Triode Frequency Conversion at FM-VHF
I am looking for good references on triode frequency conversion used in 1950's FM radios.
Some battery radios used the DC96, later some used used the DF97 pentode in triode connection. Grundig's line of 3-tube AM-FM AC-powered radios used the ECC85 dual triode with heater cathodes as a self-oscillating converter with the other triode as a grounded grid preamp.
The two battery tube types were used at the front end as self oscillating mixers, without preamplification, because RF gain at 100MHz was poor with these filamentary cathode tubes.
I have downloaded and read the english language Philips data sheet for the DF97 with extensive curves showing performance in self oscillating mixer configuration. It does not mention what frequencies are involved, just the overall conversion transconductance, which tops out around 300uS.
I have also downloaded the excellent german language article on the DF97 RM page. But one year of German I learned 25 years ago in College is not enough to get through the article.
Perhaps one of the German language members cold do German OCR on the scanned text, and post the text, so that Google-Tranlate could be used for a crude translation to any language.
The DC transconductance of the battery tubes is around 1mS, and for the AC triodes, around 10mS.
There are a few aspects about the operation of triode mixers or self-oscillating converters that I already understand, but I would like to know more.
The folowing circuit came from a Finish AM/FM Lucia radio made by Helvar in 1959. I chose this example because the schematic is particularly clear.
One notion that I think I grasped, is that the triode mixer or self oscillating converter always has an IF 10.7 MHz transformer load that looks like a very low impedance at 100MHz. So we have essentially a grounded plate for the incomming RF signal at 100MHz. This is ideal to get the maximum possible HF gain from the triode. The tube is run as a transconductor at RF: high Zin to low Zout.
At the LO frequency, the grid-to-plate transformer forms a classic Armstrong oscillator.
The RF and LO frequencies are added at the Grid. This explains the "additive mixing" name of this process. But the large LO swing makes the "addition" very non-linear, so that it results in multiplication of the LO and RF currents at the low impedance (grounded) plate.
At the IF frequency, the load tank resonates and the impedance becomes very high. I am going to guess on the order of several kOhms, given that the conversion transconductane is on the order of 300uS. A 10KOhm IF tank circuit resistance at resonance would give an overall 10dB of conversion voltage gain.
Lastly a capacitor from the other end of the IF load tank provides neutralization at the IF frequency, So that operation free from spurious IF oscillations is obtained.
I also know that triodes can be constructed to have faster transit times than other tube types. And the DF97 probably has a faster transit time as a triode, than as a pentode.
The two key aspects that I am still curious about are about the noise figure and the switching time during the mixing process.
The ECC85 is fast enough for use as a grounded grid amp. Could the noise figure have been improved with a grounded grid preamp using the DC96 or DF97?
Was the transit time of the DC96/DF97 so marginal at 100MHz that they could not help as preamps?
The low plate load impedance of the IF transformer at RF frequencies in the self oscillating mixer is much less demanding than the higher impedance present in a preamp, even with a grounded grid.
Another thought that comes to mind as a way to improve noise figure and conversion transconductance for the DC96 and DF97 self oscillating converters, would be the use of a pair of DC96, or a pair of DF 97 in push-pull.
The concept is as follows: The input grids are driven together with RF in phase, but with opposite phases of the LO frequency. The cathodes are grounded as usual. But the plates split to opposited ends of the center tapped IF transformer. The transformer is resonated with two caps from the tap to the ends. The center tap of the IF transformer then goes to B+. Two coupled feedback windings in series wiht the plates sample the LO signal at the plate differentially, and feed it back to both grids.
If this works, it should double the conversion transconductance, and improve the noise figure.
In other circumstances, where the switching or mixing device is much faster than the frequencies involved, there is little advantage for full wave mixing or detection over half wave. Proper impedance matching can make the half wave and full wave circuits have equal (nearly) conversion or detection efficiencies.
OK, another thought about frequency conversion at the RF limit of the conversion tube is the double conversion circuit used in the Meissner 8C FM tuner. Here the concept is to cut the LO frequency in half so that the relatively slow mixing tube spends less time in the switching process, and more time amplifying signal. For example, if the mixing tube can switch in 3ns, then only about 2ns of effective amplification time are available with a 10ns period at 100MHz. That is 20% duty cycle. But the same tube mixing at 50MHz would have 7ns of effective mixing time. That is an improved 35% duty cycle.
In the Meissner, the same LO converts the first variable 50MHz IF to a fixed 10.7MHz IF.
The push-pull idea and the Meissner half frequency LO method contribute some gain, but more importantly they contribute an improved noise figure, while using tubes that are barelly fast enough for FM-VHF work.
the quality of the text is too bad to make an ocr with limited errors. I would like to translate part for part. Perhaps some other people will help with some other parts so we come to an result in a short time.
I sent the first part of a english word file to you. If someone would help in doing tranlation of an other part, I send the file. Even now, the file is too big to publish it here.
( about 1,3 Mb )
Thank you very much for your translation work. I would also like to recognize publicly, the diligent efforts of Herr Knoll and Prof Dr Rudolph, who also sent me materials on triode mixing.
I have a special question for Herr Knoll, that may be of interest to the wider forum audience:
Earlier in this thread I brought up the notion of push-pull, or full wave mixing as a way to improve noise figure in mixers, along with a modest gain improvement. Recently, I saw a paper by E. W. Herold from 1946 about "Superheterodyne Frequency Conversion Using Phase-Reversal Modulation".
I have a copy of the paper I can share privately if you send me an email.
This indicates to me that push-pull frequency conversion was considered seriously. I would expect that it would be most advantageous with battery tubes that are barelly fast enough for 100MHz work. Perhaps push-pull mixing with two mixind diodes would have been advantageous to improve the noise figure of TV-UHF tuners.
The question for Herr Knoll is: Was this type of push-pull conversion considered seriously when you were designing tuner front ends?
(please feel free to respond in German, I can use google-translate)
I have been reading again, the famous thread analysing the Grundig 5040W/3D
Two additional design concepts have emerged from this reading:
The bridge concept was a frequently used design technique to correct many parasitic reactances, such as the classic Miller plate-to-grid capacitance, plate-to-cathode capacitance in a grounded grid stage and minimization of oscillator radiation to the antenna in front ends with self-oscillating mixers.
The other concept is that, even with fast VHF/UKW tubes, such as the ECC85, ECC81, or EC90, transit time effects may affect the the stability of classic Miller neutralization schemes. Transit time delays cause additional phase shift in the positive capacitance feedback path, and it may lead to instability.
Side comment on bridge neutralization: I design monolithic Analog-to-Digital converters at Linear Technolofy for a living. A central part of my work is very high speed sampling circuits, which can be thought of as mixers. Much of this circutiry is "open loop". Some of the circuit techniques I have devised to reduce distortion and improve accuracy remind me of the neutralization techniques I have been reading about.
Additional information can be found in:
Fink, D.G.: Television Engieering, 2nd. ed. McGraw-Hill, 1952, pp.650 -660
Thanks for the reference. It appears that the 6J6 was the hot tube of the day.
A while back, I tested step response of varioust tubes,and got these results.
I used this circuit
and got this step response:
The rise and fall times at less than 1ns suggest that this tube would work well abobe 100MHz. Even if the signal goes through two stages, as it does here, wiht a follower driving a grounded grid stage.
The 6J6 was also use extensivelly in external VHF signal boosters.
The differential symetry that the 6J6 makes possible is put to good use in this design. The neutralization appears quite straight forward, with simple capacitive cross-coupling from plates to grids.
It appears that the coupling from the LO is done to one input of the diff pair mixer, while the RF is coupled to the other input capacitivelly with C14-0.68pF. The grid pairs of each tube are connected with the switchable inductors in all three stages.
The 1.4pF cross-couple neutralization caps shown in the RF stage closely match the specified Cga=1.5pF.
The antenna network includes an series resonant IF trap, and an impedance matching network. If the values were known, they might show that the impedance was stepped up into the neutralized RF stage, for best noie figure.
Curious to see that the mixer does not appear to be neutralized with cross couple feedback caps. C21 is the output IF coupling cap. It appears that neutralization fell to L6+C18 series resonant pair, grounding out IF frequencies at the common point of the grid coils.
C24 and C25 are wired in the LO like cross-coupled neutralization caps, but they are 3 times larger than the Miller capacitance, so their purpose is positive feedback for oscillation. Interesting to note that Cga would probably also have produced oscillations, but Cga is much less stable than C24 and C25, and the LO must have very stable frequency.
Comments, and corrections invited,
Hello Mr. Sousa.
See here my answer in german language.
regrads Hans M. Knoll
Dear Herr Knoll,
Thank you very much for the excellent survey of FM front end designs in 1950's Germany, and for the writing in German and English!
I will add to your survey, the list of RM resources that I have been using to study front end frequency conversion. Much of this material was contributed by you:
grundig_5040w3d_circuitry_analysis_part_2.html (translated to English by Thomas Albrecht)
Die neue Technik in der UKW_BOX ab 1957_V2(0).pdf (printing disabled on this pdf, but google-translate works)
ukw_einbauteil_type_u2w_pendler.html (the German term "pendler" translates to "commuter" but it refers to the quenching action of super-regenerative detection)
Grundig_Pendler-Funktionsbeschr.pdf (printable, but google-translate failed)
I will stop the list here, but it could be continued.
Now, mixer gain.
Basic mixer theory for single phase mixing limits conversion gain between 1/pi=1/3 and 1/4 of the available DC transconductance (gm) of the tube.
A good explanation of this result is given by
E. W. HEROLD in
"A NEW TUBE FOR USE IN SUPERHETERODYNE FREQUENCY CONVERSION SYSTEMS" published in Proceedings of the IRE, February 1936. (I have a copy of this and other papers I can share privately via email k2w at philbrickarchive dot org)
For example, a DF97 wired as a triode, has a DC transconductance (translates as "steepness" from German) of 1mS, but the conversion transconductance, when used as a mixer, is only approximatelly 0.3mS, as can be confirmed in the published DF97 data sheet. The simple explanation is that the tube is effectivelly off for 2/3 of the time, and effectivelly on for 1/3 of the time, as the oscillations go positive and negative.
As Mr, Knoll confirmed, a push-pull mixer has a higher conversion gain. The simple explanation, is that, while one tube goes off, the other turns on, so the conversion gain is doubled. Now the theoretical maximum conversion gain is 2/3 the value for DC gm, or 0.6mS for the DF97 example.
A continuation of this line of thought would lead to the case where a 3 phase clock could drive three mixers in parallel for Conversion gain that is nearly the same as the DC transconductance. Using three tubes as mixers would be excessive in most situations, but using 3 germanium diodes would be reasonalble.
OK, now another question arrises:
Why go through all the trouble of a circuit with 2 or 3 mixers, just to get a little more gain?
The push-pull 2-mixer case has 6dB more gain than the single mixer, and the 3-mixer case has 10dB more gain than the single mixer. A single stage of IF could have provided more than the desired gain much more simply.
Another way to get the added gain is with a RF preamp, which as nearly universal in all AC-powered table models.
One reason for push-pull mixing was described by Mr Knoll, and shown in figure 4 above, for the double-balanced mixer Roadmaster 1961.
The reason that I am most interest in, is hiss-noise reduction at the receiver input.
One of the principal characteristics of front end design that sets the receiver hiss, is the transconductance of the input tube, as seen reflected to the antenna inputs, after some transformer or LC impedance trasformation. This is the reason why nearly all FM table radios employ a high transconductance tube in the RF preamp stage. The gain of this tube is not being chopped by oscillator action as is the case in the mixer. So full gain is available. Additionally, this tube should have uncompromised AC gain at VHF-FM. All tubes start to loose gain, if the frequency is high enough. Some tubes loose most of their gain before VHF-FM is reached, so a tube with high gm and high bandwidth is always seleted for the FM front end. Good examples are the ECC81/12AT7 and the ECC85/6AQ8. In general, triodes and especially made tetrodes can be made with higher bandwidth than pentodes.
Mixers always have worse gain and worse hiss-noise than amplifiers using the same tubes, but this problem is easily solved with the RF preamp tube in front of the mixer. The RF preamp the hiss-noise performance of the set.
OK, so why am I still interested in improving conversion gain/hiss-noise?
My interest is for two special cases: FM portable tube radios, and early UHF-TV front ends.
No FM portable tube radios and no early UHF-TV tuners have been found with an RF preamp tube.
In the case of FM portable tube radios is that there is not enough available transconductance (gain) at 100MHz in battery tubes like the DF97, or the 5678 to justify adding a preamp to reduce hiss in portable FM radios.
In the case of early UHF-TV receivers, there were no UHF tubes that were cheap enough to use as the preamp of a UHF tuner.
In both cases, the only available avenue for further noise reduction, which was particularly bad at UHF, was to use more efficient methods of mixing employing 2 or 3 oscillator phases, to improve conversion gain.
In microwave work, one common type of mixer is the Ring mixer. The Ring mixer is a form of 2 phase mixer employing 4 diodes, that also has good port isolation between RF, LO and IF. Perhaps RM member, Mr. Andreas Steinmetz, could contribute from his experience as a Microwave Engineer. (Mr. Steinmetz contributed significantly in the analysis of the Grundig 5040D/3D )
Thank you all, especially Mr. Knoll, for contributing so richly to this thread.
Comments/corrections invited, Regards.
Additional information concerning VHF-TV amplifying and mixing, found in "Noll, E.M.: Television for Radiomen, McMillan, 1955".
47. Grounded‑Grid and Cathode‑Coupled Amplifiers
Two tubes suitable for wide television application are the miniature triodes 6J4 and the dual triodes 6J6. The 6J4 miniature triode is a grounded-grid amplifier, and the 6J6 consists of two high‑gm triodes. It is an admitted fact that the advantage of the pentode in the amplification of i‑f and r‑f frequencies is its high plate impedance, which means it will amplify a small signal voltage to a much greater extent. However, in wide‑band amplification of i‑f and r‑f frequencies, the plate load impedance itself must be lowered in order to pass the wide band of frequency; consequently, the advantages of the pentode have been nullified. Thus, it is possible in wide‑band service to use a triode with a high gm to give us approximately the same gain as a normal pentode would. An added advantage of using a triode in this type of service is that its inherent noise is much lower because of the absence of additional grids. Another advantage of the triode in this service is its inherent lower impedance, which means that less severe external loading is necessary to cover a band of frequencies. In fact, the triode amplifier is linear over a substantial band of frequencies without any external loading at all. This is particularly the case when a cathode‑coupled arrangement is used.
A typical grounded‑grid r‑f amplifier is shown in Fig. 48. Note that the grid is grounded, that the cathode and the plate are above ground, and that the stage uses a tuned input circuit as well as a tuned plate circuit. Normally a tuned‑grid tuned‑plate stage would oscillate; however, in a grounded‑grid stage the grid is grounded and acts as a shield between input and output circuits.
Any feedback due to plate‑cathode capacitance is not in proper phase to produce oscillation. The advantages of the grounded‑grid amplifier are as follows:
1. Low impedance and an inherently broader bandpass characteristic.
2. Low electrode capacities, permitting high L‑to‑C ratio and greater gain tuned circuits.
3. Low tube noises and a better signal‑to‑noise ratio.
The cathode‑coupled amplifier (Fig. 49) uses the two sections of the 6J6 dual triode‑first section is a cathode follower; next section, a grounded‑grid amplifier. The gain of this stage, although it consists of two triodes, is equivalent to the gain of a good single‑pentode stage. The number of component parts are approximately the same; however, the cathode‑coupled stage has the following advantages:
1. Wide bandpass characteristics because of the inherently low impedance triode.
2. Low noise characteristics, because internal noise is low in a triode.
3. Low input and output capacities and consequent high L‑to‑C tuned circuits.
4. Grounded‑grid connection to minimize tendency to oscillate. The coupling between the two triodes is due to common impedance of the cathode inductor, which has a substantial reactance at the frequencies to be passed.
Another application for the cathode‑coupled stage is shown in Fig. 50. In this circuit, which shows a grounded‑grid r‑f amplifier and a cathode‑coupled stage following it, the cathode-coupled stage acts as a mixer. The local oscillator signal is injected by connecting a portion of the oscillator coil between the grid and ground of the second section of cathode‑coupled stage. Normally the grid is tied directly to the ground.
These special miniature tubes perform well at frequencies up to 500 megacycles and higher. Consequently they will see wide application in television relay and high-frequency transmission. At these frequencies, and for that matter at lower frequencies too, it is well to utilize a push‑pull arrangement (Fig. 51) to lower effective capacity and form a balanced high‑frequency circuit.
Many tuned circuits used in high‑frequency television circuits are so‑called linear tank circuits made of an effective section of a transmission line. For example, a quarter wavelength of transmission line shorted at the end will act as a parallel resonant circuit and can be attached as such to a vacuum tube circuit at its open end. One advantage of this type of a tuned circuit (Fig. 51) is that it can be made to resonate at the desired frequency with the distributed circuit capacity forming a high Q, high L‑to‑C ratio circuit. It can be designed to be a low‑reactance tuned circuit and minimize the effects of lead inductance and circuit capacity. It also lends itself to switching systems as its resonant frequency can be changed readily by simply moving the position of the short at end of line (shorter the effective length of line between tube and short, the higher the resonant frequency). It is not only adaptable to push‑pull circuits but is equally effective as a tank circuit for single‑ended stages, replacing the usual form of coil and capacitor. Linear tuned circuit is not always in the form of straight rods but can be in form of a loop with a moving or switched shorting bar or made of small coils added incrementally to change frequency to some lower value. Concentric tank circuits are also adaptable.
48. Receiver R‑F Sections
With the advent of miniature tubes, the r‑f section of the television receiver, consisting of r‑f amplifier, mixer, and oscillator, has become a compact, efficient assembly. The entire r‑f section of the new receivers is mounted on a highly stable turret assembly or multisection selector switch which serves as a bandswitching unit for the 12 television channels. All circuit components are a part of the assembly except signal, heater, and supply voltage lines. Needless to say, the small size of the miniature tubes makes the unit all the more compact. On this unit are generally three switch wafers or tuned sections which switch the r‑f amplifier, mixer, and oscillator inductors channel by channel. Other systems employ continuous tuning, permeability tuning, or printed circuit technique.
Fig. 52 RCA Selector Tuner (RCA 630TS TV set)
The r‑f selector tuner of an RCA model is shown in Fig. 52; [a Farnsworth version, in Fig. 53.] Schematic diagrams of same units are given in Figs. 54 [and 55]. The RCA unit is a push‑pull r‑f amplifier, mixer, and oscillator arrangement using transmission‑line sections as parallel resonant circuits. [...]
This schematic shows additional Link Coupling from RF and Osc to MIX. This is not so clear to be seen from the schematic in post 5.
Comments/corrections invited, Regards,
Mr. Sousa. Your Quenstion above was:
#My interest is for two special cases: FM portable tube radios, and early UHF-TV front ends.
No FM portable tube radios and no early UHF-TV tuners have been found with an RF preamp tube.
Hello Mr. Sousa.
I am agree with your solution, the Tubes at this time are not suitable for these high frequencies.
Dear Mr. Knoll, and Prof. Ing. Rudolph.
Thank you so much for the extensive material you have posted!
In my quest to understand FM-VHF-UHF front end design in the 1950's I have been reading all the material you have posted, along with a few classic IRE (later IEEE) papers.
Most of these papers were written or co-written by E. W. Herold and L Malter of RCA and were published in the Proceedings of the IRE (Institute of Radio Engineering, later IEEE):
1936 A New Tube For Use In Superheterodyne Frequency Conversion Systems
1942 The Operation of Frequency Converters and Mixers for Superheterodyne Reception
1943 Some Aspects of Radio Reception at Ultra-High Frequency in 5 parts
1945 Conversion Loss of Diode Mixers Having Image-Frequency Impedance
Learning about front end design from all these perspectives (RM and IRE) has been very useful.
In particular, I will share a few points from these papers that apply to this topic:
1- Crystal diode mixing was often realized with very good conversion efficiency, with losses less than 2dB, (less than 20%). This means that the Antenna could drive the diode mixer, loose no more than 20% in frequency conversion, and drive the first very low noise IF stage. This combination can outperform a stand-alone mixer, because the mixer gain is lowered to the 1/3 to 1/4 of the gain of an equivalent amplifier.
I was suspicious that single diode mixers might be more efficient than active mixers. The papers listed above confirm that. My suspicion came from some experimentation I did a few years ago with single diode half wave detection vs full wave diode detection in crystal radios. The empirical result that I got was that for large signals, equal detection efficiency could be obtained with half wave detection or full wave detection, as long as the inductor tap from the high-Q LC tank circuit could be adjusted to a lower tap for full wave and a higher tap for half wave detection. The large signal assumption means that the resistive drops through the diodes are negligible losses, and so is reverse conduction. In this case, the diodes can be though of as ideal switches. In the full wave case, the diode/switches conduct twice as long as in the half wave case, thereby presenting a load that is half the impedance of the half wave case. Either case could convert RF to DC power with equal efficiency. My experiments relied on the Hi Q of the Tank circuit to average out the switched mode loading of the diodes. A similiar behavior explains intuitively the very high conversion efficiency that is possible with a single diode, but is not possible with a single mixer tube, which has no better than 30% conversion efficiency.
Herold and Malter give a very thorough mathematical presentation about diode mixing that focuses on the bidirectional nature of the diode, as compared to an amplifier, and on the importance of input and output impedances.
2- RF vs Mixer front ends. At higher input frequencies, the gain of an RF preamp drops below the gain of a mixer, so the RF preamp is no longer useful. This is the case with portable FM tube radios that use battery tubes that don't have enough gain at 100MHz to be useful as RF amplifiers.
The following is extracted from part 4 of "Some Aspects of Radio Reception at Ultra-High Frequency" by Herold and Malter, October 1943 Proccedings of the IRE.
"Above a certain frequency (which we may call the crossover frequency) the circuit bandwidth, with no external loading, exceeds that required by the application at hand, this being increasingly
the case as the frequency goes higher and higher. As a consequence, RF amplifier gain drops off above the 'crossover" frequency. A point will finally be reached for which the amplifier-stage gain drops to unity or lower, in which case it would obviously be foolish to use such a stage, and conversion should be employed in the first stage. In fact, better signal-to-noise response can usually be obtained by immediate conversion for cases in which the amplifier gain is still somewhat above unity. Since the intermediate-frequency-circuit resonant impedance is independent of the signal frequency, the mixer gain remains unaltered as the signal frequency increases, so that the mixerstage gain may eventually exceed that of the amplifier stage. This constitutes a further argument for immediate conversion above a certain frequency, which we may refer to as the "transition frequency."
A note on triode vs. tetrode vs. pentode noise:
All tubes, like all resistors, suffer from random thermal electron agitation, which causes noise. But the noise of a tube is worse than the noise of a resistor of comparable impedance (=1/gm).
One common formula relates triode noise to the noise of an equivalent resistance as
Tetrodes and Pentodes have an additional opportunity for random behaviour in the electron flow. The random behaviour comes from the uncertainty of an electron getting trapped by the screen grid or landing on the plate. This type of noise is called partition noise, and is often several times larger than the noise of the pentode or tetrode wired in triode connetion. In triode connection, the uncertainty of loosing a plate electron to the screen is eliminated because all electrons are collected.
Partition noise can be added, as an approximation, to the simple formula for the triode as follows:
Ig2 is the screen current
Reference for tube noise info: http://www.john-a-harper.com/tubes201/#Noise
The conclusion is that pentodes and tetrodes are, in general much noisier than triodes because of partition noise, but the pentode/tetrode partition noise disappears in triode connetion.
One exception to the partition noise rule applies to especialy constructed VHF amplifier tetrodes like the 6CY5, 6EA5 and 6CV5. These tubes have screen grids perfectly alligned with the control grid, so that they take very little screen current, and thus contribute little partition noise. Being tetrodes, these tubes have very low feedback capacitance, on the order of 0.03pF from Plate to control grid. This low capactiance makes it easier to achieve higher gain with a higher L/C ratio of the resonant plate load.
My understanding of front end FM-VHF-UHF design in the 1950's has been greatly expanded by the contributions to this thread.
Thank you again Mr Knoll and Prof. Ing. Rudolph and Mr Beckmann for the excellent contributions.
Thomas Albrecht has translated the 1953 Radio-Mentor article "DC90 in additiver Mischschaltung" from German to English. Thomas based his translation on the text capture by Hans M. Knoll.
The English Version is DC90_in_Additive_Mixer_Circuitry.pdf
The original German version is DC90 in additiver Mischschaltung_v2.pdf
Mr. Knoll first uploaded this article in a thread discussing the use of triode mixers in FM front ends:
Thank you all who contributed to this effort.