AM Micro Transmitter

[dominicbeesley]

Here's an updated, hopefully clearer and correct diagram.

Added blocking cap between osc. anode and coil.
Added integrating cap in parallel with demodulating resistor in feedback loop
Added RF decoupling cap from top of tank circuit (improves output, stops RF feeding back into driver stages).

Going to have a play with aerials and matching circuits now

Dom

[kalee20]

Dom, I'm really impressed with the circuit.

I've been playing with a micro AM transmitter lately. I also wanted to use a PCL82 valve (not because I got some cheap, but because I got some free from plundering local tips in the days when TV's were dumped). I also wanted to use negative feedback, similar to how you have - you've beaten me to it! However, at the moment I'm playing with tr**s****rs and a frame aerial (10" square) to radiate.

First attempt used just an oscillator, modulated, driving the aerial. This gave good linear modulation, by plotting RF output against supply voltage. However, sound quality was bad and changed significantly as I tuned the receiving radio. I quickly found that this due to frequency shifting with amplitude - giving a weird set of sidebands which didn't fit inside the receiver's selectivity curve at all!

Adding a separate output stage, and modulating that (leaving the oscillator at a fixed supply voltage) cured that. I'm interested in your comment about having the PA stage slightly off-tune giving better treble response. I've been theorising about the output, and I think if the PA tank circuit has too high a Q then it can't respond to the modulation as quickly. So, I think you could try an overwind on your output LC circuit coil to drive the aerial with more RF voltage. Then, the extra power radiated will reduce output Q. (A quicker check would just damp the circuit with a resistor).

I wouldn't worry about the PCL82 heater-cathode voltage too much. It should be good for at least 150V - the valve is rated to 200V.

[MichaelR]

First Class thread , thoroughly enjoying the discussions and ideas

Mike

[dominicbeesley]

Originally Posted by kalee20

Dom, I'm really impressed with the circuit.

Adding a separate output stage, and modulating that (leaving the oscillator at a fixed supply voltage) cured that. I'm interested in your comment about having the PA stage slightly off-tune giving better treble response. I've been theorising about the output, and I think if the PA tank circuit has too high a Q then it can't respond to the modulation as quickly. So, I think you could try an overwind on your output LC circuit coil to drive the aerial with more RF voltage. Then, the extra power radiated will reduce output Q. (A quicker check would just damp the circuit with a resistor).

Thanks Kalee,

Yes, I tried modulating the oscillator directly in an earlier transistor experiment and could only get usefull service with very low modulation levels, due to the the thing then being an FM as well as an AM transmitter! Was good enough for "baby monitor" type stuff. Here I'm after something a bit more Hi Fi.

I think probably the reason for it "sounding better" slightly off tune is indeed that the Q of the tanks cirucuit for the final stage is too high and is peaking on the lower frequency sidebands, shifting it slightly is adding a fake pre-emphasis. The feedback circuit helps a little with this but can only do so much.

The feedback circuit could probably be improved by adding a clamp diode in parallel with the first 100k resistor, this would then take both halves into account, at the moment it just takes one half of the wave form and integrates that.

When I get the chance I'll run the aerial from an overwind, hopefully calculated to give an impedance of about 50 ohms, then I'll try a frame aerial too. How many turns do you reckon for a 10" aerial at around 400m? And should I then have a capacitor in series with the aerial? I'll also try some different resistances is parallel with the tank and see if the quality of the sound changes (with and without the feedback loop in place) and report back with results

Cheers

Dom

[kalee20]

Hi Dom,

I think I'd start by an equal number of turns for the overwind, which would double the RF voltage getting to your aerial.

I'd be interested in your results with a frame aerial, but I'm not recommending it! I'm using it as a combined tank inductor and radiator. At close range, pick-up is good with another frame aerial, but I think that's due mainly to direct inductive coupling. Reception falls really rapidly with distance, and that is consistent with the near-field pure magnetic coupling. So, my next experiments will use a vertical wire. (I need a suitable portable radio to check. I have an Ever-Ready Model K, but it needs some work first...)

I do note you have a 33mH RF blocking choke after the modulator, and your RF decoupling capacitor is 0.1uF. May I suggest dropping the capacitor to 0.01uF? The existing values have a cut-off frequency of 2.77kHz, which will really make the negative feedback circuit work hard to restore treble. Preferably also reduce the choke a bit too.

The other thing about the NFB is, that there's a direct AF path from PA anode to AF input. I can't help feeling that if you had just RF coupling then the feedback signal would be more representative of what the circuit transmits. I have in mind myself a small centre-tapped pick-up coil slipped over the tank coil, with detector diodes. You can vary the feedback just by moving the coil about. See attached sketch.

[dominicbeesley]

Thanks for those suggestions. There isn't really an AF path from PA anode round to AF input, just a bloke who keeps missing capacitors off his circuit diagrams! There should be a 100pf blocker before the feedback network to remove all AF from the signal before demodulating (I did have this problem before).

I take your suggestion about the feedback coming from another winding and had thought of something like that - I just got tired of winding coils by that point! I'll try about 50 turns on a 1" former that I can move around for best sound.

Indeed the 0.1uF cap probably is a bit fierce. I'll drop this back again, though that lead to quite a bit of RF leaking back through the choke, though it didn't seem to hurt too much as the cathode follower is pretty robust!

If I do the overwind thing will this not increase the output impedance? I want to feed this into a 75/50 ohm distribution network - I confess to being a bit baffled about how to do this....

Though here we're wandering off topic and it might be best to take the feedback bit to the other thread at http://vintage-radio.net/forum/showt...452#post164452

Cheers

Dom

[Station X]

You could tune the transmitter into a dummy load ie a 50 or 75 ohm resistor. If you can't achieve a dip in anode current you'll know that some changes to the output circuit are necessary. You can then connect the transmitter to a real feeder and aerial. If the dip occurs at the same settings you'll know that the aerial is a good match.

A half way dipole is generally said to have an impedance of 75 ohms, but it will be enormous at MW frequencies. To match correctly to other aerials you may have to change the output circuit to a PI network type or use an external aerial matching untit (AMU) sometimes called an Aerial Tuning Unit (ATU). An SWL type AMU should be OK if you're using low power. With high power the variable capacitors will flash over.

[kalee20]

Yes, a half-wave dipole at medium wave will be 200 metres long, and so it would have to be made of fairly thick wire for the resistance of this not to become an issue. Not practical for a micro-transmitter!

Regarding matching, to me it depends on how the PA stage is operating. If it's Class A (which would be ideal to stop radiating harmonics) then you just need to pretend the PCL82 is operating at AF. The optimum load is approximately the modulator quiescent output voltage divided by PA anode current, times 0.9, and you can arrange this load by a suitable RF transformer. Feeding a few metres of wire, the transformer will be step-up, but for a 50 ohm or 75 ohm load it will be step down. So, sorry, that means winding secondaries on your tank coil!

Alternatively, you could use a pi network as Station X says, but that will involve a fair bit of complex number crunching.

And, if drive level is such that it isn't Class A, then all bets are off anyway. My approach would be to use a 'scope to monitor the PA anode and optimise like that. I'd want my anode voltage to swing to near zero, but not negative (because screen grid dissipation would increase). So, if it DID go negative, then I'd adjust the turns to feed the aerial or transmission cable. But if it didn't, I'd try and increase drive first, expecting output to increase and then level off, and then reduce coupling to the aerial.

[MichaelR]

Originally Posted by kalee20

And, if drive level is such that it isn't Class A, then all bets are off anyway. My approach would be to use a 'scope to monitor the PA anode and optimise like that. I'd want my anode voltage to swing to near zero, but not negative (because screen grid dissipation would increase). So, if it DID go negative, then I'd adjust the turns to feed the aerial or transmission cable. But if it didn't, I'd try and increase drive first, expecting output to increase and then level off, and then reduce coupling to the aerial.

Not a bad idea but would it not be very susceptible to changes in aerial wire routing etc, in other words any environmental changes would affect your setup and could put you into a mode full of distortion.

Mike

[kalee20]

Only if the aerial characteristics changed. And that would affect most output stages anyway.

[MichaelR]

Originally Posted by kalee20

Only if the aerial characteristics changed. And that would affect most output stages anyway.

Of course that is the case however the practicality of resetting up your Tx with the method is what I am questioning.

The problem seems to me to come up with some reasonable method of distributing the MW RF. A decent Ax is not acceptable as there would be issues of radiating too far, anyway a decent match Ax at these frequencies would have to be outside and high up.

What about feeding into a reasonable length of inexpensive coax with resisitve load on the end. It does not matter for the end resistor to be 50 ohms the idea is to cause the unbalanced coax to radiate . If you ran the coax down a wall cavity , you could load it up with sufficient power to keep the transmission within the house.

Mike

[dominicbeesley]

Hello all,

I've just spent another pleasant evening playing around and have managed to put into practice some of the hints mentioned so far. Mainly in order to improve the AF response of the thing. I haven't got round to sorting anything out with the aerial, other than running a longer wire! More on that next week - going away this weekend....

Below is attached an updated circuit. The main changes are:

• increased gain in the first AF stage (220k collector load, 22K input resistor)
• Better filtering of feedback (includes a choke), with less attenuation of signal (smaller "smoothing" caps)
• Clamping of input to feedback demodulator
• AF decoupler at choke input to final pentode stage reduced from 0.1uF to 4n7 - much less roll off of high frequencies.

To assist my understanding and because I like a nice graphical display, I've recorded a special audio track where one channel is a 20Hz sawtooth and the other channel is a sweep of AF frequencies from 500Hz to 15kHz., with "pips" at 1kHz, 2kHz, 5kHz, 10kHz. The sawtooth is used to drive the X-Mod of the scope and makes a mini-wobbulatorish sort of thing, that graphically shows AF response curves to watch - while you tinker. (Note due to my not thinking things through the graphs are "backwards" i.e. the higher frequencies are to the left).

The two oscilloscope shots show the input waveform at the top and the signal at point A in the feedback loop, the first picture shows without feedback, the second with.

Not bad, though it was much worse before I'd made the changes suggested by kalee and done a few other tweaks.

In my next post I'll put up the results of picking the signal up through a couple of sets. If people show interest I'll post up results for all the sets I can be bothered to drag into the workshop, and maybe see what improvements I can make by "tweaking the IF" of one of the better sets.

Also attached are a couple of pics of the mess I've made, first one shows most of the body of the circuit, the second the feedback circuit.

Hope this long post wasn't too boring!

Cheers

Dom

[dominicbeesley]

Heres a couple more oscilloscope shots. This time the top trace is the signal in from the CD player, the lower trace is taken from the actual speaker of the set

First from a Perdio Transistor set (not sure what the model is).
Second a pretty ropey Sobell set driving a random speaker (its own speaker is a bit rattley!)

As can be seen, both perform fairly similarly: falling off fairly quickly above 2kHz, with the 3dB point seemingly at approx 5kHz. The picture's not come out too well, but the Sobell set seems to do all sorts of weird things above 10kHz, I think it needs a bit of an overhaul....

I have great hopes for my P76F, which sounds excellent, not quite HiFi yet but not far off!

If anybody wants it, the WAV file used to generate the displays can be downloaded from http://brahms.demon.co.uk/wobbulator1.zip

Time for bed now

Dom

[kalee20]

Have you tried displaying trapezium figures on your 'scope? You put the modulating AF signal to the X-input, and the modulated RF to the Y input. Then the screen shows a vertical line (when the AF is zero) which boadens out into a trapezium as the AF is increased. At 100% modulation you get a triangle. If the modulation is non-linear, then the sides of the figure aren't straight.

I know that this won't prove linearity of the whole system (including the receiver), but it will help you to optimise your transmitter.

The reason I suggested full-wave detection in my sketch last night, is because it allows much less filtering to be used - which is something you've given attention to anyway. I'm a bit baffled now how clamping the input to the detector, with a Silicon diode, will work - I'd have thought it would just strip off the AM? But, if you reverse this diode, you'll have a voltage-doubler type detector, which may be better.

Dom, this is a really interesting thread, and I'm fascinated by what you are doing!

[dominicbeesley]

Hi kalee,

Thanks for your continued interest. I've just tried a trapzium trace (though I didn't know that that was what they were called!) Without NFB I get a pretty ropey figure a bit bent with "hoops" to the top and bottom (about 1/8 the height of the modulation in the middle). With NFB its pretty much spot on though I can't get 100% modulation then (due to my signal source not being strong enough and the reduced gain).

I think this non-linearity without feedback is due to the audio stages being a bit badly set up. I think the first triode stage would benefit from a bit of work, or perhaps a few more anode volts, I might even add in another audio stage for a bit of extra gain, but then I'll need to buy some more valve bases and its all getting a bit pricey unless I can find some cheap ones somewhere.

The trapezium trace is good for finding non-lineaties in the output but when I tried doing this through to the speaker of a set the (are they called lissajous?) plots didn't really tell me much other than that the radios are doing all sorts of things to the phase of the signal. I'll definitely be using them when I come to do some more tweaking next week though.

The wobble plots I did are a really quick and nice way of checking the bandwidth is not affected by component changes and gives a realy nice check on radios response, but doesn't work quite so well when looking at the RF directly.

The clamp diode just makes the waveform alway be referred to 0V maximum so that on the lower troughs are always doubly modulated (poor description I know). I'll definitely try you suggestion of an extra winding on the tank circuit for the feedback and an overwind for the aerial next week, just need to have a tidy up and find the right bits and pieces for coil winding...

This is great fun for me and I'm glad you're finding it interesting....though its taking up far too much of my time!

Dom

[kalee20]

For the diode, check the connections. I understand your description, but you need the clamp diode the other way round. As your circuit is drawn, they'll both conduct one one RF half cycle, and both block on the other.

If you reverse the wrong diode, you'll get positive feedback, not negative. It looks to me like the 'shunt' is the one to reverse.

As for valveholders, I can spare a few B9A ones for this project. They're not new unfortunately, but they are good, skirted, moulded ones. Send me a PM with your address and I'll stick them in the post for you.

[MichaelR]

Originally Posted by kalee20

For the diode, check the connections. I understand your description, but you need the clamp diode the other way round. As your circuit is drawn, they'll both conduct one one RF half cycle, and both block on the other.

If you reverse the wrong diode, you'll get positive feedback, not negative. It looks to me like the 'shunt' is the one to reverse.

.

You are right about that Kalee , the diodes should be just a full wave rectifier.Do you need a 33n isolating cap to the input grid. If you removed this you would also feedback the DC component along with the audio component. could this be utilised in the design of the front end to overcome load variations. If the average RF level decreases due to changing load conditions, the DC voltage feedback decreases also (goes more negative).

Mike

[kalee20]

I like Mike's idea, but it would shift the DC operating point of the modulator. Just think, if the load increased, the RF output would fall, so the DC output from the detector would also fall, and the modulator would feed out a higher DC voltage to try to re-establish the status quo.

That would then mean that the modulator has less 'upward ' swing available before the cathode follower bangs its head on the supply rail and clips the AF. So the degree of available modulation will be less.

[MichaelR]

Hi kalee,

If the preamp was decoupled with the AF fed to the mixer via a capacitor , the operating point of this mixer valve would not be affected by dc variations.

My thinking which may be wrong was that to have some form of auto gain control on the first valve to stop posible overmodulation if the loading of the rf section reduced the carrier.

Mike

[dominicbeesley]

Sorry, my mistake again with the diode, I've drawn it upside down, just goes to show what drawing a cct diagram at 1 in the morning does!

Thanks Kalee for the offer of valve bases! PM along in a moment.!

Hi Mike, I did try something a bit like that in my earlier experiments with using ECH42 valves and it gave a sort of AGC/compression system that caused no end of confusion until I realised what I'd done.

I'm not sure how well it would work with these valves as they've got a pretty sharp cut-off but I might give it a go, via some sort of integrating network and allow it to be switched off or varied for taste?

I think I've tracked down some of my bother now is due to the first audio stage not being class A - its going into grid current quite a bit, the distortion is masked by the feedback!

I think this week I'll get a better front end designed with another triode gain stage and a pentode as my final AF gain stage before the cathode follower - this should allow me to swing right down to nearly 0 volts without anything going into grid current, the higher amplification afforded by two triodes and a pentode should also allow me to use a nicer CD source and more feedback!

I'll also bite the bullet and sacrifice one of my expensive mains Tx's and wind up a proper decent 200V power supply for the who shebang.

Looking at my loadlines I'll now be able to swing right from 200V down to 12V or so, so if I bias my final RF stage by a little more than this I can get 100% modulation with no grid current anywhere (hopefully).

Once again sorry about this diode mix up - very embarassing!

Dom