I am trying to generate an AM-signal as it would be input on my tuner if I had an antenna present instead. This, because I want to try out AM reception on an experimental board with RF input, but I have very weak AM signal reception in my current location. Looking at RF-generators on the market, the prices are not in the region of what I am prepared to pay. So, I just thought I'd ask you guys if it is possible to "roll your own" from discrete components if, as in this example you only want AM, you go for fixed modulation frequency and perhaps other simplifications?
I have bought this circuit
and successfully tried it out, but the outputs are not on RF level in terms of voltage and output impedance, I assume.
What is your budget? You can easily spend anywhere between 3 and 5 digits on RF generators... or drop down to 2 digits for something used on eBay...
You have at least a thousand times more power than you need in that circuit to drive an AM radio. :-) Don't worry about the impedance mismatch -- at the frequencies you're talking about a short piece of cable appears very "lumped" and you you just have a generation with a relatively small impedance looking at an input section of a radio that actually will tend to have a rather ill-defined input impedance, but hopefully it's somewhere in the ballpark of
Here's a cheap RF generator that does AM modulation:
-- I wouldn't be surprised if it uses a circuit similar to that XR-2206 that you posted; the unit does work, but even on an oscilloscope the sine wave looks visibly imperfect... on a spectrum analyzer it's really scary; often this doesn't matter too much for testing AM receivers, however.
(Off-topic... but if you are looking for pure sine waves, ironically enough those old HP generators that used a light bulb for their amplitude limiting mechanism tend to be noticeably more pure than even $15,000 DDS-based generators!)
No problem. With the bias network shown, and betas from 100 to 358, the collector voltage on the 2N4401 oscillator is a couple of volts or less. It won't sustain oscillation, so the power output will be zero.
Better to connect the collector directly to +5V and use the bias network shown in
Since there is no trim capacitor in series with the crystal in this circuit, the emitter resistors will need to be trimmed for the desired signal level into the differential amplifier. However, there is no longer any need to wait for thousands of cycles or more for the oscillator to settle to steady-state. Just use the .IC command as shown to set the initial current in the inductor, and use the Peak Search function to check if the oscillation amplitude is increasing or decreasing. Trim as desired.
There is little fear of exceeding legal limits on the power output with 15 volts to a 2N4401. Have a ball.
The XR2206 is Ok as you can adjust it all the way from VLF to the medium waves. All you need is to attenuate the output voltage and reduce it's output impedance. Do this by adding a couple of components as in the diagram. (To view it, use a 'fixed width' font) Adjustment of the 1k potentiometer will give a radio signal from near nothing to that that of fairly weak radio station. Would suggest though that you enclose the XR2206 and all the components in a tin box (connected to 0V), as otherwise it is easy for the XR2206 high level voltages and currents to leak through the air and straight into the radio. .------. | | | | ||10n ___ 22k |Pin 2 o----||----|___|---, | | || | | | | | | | | | .-. ___ | | | |
Thanks. Now if I could just get FB to look at it and tell me where the problems are, my life would be complete. Anyway, I think I may have stumbled across a problem with JL's optimization of the phase slope through zero. Several problems, in fact. And this leads to the conclusion that the Mercury equations in the paper are incorrect. Although stolen from numerous sources on the web, they fail to include the crystal loading capacitance and its effect on the series resonance. So if you try to plot the phase intercept throuh zero, it will occur higher than the calculated resonance frequency. And there is no way to bring it down to the correct frequency.
But we all know that already. Crystal manufacturers specify the frequency in series or parallel mode (parallel as in Wenzel's Colpitts, series as in the Clapp oscillator in my paper). So if a crystal is manufactured for one mode, and you put it in an oscillator that uses the opposite mode, it won't oscillate at the correct frequency.
A fact well-known by everyone, now confirmed with precision frequency analysis in Microcap SPICE.
So on to the next topic of frequency pulling, which as far as I know has never been run in SPICE due to the long startup and settling time. This is now trivial since the new method is roughly 2,500 times faster (please see the update in the paper if you haven't seen it - Estimating Startup Time.)
Then do Pierce, Colpitts, Butler, and overtone versions, spurious reponse,
74AC inverters, Q-multipliers, then maybe some transient analysis of narrowband crystal filters.
I went for John Jardine's setup and now I can encode and decode AM on my experimental board!
I made some tests of the bandwidth, and came to the conclusion that there seems to be a bandwidth of 5-6kHz on the receiving side. Measuring at the modulated output from the 2206, there is no problem with transmitting 12kHz, the modulated signal looks perfect.
I have, maybe naively, thought of maybe transmitting DRM at a later stage by generating a DRM signal with Spark software
and sending it with this device. But, considering the spectra for DRM default from this software is between
4.5kHz and 20kHz, this seems impossible. Any thoughts on the bandwidth issue? Can it be improved? Is there an upper limit of what can be expected?
Good signal generators require a lot of screening, which makes them expensive. I have a couple of synthesised Marconi 2019A units, they are quite old but very good quality units and weren't too expensive. We had lots of them in the lab at Racal Comms, where I worked a few years ago.
If you ever want to look at circuits where that is exploited to extremes check the vintage SSB transceiver IC-202 from ICOM. It uses pulled crystals, one per 200kHz segment (plus some margin) of the 144-148MHz band. Very stable, very nice spectral cleanliness, low power consumption versus a synthesizer. I used that set for years before giving it to my father.
That really is an extreme example. That is 0.2/144=0.00138, or 1,388ppm. The crystal probably can't be running in overtone mode, but has to be multiplied up from the fundamental. Even then, normal crystals won't go that far. There's a nice comparison of the pulling capability of overtone vs fundamental in figure 2 of
Equation 3 gives the pulling amount as a function of the crystal capacitances. I betcha the capacitances don't work out to realistic values, so there is probably some other circuit trick, such as an inductor in series with the crystal. This may work, but it probably degrades the phase noise and frequency stability, as is often mentioned in rrah whenever the topic of pulling comes up. But it may be ok for casual use.
Also, I have definitely come to the conclusion that the frequency calculation in the paper is incomplete. I need to add the series trim capacitor to the equations to bring the frequency down slightly to the expected value. This means redoing all the figures in the paper to reflect the new component values, which will take time. But I'll try to get to it this weekend.
After setting the open loop frequency to exactly 1MHz, it will be interesting to see the actual frequency that Transient Analysis predicts it will run at. And how much it changes with different operating conditions.
Since the .IC command starts the oscillation in a known phase and steady-state amplitude, it is easy to go out several hundred cycles and look at the phase of the oscillator and compare it to the calculated frequency.
These kinds of investigations were difficult or impossible to do before due to the long startup and settling time when running the crystal at full Q.
But using the .IC command makes them easy now, so there is a lot of work to do in confirming all the various modes and circuit parameters that everyone had to take for granted before.
It's multiplied but WRT to phase noise and frequency stability this set is quite excellent. I ran it with power amplifiers and if phase noise were excessive I would have gotten into a serious spat with neighboring ham radio operators. It is a really clean radio.
As for stability: A friend and I both had that same type IC202 and left them on for weeks. The power consumption was next to nothing. One could call the other and the frequency was nearly dead on. In SSB a deviation of less than 100Hz requires an adjustment. If temperature extremes would hit we'd have to do that, otherwise there was little need to adjust.
You probably meant more than 100Hz. I have a pretty good ear for tone and love classical music, so I need to adjust ssb when it gets more than about
100Hz is about 0.7ppm. Pretty good, but you don't say if that was close to the band edge where the crystal was mostly in control, or at the other edge where the LC circuit was dominating. BTW, do you happen to know how they managed to pull it that far? Regards,
Well, I mostly listened to the Stones, AC-DC, Led Zeppelin and so on. Maybe that's why :-)
Oh, and I like Blues (old style) and Country (also the old stuff).
We used it everywhere in the respective bands, adjusting to other's "favorite" standby frequencies if they wanted us to. My father has the schematic but meantime it has been posted on the web by a few folks. I will try to find one later since my wife has coaxed me into fixing the deck stairs today. Oh, do I dread that in the 95F+. But what's the saying? If the missus ain't happy, ain't nobody going to be happy...
Ok, took a break from stair case fixing. I do not like wood frame construction at all but what can I do? My arm seems to fall off now. Anyhow, one source for a free schematic of the old IC202 seems to be:
This one is kind of hard to read but you can see the concept in the south-west corner of the first page: