Why is this 5MHz TX circuit so much worse in real life ?

[MathWizard]

I designed a ~3Mhz 2 BJT, receiver and demodulator, LTSpice agreed it should have peak gain at that frequency. But in real life on a proto-board, it's peak gain was about 4.85MHz. I'm guessing it's mostly the BJT's that are different in real life, some of it could be the caps/ and inductor too.

So for a TX, I started with a low gain 4.9MHz Colpitt's oscillator, and then noticed all the FM radio stations were a lot closer to my signal strength then I wanted. I could swap or change the Colpitts caps, and get much higher gain, but I haven't.

Instead I added some RC and LC filters, between some more amplifier and buffer stages, and it works well enough in LTSpice.

But in the real proto-board circuit, the AM and FM radio stations, got stronger, and no were near as weak as in the sim by the end.

The dBVrms levels in sim and real circuit match pretty well, until the Q4 buffer emitter. The sim has about
-100dBV for 800kHz local AM
-44dBV at 4.9MHz
-165dBV at 100MHz local FM radio

The real circuit has -46dBV for 4.9MHz, but the AM and FM stations are only about 7-15dB less at Q4 emitter. So I'll have to un-solder some stuff, simply swapping the upper and lower cap in the Osc, will get me much more more gain, but I wanted a clean signal, not a strong signal, at this point.


Is this a lack of shielding over the circuit, like the green through style inductors ? Or the BJTs, Or the layout ? It's 5 BJT's , all pretty neatly packed, more or less like on a beardboard, with V+ and V- parallel and directly adjacent to each other.

The image order should be sim C3 /E4 C5, for collector and emitter, and the the real ones

[MathWizard]

This is just the RF stage of a TX. I put it in a tuna can, with the cover, and that made no difference I could see. Then I swapped the tank circuit caps and get a few times more output voltage.

But I found the real problem, the coupling cap into the base of Q5, was poorly soldered on, and broke loose tonight, so I had next to no signal there.

So now it actually is doing want I wanted, which is removing most the AM and FM radio from the RF osc. signal. Still nowhere as much as in the sim, but now there's ~40dB between 4.9MHz and the 800kHz local station.

Looks way better too, doesn't look all ropey.

[RFDx]

So now it actually is doing want I wanted, which is removing most the AM and FM radio from the RF osc. signal. Still nowhere as much as in the sim, but now there's ~40dB between 4.9MHz and the 800kHz local station.

The oscillator obviously can't produce any strange and off frequency AM/FM radio signals. You are picking up signals from local/strong radio stations because of the the long ground wire of your probe which then naturally show up in the FFT.

[mikerj]

The passive low pass filter connected directly to the oscillator tank will be loading it and hurting the Q (and shifting the frequency).  Better to buffer the oscillator output first before filtering.

[MathWizard]

I'll going to start over from the beginning, with a stronger signal, I didn't try to calculate any loading or the actual Colpitt's circuit, I just chose the tank circuit by choosing 2 caps way bigger than the Cu and Cpi. Overall, the real life output might not be much better than I started with, and adding some tunablity would be nice. 

Trying some calculations on the Colpitt's, my AC model should be resonant at 4.1MHz, which is the same as the AC model. But I designed it for 4.9MHz, which is what the BJT sim, and real life thing work at. So IDK what's going on there.


I found out 2 of my probes don't work as good as they should, or they are lower BW than some of my others. But I have 4 new probes, and they all have ground springs, so I'll have to try them.

[MathWizard]

For the colpitts oscillator, I worked out in detail, the open circuit transfer functions of the amp, and the feedback/tank circuit. And separately, the calculations match very well with the AC models, and the BJT sim. I plotted all it separately in a spreadsheet.

But for the total open loop gain A*B, all in the s-domain, so just the sum of all the magnitudes, I get roughly 27.8dBV, at 4.1MHz. But the AC model has -0.53dB, and the BJT sim has 9.4dBV, for the open loop gain.

What am I missing now ?

For the amp I have

Av=((ro||Rc) / (ro||rm)) * (1/ (s(ro||Rc)*Cu + 1)

For the feedback/tank I have

Bv=( (C2/(C2+C3) ) * (sR3*C1) * (sLG3+1) ) / (s^2LCd + sR3Cd +1)

where Cd=C1+(C2||C3) and R3 is series resistance of the inductor

So if they seem spot on separately, why isn't Av*Bv giving me the open loop gain ? And why does the AC model differ by 9.5dB from the full open loop BJT circuit ?

 CB colpitts.asc (4.61 kB - downloaded 13 times.)

[MathWizard]

Ok the model was pretty much matching the BJT sim, but I wasn't comparing the same nodes, and 1 cap in the wrong place for the calculations.

But anyways, having found both transfer functions, in more detail than needed, or anyone probably cares
 
R1=rpi||RE
rx=ro||rm
r2=ro||Rc
r3=rx||R1
r4^2=ro*rx-r2*r3
Cd=C1+Ct
C5=Cpi+C4
Ct=C2||C3
tau1^2=(ro*rx*r2*r3*Cu*C5 / r4^2)
tau2=ro*rx (r3*C5+r2*Cu)

Av = s(ro*r2*r3*C4 / r4^2 ) * (1 / ( s^2(tau1^2) + s*tau2 +1) )

And

Hv = s(R3*Ct) * (sLG3 + 1) / (s^2(L*Cd) +s(R3*Cd) + 1)

And separately their magnitudes and peak frequencies match great out to 1GHz were Av goes a bit different.

But just multiplying Av*Hs in the s-domain, and adding their magnitudes for Gv, it's not matching very well at all, on paper, but the AC model and BJT sim match fine.

And I just saw a definition, saying you just multiply cascade/series transfer functions in the s-domain, with no mention of convolution, having to take inverse Laplace transforms , or anything for finding Gv(s)

why aren't the pic's in order I entered them ??

[MathWizard]

Ok I have to work out both sets of ABCD parameters, then use matrix multiplication, then I'd get the total cascade ABCD param's.

[MathWizard]

Well I finally got a decent version working at 5.23MHz. I went down a few rabbit-holes trying L-matching with the BJT's on protoboard and making the coupling on the breadboard. And plus with the audio on a separate protoboard, and a bunch of wires going everywhere, it never really worked once it was all back on the protoboards.

In the end I ignored most my calculations and LTSpice zin, zout's, and just went with what worked as the coupling. So some of it looks bad in LTSPice's Bode plot's, but not as wacky as some of the LC-matching. For my TX antenna I just have almost a meter of wire. And so far what seems to increase the signal on my receiver was just adding a tank circuit on the TX. I'm not happy with my output buffer either, but for now...

The TX has a 5k pot after the oscillator buffer, with that I can change the diffamp output level quite a bit, especially if I put a variable cap between the RFamp and Diffamp. And the audio mixing is fully adjustable too, so I can tweak the mixing levels. The audio section could be more centered on 1kHz, not 10kHz, but it's ok for now.


Then I built an 88uH inductor, 30mm diameter and only 16mm wide. And with some L-match that works, just into 1 RF amp I get a -20dBV AM signal to work with, I just need to dump some more local AM from my signal. And I should be able to recover the audio.

Last year I made a 1.5MHz version that my radio could pick up, it wasn't very good. I hope this sounds ok. But it looks way better than last years version.


The TX is only about 2ft from the RX, I IDK what range this has but it's probably plenty.

The goal is to broadcast what my PC is playing, to the next room or basement. I might try making a BJT volume compressor circuit for the TX audio in.