Re: Precision Modulation Monitor - NOT! Author: MWeiss7401 Email: mweiss7401@aol.com Date: 1998/01/19 Forums: alt.radio.pirate _________________________________________________________________ Subject: Precision Modulation Monitor - NOT! From: mcarter@btigate.com (Max Carter) Date: Sun, Jan 18, 1998 10:36 EST Message-id: <34c2203b.2765913@republic.btigate.com> >If you plan to use the Modulation Monitor Circuit shown at http://members.aol.com/amn92/Mod_mon.HTM as a foolproof way to keep your station from over modulating, you might want to reconsider. There are other choices out there - better and simpler.< Then by all means, SHOW US. >When I first ran across the Precision Modulation Monitor Meter Circuit it just didn't "look right". To be specific, it looked as though the Circuit, while working just fine with a symmetrical modulating (audio) signal, probably would not accurately indicate when an FM transmitter is being modulated ASYMMETRICALLY. In other words, during conditions when the transmitter is being heavily deviated in one direction, say to +100 kHz, while at the same time being under-deviated in the other, say to -50 kHz. The basis for my doubt was the fact that the Circuit uses two peak-holding detectors - one for positive peaks, one for negative peaks - followed by a summing amp, instead of what seemed to me the correct configuration - a full-wave precision rectifier, followed by a single peak detector. < I think I recall having this conversation before, and pointing out that the REASON for two peak detectors is to read the negative AND the positive peaks independently, to prevent errors that occur with assymetrical waveforms. In an earlier conversation, you mentioned that you would have placed the peak hold circuit AFTER summing the outputs of the rectifiers. I don't quite follow the reasoning why this would be any better, since both circuits achieve EXACTLY the same end result. Let's see what your test results indicate and why you got the results you got: >To check my suspicion, I built the Circuit on a solderless breadboard and ran it through a rigorous test. The test setup was as follows: A 1-kHz, adjustable duty-cycle squarewave signal from a function generator was used to directly drive the paralleled inputs of a North Country Radio MPX96 stereo transmitter operating in mono mode (no pilot) and without pre-emphasis. No compression, limiting or other processing was used in the audio path. An oscilloscope was used to monitor and set the function generator. The RF output from the transmitter was fed to both a Heathkit AJ-15 FM tuner and an IFR model COM120B service monitor, which served as a measurement standard. The baseband output from the tuner was connected to the Modulation Monitor Circuit. < A questionable signal source. Might I suggest dispensing with the transmitter entirely and feeding the function generator directly into the meter circuit to eliminate any errors that might be introduced by this non-commercial quality transmitter. I'd prefer you used something capable of accurately producing square wave modulation, but let's go on anyway: >The setup was first tested with the generator set to 50% duty-cycle. The generator amplitude was adjusted to produce a range of deviation readings from 0 to +/-75 kHz. With each setting-change the DC output from the Modulation Monitor Circuit was noted. Under these circumstances, the Circuit always produced an output voltage proportional to peak-to-peak transmitter deviation.< Of course it would. >The function generator was then set to produce a reference reading on the service monitor of +/- 50 kHz and the output voltage from the Modulation Monitor again noted. The output signal from the function generator was then varied through various duty-cycle settings in the range of 10% to 90% while maintaining constant peak-to-peak amplitude. The IFR COM120B reference standard always correctly indicated the asymmetrical modulation of the transmitter in exact relation to the duty-cycle of the modulating signal - a function generator setting of 75% / 25%, for example, produced a reading of +75 kHz / -25 kHz on the IFR.< How did the IFR COM120B indicate that? Does it have two modulation meters, one for positive, the other for negative? What you discribed otherwise sounds like an RMS reading, not a peak reading. If your IFR COM120B is producing assymetrical results on ONE meter, it's not measuring *peak* deviation--it's really measuring duty cycle-a function of RMS power (not a very useful means of checking peak modulation, which is what the FCC is concerned with). I say this because I note your use of the term, "in exact relation to the duty-cycle of the modulating signal" -- a sure indication of average-responding metering, not true peak responding metering. One other possibility is that you desire a metering system that gives not one, but TWO readings, one for negative modulation and the other for positive modulation swings. I haven't seen that type of indicator in practical use for FM. It is however, used in AM modulation monitoring, because advanced AM transmitters like the BE-2.5 are capable of assymetrical modulation. So let me inject a question here to find out if we are really trying to compare apples to oranges: Do you observe constant output from my meter circuit as you change duty cycle, but expect to see varying output? If so, how could you make the readings with only one meter, since this has no polarity switch? My meter circuit doesn't differentiate between symmetrical and assymetrical waves; it merely is concerned with TOTAL peak deviation, which is apparently what every standard broadcast modulation meter, analyzer and test set seems to be concerned with. I realize that it's possible to take each rectifier output and drive separate meters for true positive and negative deviation distinction, but this information makes taking readings an arduous task and doesn't offer the kind of information the broadcaster needs to know, which is fast, accurate readings of total peak deviation. Let's go on: >The output from the Modulation Monitor Circuit produced no such indication. As changes were made the output voltage from the Monitor Circuit would spike upward for a fraction of a second - too brief to get a meter reading - but would always quickly settle to the ORIGINAL reading. In other words, the Modulation Monitor Circuit, while correctly indicating peak-to-peak deviation, was blind to sustained asymmetrical deviation!< Of course not! It produces a reading which is a function of PEAK modulation, not average duty cycle. In effect, one may even question whether the transmitter did not indeed swing in BOTH directions equally, but for different periods of time in each cycle. Irregardless, that's what the fullwave peak detectors compensate for. I chose to conduct some tests myself, using various duty cycles of a 100hz square wave from 50/50 all the way to 90/10. I could not see what the problem was, unless your version of the test circuit was itself assymetrical due to out of tolerance resistors for instance. During my testing, I provided the baseline 50% duty cycle and calibrated the meter for 100%. The output of the function generator was fed directly into my FM exciter. I also tried a 70/30 and a 90/10 duty cycle. Meter readings were within 1/2 % of eachother in all three cases. Just to confirm that my fullwave precision rectifiers were indeed necessary, I measured the output of each bank of rectifiers, before summing. With the assymetrical signals, the outputs were proportionately unbalanced, but at the summer amp, the weaker output of one bank was cancelled by the stronger output of the opposite bank, effectively cancelling errors that would have occured with a halfwave rectification system. The next thing I did, after verifying that both sides output equal voltages when a 50% duty cycle was modulating the transmitter, was to investigate a hunch that some builders might elect to use 5% carbon resistors instead of 1% metal film, so I dropped in a resistor of ten times the value across the feedback resistor in the positive precision rectifier. I went back and did more testing with this jury-rigged circuit "fault" to see what might happen. My observations were: The meter had to be re-calibrated when the reference 50% duty cycle wave was applied. Once done, I continued to assymetrical testing. The 10/90 signal gave the same results as before, but when in INVERTED it, suddenly the modulation reading dropped by 4%. This with only a 10% change in resistor value. I later removed the extra "fault" resistor and continued testing to verify that with balanced gain, the polarity of the assymetrical signal didn't matter. From this result, it occured to me that this *might* have been one of the problems you observed. Indeed, both rectifiers must be tightly matched if this circuit is to function accurately in the highly impractical "test" situation you've set up here. Of course, such tracking accuracy will not be critical to the microbroadcaster, since program material does not consist of a 90/10 duty cycle square wave (not that most microbroadcasters could pass one intact anyway). >Thinking there was probably a better way, I rebuilt the circuit, incorporating a standard full-wave precision rectifier (Fig 5-9, from IC Op-Amp Cookbook, Walter G. Jung, First Edition, Howard W. Sams, 1974) followed by a standard peak-hold detector (Fig. 5-12, same source) and a non-inverting buffer, then re-tested. [Incidentally, the standard circuits can be implemented using only two op-amps and two diodes for the precision rectifier, and two op-amps and 1 diode for the peak-hold detector/buffer.] As expected, the standard circuit responded precisely to any combination of modulating signal amplitude and asymmetry with an accurate indication of peak deviation.< I'd like to understand just WHAT exactly you had in mind here, because I suspect we're talking about two different types of monitors. Of course if you have ideas on simplifying this circuit without compromising it's accuracy on reporting TOTAL deviation with assymetrical waveforms, then perhaps it would be good if you shared it with the group. >So, don't believe everything you read on the Internet. I suppose it could be argued that real-world conditions are different, that asymmetrical modulation only takes place intermittently, not continuously, as in my tests. But, you see, the word "precision" implies "foolproof" and the Circuit, as presented, isn't. Not to say it isn't useful, just that it isn't precise, and it certainly isn't as simple as it could be. A single AC-coupled half-wave peak-holding detector would probably work as well and be a lot easier to build. The foolproof, standard circuit is also easier to build.< Totally incorrect here. I tested that concept with assymetrical waveforms too, by disabling one of the rectifiers and found that it became polarity sensitive! In one direction, the reading would be 100%, but in the other, it would drop to somewhere below 30%! Hardly and accurate system! (As a matter of observation during my tests, I noted that the average-responding level meters on my Akai tape deck, which I leave on and set to a reference level to give me an idea of my station's average "loudness" gave a 6dB difference in reading depending on polarity of the 90/10 square wave. When it was positive, the meters read +3dB. When I inverted it, the meters on the Akai read -3.5dB. Must be a halfwave rectifier in there. :-) >>I also disagree with the author's advice to use an electromechanical meter movement, with its unknown ballistic characteristics, as the indicating device in this application. A better choice would be an LED display driven by one or more LM3914 LED driver chips. This would allow the ballistics of the meter to be completely specified in the electrical domain. An unbeatable combination, I think, would be two peak-holding detectors (one for positive and one for negative), each driving its own LM3914 display. Sincerely, Max Carter<< On this last one, I'll give you concession, because picking and choosing meters and matching the ballistics properly is a task for those intimately familiar with such meters and how to test their response times. I chose a particular meter, measured it's response and damping and then used a 50mS pulse at 2-second intervals as my goal for 90% deflection from zero. I modeled the ballistics of this metering system after the modulation meter in the Harris MS15 exciter. After considerable tweaking, I'm pleased to say it keeps right up with it, and is MUCH faster with short peaks than the meters in the QEI 691 that I frequently work with. That unit REQUIRES the peak hold circuit with a lamp that lights up for 2 seconds after each 100% peak is registered. Whether you believe it or not, based on your results, the meter system *I* constructed will respond to signals as short as a record click. And I did have extensive time to compare it with a Inovonics modulation analyzer which does use the LM3914 (3 of them per meter) and bar/LEDs. The thing I didn't care for about it was a lack of resolution to less than 1%. But other than my comments, the two meters provided virtually identical readings of peak modulation, with various program material. So I decided that a peak LED was not necessary, because the meter responds so fast as to make LEDs redundant. Of course there's no harm in having one either, I just felt it added no additional function. Now you idea of having two meters, one for negative swings and one for positive, is a pretty neat idea, but I'd go a step further and provide a TOTAL modulation readout on a third meter, that way the total modulation could be read at a glance, but if a broadcaster wanted to know how much that signal was divided between positive and negative deviation, then the other two meters would provide that additional information. Since no station I have ever worked for employs such a reading for FM use, I avoided it altogether. My goal is to provide reasonably-simple, cost-effective circuits that people can construct and use with excellent results. The meter is called "precision" because of the high sensitivity of the rectifier circuit, which allows accurate readings of signals at or even below 1mV. It will accurately read the pilot level when no audio is present, without switching the meter to a dedicated full-scale filtered mode. Even the QEI can't do that in normal L+R mode. I would have like to see exactly what you did with this circuit and made measurements myself to find out if any of the possible circuit faults that I intentionally introduced above to test my hunch, in fact exist in your version of the circuit. The version of the circuit in use here, works and works extremely well. It will spot transmitter modulation overshoot that the FCC type-accepted units don't resolve. Part of this need to observe tiny amounts of overshoot became necessary for me to measure the performance of my air chain, in which I had spent considerable effort making very "tight" with respect to peak control. And only this meter sees a big difference between the overshoot of the local rock station, and the relative lack thereof on my station. The Inovonics, with its LEDs, can't resolve under 1%. The QEI is too slow, and sitting there counting 100% peaks per minute is not as interactive a process as I prefer. Watching this meter respond, lightning fast, to every nuance of the envelope, tells me a lot more in a way I can directly apprehend, than the other two instruments do. I'm already thinking of building a portable version to carry with me when I do translator setup work for the stations I am employed by. After you've hauled a QEI 691 up a few flights of stairs and down some long hallways every week for a few years, the mind seeks a better way. Mark The "Peg-legged" Bass Pig ^ ^ _o^O-/ \ (oo)_ ?) _/ ) \_ SPAM-Pruf http://users.aol.com/amn92/amn.htm "I support the micropower broadcasting movement and freedom of the airwaves" _________________________________________________________________