Call Search
     

New to Ham Radio?
My Profile

Community
Articles
Forums
News
Reviews
Friends Remembered
Strays
Survey Question

Operating
Contesting
DX Cluster Spots
Propagation

Resources
Calendar
Classifieds
Ham Exams
Ham Links
List Archives
News Articles
Product Reviews
QSL Managers

Site Info
eHam Help (FAQ)
Support the site
The eHam Team
Advertising Info
Vision Statement
About eHam.net

donate to eham
   Home   Help Search  
Pages: Prev 1 [2] 3 4 5 6 7 ... 18 Next   Go Down
  Print  
Author Topic: Designing & Building a High-Peformance Subminiature-Tube Regenerative Receiver  (Read 177343 times)
KB1WSY
Member

Posts: 1309




Ignore
« Reply #15 on: May 06, 2015, 12:09:27 PM »

Deciding on how many turns for the plate coupling coil is another matter. One can go for a large coupling coil approximately self resonant  but loosely coupled to the detector coil, or for a smaller coil  more tightly coupled. Now you are currently running with a 5 microhenry coil: I would go for 10 microhenries and less capacity. This looks like around 27k with a Q of 60. Now have about a half as many turns on the coupling coil and the plate load for the RR stage is 27/4 kohm and the gain is about 6 to 9. The transformed plate resistance is thrown across the tuned circuit: that will have little effect. So there can be reasonably tight coupling between the coupling winding and the detector.  I wouldn't wind the coupling winding over the tuned winding but put it 1/8 inch away and close wind it.

Thanks, lots of stuff to ponder, as I re-work the coils.

It's interesting that there are two fundamental skills in radio design: the applied mathematics; and the practical knowledge derived from actually building radios and seeing how the theory works out in practice. You possess both in abundance, I am in awe!

Here's what it should look like after the re-work:



Edited to add a quick question: I dithered about the polarity for the 1µF electrolytic on the RF Gain wiper. I've now "fixed" it so that the positive plate goes to ground ... which I think must be correct since the bias is negative?

MDM

73 de Martin, KB1WSY
« Last Edit: May 06, 2015, 12:26:12 PM by KB1WSY » Logged
KB1WSY
Member

Posts: 1309




Ignore
« Reply #16 on: May 06, 2015, 12:48:52 PM »

I wouldn't wind the coupling winding over the tuned winding but put it 1/8 inch away and close wind it.

Another question, something that's always confused me.

The tickler is currently wound on the bottom end of the coil form, underneath the "cold" end of the tank coil.

If I put the new primary winding (the plate coupling coil) above the detector tank coil, winding it (as usual) in the same direction as the tank coil, should the "cold" end of the coupling coil be immediately adjacent to the top of the tank coil, or should that be the "hot" end?

I suppose the same question also arises for the antenna coupling coil at the front of the RF stage....

73 de Martin, KB1WSY
Logged
WB6BYU
Member

Posts: 18535




Ignore
« Reply #17 on: May 06, 2015, 02:05:37 PM »

Generally I'd put the grounded ends of the coils together if I can - that reduces capacitive coupling
between the windings.

The direction of the winding doesn't matter for any of the coils except the tickler coil - that one
needs the proper phase to get the tube to oscillate.
Logged
KB1WSY
Member

Posts: 1309




Ignore
« Reply #18 on: May 06, 2015, 09:10:22 PM »

Spent much of day rewiring the detector: changing the capacitor values in the tuned circuit and rewinding the tank coil to try to get Peter/G3RZP's suggested 10µH/50pF LC ratio for the 40m band. (Haven't yet built the RF stage; decided to alter the detector stage first.)

Am rather discouraged. With antenna connected in the usual way, receiver is overloading drastically on almost every signal. Altering value of the series capacitor in the antenna connection doesn't help. Also, regeneration control is noticeably rougher and very "touchy" (I tried changing the tickler's turns and position, but it didn't help). Just much worse all around.

Tomorrow I will probably experiment with the tank-coil form factor. As of now, it is 15 turns that are 0.7" high, wound on a 1.25" pill bottle. That is nowhere near the 1:1 height/diameter ratio that Peter suggested for maximum Q. Looks like I should try a 1" or 3/4" form. On the other hand, the current symptoms (serious overloading) imply, don't they, that the Q has *increased* compared to the previous LC combination? Won't even higher Q make it even worse?

Another thing that's puzzling is that my calculations for number of turns was way off. To get 10µH it should have been 18 turns, not 15 (15 turns is supposed to be more like 7µH) and yet the only way I can bring in the desired band (6950-7150) is with the 15 turns (I don't have an inductance meter). I've done the calculations for the capacitance network over and over again, and believe them to be correct.

Edited to add: I've just measured the frequency coverage from one end of the dial to the other and it's only about 140kHz when it's supposed to be at least 200kHz. That tells me there's probably something screwy about the capacitor network (and also that the 15-turn coil is probably closer to 7µH than 10µH). It's time to go to bed -- I will check my wiring and calculations again tomorrow!!

The coil is also extremely fiddly: very small changes in the spacing of the windings yield very large changes in the resonant frequency of the tuned circuit.

As I understand it, Peter's rationale for the change in the LC parameters is to provide a better load/impedance match when the RF stage is added. My question therefore is: Should I worry about the sharp deterioration in the detector, as a stand-alone detector? Is everything going to be OK when the added load of the RF plate coupler winding and RF stage is added, will it just dampen all that overloading?

73 de Martin, KB1WSY
« Last Edit: May 06, 2015, 09:29:45 PM by KB1WSY » Logged
G3RZP
Member

Posts: 1324




Ignore
« Reply #19 on: May 07, 2015, 03:26:11 AM »

You could try the experiment of damping the Q with a shunt resistor to see if it's getting too high. Also the 2.2 megohm grid leak may well be a bit on the high side - I always used about 470k to 1 Meg. However, having said that, the 1927 ARRL handbook suggests values up to 7 Megohm with 201A or 199 tubes: the 199 data sheet suggests 2 - 4 Megohm. I found it easy to get squegging with such high values, but I was using tubes with an appreciably higher transconductance than they had. Typical transconductances for tubes back then was around 500 micromhos or so. One argument for the higher grid leak values is that they mean less damping of the tuned circuit, but if we assume a Q of 100 - which could be hopeful - that gives a dynamic resistance of 44kohm, so even a 470k will have little effect. The RSGB 1938 handbook suggests 1 - 2 Megohms. Sowerby's 1938 'Foundations of Wireless' likes higher values to avoid damping but he is concerned more there with the broadcast band where higher dynamic resistances were the norm anyway.

I have a feeling that some experimentation with tickler coupling and seeing where the oscillation starts as it is varied could be useful.

Interestingly, the Admiralty handbook of 1938 likes the idea of the detector not actually oscillating and a separate oscillator being used for the reception of CW. As I recall, this was done in the pre-WW2 Marconi CR200 receiver which covered something like 15kHz to 1MHz or so.

The 1927 ARRL handbook suggests an antenna coupling capacitor of two pieces of brass, 1/2 inch square, separated by 1/8 of an inch - I calculate that as 0.45pF, and too big a capacity there could affect where regeneration starts. Certainly, too big a capacity will affect the overload situation.

Terman (Radio Engineering, McGraw-Hill 1936) says that the feedback should be such that oscillation stops with a screen voltage between 20 and 40 volts for smooth regeneration, although it should be noted that he's talking of plate potentials of the order of 150 volts or so. If too low, the detection efficiency falls off because the tube gm is too low, while if too high, the oscillations start quite abruptly.
Some authorities like the idea of returning the grid leak to the positive side of the filament....

The idea of running with a higher L/C ratio is to get a better step up from the antenna.

Harmsworth's Wireless Encyclopedia (1924) doesn't go into optimum LC ratios, but likes the idea of a 500pF tuning capacitor for the medium wave (broadcast) band, which suggests the 50pF at 7 MHz is in the same ratio. The regenerative detectors in the ARRL handbooks use a wide range of LC ratios, tending to get smaller as the frequency goes up -presumably to take account of the effects of stray capacitance becoming more dominant.

Incidentally, what material is the coil former? Even at 7MHz, some plastics aren't very good, especially coloured ones. A quick and nasty check is to put a glass of water and the coil former in the microwave oven and zap it for 30 seconds or a minute or so to see if it gets warm. If it does, that tells you that it's lossy at 2.4GHz: if it stays cold, you can be pretty certain it's not lossy at 7MHz.

I think you need a capacitor to ground between the junction of the tickler coil and the RF choke. Something like 100pF.

So there has been an extensive trawl through my library for this. Interestingly, Scott-Taggart in 'Thermionic Tubes in Wireless Telegraphy and Telephony' (Iliffe, 1924) is very quiet on the subject.
Logged
KB1WSY
Member

Posts: 1309




Ignore
« Reply #20 on: May 07, 2015, 04:48:55 AM »

You could try the experiment of damping the Q with a shunt resistor to see if it's getting too high.

Across the coil, right?

Also the 2.2 megohm grid leak may well be a bit on the high side - I always used about 470k to 1 Meg. However, having said that, the 1927 ARRL handbook suggests values up to 7 Megohm with 201A or 199 tubes: the 199 data sheet suggests 2 - 4 Megohm. I found it easy to get squegging with such high values, but I was using tubes with an appreciably higher transconductance than they had. Typical transconductances for tubes back then was around 500 micromhos or so. One argument for the higher grid leak values is that they mean less damping of the tuned circuit, but if we assume a Q of 100 - which could be hopeful - that gives a dynamic resistance of 44kohm, so even a 470k will have little effect. The RSGB 1938 handbook suggests 1 - 2 Megohms. Sowerby's 1938 'Foundations of Wireless' likes higher values to avoid damping but he is concerned more there with the broadcast band where higher dynamic resistances were the norm anyway.

Good idea. I did some experiments with grid leak value early on in this saga, but haven't touched it since. For what it's worth, back then I tried various values between 2.2M and 4.7M. There was no audible difference, from what I remember. But I didn't try *lower* values. I will try a wider range: from 470K through 4.7M 6.8M.

I have a feeling that some experimentation with tickler coupling and seeing where the oscillation starts as it is varied could be useful.

The tickler coil was originally 2 turns, tightly coupled (close-wound directly below the tank, with no space in between). I tried 1 turn, which didn't work (hard to get any oscillation at all). I tried 3 turns, tightly coupled in the same manner, and this was slightly better than 2 turns although the difference is not great. In all cases (1, 2 or 3 turns) the tight coupling was necessary in order to get any oscillation. Moving the coupling even a few millimeters away from the tank killed the oscillation.

Incidentally, what material is the coil former? Even at 7MHz, some plastics aren't very good, especially coloured ones. A quick and nasty check is to put a glass of water and the coil former in the microwave oven and zap it for 30 seconds or a minute or so to see if it gets warm. If it does, that tells you that it's lossy at 2.4GHz: if it stays cold, you can be pretty certain it's not lossy at 7MHz.

It's an orange-colored pill bottle. I was wondering about that! I knew about the microwave trick but have never tried it -- will do so today (after removing the wire!!!!). By the way, the inductance-calculation errors also happened with the old 5µH coil: to get that value it should have been about 12 turns, 0.5" tall, but I couldn't get into the 40m tuning range without removing a couple of turns -- and theoretically 10 turns is only about 4µH.



The next thing I'll try is an old, trusty Millen 45005 coil form: 5-prong, 1" OD. I'm usually reluctant to use them for experiments as they are hard to find, but I have a couple that are a bit messed up (drilled and scratched up by a previous owner, and with pins loose from excess soldering heat) and I'll play with those.

On a 1" form and with 20AWG wire (wire diameter 0.032"), 23 turns close-wound is about 0.92" long (allowing for the magnet-wire insulation thickness) and is supposed to yield 10.35µH. It comes close to your ideal 1:1 dia/length ratio.

Edited to add: however those Millen forms aren't very tall. They only have about 1.5" in useful height for winding. If the tickler and tank take up, say, 1.1" using 20AWG magnet wire, then there's hardly enough room at the top for the plate coupling coil from the RF stage, assuming that the coupling coil is half the height of the tank coil. At the very least the coils will be very close together. So perhaps I need to find another 1" form; or I could experiment with thinner magnet wire, to increase the number of turns available with a given coil height.

I think you need a capacitor to ground between the junction of the tickler coil and the RF choke. Something like 100pF.

Will try it.

So there has been an extensive trawl through my library for this. Interestingly, Scott-Taggart in 'Thermionic Tubes in Wireless Telegraphy and Telephony' (Iliffe, 1924) is very quiet on the subject.

It's clear that the old-timers managed to get excellent performance out of their regenerative sets, but it was hard work and very fiddly!

For what it's worth, monitoring the band this morning, the overload is much less. Some signals have no overload at all. So among other things, band conditions last night may have been a factor (in any case, nights are more propitious to strong-signal overload in my experience). I am hoping that the alterations that you suggest -- and changing the coil form factor -- will yield progressive improvements.

Edited to add: I also assume that you agree with me that it's useful to get the best possible performance out of the detector working in stand-alone mode, prior to adding the RF stage. One thing at a time....

73 de Martin, KB1WSY
« Last Edit: May 07, 2015, 06:24:10 AM by KB1WSY » Logged
KB1WSY
Member

Posts: 1309




Ignore
« Reply #21 on: May 07, 2015, 05:29:24 AM »

I have a general question: how does a grid-leak detector work?

I've looked this up in several books but unfortunately they just tend to present you with a schematic and say, "this is a grid-leak detector" without further ado. This is often in the same chapter that explains diode detectors and triode/plate detectors -- which are easy to understand and usually explained in detail.

Why should a resistor, shunted with a capacitor, inserted into the grid circuit turn an amplifier into a rectifier/amplifier? The grid leak resistor is a standard technique for obtaining grid bias ... so it must have something to do with the addition of the shunt capacitor, huh?

Incidentally I love the way that the older British books describe it as the "leaky grid" detector. Sounds like a case of incontinence!

73 de Martin, KB1WSY
Logged
KB1WSY
Member

Posts: 1309




Ignore
« Reply #22 on: May 07, 2015, 06:10:31 AM »

For the coil form, I just ordered a length of this stuff: 1" OD polycarbonate pipe (trade name Lexan). It will be useful if the Millen forms turn out to be too short to be useful; and will be a less destructive medium for experiments (I hate to use those Millen forms for anything other than a final product).

http://www.amazon.com/Polycarbonate-Tubing-Wall-Clear-Color/dp/B00193SSPG/ref=pd_sim_sbs_indust_2?ie=UTF8&refRID=04TKT98TWZRMBWZZZ12P

73 de Martin, KB1WSY
Logged
JAHAM2BE
Member

Posts: 319


WWW

Ignore
« Reply #23 on: May 07, 2015, 06:37:50 AM »

I have a general question: how does a grid-leak detector work?

A simple explanation can be seen starting on p.2 here:

http://www.inictel-uni.edu.pe/sites/default/files/archivos/2015/publicaciones/04/the_modern_armstrong_regenerative_receiver.pdf
Logged

KB1WSY
Member

Posts: 1309




Ignore
« Reply #24 on: May 07, 2015, 07:06:32 AM »


Perfect, thanks; even I can understand it now! Also, thank you for stimulating my extended interest in regenerative radios, a couple of years ago -- when you gave me quite a lot of help with my then-balky ARRL regen build and its unusual Vackar oscillator.

Edited to add: incidentally, the article you link to has the alternative circuit in which the grid leak is connected between grid and ground, instead of being in series with the tuned circuit (while the capacitor remains in series). I guess I'll give that a try, too....

I will also be experimenting with a throttle capacitor (instead of the adjustable screen voltage) for regeneration control. But that comes later....

73 de Martin, KB1WSY
« Last Edit: May 07, 2015, 07:14:22 AM by KB1WSY » Logged
G3RZP
Member

Posts: 1324




Ignore
« Reply #25 on: May 07, 2015, 07:14:18 AM »

Martin

I'm having trouble with some of your numbers.

Quote
On a 1" form and with 20AWG wire (wire diameter 0.032"), 23 turns close-wound is about 0.92" long (allowing for the magnet-wire insulation thickness) and is supposed to yield 10.35µH.

0.004 inch thick insulation seems a lot.  Looking at my wire tables, I see ARRL quote 29.44 turns to the inch: Scroggie, for 21SWG (which is 0.032 inch) quotes 28.96 turns to the inch. That averages to 29.2 tpi or 0.03425 inches, giving a enamel thickness of 0.00123 inches, which sounds more likely. Thus 23 turns will be 0.79 inches long and have an inductance of 10.94 microhenries.

Not sure how you get 10.35 microhenries. I make 23 turns, 1 inch diameter and 0.92 inches long to have 9.3 microhenries.

I use Wheeler's formula  L = (a2 n2)/(9a + 10b), where a is the radius in inches, b is the length in inches and n is the number of turns.

How is the coil mounted? Is there a metal chassis or something that would pass as such? The coil needs to be at least one and preferably 2 diameters away. If you have a metal plate on which the coil is mounted, you could well get problems if either the tuning coil or the tickler get too close to the metal.

In the 6SN7 regen in the ARRL handbooks from about 1947 to 1957, for 40m, they use  4.3 microhenries from 13.5 turns and a tickler of 1.25 turns spaced 1/4 inch away: the plate and grid coil ends should be at opposite ends of the form. For 80m, they use 17.6 microhenries from 25 turns of #26, close wound, with a tickler of 4 turns, spaced 3/8 of an inch.

From which you can see quite a variation in L/C ratio, and surprisingly small tickler coils.

Shunt the detector coil with decreasing values of resistor, starting with 470k and see if it smooths the regeneration onset.

I also feel that you need a smaller coupling capacitor to the antenna.

Quote
For the coil form, I just ordered a length of this stuff: 1" OD polycarbonate pipe

It is about 1000 times more lossy than polystyrene, polyethylene or Teflon, and about twice as lossy as the mica loaded phenolic Millen forms. Slightly better than PVC, and about half as lossy as most woods. If it is black and carbon loaded, it is likely to be worse than that figure, while it's about 2/3 of the loss of FR4 printed circuit material.

Logged
KB1WSY
Member

Posts: 1309




Ignore
« Reply #26 on: May 07, 2015, 07:31:31 AM »

0.004 inch thick insulation seems a lot.

It was just a wild guess, so I'm not surprised I'm wrong.

Not sure how you get 10.35 microhenries. I make 23 turns, 1 inch diameter and 0.92 inches long to have 9.3 microhenries.

I use Wheeler's formula  L = (a2 n2)/(9a + 10b), where a is the radius in inches, b is the length in inches and n is the number of turns.

I used an online calculator which, itself, claims to use Wheeler's formula (the calculator is here: http://www.66pacific.com/calculators/coil_calc.aspx). Before building any more coils I'll do the calculations myself, which is a good exercise in its own right!

How is the coil mounted? Is there a metal chassis or something that would pass as such? The coil needs to be at least one and preferably 2 diameters away. If you have a metal plate on which the coil is mounted, you could well get problems if either the tuning coil or the tickler get too close to the metal.

It is mounted with a single brass screw in the middle of the base, onto a piece of wood. The shaft of the tuning capacitor is about 1.7 diameters away, but otherwise there is no metal.

...the plate and grid coil ends should be at opposite ends of the form.

Can you elaborate? You mean the "hot" ends right, with both of the "cold" ends in the middle? And the tickler underneath the grid coil?

Shunt the detector coil with decreasing values of resistor, starting with 470k and see if it smooths the regeneration onset.

I also feel that you need a smaller coupling capacitor to the antenna.

Will experiment with both.

It is about 1000 times more lossy than polystyrene, polyethylene or Teflon, and about twice as lossy as the mica loaded phenolic Millen forms. Slightly better than PVC, and about half as lossy as most woods. If it is black and carbon loaded, it is likely to be worse than that figure, while it's about 2/3 of the loss of FR4 printed circuit material.

Phooey. OK, will try to find something less lossy. BTW I have a large collection of Air Dux style coils, in all sorts of diameters and winding pitches. Presumably those would be good in this application, but "adjusting" them is a destructive process and I hate to do that until I have a better idea of the optimal inductance and configuration in this application. I also have a nice 1" grooved ceramic form but I was saving that for my future transmitter VFO.

Edited to add: Also have a sizable collection of original Amphenol 1.25-inch polystyrene forms, but those are too precious to waste on experiments....

73 de Martin, KB1WSY
« Last Edit: May 07, 2015, 07:37:24 AM by KB1WSY » Logged
G3RZP
Member

Posts: 1324




Ignore
« Reply #27 on: May 07, 2015, 07:56:39 AM »

Quote
Can you elaborate? You mean the "hot" ends right, with both of the "cold" ends in the middle? And the tickler underneath the grid coil?

Yes.

Quote
I used an online calculator which, itself, claims to use Wheeler's formula

Wheeler's formula has some constraints, but for coils with reasonable L/D ratios and not too small a diameter, it should be within 5%. Connecting leads, as a rule of thumb, are about 20 - 25nH/inch. We always reckoned on 1nH/mm in integrated circuit packages which is 25nH/inch.
Logged
KB1WSY
Member

Posts: 1309




Ignore
« Reply #28 on: May 07, 2015, 08:23:20 AM »

Thanks Peter. Must get back to my "day job" for at least a few hours but for the time being one last question.

While I cannot claim to have much to do with the design of the current experiment, I am taking pride in understanding exactly the purpose of every sub-circuit, and the function of every component.

So now let me make sure that I understand the RF amp biasing circuit that you designed. Here is my "take":

It's a directly heated cathode, therefore we cannot use the usual cathode biasing resistor, nor can we get enough bias with a grid-leak resistor. Therefore we need a source of negative potential, but the ground bus is already the negative side of the receiver's power supply. For that reason, we need to "float" the LC circuit. The 2000-5000pF capacitor at the bottom of the coil functions as a RF bypass (but why is it necessary, at that point in the circuit?). The 1µF on the pot wiper helps smooth the DC bias. The pot varies the grid bias to provide adjustable gain.

Correct?

(Concerning the -4.5V bias, this will either be a separate battery, or it will be tapped with a series resistor at the bottom of the B+ battery -- this is similar in general concept, I guess, to the taps that are sometimes used on HT transformers to provide bias voltages.)

73 de Martin, KB1WSY
Logged
G3RZP
Member

Posts: 1324




Ignore
« Reply #29 on: May 07, 2015, 09:19:35 AM »

Quote
The 2000-5000pF capacitor at the bottom of the coil functions as a RF bypass (but why is it necessary, at that point in the circuit?).


That point needs to be at RF ground potential. If you let it float, RF wise, you aren't providing an input to the RF stage between filament and grid.

 
Quote
The 1µF on the pot wiper helps smooth the DC bias. The pot varies the grid bias to provide adjustable gain
.

Pretty much. The electrolytic bypasses noise from the pot as it is rotated.
Logged
Pages: Prev 1 [2] 3 4 5 6 7 ... 18 Next   Go Up
  Print  
 
Jump to:  

Powered by MySQL Powered by PHP Powered by SMF 1.1.11 | SMF © 2006-2009, Simple Machines LLC Valid XHTML 1.0! Valid CSS!