Here is another way.

The Grebe CR-5 receiver has no visible grid leak. It's a paper strip painted with graphite hidden behind the coupling mica condenser. I have used the same method to repair them. Once the desired resistance is obtained, the pencil trace is protected with lacquer or super glue.

It is important that the grid leak value in conjunction with the capacitor, is not too large. Although it results in higher gain, the upper audio frequency response becomes severely limited and distorted (usually negative peak clipping) because the discharge of the capacitor cannot track the envelope of the modulation quickly enough when the modulation frequency goes over a certain value. In other words, important that the RC time constant is not too long.

Terman resolved the problem using the AC impedance of the detector divided by the DC resistance being less than the modulation m, which resolves to a formula of f = 0.275/RC, where f is the upper frequency where distortion sets in with m=0.5. Using two other methods I resolved it to 0.32/RC or 0.5/RC.

In the post war period where fidelity became an issue, RCA moved to 100k & 250pF and the detector was good to 12.8kHz when the modulation was 0.5. With a 3Meg grid leak and a 250pF capacitor in a 1920's radio it was pretty poor and distortion set in at around 1kHz.

One way to improve the detector performance would be to have the capacitor discharging via a constant current sink, rather than a resistor.

Could you point me to a reference for that formula?

Hi Folks!

I was re-arranging various info files on my primary computer as well as researching Cornell-Dublier products, when I paged through the pamphlet for Dublier 601's. Many times I have posted to and about, "Where's the grid leak?". The posted attachment was Dublier's answer...

I will post more in time on many varied Wireless & 20's radio topics.

I am still searching for "How to Bank-Wind coils". I suspect that has yet to be scanned from my paper library .

Certainly much more is stored at World Radio History but so long as it resides there it cannot stimulate participation as it does here at ARF...

On topic, I have a Paragon RA-10 & DA-2, it does not have a definable grid leak resistor. FWIK it depended on the poor quality of tubes in those times for the needed leakage. I had often wondered why the rough ceramic bases of the early tubes could not be penciled from grid to filament (+) as a leak. But then, they were already leaky

Chas

Terman covers detector distortion briefly in Radio Engineer's Handbook (1943) starting on page 554, and in more detail in his earlier Radio Engineering 2nd Edition (1937) starting on page 426. In both cases, he treats only simple series diode detectors, not grid leak detectors, and is considering the total audio and DC loads on the detector. He later covers grid leak detectors and discusses the grid leak resistor/capacitor combination as one part of the overall detector load, but only addresses the frequency distortion that can be introduced if the grid leak resistor value is too high to allow the capacitor voltage to follow the modulation.

Terman attributes the original research to Kilgour and Glessner "Diode Detection Analysis" Proc. IRE, Vol. 21, p. 930, 1933. I haven't read the Kilgour and Glessner article.

Ben Tongue also addressed the situation of unequal audio and DC loading on the detector in a crystal set application (again just a simple series detector, not gridleak), and hence the invention of the "Benny" for compensating for the difference:
http://web.archive.org/web/201612211349 ... xtlsd.html

I wrote this up over 20 years ago. I derived the 0.275/RC from Terman's remarks about detectors (explained in the article, it is not a good copy, it got a little disordered with the order of the paragraphs, it was published in the OTB).

But I also explained two other ways to get the approximate formula:

https://www.worldphaco.com/uploads/Radi ... 920_s..pdf

By the way, Terman was particularly good at summarizing a problem with a few sentences. One other example of this is his remarks and equations for coupled resonant circuits. But, sometimes you need the result in a different form, for example how those results give the relative frequency response of a vintage tube radio inter-stage transformer. If you look at pages 11 on in this article, you will see how it is possible to use Terman's equations to find the result:

https://www.worldphaco.com/uploads/THE_GREBE_MU-1.pdf

Coincidentally if you look on page 21 you will see binhbui's Avatar, Dr. Mu. He was a fictitious doctor invented by Grebe, to market their radios, as if life wasn't stranger than fiction. Dr. Mu would quote Chinese Philosophers and link the wisdom with the quality and value of Grebe radios. I liked it.

I came across this in an early Clapp-Eastham catalog. It looks like the one in the Grebe, above.

Bob, thank you for the references.

Hugo, thank you for the articles. It's interesting to visualize the grid-leak detector as a diode detector followed by a triode amplifier.

The frequency distortion due the long time constant of the grid leak and capacitor is valid if grid rectification is considered alone. Ballantine showed that combining grid and plate rectification can compensate the frequency distortion in Proc. I.R.E. May 1928 p.605.

For tubes with directly heated cathode, the cathode is not at the same potential. In a detector circuit using such a tube, the grid leak is returned to the positive terminal of the filament causing the grid to conduct even without any signal. Since the grid components in series with the grid resistance establish a relative negative bias, plate rectification is not insignificant.

In the Bell System Technical Journal vol. V 1926 p.447, Llwellyn gave a generalized transfer function for the detection component of the plate current (Eq. 42) which can be used to calculate the audio frequency response of the grid-leak detector. From there he gave a formula for calculating the optimum grid-leak capacitor for an audio frequency. Besides the grid leak, the formula takes into account the carrier frequency and the grid resistance.

https://worldradiohistory.com/Archive-I ... 928-05.pdf
https://www.google.com/books/edition/Re ... g=RA25-PA1

Decades ago, I worked at a company that made hybrid power electronics circuits. We used a ceramic substrate and laser-trimmed thick film resistors.

Resistors were designed to be increased in value, never reduced. Adding extra material, like a graphite pencil mark, increases the chances the resistor will change value with temperature and humidity.



Without a laser, one could still grind away some of the resistive material (Dremel tool?) while keeping the resistor attached to an ohm meter. Grind and measure until you get the desired value.

Rich

Here's another reasoning for deriving the highest audio frequency before frequency distortion occurs.

Let's assume the grid-leak capacitor is charged to 1V before it starts decaying. Its voltage exp(-t/RC) has a maximum negative slope of -1/RC when the decay begins (at time t=0).

Let's assume the audio modulation signal is -m*sin(w.t) where m is a modulation index (0 - 1) and f = w/2pi is the audio frequency. Its maximum negative slope is -m*w.

For the audio signal to be reproduced without distortion, the maximum rate of decay of the capacitor voltage must equal to or larger than the maximum signal slope. Thus 1/RC >= m*w or f <= 1/(2*pi*m*R*C). For m = 1, f reduces to the familiar -3dB cutoff frequency of an RC lowpass filter which is the most conservative formula to use. For m = 0.5, f <= 1/(pi*R*C) = 0.318/RC which is the same as your second formula.

I once had a 30's era, Triplett VOM, one of its resistors, carbon "dogbone", had a neat rounded bottom notch cut into the body, painted with a sealant. The multiplier dot was on the other side of the notch.

I mentioned in my article though, that anode bend detection preferentially amplifies the positive going part of the input carrier, but grid leak detection amplifies what amounts to a voltage doubled negative half of the input carrier. They are opposing effects and in the boundary between the two types of detection, no demodulation is possible.

It may well be that a trace of anode bend detection could help linearity, but too much would disable the grid leak detector. And it is not the fundamental functional principle of the grid leak detector.

I also explained the voltage distribution along a filament of a directly heated tube, how it affects the average bias and demonstrated how that can be modeled as a chain of indirectly heated tubes.

Since this post is about how to make an adjusted grid leak, I'd like to discuss the need for adjusting a grid leak without overwhelming technical details. The following discussion is applicable to directly heated tubes such as the 201A. It is assumed that the grid leak returns to the positive side of the heater.

Very Large Grid Leak

When the grid leak is very large such that it forms a very long RC time constant with the parallel capacitor, the capacitor voltage approaches the heater voltage and varies very little with the signal AM envelope. The grid is negatively biased heavily and grid rectification is thus very small. Any grid rectification taking place is "crushed" by the large RC time constant. If the grid leak is reduced enough to place the operating point just past the start of the bend in the transconductance (plate current vs. grid voltage) characteristics, plate rectification takes place. People did this in the past to detect weak signals (in addition to adjusting the plate voltage).

As far as the audio frequency response is concerned, the large RC time constant is deceptively irrelevant. Its role is to maintain an almost constant capacitor voltage to shift the signal down to a negative offset for a satisfactory plate detection. What becomes relevant is the time constant of the capacitor and the grid resistance which form a high-pass filter. When the grid starts to draw current continuously, the grid resistance can no longer be ignored and must be taken into account when determining the audio frequency response. Since the high pass filter emphasizes high frequencies, the audio frequency response tilts up after plate detection and before any post filtering.

Very Small Grid Leak

When the grid leak is very small, the grid is positively biased too heavily. Since the grid conducts too much, grid rectification is small except for strong signals. The operating point is placed on the straight portion of the transconductance characteristics and therefore plate rectification is also small. Any occurring rectification is due the non-linearity of the characteristic curves.

Intermediate Grid Leak

As the grid leak is decreased from a value that mostly favors plate rectification, the grid conducts more and more. Grid rectification comes into play more and more until it cancels out plate rectification. As shown graphically by Ballantine a long time ago and pointed out more recently by Hugo, these two modes of rectification are practically out-of-phase at low frequencies and they subtract from each other when combined.

If the grid leak is decreased further past this cancellation point, there exists a value where the plate rectification is one half of the grid rectification. When this happens, the low-pass response of the grid rectification is exactly compensated by the high-pass response of the plate rectification (see Ballentine article referenced above).

Decreasing the grid leak still further reaches an optimum value which mostly favors grid rectification. This can be easily seen by subtracting the envelope detected by grid rectification from the original AM signal to obtain the signal input to the grid. As the grid rectification increases to its limit, there will be little positive envelope left for the plate to detect while the negative envelope is almost doubled.

Summary

It was well known that the negative grid bias varies with signal levels. The grid leak needs to be adjusted to give a satisfactory detection for a signal condition. It can be adjusted to choose plate rectification only for weak signals, grid rectification only for strong signals, or a mix between between the two modes of rectification to reduce distortion and to obtain a flatter audio frequency response.

It is an unfortunate thing that tubes like 201A's make for very poor grid leak detectors, though they do work of course.

The main reason is that the widely spaced grid to filament (cathode) of the 201A tends to behave like a diode with a high forward resistance when the grid-filament is forced into forward conduction by signal. This is completely unlike a semiconductor diode, or some smaller indirectly heated triodes where the cathode-grid makes for a much more efficient diode. As it does of course in local oscillator circuits which rely on this method for bias.

So, what happens is, the incoming carrier is poorly "clamped" and the positive going carrier voltage at the grid rises and falls with modulation, decreasing the efficiency of the detector because this reduces the average plate current excursions corresponding to the modulation. Though there is still enough asymmetry to make it work as a grid leak detector.

If it was a silicon diode or a BJT's B-E junction the carrier peaks would be hard clamped to about 0.6V above common, even for the modulation troughs, if the RC time constant was not too long. I pointed this out in the diagram in my article.

A 6J5 makes a much more efficient grid leak detector or a general detector than a 201A. Philco realized the grid-cathode of the 6J5 was a particularly good detector and took advantage of this in some of their radios.

One thing you can do, though it is somewhat anachronistic, is you can add a silicon signal diode across the grid and one filament connection of a 201A grid leak detector to boost the audio output. Or one other way is to use a tube diode, designed for signal rectification such as 1/2 of a 6AL5 which is very good, or even one of the small signal Acorn like tubes, such as the EA50 which was commonly used in early TV & Radar work in signal clamping at the grid of another tube, for example in a sync separator circuit.

There is a photo of the interesting EA50 on page 7 of this article. I first saw one of these when I was about 7 years old. I tried to use it in a crystal set and got poor results due to the very low signal level, until my brother suggested forward biasing it a little with a battery and a 10 Meg resistor, then it worked spectacularly well.

www.worldphaco.com/uploads/ARGUS.pdf