Why is this AM TX circuit so different than the sim ?

[MathWizard]

On a breadboard, I made the circuit below, with a ~30cm wire attached.

For now just using a signal gen, I used 50mVpp RF, and could apply 1.2Vpp AF, to get a fully modulated signal out.

I had about 30mVpp AM output, about 600mVpp with an added CE amp, and the radio can pick it up ~15ft away, in the kitchen where I want it. The real circuit works best at 1.3MHz, but the radio is better set at 1.15MHz.

I used BC547B's with a measured gain ~205. In LTSpice, using BC547B w/300beta, or 2N2222's with only 200Beta, and similar Is, the circuit can't even take 400mVpp AF, without looking bad, and it doesn't look well mixed to start with.

But the real circuit looks great. So why is LTSpice so different at this, is it the real BJT's are so different from the model's, and how much difference does the breadboard make ?



When the radio tunes onto a local station, most the background noise goes away. But with my signal, the 1kHz tone volume was ok, but the noise was there at about the same level, if not a bit higher. So is the AM TX still not high enough, or is a bunch of noise get mixed into my TX output aswell ?

Using a electret mic, and a bad 2 stage amp, I could hear my computer speakers, from the radio in the kitchen, but not clearly. So now to remake that part and add some bandpass filter to the input aswell I guess. And try something besides a 1N4001.



the red is CC stage and the yellow is the emitter of the diff pair

[807]

First thing I would try is increasing the length of the aerial. 30cm is way too short. Try a couple of metres or more. The noise on the radio is almost certainly due to the low signal strenth.

I'm not sure what that 1N4001 is doing in the circuit. What happens if you replace it with a link?

I can't explain the LTSpice naff output, but what matters is the actual real world results & the modulation looks excellent.

[LM21]

Yes, the diode may even be harmfull there. If it is fast, it may form a peak hold circuit. L1 and C6 form a resonant circuit, what is its resonant frequency. Do you really need it. So, connect C5 to C7, and remove all parts between them. You don't need both C5 and C7.
You may get a little more power, if you use a smaller R9, but the antenna has more effect, I think.
If you are worried about harmonics and other unwanted frequencies, you can use lowpass filter, not resonant circuit like here, L1 and C6. But I don't think this is your biggest or first problem.

[MathWizard]

Ok I never tried a longer wire or higher Vcc either. I tried a Germanium 1N60, and a 1N4218, and I wasn't impressed, it look distorted, but not with the 1N4001.

I read the diode was there to only provide 1/2 wave pulses to the tank circuit, and that the inductor will create the negative 1/2 wave. I'll try without too.

Yeah the tank circuit acts as a bandpass filter, if nothing else. I had already made a Colpits osc. for the RF signal, I need to use it again, and not the sig.gen. My circuit had peak magnitude at ~1.3MHz. Later when I make a receiver, and get that working, I'd go to a higher more obscure frequency.

Yes I need C5 and C7, or else something would be shorted by that inductor

Using a single CE stage for the mic, (not the tail BJT) 15mVpp in, gives full saturation w/ 29mVpp o.c. from the tank. And if I whistle, I can easily over modulate the signal. So it is sensitive enough.

When I played a 1kHz tone on my PC, and TXed it too the radio, I can make that out. But dialog, I can't really make out at all. I think the CE stage I added on the output, makes the output look a lot more jittery.


 

Now I don't plan to use the mic anyways, I'm going to try with opamps, to buffer and mix the L/R and centre channels, from the stereo input, or output. I could run a wire and speaker to the kitchen too, but that's not as fun.


So at 1.15MHz, that's 260m wavelength. I have tons of wire, I'll run it along a wall.

[radiogeek381]

I can't explain the LTSpice naff output, but what matters is the actual real world results & the modulation looks excellent.

With respect, the modulation is awful.  You are over modulating quite a bit. to the point where the sine wave is crossing over (inverting the phase) of the carrier.

You can see it in the nearly flat tops between peaks.  See the little bump?  That's from the "minimum" value of the modulating signal. 

Back off on the modulation and you'll see an improved tone, at least.

[MathWizard]

So that's also too modulated, ok. Yeah the tone is pretty muffled, but it's also way too garbled, last I tried. Since I'm using the sig.gen, I'll try the modulation equation.

[MathWizard]

OK so with 20ft or so of wire, and 1.2Vpp on a CE output, I can pick it up no problem, the radio locks onto the RF and the noise pretty much goes away. And now with an mp3 player speaker 1" from the mic, I can listen to Die Hard 1 in the kitchen without much problem. It's still not perfect, but yeah it's more a tone thing than anything else when listening to voices.


Now I have a few little laptop speakers, so if I was going to tap off my PC/stereo, I could make a TX/RX for L/R and center speakers. Some day I'll try FM.


That diode would help in circuits that don't have nice sine waves in, like some circuit stage that gets saturated a lot and look more like a bunch of pulses, or square-waves. It will then create the the 1/2 cycle, it's from old circuits.


Here's where the police lady says something about an FCC violation, in Die Hard 1


I guess the audio is still too strong, I added a resistor divider on the mic, to turn it down a bit.

Here's some gunshots in the movie, fairly loud

[807]

I can't explain the LTSpice naff output, but what matters is the actual real world results & the modulation looks excellent.

With respect, the modulation is awful.  You are over modulating quite a bit. to the point where the sine wave is crossing over (inverting the phase) of the carrier.

You can see it in the nearly flat tops between peaks.  See the little bump?  That's from the "minimum" value of the modulating signal. 

Back off on the modulation and you'll see an improved tone, at least.

Perhaps "excellent" was a bit generous, as the positive peaks are a bit rounded, but I wouldn't call it awful. I can't see where it's showing "quite a bit" of overmodulation either. Surely phase inversion only occurs when the negative peaks go below zero, which I can't see in that scope capture.

I simulated the same conditions with my sig gen. Looks like a mod depth of around 88%.

I've attached scope pics at 88%, 100%, & 110%. You can clearly see the phase inversion at 110%.

[807]

...I guess the audio is still too strong, I added a resistor divider on the mic, to turn it down a bit.

Here's some gunshots in the movie, fairly loud
(Attachment Link)

Yes, that 2nd attachment certainly shows overmodulation. You can see the extended "flat spots" quite clearly.

[MathWizard]

So what sort of power level, should I be expecting at the input to some wire or ferrite coil antenna and tank circuit ?

[807]

If you search for the Wenzel modulation method, you will find some info on this circuit.

http://techlib.com/electronics/amxmit.htm

https://www.geojohn.org/Radios/MyRadios/AMXmitr/AMXmtr.html

http://techlib.com/electronics/amxmit/transmitterpaper.htm

Many years ago I bought a little AM TX called the Gizmo. Can't remember if I did any scope testing with it, but that used the differential amp (long tailed pair) modulation as well.

http://www.geocities.ws/raiu_harrison/mwa/tech/circuits/gizmo.html

[807]

...I guess the audio is still too strong, I added a resistor divider on the mic, to turn it down a bit.

Here's some gunshots in the movie, fairly loud
(Attachment Link)

Yes, that 2nd attachment certainly shows overmodulation. You can see the extended "flat spots" quite clearly.

I also notice that the modulation isnt symmetrical. Ideally the unmodulated RF should double it's amplitude on positive peaks & reduce to zero on negative peaks at 100% modulation. So a 1v p-p unmodulated signal should vary between 2v p-p & zero at 100% modulation. Your modulation is mostly negative.

[MathWizard]

Ok I have to learn up on that stuff too. I just got a Signals and Systems textbook too, there's a bunch of basics in their I should do ASAP.

Yeah after measuring the beta of some BJT's I was using, no wonder the sim was so different, and now with all that measured and some higher Beta BJT's, it's looking good.

I worked out the equations for the ABCD parameters of the mic-CE amp. And for the diff.pair+tail current BJT, as a whole, or maybe just the A,C of that.

And today I made up a common collector Colpitt's oscillator, and worked out it's ABCD parameters in the s-domain, of the BJT part, and then the feedback tank. Without choosing C's. But I never worked out the total yet, huge equations already just for that. But I also did the ABCD-param's of the whole C.P. osc, if the loop was open. If I can pull off closing the loop, with those eqn's, I'll be happy.

IDK if I can get all that to really work right yet. Or working out the inverse Laplace transforms, but I can do the 2ndO.D.E's for the LC part, but I never tried with a BJT added in before. I was reducing the equations down to a few core constants and RE and IE and beta. Hoping to see what makes for a good or bad circuit. I just chose an A-param of the amp, RE, the BJT, Rbb, and it all seemed good enough, for a lot of C1/C2.

For instance, A of the CC amp is A = 1 + beta*Vt/((beta+1 )*Va) +Vt/(RE*IE)

Mine is A=1.02450V/V on paper, and very close if I make the same AC model in LTSpcie, and pretty close to the BJT, and the real circuit.

But anyways the CP osc. circuit works no problem on the breadboard, with a wide range of C1 and C2 combo's. In LTspice, it works with C1/C2 or C2/C1, so I'm not sure yet why that is from the eqn's. Once I settle on a better frequency, I'll trying to decrease the gain by splitting the emitter resistor, and make a nicer sine wave with less current spikes from C1 and less vbe I guess, flattening out the osc. when it's over powered.

With it applied to the mixer circuit and it's LC tank, I get some extra noise frequencies that make it fuzzy and look worse. Higher and lower freq's. IDK how much 60Hz get into all this as well, maybe that's my fault.

I haven't measured my LC parts on the LCR meter yet, nothing has been at the calculated freq's for either tank circuit alone, on the breadboard. Soon I'll solder up some of it.


Here's osc' output, w/ just a 2.2k load. It's a bit fuzzy, and tilted to the right. I barely played with the cap values on the BB. Here's basically the circuit in LTspice, but on the BB it more like 2MHz, not 1.4MHz. Once I measure the parts, then I'd worry about trying to calculate the difference. When I make it on a copper clad board, I hope it's a lot closer to basic calculation.

And then the yellow osc. connected to the mixer and other LC tank and 2 other BJT stages, the red trace.

See all the higher frequency ringing.



Ok the osc will work fine at about the correct freq, it was the other tank. So now matching it better to the 1st, and just passing the osc signal through a 15p bypass cap, the mixer/tank output looks a lot better.

[radiogeek381]

Re: my early quote about bad modulation, I was reacting to the yellow trace from January 23, 2023, 12:49:01

The latter traces look better, but the gunshot stuff still over modulates a bit. 

As a matter of taste and good neighborliness, AM modulation should be kept to less than 80%.  It will result in cleaner sidebands as you'll avoid the sudden on-off transitions.  There's a reason for that extra "1" in the term (1 + A(t)) sin(2 pi f t)  |A(t)| < 1  .  There are ways to push the limit, but they really aren't necessary here.

Yes there are ways to recover signals beyond 100% modulation, but why bother.

[807]

...There are ways to push the limit, but they really aren't necessary here...

Yep. For a small tx that only transmits a few mW, then a basic design is OK.

Once you get addicted to AM (like me), and maybe thinking of a higher power tx, you start to look at ways to improve your "loudness" on the band. Trying out things like asymmetrical modulation to get over 100% without clipping, compressing the audio to get the most out of the modulating signal etc.

[MathWizard]

Well what I built was not quite powerful enough, not for how I was adjusting it anyways. So this time I chose 20mWdc in the output resistor. Then worked back and tried to match the power between output stages better. I should do it more for the rest too...but I'll have to do it all again for considering time constants. It's hard enough getting the DC to match the AC with out caps. And in the end, I still just went with closest common values to my calc's, for the most part.

And now it's working better than ever, and without even needing some bypass and feedback caps like yesterday, today the signal output is way way cleaner.

 The audio is not bad, a bit muffled right where the noise is the least when tuning.

I added a 100k pot in 3 places, to tune the RF, AF and overall mixer output levels, into the overpowered amp stage. It needs more fine tuning, but until I make in on a copper clad board, it's great.

Never mind that the mix levels in the schematic, but those are all the parts/values


If any ones tries it, my BJT tester said my BC547B had a beta of 431. I measured 330, with my DMM, at ~10mA =Ic
Then used Is=1.96fA, and Va=112V of a BC547B, from LTspice, and found that if I used n=0.848850, it matches my VBE and IB pretty well. But in LTSPice, a BC547C seems to match the current of my real BC547B's

I need to make a real BJT tester jig, DC and AC, I never tried to measure their AC properties closely anyways.

[MathWizard]

Yeah that was the wrong version I attached yesterday, here's the working schematic, and my layout. I built almost all of it on a proto-board, then realized I forget 2 resistors and a trimmer, so I had to squeeze them in. IDK how bad my layout is. I tried to keep the audio on 1 side, the osc. and the other side, the output above that, and the mixer in the middle.

Now to let my proto-board dry and see if it works. Maybe add a few 100n caps, I forgot them. And before I try making another better mini-TX, I'll try designing a receiver.

[MathWizard]

Well I did make the tanks wrong by accident like that. I know they aren't like that at AC. But the added tank right on the emitter side of the osc, was working for ages on the breadboard with all sorts of cap combo's. It also works pretty well in LTSpice.

But on the proto-board, it didn't want to work. So after a day wasted, I removed that. And still, compared to the breadboard, I was having a lot harder time getting a stable, usable oscillation, with many cap values I tried. In the end I'm using a smaller emitter resistor.

I made a spreadsheet calculator, for doing all the phasor math. And did find the ABCD params, or the amp and the tank. And with common values, my gain of the tank w/ 682R output load, is still 9 and under 1deg phase.

And the amp's gain is now closer to 98%. But still, I'm having a hard time finding sweet spot without getting pops and wobbles on the radio station.

I haven't tried a model of BJT w/ capacitance. I'm guessing the Miller effect of a CC amp does not magnify the Cbe. And I read it's just a few pF and the external Ce to gnd should just be way bigger.