Comments:

The TV I.F. filter, at home atop N1LF's R-7000.
Click on the picture for a larger version.

• The author of this article (KA7OEI) does not own an Icom R-7000 or have access to one with a TV demodulator.  Testing of this filter has been done by Les Rayburn, N1LF, who lives clear across the country.  The construction of this filter was done at the request of Les who happened to read another of my web pages, IF Bandpass filtering of AM TV signals, that describe using "narrow" IFs on TV demodulator IFs to receive signals that would (possibly) not otherwise be detectable.
  
• With the demise of high-power analog TV transmission in the U.S., these techniques are no longer generally applicable in the U.S. as the R7000 and its TV demodulator are not useful in demodulating digital TV signals:  Attempts to similarly filter digital TV signals will simply render them unusable to a digital TV receiver.  This information may still be useful for receiving amateur radio TV, signals in or from those countries where analog TV is still used, and for those who are trying to receive the still-extant low-power U.S. analog TV stations and translators that weren't affected by the U.S. digital TV changeover.
  

About "Narrowband" TV I.F. filters:

The nominal bandwidth of the I.F. of a TV is 6 MHz.  In reality, not all of the energy to be found in this bandwidth finds its way into the demodulated video signal:  Typically, the video response is limited to 4.2 MHz or less, and the filters in the IF are typically somewhat narrower than this, to provide some "breathing room" between adjacent channels.  The "active" bandwidth of a typical receiver, for video, is typically 5.2 MHz or so - more or less.

Communications theory states that with a given modulation method, the wider the bandwidth, the more information that may be sent on that signal.  This does not occur without cost:  The more bandwidth you have, the more noise your receiver intercepts.  In order to maintain a given signal quality margin, the more bandwidth your signal occupies, the more power you need to run to maintain a given signal quality.

Analog TV signals pose an additional challenge:  The vast majority of the power in the video signal may be found within the first 1 MHz of the video carrier.  This means that only a small portion of the video energy may be found in all of the rest of the signal, and it is in this minuscule power that the fine resolution and detail of the picture may be found.

What this means is that when signals get weak, the first information to be lost is the fine detail.  To make matters worse, the noise that appears in the signal (which is, by definition, in the "high frequency" parts of the video signal - because that is where the transmitter power isn't) further obscures fine detail.  In this situation what happens when you simply remove some of the higher frequency components?  Since these video components are, presumably, already lost in the noise, you aren't really losing anything by cutting them out.

The anatomy of an RF video signal:

Video is, when analyzed in detail, an extremely complicated signal in its baseband form.  It follows, therefore, than a carrier modulated by such a signal is also very complicated - and it is the RF signal that we are concerned with.  Rather than looking at it with a microscope, we'll stand back at some distance and look at the big picture.

The spectrum analyzer plot:

A spectrum analyzer plot of an off-air TV signal.  See text for details on the sweep.  Note that the vertical scale is 10dB/div.
Click on the picture for a larger version

Looking at the spectrum analyzer plot to the right, an actual off-air (from KUED, channel 7) is shown.  The "upper" line on the plot is a result of 30 minutes of accumulating peaks (e.g. the "peak hold" function) of the video signal while the "lower" line on the plot is an average of 100 "sweeps" of the analyzer and represents a shorter-term average.  The vertical divisions are 10dB while the horizontal divisions  represent 1 MHz.

The 6 MHz "mask" of the video signal can be clearly seen in this 10 MHz wide plot:  At 2 MHz from the bottom and 2 MHz from the top the video signal has been well-filtered.  The largest signal (3.25 divisions from the left) is the actual video carrier, while the second largest signal (2.25 divisions from the right) is the audio carrier.  The third peak (about 3.25 divisions from the right) is the chroma (color) signal.

Typically, any AM signal is symmetrical (those "extra" signals above and below the the carrier are referred to as sidebands) about its carrier, but why not video?  Looking at the analyzer plot, you'll note that the energy below the video carrier extends downwards by only 1.25 MHz before it drops off.  This omission is not accidental - in fact, quite a bit of effort is gone through to make sure that the signal is appropriately removed.  Why do this?  If the video signal were unfiltered, it would be symmetrical about the carrier.  If you think about it, with any AM, signal the sidebands above and below the carrier are absolutely identical in the content that they carry.  If they are identical, why then, would you even need to transmit both of them - as that would be a waste of power?  After all, SSB uses just one of the sidebands and it works just fine.

SSB TV is possible, but doing so would have greatly complicated the design of early TVs and transmitters.  Having both sidebands in their entirety would have meant that the TV channel would have been about 9.5 MHz wide instead of the current 6 MHz.  A compromise was reached and instead of having the entire lower sideband of video removed (which would have been difficult to do, technically speaking) a portion of it is retained -  and this also makes design of receivers a bit easier in terms of filtering.

One point of interest is to note that the plot clearly shows that the vast majority of the energy is concentrated within a few hundred kHz of the video carrier.  (Note:  If the plot were done using a linear scale, one could determine the power over the bandwidth by integrating the area under the plot.)  Comparatively speaking, only a fraction of the energy is contained in the rest of the signal.  As you might notice, the majority of video energy, being close to the video carrier, is transmitted on both sidebands while the higher frequency video signals (such as fine detail and color) is only transmitted via the upper sideband.

The R-7000's I.F. - as in video:
 

A "map" of the I.F. of the Icom R-7000 receiver.  The "nominal" audio FM signal may be found near 10.7 MHz.  For video, the bandwidth extends above the audio carrier.  Note that this "map" represents NTSC signals.

In the following paragraphs, it is important to remember that the R-7000 as well as conventional TVs use "high-side" local oscillator injection.  The result of this is that the spectrum of the video signal is inverted.  That is, components such as chroma (color) and sound - which are, over the air, higher in frequency than the video carrier, are now lower that the video carrier inside the radio.  Yes, I know it can be confusing, but hopefully it all makes sense...  Also note that for this discussion, we are referring to NTSC signals.

The nominal IF of the R-7000 (at least in wide FM mode) would seem to be a standard 10.7 MHz.  Clearly, the IF bandwidth for a wideband FM signal is too narrow for video.  Apparently, the "video IF" output is taken from a point within the R-7000 prior to any bandpass filtering and the "10.7 MHz" IF extends up to at least 16 MHz.  Refer to the drawing to the left.

When a video signal is tuned in using the R-7000, the audio carrier is set at 10.7 MHz and the receiver itself is used for reception of the audio signal.  Because the R-7000 uses high side local oscillator injection, the video carrier is 4.5 Mhz above  the audio carrier - or at 15.2 MHz.  It is a video signal at about 15.2 MHz that is demodulated and it is around this frequency that additional IF bandpass filtering should be done.

Note that with the high side local oscillator, the spectrum of the IF is inverted - that is, the "higher" video frequencies (such as chroma and audio) are below the video carrier. Keep this in mind for the later discussions.

Notice:  Because this filter is inserted in the signal path between the R-7000's IF and the TV demodulator, it has no effect at all on the operation of the R-7000's internal demodulator.  When using this filter, you would use the internal FM demodulator to receive the TV audio.  Since both the "1 MHz" and the "300 kHz" filters remove the sound signal, don't expect to hear it using the TV demodulator box.
 

A spectrum analyzer plot of the "1 MHz" bandpass filter.  Note that it is really about 1.5 MHz wide at the -3dB points - see below for an explanation.  Note that the vertical scale is 5dB/div and that the span is 500 kHz/div.
Click on the picture for a larger version

The "1 MHz" filter:

Having done some experimentation in the past, the first logical reduction of bandwidth would allow 1 MHz of video bandwidth to be passed to the video demodulator.  As you might expect, this means that all color information is removed (which you probably couldn't see anyway...)  Also, much of the fine detail is lost - but how much?

Perhaps surprisingly, 1 MHz of video bandwidth allows quite a bit of detail to get through - more than enough to see even "medium-sized" text, and more than enough to see full-screen IDs as demonstrated by the picture.
 

A signal as viewed through a "1 MHz" filter.
Click on the picture for a larger version

Looking at the analyzer plot, you'll note that the so-called "1 MHz" filter is really about 1.5 MHz wide.  Why call it a "1 MHz" filter then?  You'll notice that the markers on the analyzer plot are set showing that the lower edge of the video bandpass filter is 1 MHz below the video carrier (or, taking into account the spectral inversion, it is 1 MHz "above" the actual off-air video signal.)  Doing this allows video components of up to 1 MHz to be passed through the filter.

Why make it 1.5 MHz wide rather than 1 MHz wide?  It would be impractical to build a filter that cut off "instantly" on the high side of 15.2 MHz (just past the video carrier.)  Also, keep in mind that most of the video energy is contained close to the video carrier - and we are still passing that portion of the video signal that contains significant energy.

How much improvement do we get by reducing the bandwidth to "1 MHz" as compared to the full-bandwidth signal?  In terms of power versus bandwidth, we will gain approximately 5.6 db - which is as if the power of the transmitter increased by a factor of 3.5.  In reality, such an improvement is hard to quantify simply through visual observation of the signal.  For example, with a given "noisy" signal, an experienced TV DXer will likely be able to pull a callsign out of the noise well before the casual observer could owing to "gray matter" (i.e. your brain) integration.
 

A spectrum analyzer plot of the "300 kHz" bandpass filter.  Note that it is really about 500 kHz wide at the -3dB points - and for the same reason as the "1 MHz" filter.  The markers show the -3dB point 300 kHz "above" the video carrier.  Note that the vertical scale is 5dB/div and that the scale is 200 kHz/div.
Click on the picture for a larger version

The "300 kHz" filter:

If the signals are really weak, you might want to try to reduce the bandwidth even further.  You would do this when the signals are so weak that you have nothing to lose by knocking the IF bandwidth down a bit more.  Of course, reducing the bandwidth will sacrifice resolution - but it will also reduce the noise somewhat, maybe making an ID possible.
 

A signal as viewed through a "300 kHz" filter.
Click on the picture for a larger version

As in the case of the "1 MHz" filter, the "300 kHz" filter is really wider than 300 kHz - in this case, about 500 kHz wide.  This is done for the same reason as for the "1 MHz" filter:  It would be impractical to build a relatively simple filter with a "brick wall" cutoff.  In looking at the picture you'll notice that details are considerably softened.  This does not mean that the finer details are permanently lost, however.  Because the filter doesn't have "brick-wall" response, those picture details at, say, 600 kHz away from the video carrier are attenuated by about 20 db.

Some tricks with the "300 kHz" filter:

Les has done some experimentation with a video processor and he reports that appropriate "sharpening" can resurrect some of the "lost" resolution.  But why would this work?

Because the filter doesn't roll off in a "brick wall" fashion, many of the filtered components of the signal are still present - but attenuated such that they are not easily visible.  A video "sharpener" (preferably one with an adjustable "knee" frequency) can bring back some of these filtered components to the point that they can (optionally) contribute to the picture quality.

This trick also works with tape-recorded (but filtered) video:  Because even a cheap VHS VCR will have over 40dB signal/noise ratio, even bringing video content up by 20dB will still leave the tape machine's intrinsic noise down by over 20dB.  What this means is that you could record a signal with the 300 kHz filter now and do some enhancing later.

There are limitations to this technique that go in both directions:

• It is only practical to boost attenuated signals by so much.  20dB of high-frequency boost is about all you could reasonably expect to get.  This means that video components that are farther down (such as those at 1 MHz) are probably pretty much lost.  You'd have to judge the quality of the signal to be able to decide which filter to use.
• An advantage of having recorded signals already filtered by a 300 kHz filter is that there is less noise in the signal that might cause the recorder or monitor have stability problems.  Also, having a narrower bandwidth will also prevent the AGC in the demodulator from keying excessively on noise - something that could make the recovered video look "dark" or excessively washed out.

The schematic and partial parts list for this filter may be found here.

Note that this schematic should be printed in landscape mode.  Unless your browser automatically resizes images for printing, it may not properly fit on a single sheet of paper.

Electrically, the filter is quite simple:  Q1 and Q2 are amplifiers that provide gain (to make up filter losses) as well as provide consistent source and termination impedances.  The filters themselves are 3 pole filters, tuned appropriately to provide the desired bandpass response.

Tuning the bandpass filter:

It must be emphasized that, due to the rather precise nature of the bandpass filter, it is not practical to build this type of filter using fixed, off-the-shelf components and have it provide a predictable, repeatable bandpass response.  Instead, variable inductors are used to allow standard capacitors to be used.

The use of variable components means that, once the filter is constructed, it must be properly adjusted and this can only be done with the proper equipment.  Some examples of the sort of equipment that might be used are:

• Sweep generator and oscilloscope.  A marker generator makes precise adjustment somewhat easier, as does a simple diode RF detector.
• Spectrum analyzer and noise generator or tracking generator.
• Network analyzer

A view of the interior of the I.F. bandpass filter.
Click on the picture for a larger version

If you don't have any of this equipment - or access to this equipment - I'm afraid that it will be extremely difficult to properly adjust this filter:  It cannot be properly done just by watching the video passing through it.  Sorry...

"Will you build a filter for me?"

The quick answer is No.  The filter depicted is a hand-built prototype - and I really wouldn't care to build very many of these, and it is likely that no-one would really want to buy them if I were to charge enough for them to be worth my time.  While a circuit board would decrease the amount of time required for construction, alignment would still be required.

"What about a complete parts list?"

I hate to say it, but if you need parts numbers for the individual capacitors and resistors you should probably seek the help of someone with RF experience to help you construct the filter...

An as-built filter:
 

Another view of the interior of the filter.
Click on the picture for a larger version.

A picture of the filter may be seen to the left.  This filter is built using point-to-point construction, some of it being "dead bug" construction, with all components being mounted on copper printed circuit board material.  Because the frequencies involved are only on the order of 10-20 MHz, wiring is not extremely critical, but reasonable care should be taken to avoid things like bundling or paralleling input and output wires to avoid degradation of filter performance.

Additional considerations should be taken to making sure that the coils are shielded and well-secured.  In the prototype, the shield cans were soldered directly to the circuit board material.  This provides a solid base to which the other components that comprise the filter are mounted.

The upper picture shows the "1 MHz" filter section mounted in the rear on a piece of circuit board material that is soldered to the main board, while the "300 kHz" filter components are in the foreground.  The "input" of the filter is to the right, and the output is to the left.  Some of the larger components and some of the wires are shown being held in place using RTV (silicone rubber.)  After proper adjustment of the inductors, it would be a good idea to put some blobs of RTV on the cores to prevent them from being accidentally readjusted, or from drifting due to vibration and temperature changes.

The circuit board is mounted using standoffs inside a standard aluminum utility box from Radio Shack with the RCA connectors being mounted on the ends of the enclosure.  In the lower picture, part of the "1 MHz/300 kHz" selector switch may just be seen under the upper right-hand corner of the circuit board.

Using the filter with the R-7000:

Thanks to Les, N1LF for his input on how the filter is used with the R-7000.

The first thing to be aware of is that when this filter is enabled, you will no longer be able to receive sound through the outboard TV demodulator.  In reality, this isn't much of a problem because the performance of the R-7000's internal demodulator is superior to that in the TV module.  Note that the frequency display on the R-7000 will nominally be that of the sound carrier - and if you do much TV watching with the R-7000, you will already have a chart for this.
 

"Will this filter work with PAL/SECAM signals?" The answer is YES.  When stripped of their color information, the spectrum of a PAL, SECAM, and NTSC signals are, for all practical purposes, identical.  The only difference (and it is irrelevant in this discussion) is that the vertical interval signal occurs less-often (e.g. 60 times a second for NTSC and 50 times a second for PAL/SECAM.)  Because the horizontal rates of these three TV systems are nearly identical, the effect of the filter on the systems will also be identical.  That is, with a 1 MHz filter, you will lose as much detail on an NTSC image as you would on a PAL/SECAM image and vice versa.  (We're talking about monochrome images, remember!)

There are several techniques to determine when weak TV signals are potentially available to receive:

• You start seeing weak sync bars.  This is a sure sign that some video is starting to get through.
• You start hearing TV audio.  Another sure sign - but you may be able to hear audio long before you start seeing any video - if you ever see video.
• If you monitor the video carrier frequency - especially in USB or LSB mode - you can hear video carriers long before they are strong enough to "watch."  This method has the advantage (and disadvantage - in some cases) that you can select whether you are listening for TV signals with a "zero" offset, a "positive" offset, or a "negative" offset - or even the precise video carrier frequency - something which can help you identify whether a particular station is or is not what you think it might be.  Additionally, some video transmitters have a unique  "sound" to them that, through experience, can help you identify it even before its signal becomes strong enough to be visible.

If the signal is still too weak (that is, you see only sync bars when in the "bypass mode" it will certainly be worth going to the "300 kHz" filter to see if further noise reduction can help identify the image.  While this narrow bandwidth reduce reduce noise, this benefit will have to be weighed against the loss of resolution and detail.  If the signal was extremely weak to begin with, you may not have anything to lose.

Other comments and possible receive techniques:

Much TV DXing is done on the Low-VHF band (channels 2-6) and when a band opening occurs, it inevitably starts with the lower channels - that is, if there are lower channels to watch in the area from which there is a band opening.

As the MUF (Maximum Usable Frequency) goes up as the band opening "improves" the higher channels will start to come in.  In many cases, one will see some video - but the MUF hasn't gotten to the point where the sound carrier is being propagated.  In these cases the video bandwidth is already being compromised because the MUF isn't adequate to carry some of the higher frequency content anyway.  In these cases, the narrower filter will certainly be beneficial, as it will remove noise from the higher video frequencies that aren't "there" anyway.

"Offset" tuning to optimize reception:

Another possible technique is to adjust tuning of the receiver.  Nominally, the demod is "looking" at a video carrier frequency of 15.2 MHz, but if you tune the receiver up and down, the filter stays put, but the video carrier is "shifted" within that passband.  In extreme cases of "MUF Video bandpass limiting" you may be barely receiving the video carrier, but there may be the vestigial component of the video that is getting through.  Often, the receiver and demodulator will be removing some of this energy - and the "1 MHz" and "300 kHz" filters will too.  To counteract this, you may try tuning a few hundred kHz lower in frequency than you normally would to see if the picture is any better.  You'll probably lose sound - but if this effect is, in fact, occurring, you probably wouldn't have had any sound anyway.

One problem that might pop up when doing this has to do with the video demodulator itself.  The R-7000's TV demod is a synchronous demodulator - and this type of demodulator has a significant performance advantage (6dB or so) over the older-style "envelope detectors" in weak signal conditions.  With this scheme, the demodulator itself locks on to the video carrier and regenerates a "pristine" local version of it:  It can do this - even on weak signals - because it need only lock onto the carrier itself which is something that can be done using narrowband techniques which are beyond the scope of this document.

What can happen is that, with the limited "lock range" of the synchronous carrier regenerator, the demodulator may not track very far when you tune around"  The fact that a very low I.F. frequency of 15.2 MHz is being used (instead of the "standard" 45.75 MHz) means that the "lock range" is reduced even further.  A possible addition to an R-7000's demodulator may be an extra control to allow "tweaking" of the demodulator.

Some possible accessories:

For the avid TV DXer, I can think of a few accessories that might help one pull out the weak video signals:

• It should be practical to build a continuously variable bandwith TV filter.  There are several approaches that one could take:
• Use an approach similar to the "Variable Bandwidth Tuning" (VBT) found in many amateur rigs.  In this scheme, there are two I.F. filters in different I.F. stages.  When they are "lined up" with each other, the resulting bandwidth is that of the narrower of the two filters.  If one "slides" one of the filters (by twiddling the inter-I.F. conversion frequency for example) then what you get is a passband in which the upper edge is defined by one filter and the lower edge is defined by another.  Theoretically, one could adjust to as narrow a bandwidth as you wanted.  Practically speaking you can't really get good performance when the filter is "narrow" (as owners of rigs with VBTs will attest) before the effects of the (lack of) sharpness of the edges of the filters starts to become a problem.
• Have a varactor-tuned filter with a design similar to the one above.  This is actually a fairly attractive approach, but is somewhat limited in its range.  For example, one could build a filter that could be adjusted from, say, "100 kHz" to "1 MHz" in width.  The "gotcha" here is that it would be far more complicated than having the varactors tuned with several potentiometers:  The precise voltages for tuning the various sections for certain bandpass responses would be rather awkward.  Theoretically, one could have a potentiometer for each varactor stage and a multi-ganged switch and have a fairly broad selection of "filters" to chose from.  If I were to do it, I'd probably use a simple microprocessor (like a PIC) and have it generate control voltages - and then program various filter settings into it.  One could then have the PIC do an interpolation between the various settings.  (That would be another entire project, though...)
• Have a high-pass video enhancing filter with an adjustable "knee" with the "knee" being that frequency above which the emphasis occurs.  Being able to adjust the frequency at which emphasis begins (as well as the magnitude of the high-frequency emphasis) allow maximum flexibility under the various conditions.  (That is, being able to twiddle knobs until the signal looks the best...)  This box would also be capable of the reverse - that is, attenuation of frequencies above the "knee" frequency.

Note:

The pictures (above) that show a signal received through a "1 MHz" and "300 kHz" filter were done using the filter described on the "IF Bandpass Filtering of AM TV Signals" page - which is not this filter.  As soon as Les is able to get some actual off-air pictures, they will be posted.
 

This page is a work-in-progress.  It will be completed as time permits.

Acknowledgments:  I'd like to thank Les Rayburn, N1LF, for helping me test this filter.  (I would never have built it unless he'd asked me...)

Other related web sites:

• The "IF Bandpass Filtering of AM TV Signals" page has more insight on this method of receiving weak/degraded TV signals.
• Bandwidth reduction of DXTV Video - This page shows another method of reducing vision bandwidth in a more conventional set.

Any comments or questions?  Send an email!