Background
"Boatanchor" communication receivers are quite popular and have certain appeal. They are large, imposing, solid, create an impression of robustness, offer traditional flywheel assisted tuning, straightforward controls, etc. Their performance is rather poor by modern standards though, particularly in respect of high noise and narrow dynamic range, thanks to almost universal use of noisy multigrid (hexode, pentagrid) mixers. These issues are inherent and users have to put up with them. However, there is one common shortcoming of almost any boatanchor which can be easily fixed -- quality of SSB reception.
BFO injection
For CW reception majority of the tube communications receivers simply employ BFO injection into AM envelope detector. Eddystone 680X receiver, is not an exception. A partial (redrawn) schematic of the original Eddystone 680X, covering AM detector, AGC detector, noise limiter and BFO is presented in Fig. 1. (Here one can find full schematic diagram of Eddystone 680X. )
Fig. 1. AM, AGC detector, noise limiter and BFO in Eddystone 680X communications receiver.
In CW mode, a Hartley BFO (V6) is activated, injecting the carrier into AM detector T4 IF transformer. To allow for clean undistorted audio, injected BFO must be higher than the IF signal. Therefore injection level at T4 is rather high -- up to 10V. Coupled to the primary side of IF transformer T4, it creates some undesirable AGC voltage of several volts unrelated to signal strength. Therefore in CW mode ACG can not be used and should be switched OFF (AGC switch is not shown).
And there are more issues. Firstly, any signal stronger than BFO injected level would be demodulated directly, creating noise, distortion and interference. The same applies to impulse interference and noise. Secondly, due to interplay between coupled LC tanks in T4, BFO injection level depends on the BFO frequency (pitch), which is an inconvenience. Lastly and most importantly, due to relatively strong direct coupling (via C107) between IF transformer T4 and BFO tank T6, BFO can be pulled (force synchronised) by a strong IF signal. That adds more distortion while copying SSB transmissions and makes listening to AM stations on "zero beats" painful -- BFO is either synchronised (locked) or loses synch with coarse harsh buzz instead of expected "velvet" hum. As a result, the Eddystone 680X receiver, otherwise quite advanced, becomes unsuitable for today's SSB listening on ham bands.
The above problems can be fixed by fitting a product SSB detector instead of BFO injection method. Such techniques are covered for Lafayette HE-30, Trio 9R-59DS and Electrosound radios, while this article covers Eddystone 680X.
Pentagrid based product SSB detector
Conversion of Eddystone 680X receiver to product CW/SSB detector is illustrated by Fig. 2.
Fig. 2. Product detector in Eddystone 680X receiver. Added and modified components are highlighted green.
Here a pentagrid V6 6BE6 works in a standard Hartley self-oscillating configuration, its output being audio frequency. In SSB mode AM detector gets completely disabled by 14V reverse bias of diode V7 through RbRa divider. Audio is fed through RfCd merging at the volume control. Connection can be made at either side of noise limiter switch SW13 (as shown in dotted green), but SW13 shall be closed anyway to avoid distortion caused the noise limiter diode V13. In AM mode circuits RaCa and RfCd do not significantly affect AM detector operation. Cd is chosen relatively small to cut down residual hum, 6BE6 noise and minimise the risk of "motorboating" of the audio amplifier through possible parasitic supply rail coupling.
IF is coupled to the signal grid of V6 via the incumbent C107. For better linearity, IF signal level at the signal grid 3 of V6 should be as small as practical. For that reason dividing capacitor Cc is rather large. Smaller IF signal level improves linearity and dynamic range of the product detector, but 6BE6 noise and microphonics might become more noticeable. With too large IF signal at grid 3, overloading and undesirable direct AM demodulation is more likely to occur at strong signals. For the above reasons, C108 is to be selected on test for optimum results.
Oscillation amplitude of V6 should be moderate -- just enough to get reliable oscillations and modulate cathode current down to cut-off. About 3...4V peak-to-peak at grid 1 is probably sufficient. Excessive amplitude is undesirable, which will be explained below. Therefore, R61 should be selected as large as practical to obtain reliable, but not excessive oscillation. Once R61 value is determined, heptode plate load resistor Rc should be selected so that plate voltage is about the same, not lower than screen voltage at grids 2, 4. Smaller plate resistor will result in lower overall gain, larger plate resistor will cause tube saturation and distortion with no gain advantage.
Due to perfect isolation of BFO from the IF transformer T4, AGC can operate normally in SSB/CW mode. If AGC time constant feels too small (particularly for "attack"), then C97 can be increased to 0.1...0.22uF.
"Balancing" feature of the detector by Rd is crucially important. In effect, RdCb is nothing other than a way of creating negative bias to grid 3. Note however, that it does not affect grid 1 bias and hence the oscillation conditions, as R60 is disconnected from ground. If the negative bias is absent (Rd = 0), then, in respect of IF input, V6 grid 3 tends to works similar to a grid leak detector -- plate current would decrease with increasing IF signal envelope. If grid 3 negative bias with respect to cathode is high, then the valve would be close to cut-off and exhibit anode bend mode of envelope detection -- plate current would increase with IF signal amplitude rising. Obviously, there exists some middle bias value at which both phenomena cancel each other (at least in the first approximation). This is a point of "balance" where undesirable direct amplitude envelope demodulation is suppressed to the maximum extent. It is recommended to use a multi-turm potentiometer for the balance adjustment resistor Rd.
Because V6 cathode is tied to a tap of T6, some RF BFO voltage is present at the cathode of V6. Since cathode - grid 3 section effectively represents a diode, this cathode RF can be rectified and create negative bias at grid 3. If oscillation amplitude is too high, this rectified voltage will be relatively high as well, and may provide substantial negative bias to grid 3 even if the balancing trimmer Rd is set to zero. In such situation it might be impossible at all to reach the point of optimum balance. That is why excessive oscillation amplitude is not desirable, as was mentioned above.
Note that circuit in Fig. 2 requires the "cold" end of BFO coil T6 disconnected lifted from the chassis. If it is physically difficult, for example in case the whole assembly is fully enclosed in the shield can, then an alternative solution (Fig. 3) may be considered.
Fig. 3. Alternative arrangement of Hartley oscillator with balancing bias circuit.
In this configuration only R60 should be disconnected from ground and tied to cathode of V6. Since now Rd Cb circuit carries some RF, it should be mounted robustly mechanically and in a compact manner to reduce BFO radiation and avoid frequency wobble when shaking the receiver.
Conclusion
Product SSB detector greatly improves listening pleasure of a boatanchor. Product detector installation method, described above, and particularly the way of electronic switching between AM and SSB detector by only a single pole switch may be applicable to other receivers with similar design. Detailed discussion of this Eddystone 680X example can serve as a guide to converting other receivers.