Ben Tongue on VHF and UHF design

Fellow radiophiles,

Ben Tongue has sent me some very interesting comments and perspectives on his approach to the design of his pioneering broadband VHF low noise amplifiers. The following italicized blue text is quoted with Ben's permission, from an email dated April 26th 2010, that he sent me.

Regards,

-Joe

---

JS- [...]Now that I have your ear, there is a question about tube noise that I have been dying to find out about. I know that partition noise is a big problem with Tetrodes and Pentodes. But several Tetrodes and even some Pentodes were designed for very low noise use as VHF preamps. Apparently, you always favored triodes for noise figure.

At the end of the tube era, the triode cascode became the favored front end for FM radios. It had the higher impedance of a pentode and the noise figure of a triode. I am stating this quite loosely. I wonder to what extent a pentode could hide it's noise figure simply by being able to be driven by a higher stepped up impedance. Then there is the resistive loading of the space charge transit time. I just have a whole bunch of hear-say that has never been crystallized into a coherent understanding of the real tubes available in the 1950's.

I would love to hear your experience on triode vs pentode tube noise.
 

BT-[...] re RF tube noise: A pentode always has partition noise so I always ruled it out when looking for a "best" low noise figure solution for a TV frequency front-end. That leaves us with triodes, their main problem being transit-time resistive loading from grid-to-ground. One solution here is to select tubes with smaller grid-cathode spacing and frame grid construction.

The big deal with the cascode connection, as I see it, is the ability to get pentode-like input and output impedances and high gain without the problems of neutralizing a high gain stage. If neutralization of the first cascode stage is found necessary, it normally has a voltage gain of only unity, making neutralization non-critical. Another benefit of the cascode connection (which I often used) is the ability to increase gain by adding some mild impedance transformation between the two stages.

At high frequencies noise performance deteriorates as frequency increases, mainly because the shunt resistive component due to cathode-grid transit-time effect drops sharply. I think it goes down inversely proportional to frequency squared. Its equivalent noise temperature is cathode temperature, so it's a noisy beast. In setting impedance matching for best noise figure, one always finds that driving the tube grid from a transformed source transformed to a value lower than a perfect match improves noise results. This mismatch tweak tends to short out the transit-time noise voltage appearing on the grid to a greater extent than the reduction of signal voltage on the grid.
[...]

Another excerpt from Ben's email shares interesting perspectives on the design of UHF diode mixers:

BT-[...]I wonder if you might have access to a General Electric App. Note written by Erich Gottlieb around 1950. The subject was how to use the G.E. 1N72 germanium diode as a mixer for TV UHF frequencies, as I recall. GE had determined that the mixer diode performance optimized when the local oscillator exitation produced a rectified DC current of 1 mA. Erich suggested DC biasing the diode in the forward direction from a 0.25V source having an internal resistance of 250 ohms. Now, if the oscillator exitation was not changed, the quiescent bias would be zero volts and performance would not be changed. In a real world UHF converter the rectified current would usually change as a function of tuning. This degraded performance because the RF and IF resistance of the diode would change, upsetting impedance matching. The 0.25V, 250 ohm DC bias kept the RF and IF resistances of the diode approximately fixed even when the local oscillator power level changed, when it was tuned. This especially improved performance at the high end of the UHF band when local oscillator strength dropped off.If you can get a copy of the G.E. App. Note, I would appreciate receiving a copy.

I never found this App note, perhaps an RMORG member has this note. Some of my response to Ben:

JS-[...] Since our last email, I have looked up more info on partition noise. I attached a 1948 paper by William Harris in the RCA Review with a survey of equivalent input noise resistance and input conductance at high frequencies for various tube types. The product of these two is pointed out as a figure of merit for the tube noise.

One interesting tidbit from this paper is that the 6AK5 at 100MHz, as a pentode has Req=1.9k and gin=125uS=1/8KΩ, while, as a triode it has Req=0.38k and gin=125uS=1/8KΩ

Partition noise quadruples the noise resistance, which is to say, doubles the noise voltage.

I find your remark about driving the triode grid with more conductance than gin to shunt out the transition noise very interesting. It kills more noise than signal.

You also touch on noise temperature of the space charge. I was amazed to realize that the averaging effect of the space charge of the unruly thermionic cathode emission to be equivalent to a cooling of the space charge temperature. Your remark about the space charge temperature looking like cathode temperature helps to explain the space charge temperature concept.

On a related thought about noise, I was thinking about how the noise of each of the various stages of your double-hump VHF-HI&LO broadband preamps work. Each stage has it's complementary pair of humps for the HI and LO bands. I have to assume that the off-band stages either have enough gain outside their respective band to keep from adding noise until the appropriate stage is reached, or that outside each of the bands, the stages turns into a quiet pass-through. Perhaps a combination of both happens.

With this in mind, I would assume that the stage involved in the highest frequencies would be at the input.

You confirm my understanding of the merits of the cascoded front end. I still find it amazing that so many pentodes/tetrodes were developed for VHF front end use, like the dual tetrode 6C9 or 17C9. Perhaps these tetrodes were worth the noise loss because of the higher possible gain.

About a year ago, I read a fair amount about diode converters and how good their conversion efficiency could be. If surrounded by the right kind of electric fly-wheel inertia, they could give conversion efficiencies of 90%. While this explains why they were used for UHF converters, it makes one wonder why they were not used at FM or VHF.
I can't imagine that would be more likely to radiate than a carefully balanced triode self-oscillating mixer front end.[...]

BT-[...]re the zero capacitance probe, I think my problems were cause by using a tube with its large associated capacitances (FETS didn't exist at the time). If I had loaded down the cathode-to ground with a lower resistance, I probably could have stopped the oscillation. Of course then, that would cause the cathode-follower gain to depart further from unity, reducing the capacitance reducing benefit. What does 'PFB path' mean?

re the 1945 Article by Herold on Mixer Diode conversion loss: It is intuitive that if the diode sees an impedance having a resistive component at the image frequency, that the conversion noise performance will suffer (the LO will convert thermal noise in the resistive component to IF). I found, in my design work on Blonder-Tongue UHF converters, that it was desirable to have the mixer diode see as high an impedance at the sum conversion frequency as possible (RF + LO). This minimized conversion loss. How? This current at the RF + LO frequency, produced by mixer action with the RF signal, flows through diode diode series junction resistance and dissipates power (power that came from the input signal). Dissipating some power that comes from the input signal leaves less power for output. Conversion loss is increased and noise factor is degraded.

The early preamps using the double-hump VHF-HI&LO setup had 3 peaks in the low band and three in the high.. Let us number them 1-6, starting at 54 MHz and ending around 216 MHz. Let us assume my first design, the HA1-L which had 4 cascaded tubes and therefor 3 double hump interstages. The first interstage had humps near 54 and 216 MHz. The second interstage had humps near 88 and 174 MHz. The third had humps near the center of the low and hight VHF bands. The improvement introduced on the CA1-M used a four hump quadruple tuned circuit for the first interstage with humps near 54, 88, 174 and 216 MHz. The second interstage had two humps somewhat above 54 MHz and somewhat below 216 MHz. Your guess there the third interstage humps were located would be correct.

I think that general cascode use in TV tuners waited for the development of the frame grid tube, with its close grid-cathode spacing, to improve Req. The introduction of turret tuning also helped with the control of parsitics and stability. Before the introduction of cascode tuners, and aside from RCA's 6J6 tuners, the most popular tuner architecture was switch-tuned with a pentode RF stage in an odd configuration, as I recall. A bifilar-wound center-tapped ferrite-cored inductor was connected with the tap at ground, the grid connected to one end and the cathode connected to the other end. Voila, a quasi-balanced input impedance for the twin-lead antenna wire.

When you mention "the right kind of flywheel inertia", I think you are referring to the diode seeing an impedance pole at the important higher order conversion frequencies as well as at the sum frequency conversion (in a converter having the IF at the difference of LO and RF).

I have added more B-T product info in Section E of my Site. I hope it is interesting.

[...] re yesterday's Email, the quadruple-tuned four hump circuit may be found in patent #2,710,315 listed in Section E of my Site. You may also be interested in the schematic for the 'Tunnel Diode' UHF converter (BTD-44) in the link in the Table entitled 'Some early Blonder-Tongue product literature' in Section E.

Best regards,

Ben

 

To thank the Author because you find the post helpful or well done.