RF from the same 50 Hz tone into an "ESSB" single sideband transmitter audio line input and see what happens. This is with signal level right below threshold:

Scope has 10x probes. Multiply scale by ten.

Scope is now set at 20 mV/div  for RF trace with zero at very lowest scope graticule, and  .2V/div for RF trace with zero at the center graticule.

This makes bias range 16 volts full scale, and RF range .8 volts full scale both + and - the zero center.

Bias is about 15 volts (upper bright line) with .6 volts peak RF, or 1.2 V p-p RF   (.0036 watts into 50 ohms). This is bias HIGH (PA biased just beyond class B) at 36 mW of RF.

Next we increase RF drive until bias turns on:

Now we are just barely above .6 volts peak, probably right around .64 volts. This is about 41 milliwatts into 50 ohms. At this point the bias is clamped low for full anode current. With just a 5 mW change, the EBS-1 went from standby bias mode to linear amplification mode.

This is the bias system voltage with the RF envelope below the turn-on threshold. In this case the transmitter power would be 3.3 volts peak, or 108 milliwatts PEP into 50 ohms.

The scope is at .1V/div, with 10x probes.

 Bias is at 35 volts.

This very slight increase in drive to 130 mW PEP pulls the bias solidly low, and holds it there through a two-tone 50 Hz and 2500 Hz audio into the transmitter.

This slight apparent increase in drive level is actually caused by most of the envelope being below turn-on threshold, not by an actual increase in required switch level. There is a fractional millisecond time delay for any waveform, with a constant threshold of about 50 mW at 50 ohms (3.8 MHz with 50 Hz and 2500 Hz tone modulation in this test). The EBS actually is not trying to turn on until the dotted scope lines are reached.

In the previous trace where bias stays high, there just is not enough envelope  time above level threshold to activate.

Frequency Response

The following shows frequency response of an unmodified EBS-1 board.

The total turn on delay time from full bias to full clamp (minimum bias) is 0.65 ms with a 20 dBm (100 mW) 3.8 MHz signal. 

This means if we drove an amplifier with marginal signal levels that were above turn-on threshold, the actual bias delay time to stable bias would be 0.65 ms.

 The actual bias scale in the photo is 1 volt per division. Zero is set at the white bar at the left, just above the third graticule down from the top. Low bias voltage is about 1 volt @ 500 mA. Highest bias voltage reached with this 50 Hz square wave is about 2 volts, still well within the linear region of any amplifier tube. Cut-off bias does not fully recover between 50 Hz "pulses" in this photo, so the amplifier bias remains almost as if it was never switched off.

Ramped Input Threshold Response

This test shows the threshold under a very slowly rising modulation envelope waveform. This is a 2 Hz modulation envelope at 200 mW PEP on 3.8 MHz..

Sweep is 100 mS per division.

The EBS-1 goes from full bias to zero bias at about 50 mW.

It recovers to full bias in about 2-1/4 divisions, or 1/4 second. This is the on-bias "hang time".

The response is .5 mS on delay, with .25 second off delay (hang) time. Turn-on threshold is about 65 mW.

electronic bias system amplifier

The EBS-1, contrary to incorrect spice models on one website, works just fine. Any single tone RF signal over 50 mW (1.24V peak-to-peak) into 50 ohms on 3.8 MHz would turn it on. Even a pure 50 Hz tone driving the radio! Any multi-tone signal over 100 mW also turns on and solidly clamps the bias low. It stays low even during modulation minimums where power virtually reaches zero. This is because the EBS system has fast attack response with a slight hang time.

Suggested Changes or EBS EBS-1Modifications

Early in EBS1 production, I made a mistake. I derived the EBS switching voltage from the RF Input port side of the input pi-network. This caused a problem on some bands on some frequencies, most notably on ten meters. This is because the input network, with an "open circuit load" caused by the tube being biased into cutoff, loaded the exciter drive down. This raised the switch on voltage to several volts, instead of the desired 1-2 volts.

The cure for this problem was to relocate the EBS RF sample point after the input circuit, on the tube side of the input system. Early production amplifiers may have the EBS RF sample taken from the RF input side of the input network, and this can cause problems on some bands with some exciters.

After testing in a few amplifiers, I've recommended Ameritron add a 2.2 K 2-watt metal film resistor across the switching transistor. This is to improve slower speed CW signal response through the EBS system. This keeps the tube very slightly into conduction so slower CW speeds through slow rise time transmitters are amplified without leading edge truncating and sharpening of the dot or dash elements. It isn't a good idea to remove the first .5 ms of the leading edge of every element. This does not affect sound on SSB, it does not affect cooling of the PA tube.

For now, I'd appreciate any technical feedback on this article to my e-mail address. I do not always respond because of time restraints and the number of emails I get, but I do read everything sent from other Hams.