Rich is quite correct that it makes no sense to provide greater audio bandwidth at the transmitter than the receiver is capable of providing, particularly because wide audio bandwidth, in principle, reduces transmitter and antenna efficiency if the electrical length of the antenna is small.
What appears to happen with AM receivers is not that the higher frequencies, like 5 kHz and 10 kHz (if present in the modulated signal), are absent, but that the lower frequencies are enhanced. In one of my receivers that I tested, the audio output at 1 kHz modulation is 3 dB lower than for 400 Hz modulation.
At 2.5 kHz audio bandwidth, if the cutoff is not sharp, the articulation index is high, and music still sounds like music, although it does not sound very good. The theoretical efficiency of the system is twice as high at 2.5 kHz audio bandwidth as at 5 kHz audio bandwidth.
With the higher Q at 2.5 kHz, the trick will be to keep the carrier frequency at the peak of the antenna response.
The subject line for this post is only a slight exaggeration. Bandwidth certainly does matter in Part 15 AM, but I recently found out that it is not as critical as I previously thought it was.
In several of my posts, I pointed out the physical fact that the theoretical efficiency of of an electrically short antenna depends upon the bandwidth of the transmission. I pointed this out in considerable detail in a previous post in this thread. I said that a short antenna with enough Q to allow only the carrier to be transmitted can have nearly 100% efficiency. However, in order to transmit intelligence, it is necessary to lower the antenna Q so that sidebands corresponding to the audio modulation can be transmitted. This lowering of the Q to increase the RF bandwidt causes the efficiency of the antenna to decrease.
Now, the question remains: How small a bandwidth is acceptable for AM? I've taken it for granted for a long time that about 3 kHz audio bandwidth is needed for speech, and about 5 kHz is needed for music. However, Rich responded with a link that said that AM radios rarely have as much as 5 kHz audio bandwidth. (I will say once again that the RF bandwidth is twice as wide as the audio bandwidth because there are two sidebamds.) I did some testing, and found that AM radios do, indeed, have awful bandwidth. But then, AM radios, while not great, don't sound too bad. How can this be?
I remeber reading about this subject in Hermann Hesse's 1927 novel, Steppenwolf:
"Radio Munich is on the air, playing Concerto Grosso in F major by Handel. That damned tin trumpet, the loudspeaker, puts out noise that owners of phonographs and radios call music. Behind the static and the distortion, there is, like an old painting by a great master under a layer of dirt, the noble outline of that divine music. Radio cannot destroy the original spirit of the music. Handel, disfigured by this most ghastly radio apparatus, is still divine."
Hermann Hesse says that, while AM radio is not nearly as good as a live performance, it is still worth listening to. In Connecticut, there is a classical music station, WCCC, not on FM, but on AM. AM is still able to do justice to the music it presents.
I did an experiment to determine how narrow the bandwidth of a short, Part 15, antenna can be and still give acceptable results. An electrically short antenna forms a two-pole bandpass filter that translates to a single-pole low-pass filter after AM detection. A single-pole filter does not have sharp cutoff. It transmits frequencies much higher than the half power frequency. I connected a high-fidelity music source to a single-pole RC filter connected to a buffer amplifier. I listened to the output of the buffer. With 3 kHz bandwidth, the music sounded fine. Even with only 1 kHz bandwidth, the music still sounded fine. Below 1 kHz, there was still enough bandwidth for the music to sound okay, and for speech to be intelligible, but the amplitude was decreasing as the bandwidth became lower. The lowest frequency I tried was 150 Hz. The music was still fine, the speech was intelligible, but the amplitude was low.
1 kHz audio bandwidth corresponds to 2 kHz RF bandwidth. At 1700 kHz, the Q is 850. This is probably as high a Q that is obtainable with a Part 15 AM antenna. So, to get the best efficiency, the Q should be as high as possible. There is no risk of getting the Q so high that there would not be sufficient bandwidth to pass an acceptable AM signal.
That is a very interesting report that you present. Thanks for sharing your thoughts and experimental observation.
My experimental observation of the use of a high Q antenna was not as encouraging. I connected a homebrew link coupled base coil loaded antenna (where the loading coil is tuned by a parallel cap) to my SSTRAN and found that the audio was severely degraded. It sounded as if it was BW limited such as listening with cotton in my ears. I assumed that this was due to the narrow antenna BW and lowered the Q by placing a resistor (the one that SSTRAN suggests removing) back into the output circuit of the transmitter. The audio is now excellent.
There is a confounding effect which may have caused the unacceptable audio which I might have mistakenly attributed to the antenna Q. From an Electronics Workbench model of the circuit with a series RLC "antenna" I observed a badly distorted RF waveform as well as nonlinearity in the envelope of the modulated signal. This was confirmed when I observed the actual RF signal radiated from my antenna on a scope as I compared the "high Q" and "lowered Q" antenna performance. With the resistor in place, both the model and measurement yield a "text book" waveforrm. Since I was able to get excellent audio, I did not pursue this further.
What this suggests is that my unsatisfactory experience with the audio may not have been due to just the Q but may also have been due to presenting a load to the SSTRAN which it could not properly feed. Remember that the antenna I use is not the SSTRAN design and this problem could be the result of my deviating from the SSTRAN plans.
Neil
Ermi,
Interesting test. How would you expect the narrowband IF of an AM receiver to change your conclusions about the minimum bandwidth for Part 15 AM? Probably those IF filter skirts (and passband) are not as well-behaved for amplitude and phase distortions as those you tested.
Rich
My test was of the sound of narrowband audio using a single-pole low-pass filter, not a test of an actual narrowband antenna with the RF bandwidth at twice the audio bandwidth. For my test to be equivalent to the results with an actual narrowband antenna, the carrier frequency has to be at exactly the peak (i.e. center) of the antenna tuning curve. If the carrier is not exactly at the center of the tuning curve, the attenuation of the two sidebands will not be balanced, and distortion will occur. In practice, it is difficult to match the transmitter carrier frequency to the antenna peak frequency of a very narrowband antenna. I mentioned this fact in a previous post in this thread.
One of the problems with Part 15 AM is that the combination of the transmitter and antenna is theoretically very inefficient, even under ideal conditions. It happens that, assuming an electrically short antenna, the theoretical efficiency is inversely proportional to the bandwidth. In principle, an RF bandwidth of 2 kHz gives ten times the efficiency of an RF bandwidth of 20 kHz. Prior to performing the experiment with the single-pole audio filter, which I reported here, I thought that the use of 2 kHz of RF bandwidth was out of the question, since the use of one kHz of audio bandwidth is simply too small. I now know that, if the audio frequency response has only a single pole, a 1 kHz cutoff frequency is not too narrow.
The practical problem of using a very narrowband antenna is that of frequency stability. The crystal oscillator that establishes the transmitter frequency will be very stable compared to the frequency stability of the antenna. Changes in temperature will change the antenna tuning. As a practical matter, it may be necessary to add to the losses of an antenna simply to keep it in tune. However, as my test shows, there is no need to keep the audio bandwidth wide.
Ermi Roos wrote: For my test to be equivalent to the results with an actual narrowband antenna, the carrier frequency has to be at exactly the peak (i.e. center) of the antenna tuning curve.
For greatest applicability, wouldn't the complete r-f system need to be tested, ie, should it not include the audible affects of a "typical" narrowband AM receiver including its lumpy passband response and sharp cutoff IF skirts, and not limit itself just to the r-f bandwidth of the transmission system?
//
The intention of my test was to discover what the subjective effect to audio is due to a single pole, or at least a single dominant pole. This is only a first step to determine if a single pole at a low frequency so distorts the audio that further tests are not necessary, or if the results are good enough to warrant further effort. I found that a single pole at a low frequency leaves the audio acceptable.
I think it is possible that a single pole at a sufficiently low frequency will dominate the other poles of the passband response, and the multiple poles of the shap cutoff IF skirts. In one variation of my test setup, I used an AM receiver with poor bandwidth to supply the audio input to my test fixture, which comprises an RC low-pass filter and a buffer amplifier. The addition of the low-frequency pole to the frequency response of the AM receiver still left the audio acceptable.
Certainly, if an AM system with a very high Q antenna is constructed, the complete system should be tested. The source of at least two more poles to the complete system is the tuning circuit that couples the transmitter to the antenna. This tuning circuit is often neglected in determining the efficiency of a Part 15 AM system. In a previous post by me in this thread, I explain how the tuning circuit affects system efficiency.
Ermi,
I understood your description of the test you did with an audio source and filter and my last post here was intended to provide another explanation of a perceived problem which I once attributed to antenna Q which may not be the true cause of the observed effects.
My models and observations indicated a distortion of both the RF waveform and the modulated envelope as I described. The observed distortion of the expected sine wave of the unmodulated RF carrier was not due to antenna Q since it was not sinusoidal. Pending further tests it appears that the SSTRAN, when presented with the load of my antenna system, produces distortion most likely due to a load with which the circuit cannot function properly. This does not appear to be an issue involving antenna Q.
Using the generally accepted values for antenna capacitance (30 pf.), resonant loading coil inductance (300 uH.) and antenna system resistance (20 ohms including ground) I calculate a Q of 150. This yields a bandwidth of 10 kHz. at 1670 kHz. Unless I have an atypical antenna and ground system in terms of low resistance, it is not probable that high Q caused the effect I reported.
Your report of the effect of single pole low pass filtering on an audio signal is duly noted and appreciated. My report indicates that with this particular transmitter and with my antenna system that bandwidth is not the cause of the problem that I observed. The excellent performance I have has resulted by installing the 820 ohm resistor in the circuit. PhilB has posted that this is approximately the load which the circuit is designed to drive and it leads me to conclude that my antenna system is nowhere near presenting an appropriate load to the transmitter rather than being a problem with antenna Q.
I would appreciate your comments since I might be overlooking something.
Neil
I have not seen the SSTRAN circuit, but I know that matching a low-power transmitter to a short antenna is a big circuit design problem. All of the circuit designs I have seen are very poor, and nobody, as far as I know, has come up with a good one. This is something I am working on, and this is why I am interested in this thread, which originally dealt with using vacuum tubes for Part 15 AM. I like the low capacitance of tubes, and the high output impedance of pentodes. Pentodes, however, are unsuitable for Part 15 AM, because the screen power counts toward the 100 mW input power allowed for the final stage. This leaves triodes, which might work allright if the conduction angle can be made small enough. I will probably stick with the JFET, which, however, has high output capacitance. I have mentioned in this thread that capacitance in the coupling between the transmitter and antenna is bad for efficiency.
In the design of the coupling circuit, there are several bad choices to choose from. Using a toroid looks good on paper, but there are unpredictable core losses to deal with. I like a large diameter multiturn loop, which has no core losses, low copper loss, and has a high coefficient of coupling for effective impedance transformation between the output of the final stage and the antenna. Wire made up of a bundle of fine, insulsted, strands is supposed to reduce skin effect, but I am not completely sure about that. The proximity effect in a wire bundle causes something like skin effect, where most of the current flows in the outer strands. The coil capacitance of the multiturn loop is high.
As for the specific problem you wrote about, it seems to be a design problem that has to be beaten to death on the test bench. As you probably know, most of circuit design is not all that scientific. It mostly consists of fiddling around until you get something to work. I know I have added resistors or capacitors to circuits without any idea why they were there, or how they worked.
Ermi,
There is no question that designing an AM tx. to match the variety of loads presented by part 15 AM antennas is a challenge due to the unpredictability of the load Z. Thanks for running some ideas by us here, and I hope Lee (original poster) will forgive us for getting off topic, but maybe we're coming back a bit.
Just to help you follow some of what I posted, the SSTRAN circuit is very similar to the first Wenzel circuit at this link:
The Ramsey AM-25 uses a Tmosfet (MTP-3055) as the final in a class C circuit followed by a seven pole LPF (which may also be the matching network). The maximum efficiency I have measured on my unit is 31% into a bench 24 ohm resistor. This interests me since the feedpoint Z of a resonant SSTRAN type antenna is in this range of Z.
As you think about Litz wire for inductors in the circuit, it seems to me that the R of the ground system may swamp this resistance advantage over regular magnet wire but perhaps the numbers need to be crunched to find out for sure.
I hope this helps somewhat as you ponder a circuit. Your comment about fiddling with parts rings true with me though I usually know why they are there....just don't always know their values. I recall a VHF amplifier I built on a ground plane where the dress of the capacitors in the output network made a hugh difference in the gain. I tuned it by bending them toward or away from the board.
Neil
Some Part 15 AM txs have a "loading coil" inside the tx enclosure (ex: Rangemaster), and some require an external loading coil (ex: SSTRAN).
Once the capacitive reactance of the short antenna is "zeroed" (brought to resonance) by the necessary inductive reactance of the loading coil, then the transmitter will see a non-reactive load -- which can allow for maximum power transfer to the load from the tx output circuits, which produces the maximum amount of r-f current flow on the antenna, which results in maximum radiation for the available power.
But it isn't enough for the Part 15 AM antenna to have zero reactance at the feedpoint, if the tx output circuits are unable to accommodate the variation possible in the resistive term of the load impedance. And the largest contributor by far to that load resistance term is the resistance of the r-f ground path.
This r-f resistance in the ground path is not the DC resistance of the conductor(s) leading to whatever is considered "ground," but the resistance of the r-f ground itself plus the conductors connecting it to the tx, at the operating frequency.
Consider a tx designed for a 50 ohm, non-reactive load. When its output is connected to a load that is non-reactive (resonant), but because of a poor r-f ground has a resistive term of 150 ohms instead of 50, then the load SWR seen by the tx will be 3:1, and the return loss from the load (antenna) will be 6.02 dB. The power delivered to the antenna feedpoint will be around 25% of the power that the tx could deliver there, if it was capable of matching into that 150 j0 ohm load. Most of that "missing" power would be lost as heat in the tx output networks. And this doesn't include the power lost in the r-f ground connection, which is even worse for these conditions.
So a Part 15 AM tx needs to be able to match into a fairly wide range of load resistances, even if those loads have zero reactance (are resonant).
The r-f resistance of the "ground" path in a typical Part 15 AM setup might vary from a low value of 10 ohms to a high value of several hundred ohms.
//
I had previously seen the second of Wenzel's circuits (Wenzel 2), shown in Neil's link, in "The Low Power AM Broadcasters Handbook," page 61. I had downloaded this handbook from the part15.us web site some time ago. Wenzel 2 is a very efficient circuit, primarily because of the autotransformer in the potcore that greatly increases the impedance of the transmitter output stage. This impedance transformation significantly improves the efficiency of the transmitter.
Wenzel 2 uses the antenna as a high-resistance load. The autotransformer inductance is in parallel with the antenna capacitance. The antenna parallel-equivalent antenna resistance, which is high, is the load. For best efficiency, the tuning capacitor connected to the antenna input should be as low as possible.
Wenzel 2, quite properly, drives the antenna from the collector of the final stage. A high resistance load should be driven from the collector.
The first Wenzel circuit in Neil's link (Wenzel 1), which Neil said is similar to the SSTRAN, is quite different from Wenzel 2, and is a much less efficient circuit. It's really unfortunate that the SSTRAN design is based on Wenzel 1 rather than Wenzel 2.
Wenzel 1 uses the antenna as a low resistance load. A loading coil in series with the antenna tunes out the series-equivalent antenna capacitance, which leaves the series-equivalent resistance, which is low. This low resistance is still driven by the collector, and it would be better if it were driven by the emitter. The Panaxis AM 100, in contrast, uses an emitter follower for driving the antenna. Wenzel 1 has a resistor connected to the collector of the final stage that absorbs some of the RF power, but that is not the primary cause of the inefficiency. The inefficiency is caused mostly by an improper driving impedance of the transmitter.
Wenzel knew exactly what he was doing. Wenzel 1 was designed for ease of construction and Wenzel 2 was designed for high efficiency.
The resister is R18. Just for the heck of it I temporarily installed it. I did notice what seemed like a bit more dynamic range and definitely more bass. However my range was severely cut down. Remember I am using the Valor antenna and my setup is kinda weird. I am not using any of the SSTRAN's inductors. Rather I have a small, I think more efficient and/or better value inductor soldered on the antenna plug between the ground and antenna. I do get some more range by having the antenna lead connected to ground, and the ground lead connected to the antenna. I don't know the value of the inductor off hand, but it works well no matter which way I have the antenna and ground hooked up. I think with an even more efficient, or correct value inductor, I could get more range and possibly be able to run with R18 installed. A variable inductor is what I need to really fine tune this thing.
Hi Michael,
It is really good when folks post here about their experiments and observations. Thanks for sharing yours.
It is not surprising that you observed a decrease in range with R18 installed. There was quite a discussion about this between PhilB and others in a blog a few years back. Unfortunately, my bookmarked link to this is dead.
PhilB, the SSTRAN designer and manufacturer, posted here that the unit is intended to see a collector load of about 800 ohms and with the SSTRAN design antenna this is accomplished by an impedance transformation using the variable cap to "tune" the circuit. I don't have the link handy but I will edit it in here later.
I found it. You might also want to peruse the whole thread:
PhilB's comments about the SSTRAN antenna matching
Given this information, it can be seen that my antenna did not present the proper load and I needed R18 to get a good waveform. Unfortunately, R18 is probably getting more power than my antenna.
My problem, as I mentioned, seems to be due to distortion rather than antenna bandwidth. If you can gain access to a scope, it would really be helpful for you and us if you can compare the output waveforms with and without R18. The audio enrichment may be due to decreasing distortion rather than increasing bandwidth.
With the wide disparity among antenna systems we all use it is a wonder it works at all. Congrats to Phil for what he has done with his design.
Neil
Hi Neil,
I know I need to get a scope. I am scouring Ebay for one. I will report back when I get it and can see the waveform. Thanks for posting the link. I had read through the thread, and read through it again. This all becomes a bit clearer for me as I go along. I know a scope would make it very clear....lol