RF signal path in this circuit...

[G0HZU_JMR]

In case anyone is interested to see the spectrum of a typical MW broadcast station I uploaded a quick youtube video of 1215kHz.

These days many people have SDR receivers that can do this stuff really well but the Tek analyser I have here is still quite fast and the camera and youtube don't do it justice. The display is really fluid when viewed in person.

The total display span is 20kHz and this means 2kHz/div so you can see the BW is about 12kHz and this means the AF BW is limited to 6kHz.

https://www.youtube.com/watch?v=RrPYT69Im6k

This analyser has some snazzy demodulation options fitted and so it can demodulate AM and display info like modulation depth vs time. I'll have a play with this and post up a couple of screenshots that show how the radio station uses compression/clipping to keep the modulation peaks close to 100% without actually hitting 100% and causing distortion/BW issues.

[Al (astral highway)]

Originally Posted by G0HZU_JMR

In case anyone is interested to see the spectrum of a typical MW broadcast station I uploaded a quick youtube video of 1215kHz.


The total display span is 20kHz and this means 2kHz/div so you can see the BW is about 12kHz and this means the AF BW is limited to 6kHz.

Thank you, this is really relevant and I find it very interesting.

[Al (astral highway)]

Originally Posted by G0HZU_JMR

If the Q was just 20 at 1.5MHz then the choke will look like an inductor with a 4150 ohm resistor in parallel at 1.5MHz. This resistance is getting significant wrt the target load resistance of about 1200 ohms at the modulator output.

Right! I get this, but how do you/ did you calculate to establish the value of the parallel damping resistor in the lumped circuit? What's the formula? I'm interested in getting to the heart of this and doing the same calc for my 220uH choke in this position and at 1 MHz. It's a high Q component, large, designed to handle 5A and on a ferrite core.

Thank you

Originally Posted by G0HZU_JMR

In other words, this choke would automatically add damping to the system and maybe 'too much' damping (if the loop is already over damped due to its own losses in the 8nF caps) and it would prevent the modulator from reaching full voltage swing at the output.

That all makes perfect sense. It's interesting with regards to the other circuits CW posts on the same page that only the magnetic loop antenna does not have the collector of the output transistor tuned with a tank circuit - the other two, of course, designed for use with short antennas and matching circuitry.

[Al (astral highway)]

Originally Posted by Argus25

.

One problem is the wide range of audio levels from mp3 source files. At the radio station it is all well controlled and the levels both within and between songs appear uniform, almost as though a human being is controlling the listening level.

To replicate this I started with the NE571 compander IC. Much to my surprise, even though it was taking the square root of the time averaged input signal the compression still wasn't adequate. It actually requires two in cascade, to take the 4th root, to come near the uniform listening level of a typical radio station.

Ah, now we're talking! I can see just how much is involved, now, beyond merely ensuring there's enough bandwidth! I find it fascinating and appreciate the time you've taken to make the technicalities clearer. Thanks again, Argus.

Originally Posted by Argus25

I have attached a circuit of the compressor.

Also I have attached the circuit of the driver of an AM transmitter I made.

These are great resources, thank you. There is obviously a great deal of subtle design here, far more than in the simple design posted by CW.
Of course it's now clear after others have picked this up that he really did intend his version for voice, rather than music. I hadn't clocked the scale of the challenge to get music right, but I definitely know now.

I will definitely work up a more music-suitable version of the pantry transmitter in due course.

But for now, I'll finish what I started and just see what range and quality I get from my newly-wound magnetic loop antenna - granted, with lower expectations now. I'll publish this just for the record in case anyone else is interested in working with CW's design, so people can be made more fully aware of its merits as well as its limitations.

[G0HZU_JMR]

I can offer some basic theory about the circuit operation if this helps. I've only looked at this casually and written the analysis fairly quickly so it may contain nuts...

Looking at the first diagram below... if you work out the voltage at the base of the AF (modulation) transistor TR3 it is 2.9V DC. This means the emitter will be at 2.2V and therefore the total current for the differential amplifier will be 2.2/180R = 12mA or 6mA for TR1 and 6mA for TR2 with no modulation at TR3 base.

The differential amp TR1 and TR2 is biased at about 6V at each base. If we ignore the 56p cap for now the voltage at the base of TR2 will be fixed at 6V. In order to get a full 12mA/0mA swing in current sharing between TR1 and TR2, the RF drive level into the base of the diff amp at TR1 needs to be approx 200mV pkpk. This is based on simple diff amp theory. It's OK for the drive level to exceed this, the only penalty will be the generation of high odd order harmonics. But the diff amp will 'alternate' 12mA/0mA peak between both TR1 and TR2 transistors at a 1.6MHz rate and TR2 will therefore act as a 12mA pkpk current source at 1.6MHz into the load/antenna when there is no modulation. See the first image.

If modulation is applied such that 100% modulation is achieved, TR3 needs to be fed enough audio drive such that the current source doubles to 24mA pkpk in order to achieve the required doubling of RF voltage at the modulation crest. So this now gives us enough information to work out the ideal load for the modulator output. This is the load presented by an ideal loop antenna at resonance.

The power supply is 24V and the voltage at the TR1 and TR2 emitters of the diff amp is about 5.3V. So to get the best possible voltage swing without hitting this 5.3V level at the low point of the waveform the collector needs to (optimally) swing +/-17V or 34Vpkpk when driven by the 24mApkpk current source output of the diffamp/modulator. So the ideal load resistance will be approx 34/0.024 = 1400 ohms. See the second image.

This means that the optimal load resistance looking into the capacitive tap of the loop will be about 1400 ohms. So the loop antenna has to be designed to present <= 1400 ohms to the modulator at resonance at 1.6MHz if you want to get the max swing from the modulator without distortion etc.

If we look at the the thick copper pipe used in the original design it will probably have a resistance of <0.03 ohm up at 1.6MHz. If the author used low loss caps in parallel to make up the capactance in the capacitive divider then the total ESR of the cap network is probably going to be <0.05 ohm. The reactance of the 2.5uH inductor will be about 25 ohms at 1.6MHz so the Q will be about 25/(0.03 + 0.05) = 312. So the bandwidth would be just 1600/312 = 5kHz. This is too narrow.

If you look at the notes in the article the author suggests that the Rp of the loop needs to be about 4000 ohms at resonance to get the optimal modulation swing. But the design so far has an Rp of 312 * 25 = 7800 ohms and this is too high. So he suggests adding a 10k resistor as a typical damping resistor to get an Rp close to his target of 4000 ohms. If 10k is put across 7800 ohms you get just under 4400 ohms and this is close to the design target.

The load resistance looking into the symmetrical capacitive tap will be a quarter of Rp and so it will be 4400/4 = 1100 ohms at 1.6MHz. This is close enough to the ideal 1400 ohms and it should be slightly less than the ideal anyway if you want to prevent flat peaks in the modulation due to overdrive.

The bandwidth will now be about 9kHz with the 10k resistor added and this will be just about enough for talk stations and the modulator will give close to the optimal swing at the output without adding too much distortion. So I think Charles' design is OK here as long as you use it for talk stations

[G0HZU_JMR]

Of course it's now clear after others have picked this up that he really did intend his version for voice, rather than music

One other thing to complicate the case for music is that a wider bandwidth means the possibility of more musical 'tones' (at different frequencies) appearing together at the same time. If they beat together in phase then they can build up a very sharp peak in the audio envelope. So this means it is much easier to hit the 100% peak modulation level with wider bandwidth.

So this is partly why it is wise to have some form of compromise on the bandwidth and also it underlines the need for some clever processing (clipping/compressing?) to prevent any ill effects from all this.

I watched the AM station on 1215kHz for some time on the Tek analyser with a modulation depth display and it always fell just short of 100% despite a huge variation in sound types. eg the presenter's voice, the sound of adverts and also several different music tracks.

However, that doesn't mean you 'have' to do all these clever tricks. I think some listeners will be happy with a basic pantry transmitter and AM signals from commercial stations suffer from propagation based distortion anyway... especially in the evenings. So most MW fans will be used to some degree of distortion or selective fading anyway.

[Al (astral highway)]

Originally Posted by G0HZU_JMR

So this now gives us enough information to work out the ideal load for the modulator output. This is the load presented by an ideal loop antenna at resonance...

Thanks, Jeremy, neat analysis. I'd seen differential amplifiers before but not in an RF/ modulation role, so great to have the dots joined up with your technical analysis there.

Now, I know I keep going on about it, but come on, what about the (specific, mathematically defined) role of that inductor in the collector of the output transistor!?

(I'm trying to model the effect of mine being 220uH, not 22uH to 100uH suggested by CW, AND, the additional result of it seeing 1MHz, not 1.6MHz.

This is so I can get a handle on the configuration I'm using, which deviates from CW's idealised version for the top of the AM band.

Cheers

[G0HZU_JMR]

The ideal inductor for that role would have a reasonably high Q and a high inductance so that is acts as a choke and it wouldn't change the frequency of the loop in any significant way and it wouldn't load the loop very much.

But in your case (at 1MHz) the 22uH choke has a reactance of just 138 ohm. If you had 2 x 20nF caps in the loop (rather than the 2 x 8nF that Charles used up at 1.6MHz) then the 22uH choke will effectively change the capacitance of one of the caps. The 20nF caps will each have a reactance of -8 ohms at 1MHz without the choke present.

Putting the 22uH choke across one of the caps will change the reactance of that cap to -8.4 ohms at 1MHz.

This means the capacitance will appear smaller (18.8nF) so the loop resonant frequency will go up slightly (up by 15kHz to 1.015MHz?) with the 22uH choke as opposed to a perfect/invisible choke. However, if you swapped across to the 100uH choke the change in capacitance away from 20nF would be less (19.7nF) and this would be resonant at 1.004MHz. With the 220uH choke the resonant frequency would be about 1.002MHz.

Note that the choke will also add a second resonance at the modulator output. Using a choke up at 220uH with the 20nF caps and the 2.5uH loop would add a resonance at just under 60kHz. Not sure if this is a good thing to have as this frequency is quite low... getting close to AF frequencies!

The parallel loading resistance of the 100uH (or the 220uH) choke will be much higher too meaning it will load the modulator less compared to a typical 22uH choke. Hope I've done the sums OK

[Al (astral highway)]

Great stuff, thanks. Now I recall how my calculations for the first loop I tried out here were complicated...

... by the fact that in order to stick to the 8nF caps designated in the capacitative voltage divider shown (and subsequently ordered by me), I designed two carefully precision hand-wound inductors to put in series with the big single turn loop, to get resonance bang on 1MHz instead of 1.6MHz. Fiddly stuff, but it worked. This is now of course of historic importance only.

Just one mystery left, I think. Any informed guess on the impedance presented to the collector of the AF transistor by the diff amplifier pair? I'd like to be able to model the effect on gain of changes in the value of the emitter resistor/ bypass combo...

[G0HZU_JMR]

The AF transistor (TR3 in my earlier schematics) is a current source so for AC waveforms you need to know the emitter voltage swing under modulation and also the parallel equivalent of the 180R resistor and the bypass RC network (typically = 19.6 ohms for 22R and 180R?). This will give you the swing of the current wrt modulation.

For example, if there is 0.2V pkpk voltage swing at the emitter with modulation then the 12mA current (set by the 2.2V DC voltage at the emitter and the 180R resistor) should change by 0.2/19.6 = 10mA pkpk or 5mA either side of 12mA = 7mA to 17mA in absolute terms.

So to get +/-12mA on top of the 12mA standing current (to get 0 to 24mA for 100% modulation) there needs to be 0.47V pkpk at the emitter of TR3 with the standard circuit.

The impedance of the diff amp doesn't affect the performance of the current source so I don't think you need worry about this.

[G0HZU_JMR]

In case any readers are still a bit unclear how to view this circuit, then the easiest way to visualise it all is if you think of the output transistor TR2 as being a current 'gun' that tries to fire 12mA pkpk (at 1.6MHz) into whatever load you present to it as an antenna when there is no modulation.

If you add enough modulation to double the 12mA to 24mA (to achieve max/100% modulation swing) then TR2 becomes a 24mA current gun on modulation peaks.

Case 1:
If the antenna load was a mere 500 ohms at the capacitive tap this would mean 0.024 * 500 = 12Vpkpk at the tap on mod peaks or 24V pkpk at the loop itself (because the voltage gets doubled by the capacitive tap)

24Vpkpk is well short of the optimal 68V pkpk that could be delivered across the loop. The load could easily sag to 500 ohms if the loop was built with lossy class 2 (X7R) ceramic caps or certain poly caps.

Case 2:
If the antenna load was 1200 ohms at the capacitive tap this would mean 0.024 * 1200 = 29Vpkpk at the tap on mod peaks or 58V pkpk at the loop itself (because the voltage gets doubled by the capacitive tap)

This is very close to the optimal 68V pkpk that could be delivered across the loop. This would give close to optimal performance.

Case 3:
If the antenna load was up as high as 2000 ohms at the capacitive tap this would mean 0.024 * 2000 = 48Vpkpk at the tap on mod peaks in theory. However, the system can't deliver this much swing because of the supply and biasing limitations so the modulation envelope would show signs of flattening and this would cause distortion products.

So think of the modulator as a current gun and this will perform best into a 1200-1400R load as this will give the best swing without distortion with a 24V supply.

[Argus25]

Astral Highway,

One point about the modulator transistor. The linearity of this circuit is better as the size of the emitter resistance (to audio frequencies) is higher. The transistor's base voltage is more faithfully converted to an emitter (and collector) current which exactly mirrors the input audio voltage at the base divided by the emitter resistance, because the transistor has a finite hfe. As this resistance is smaller the transistor's non linearity is more significant. Another way to look at it is that the input impedance of the modulator transistor is higher with a higher emitter resistance and there is more degeneration and less distortion. The 180R sets the DC conditions, but its also a reasonable value for low distortion from the AC perspective.

Obviously with just the one modulator transistor they were short of gain for the audio level they were using and boosted it up with the 220uF and 22R bypassing the 180R emitter resistance at audio frequencies.

My guess is that if the 220uF cap and 22R resistor were removed, and instead an additional stage of audio amplification used to recover the modulation gain, the modulator linearity would be better, with less distortion, perhaps more suited to music.

[G0HZU_JMR]

I'm not sure what levels of distortion are targeted here and I guess only AH's ears matter in this respect. But I think this circuit can deliver about 5% distortion based on a few initial simulations. To get much better than this will probably mean a complete design change.

I'm in the process of making a loop antenna for Charles' circuit and I'm going to see how far it will work. I'm going to test every part in the antenna on a VNA and also measure the final antenna on a VNA before trying it for real. I've got an old homebrew 14MHz loop antenna in the shed which is about the right size but I don't want to mess with it. So the only other material I have is some 2mm OD copper wire. It's just over 2m long an I don't want to trim it so I'm making a crude loop from this antenna. It should have an inductance around 2.5uH and an ESR somewhere around 0.1 ohm at 1.5MHz in theory. So it is already quite lossy and I expect to have to increase the current from the modulator slightly to get max performance in terms of voltage swing at the loop.

I've got various capacitor types here and they are all going to be quite lossy. The best I've found so far are some old Suflex film caps. These are really old and they are a bit marginal on voltage rating but they will be quite stable and should be OK for a quick trial etc.

[G0HZU_JMR]

For the range trial, I was just going to make the TR2 section of the modulator and feed it with AM at 1.5MHz but I've actually made the basic modulator (less the 56p and 27p caps) and done a few tests

I guess for now most people will be interested in the main stats... how far did it transmit and how much bandwidth and how much distortion? The loop used just over 2m of copper wire in a loop (2.5uH) and I optimised the modulator to deliver the full swing to the loop from a 24V supply. The transmit frequency was 1.49MHz.

I used my little old RGD Flirt MW receiver that my dad gave me when I was a boy and the loop was upstairs in the front of the house. I could still hear the signal quite strongly in the shed at the bottom of the back garden on the little Flirt radio. So probably about 60 feet in range from this single turn loop made from copper wire? It would go a bit further but I ran out of garden

The bandwidth was about 10.5kHz so the AF BW was half of this. I used a Racal 9008M to demodulate the AM with a 1kHz test tone and just under 90% modulation. The demodulated AF output of the Racal was fed to an analyser and this showed the harmonics were -35dBc. So just under 2% distortion. This is pretty good and this initial lashup doesn't have the 56p and 27p caps fitted yet. I'll try them another day.

See below for the test setup. You can see the top of the loop hanging off the side of the wooden bench. The Tek 465 scope in the background is connected to the capacitive tap via a x10 probe and is effectively displaying 5V/div. So there is just over 30Vpkpk here at the capacitive tap feed and therefore just over 60Vpkpk at the loop itself.

[Al (astral highway)]

Hey Jeremy, great stuff. Should be interesting to see what comes out of your network analyser!

Shame you can't use 15mm tube as CW does. If you're going to use 2mm wire (I used 3mm), it might be a good opportunity to try the magnetic loop at the same frequency I'm using - 1 MHz, with appropriate adjustments in the capacitative divider and inductance. (I used 2x10nF and made up a 0.5uH inductor from two precision hand-wound coils. I put these in series with the antenna, which in my case was made of a hexagonal loop 2.02 metres in circumference. Total inductance: 2.5uH, resonant frequency bang on 1MHz.)

Otherwise you're changing one variable -- but CW's circuit and mine and yours will each only have one variable in common.

I also used Suflex caps on the magnetic loop, rated to 350V. They were also old and quite fragile.

How will you increase the current from the modulator? And what power in will you be using?

Cheers

[Argus25]

I think you may be disappointed with the range here, with a single turn loop, with drive powers less than 500mW average. I expect that you will get a reasonable result in the very near field to the antenna, say 5 or 6 meters away in the same room, but it will be very poor say if the loop is in your house and you take the radio out to the back yard or are more than 15 to 30m away.

You can compensate with more power drive to the loop, though for a single turn loop (unless large size) even with 1 watt of average power (4w pep), a single turn was inadequate to cover the whole house and fairly small back yard.

My experiments indicated that a single turn 1m squared area loop was about the same as a whip antenna (which was matched in with a loading coil) about the size of a long car radio antenna around 2.5m.

For something like a 100mW legal range area pantry transmitter, you will require a multi-turn loop for good results, in my opinion. But of course it is good to do the experiments yourself to be 100% confident in what you will need.

Remember for the same sized loop, doubling the turns increases the radiation resistance by a factor of 4. If you have a 5 turn loop its 25 times better than a single turn loop for the same sized loop.

If you increase the loop's linear geometry by a factor of 1.4, you double its cross sectional area. If you double the area, the radiation resistance increases by a factor of 4. So a relatively small change in the loop's linear dimensions has a big affect on its performance because the radiation resistance turns out to be related to the 4th power of the loop's linear geometry.

Likewise if you run the transmitter at 1400Khz its much better than 700kHz because the radiation resistance for the loop has a 4th power relation with the frequency. So a loop of the same size is 16 times better at 1400KHz than 700kHz.

[Al (astral highway)]

I think you may be disappointed with the range here, with a single turn loop, with drive powers less than 500mW average. I expect that you will get a reasonable result in the very near field to the antenna, say 5 or 6 meters away in the same room...

Remarkable cross-post there! So where are we now ?

Well,Argus25, Jeremy's experiment yielded good results at 1.5MHz and with skinny wire. Mine did not. Nobody has yet replicated a trial at 1MHz...

Jeremy, did you use an iPhone as the audio source to mimic my trial ? Otherwise what did you use and how many mV P-P was the audio source ?

News from here : , I have now tested my *new* loop with 20 turns x 1.5mm over the same hexagonal 2.02m circumference snooker triangle former. It is tuned to resonance at 1MHz by a parallel 220pF variable cap. For all the reasons Argus25 has explained, this ought to perform rather well when compared to the initial one I described way back at the start of this thread.

However, the range is now approximately just 1 metre!

All is completely well with the transmitter itself, wonderful output observable on the 'scope. Voltage is just over 50V peak to peak at the unloaded collector as I reduced the emitter resistor in the current source transistor to 150R. There is no emitter bypass now. The capacitor subject of controversy in the output transistor's base is now omitted.


I have to suspect the construction of my flat, which is in essence a steel cuboid with depth 3metres and width 7 metres between steels, replicated over and over in the development. This is the only credible explanation - a 1/10 wavelength and approx 1/20 wavelength Faraday cage at 1MHz!!

I suppose you could cheat and use the loop as a 'bit of wire in the air' instead, it would still look pretty.

[Al (astral highway)]

To put this in context , there is a maximum vertical displacement (aka, 'gap', 'space' ) of 1.5 metres between steels , throughout the whole flat .There is a displacement of just 50cm when the antenna is placed in the most convenient place. There is one window with a reasonably clear line of sight but most windows overlook more steel buildings and or balconies!!

[Al (astral highway)]

Originally Posted by merlinmaxwell

I suppose you could cheat and use the loop as a 'bit of wire in the air' instead...

Can you explain please, MM?