Message	Date	From	Subject
5119	2010-09-16 02:31:03	davidpnewkirk	A fascinating cascode-triodes-detector finding--and more (long)
As I've posted earlier, the cascoded-triodes regenerative detector is a actually synthetic tetrode--a screen-grid tube synthesized by stacking two triodes between ground and B+. I've also posted that experiments have shown that using a cascoded-triodes synthetic tetrode as a pin-for-pin replacement for a 12SJ7 pentode detector already optimized to resist pulling on CW signals received at low beat notes gives performance equal to or better than the 12SJ7. One aspect that I did not investigate further until now involves my unusual initial finding that with the first dual-triode tube I used--a 13DE7 dissimilar dual triode wired with its higher-mu section closer to ground (connected to the Hartley tuned circuit) and its lower-mu section closer to B+ (connected to the wiper arm of the REGENERATION control such that its grid acted as the tetrode screen grid)--would, in conjunction the existing Hartley-tapped tuned circuits ( http://mysite.verizon.net/dpnewkirk/ej/goodman_receiver/#12SJ7_detector ) in two different receivers: -- go into oscillation with only 0.068 V on its "screen" in my 1.7-MHz detector circuit -- already be oscillating with O V on its "screen" (that is, with the "screen" connected directly to ground) in my 455-kHz detector circuit This is a fascinating finding. In a true tetrode or pentode Hartley oscillator/detector in which the cathode is above ground and connected to the tuned-circuit tap, the screen grid, usually connected to the wiper arm of the circuit's REGENERATION control, is grounded for ac and acts as the oscillator anode. To act as the oscillator anode, the screen grid must be more dc-positive than the cathode; connecting it directly to ground (turning the REGENERATION control "all the way down" [assuming that the "lower" end of the control stator is also connected directly to ground]) is guaranteed to stop oscillation. Although the performance of the 13DE7 was more than good enough for communication use, I sought to replace it with with a more conventional dual-triode tube for two reasons: -- its heater dissipated considerable power (over 5.5 W), (especially in my BG-3 receiver) resulting in considerable additional component heating as a result of the receiver's relatively small enclosure. -- I wanted a REGENERATION control capable of adjustment well below, through, and well above oscillation. Before wiring out the 13DE7 and acting on a hunch, I made one new measurement: After short-circuiting the Hartley tuned circuit ("top" to ground, leaving the grid grid-leak biased) so the detector could not oscillate, I measured the voltage at the interconnection between the upper-triode cathode and the lower-triode plate with the "screen" grounded as 13.54 V. Here, then, was why the 13DE7 cascoded-triode tetrode oscillated (in my modified Allied A-2516's 455-kHz detector) or was so close to oscillation (in my 1.7-MHz BG-3 detector), with the electrode I thought was the oscillator anode connected to 0 V/ground: The actual anode of the detector/oscillator--the junction of the upper-triode cathode and the lower-triode plate--was still at 13.54 V with the "screen" at 0 V! (I'll refer to the cascoded-triodes tetrode's upper-triode-cathode-lower-triode-plate interconnection as the circuit's virtual anode for the rest of this post.) So I rewired the tube socket to take a conventional, 9A-based identical dual triode: A 5814A (premium 12AU7A) or a 5963 (a premium 12AU7like design optimized to tolerate long periods of plate-current cutoff in computer use), both of which I already knew would work well in this detector circuit. Both of these tubes' heaters draw about 2 W--less than half of that drawn by the 13DE7. With a 5963 in place, further measurements gave a more complete picture of the decoupling between the voltage at synthetic tetrode Hartley detector's virtual anode and the voltage at its "screen" (upper-triode grid): - With the "screen" at 0 V/ground, the virtual anode voltage was 8.84. - With "screen" voltage adjusted for critical regeneration (1.4 V), the virtual anode voltage was 9.64. - With the REGENERATION control asjusted to the level I commonly use for communication, the "screen" voltage was 4.3 and virtual anode voltage was 11.1. - With REGENERATION control at maximum, the "screen" voltage was 23.5 and the virtual anode voltage was 23.0. All measurements were made with the Hartley tuned circuit shorted ("top" to ground) such that the lower-tube was not oscillating but still grid-leak biased. One pleasant aspect of the decoupling between the "screen" control voltage and the virtual anode voltage that the measurements don't communicate is that regeneration control with the synthetic tetrode is very smooth, with (at least in my detectors) little frequency shift. In effect, the decoupling between the synthetic-tetrode "screen" and the detector virtual anode makes the REGENERATION control act as if I'm adjusting it through a reduction drive. (There is no detectable hysteresis at critical regeneration, BTW. That is, oscillation onset as I turn REGENERATION up and oscillation stoppage as I turn REGENERATION down both occur at the same control setting.) With a true screen-grid tube--a 12SJ7 pentode or (as I tried more recently, a 6417 [12-V 5973] beam-power tube)--the transition through these regions is much sharper (occurs across far fewer degrees of travel of the REGENERATION control). Some speculations and further ideas for future experimentation occur to me as a result of these early experiences with the cascode-triodes: 0. Measuring the "screen" current versus and virtual-anode voltage across the REGENERATION control range could well be instructional. 1. The "screen"/virtual-anode decoupling looks to be straightforwardly explainable, at least in part, by mere voltage divison across the cascoded tubes. (If both tubes are biased such that their internal resistances are equal, we would expect the voltage at the virtual anode to be half of that at the upper-triode plate--neglecting contact-potential effects, that is. The 0.5-V differential between the "screen" and virtual-anode voltages at maximum REGENERATION is on the order of contact potential difference.) 2. Many variations on the cascoded-triodes idea are possible. More than two tubes can be cascoded; and the tubes cascoded need not be triodes. How might variable-mu v sharp-cutoff cascode components compare in a given application? 3. The relative cumbersomeness of their directly-heated cathodes notwithstanding, it would have been straightforward to construct a synthetic tetrode from available triodes (UV201, UX201A, and so on) long before true screen-grid tubes were available. Such an invention might have significantly affected the history of a class or three of radio-technology development...or not. :-) So now I'm wondering if anyone did so. 4. I look forward to experimenting with cascoded-triodes tetrodes as crystal oscillators in the tri-tet and grid-plate configurations, among others. Will the buffering between the cathode/lower-grid/virtual-anode triode and the upper-triode plate be better or worse than, or the same as, that exhibited by a true tetrode or pentode in the same circuit? 5. The exposure of the upper-triode-cathode/lower-triode-plate interconnection allows us access to the synthetic tetrode's electron stream in a way that's impossible with an integral true tetrode or pentode. What might we gain or lose in a given circuit by externally dc biasing that point and/or by adding band-limiting (filtering) at that point? 6. Experimenters who synthesize dual-gate FETs with cascoded JFETs sometimes run into stability problems not present with dual-gate parts. The possible contributions of lead inductance and capacitance in the discrete parts aside, might this problem be caused by, or related to, the solid-state analog of the "screen"/virtual-anode decoupling I report for the vacuum-tube synthetic tetrode? Best regards, Dave amateur radio W9VES

As I've posted earlier, the cascoded-triodes regenerative detector is a actually synthetic tetrode--a screen-grid tube synthesized by stacking two triodes between ground and B+. I've also posted that experiments have shown that using a cascoded-triodes synthetic tetrode as a pin-for-pin replacement for a 12SJ7 pentode detector already optimized to resist pulling on CW signals received at low beat notes gives performance equal to or better than the 12SJ7.

One aspect that I did not investigate further until now involves my unusual initial finding that with the first dual-triode tube I used--a 13DE7 dissimilar dual triode wired with its higher-mu section closer to ground (connected to the Hartley tuned circuit) and its lower-mu section closer to B+ (connected to the wiper arm of the REGENERATION control such that its grid acted as the tetrode screen grid)--would, in conjunction the existing Hartley-tapped tuned circuits ( http://mysite.verizon.net/dpnewkirk/ej/goodman_receiver/#12SJ7_detector ) in two different receivers:

-- go into oscillation with only 0.068 V on its "screen" in my 1.7-MHz detector circuit

-- already be oscillating with *O V* on its "screen" (that is, with the "screen" connected directly to ground) in my 455-kHz detector circuit

This is a fascinating finding. In a true tetrode or pentode Hartley oscillator/detector in which the cathode is above ground and connected to the tuned-circuit tap, the screen grid, usually connected to the wiper arm of the circuit's REGENERATION control, is grounded for ac and acts as the oscillator anode. To act as the oscillator anode, the screen grid *must* be more dc-positive than the cathode; connecting it directly to ground (turning the REGENERATION control "all the way down" [assuming that the "lower" end of the control stator is also connected directly to ground]) is guaranteed to stop oscillation.

Although the performance of the 13DE7 was more than good enough for communication use, I sought to replace it with with a more conventional dual-triode tube for two reasons:

-- its heater dissipated considerable power (over 5.5 W), (especially in my BG-3 receiver) resulting in considerable additional component heating as a result of the receiver's relatively small enclosure.

-- I wanted a REGENERATION control capable of adjustment well below, through, and well above oscillation.

Before wiring out the 13DE7 and acting on a hunch, I made one new measurement: After short-circuiting the Hartley tuned circuit ("top" to ground, leaving the grid grid-leak biased) so the detector could not oscillate, I measured the voltage at the interconnection between the upper-triode cathode and the lower-triode plate with the "screen" grounded as *13.54 V.*

Here, then, was why the 13DE7 cascoded-triode tetrode oscillated (in my modified Allied A-2516's 455-kHz detector) or was so close to oscillation (in my 1.7-MHz BG-3 detector), with the electrode I *thought* was the oscillator anode connected to 0 V/ground: The *actual* anode of the detector/oscillator--the junction of the upper-triode cathode and the lower-triode plate--was still at 13.54 V with the "screen" at 0 V! (I'll refer to the cascoded-triodes tetrode's upper-triode-cathode-lower-triode-plate interconnection as the circuit's *virtual anode* for the rest of this post.)

So I rewired the tube socket to take a conventional, 9A-based identical dual triode: A 5814A (premium 12AU7A) or a 5963 (a premium 12AU7like design optimized to tolerate long periods of plate-current cutoff in computer use), both of which I already knew would work well in this detector circuit. Both of these tubes' heaters draw about 2 W--less than half of that drawn by the 13DE7.

With a 5963 in place, further measurements gave a more complete picture of the decoupling between the voltage at synthetic tetrode Hartley detector's virtual anode and the voltage at its "screen" (upper-triode grid):

- With the "screen" at 0 V/ground, the virtual anode voltage was 8.84.
- With "screen" voltage adjusted for critical regeneration (1.4 V), the virtual anode voltage was 9.64.
- With the REGENERATION control asjusted to the level I commonly use for communication, the "screen" voltage was 4.3 and virtual anode voltage was 11.1.
- With REGENERATION control at maximum, the "screen" voltage was 23.5 and the virtual anode voltage was 23.0.

All measurements were made with the Hartley tuned circuit shorted ("top" to ground) such that the lower-tube was not oscillating but still grid-leak biased.

One pleasant aspect of the decoupling between the "screen" control voltage and the virtual anode voltage that the measurements don't communicate is that regeneration control with the synthetic tetrode is very smooth, with (at least in my detectors) little frequency shift. In effect, the decoupling between the synthetic-tetrode "screen" and the detector virtual anode makes the REGENERATION control act as if I'm adjusting it through a reduction drive. (There is no detectable hysteresis at critical regeneration, BTW. That is, oscillation onset as I turn REGENERATION up and oscillation stoppage as I turn REGENERATION down both occur at the same control setting.) With a true screen-grid tube--a 12SJ7 pentode or (as I tried more recently, a 6417 [12-V 5973] beam-power tube)--the transition through these regions is much sharper (occurs across far fewer degrees of travel of the REGENERATION control).

Some speculations and further ideas for future experimentation occur to me as a result of these early experiences with the cascode-triodes:

0. Measuring the "screen" *current* versus and virtual-anode voltage across the REGENERATION control range could well be instructional.

1. The "screen"/virtual-anode decoupling looks to be straightforwardly explainable, at least in part, by mere voltage divison across the cascoded tubes. (If both tubes are biased such that their internal resistances are equal, we would expect the voltage at the virtual anode to be half of that at the upper-triode plate--neglecting contact-potential effects, that is. The 0.5-V differential between the "screen" and virtual-anode voltages at maximum REGENERATION is on the order of contact potential difference.)

2. Many variations on the cascoded-triodes idea are possible. More than two tubes can be cascoded; and the tubes cascoded need not be triodes. How might variable-mu v sharp-cutoff cascode components compare in a given application?

3. The relative cumbersomeness of their directly-heated cathodes notwithstanding, it would have been straightforward to construct a synthetic tetrode from available triodes (UV201, UX201A, and so on) long before true screen-grid tubes were available. Such an invention might have significantly affected the history of a class or three of radio-technology development...or not. :-) So now I'm wondering if anyone did so.

4. I look forward to experimenting with cascoded-triodes tetrodes as crystal oscillators in the tri-tet and grid-plate configurations, among others. Will the buffering between the cathode/lower-grid/virtual-anode triode and the upper-triode plate be better or worse than, or the same as, that exhibited by a true tetrode or pentode in the same circuit?

5. The exposure of the upper-triode-cathode/lower-triode-plate interconnection allows us access to the synthetic tetrode's electron stream in a way that's impossible with an integral true tetrode or pentode. What might we gain or lose in a given circuit by externally dc biasing that point and/or by adding band-limiting (filtering) at that point?

6. Experimenters who synthesize dual-gate FETs with cascoded JFETs sometimes run into stability problems not present with dual-gate parts. The possible contributions of lead inductance and capacitance in the discrete parts aside, might this problem be caused by, or related to, the solid-state analog of the "screen"/virtual-anode decoupling I report for the vacuum-tube synthetic tetrode?

Best regards,

Dave
amateur radio W9VES