Question: RF Speech Processing
WB3U
inside a transmitter. They clip this signal, then filter it with
a second IF filter. In a 9 MHz IF, the harmonics generated by the
clipping process fall around 18 MHz, 27 MHz etc., so they're easily
removed by the second filter. I'm beginning to wonder if this is
really necessary though, particularly if the processor is an outboard
accessory.
For example, say the outboard unit used a 6 MHz carrier oscillator,
fed to a balanced modulator, in order to produce a 6 MHz double
sideband signal. If that signal was clipped, won't the harmonics
fall on 12 MHz and above? If so, the clipped signal could be
filtered with an L/C low-pass filter, then converted back to
baseband (audio) using a product detector and the same 6 MHz
oscillator. This would eliminate the cost of two expensive filters.
I've never seen this done, so maybe I'm overlooking something.
Anyone know of a reason it won't work?
73,
Jack WB3U
Wes Stewart
>From: James Garland <gar...@ohstpy.mps.ohio-state.edu>
>Subject: Re: Question: RF Speech Processing
>Date: 5 Jan 1996 20:47:33 GMT
>That's a great idea, Jack, but I'm afraid you're not the first to think
>of it. As I recall, there were several audio processors on the market
>during the 70's which used the basic idea: mix an audio-derived DSB
>signal up to an IF (usually 50-500KhZ), clip it, then mix it back to
>audio. ETO sold one of these units, if I remember correctly, as did
>VOMAX. I also recall reading a do-it-yourself article on the concept in
>one of the ham mags -- probably Ham Radio.
>Jim W8ZR
The Comdel CSP-11 was another one.
There are some intermodulation products to contend with tho. Maybe Bill Sabin
will help us here.
Some references:
"R.F. Clippers for S.S.B." by William Sabin, W0IYH, QST, July 1967.
"Ordinary and Processed Speech In S.S.B. Application" by Harold Collins,
W6JES, QST, Jan 1969.
"Speech Clipping in Single-sideband Equipment" by Walter Schreuer, K1YZW, ham
radio magazine, Feb 1971.
"Performance of RF SPeech Clippers" by Leslie Moxon, G6XN, ham radio magazine,
Nov 1972.
"Split-band Speech Processor" by Wes Stewart, N7WS, ham radio magazine, Sept
1979.
W8JI Tom
ja...@pinetree.microserve.com (WB3U) writes:
>
>For example, say the outboard unit used a 6 MHz carrier oscillator,
>fed to a balanced modulator, in order to produce a 6 MHz double
>sideband signal. If that signal was clipped, won't the harmonics
>fall on 12 MHz and above? If so, the clipped signal could be
>filtered with an L/C low-pass filter, then converted back to
>baseband (audio) using a product detector and the same 6 MHz
>oscillator. This would eliminate the cost of two expensive filters.
>
>I've never seen this done, so maybe I'm overlooking something.
>Anyone know of a reason it won't work?
>
>73,
>Jack WB3U
It will work.
Gary Coffman
>
>Rf speech processors normally use the SSB signal already present
>inside a transmitter. They clip this signal, then filter it with
>a second IF filter. In a 9 MHz IF, the harmonics generated by the
>clipping process fall around 18 MHz, 27 MHz etc., so they're easily
>removed by the second filter. I'm beginning to wonder if this is
>really necessary though, particularly if the processor is an outboard
>accessory.
>
>fed to a balanced modulator, in order to produce a 6 MHz double
>sideband signal. If that signal was clipped, won't the harmonics
>fall on 12 MHz and above? If so, the clipped signal could be
>filtered with an L/C low-pass filter, then converted back to
>baseband (audio) using a product detector and the same 6 MHz
>oscillator. This would eliminate the cost of two expensive filters.
>
>I've never seen this done, so maybe I'm overlooking something.
>Anyone know of a reason it won't work?
I haven't worked through the math, but I think the problem is
with clipping DSB. If the clipping isn't perfectly symmetric,
IE both positive and negative frequency, some horrible *in-band*
distortion products are likely to be produced.
Gary
--
Gary Coffman KE4ZV | You make it, | gatech!wa4mei!ke4zv!gary
Destructive Testing Systems | we break it. | emory!kd4nc!ke4zv!gary
534 Shannon Way | Guaranteed! | ga...@ke4zv.atl.ga.us
Lawrenceville, GA 30244 | |
WB3U
>The Vomax was a split band audio processor which is superior to RF
>processing. I still use a modified VOMAX with my 751A.
Tom, could you elaborate on this? I was surprised by your statement,
given the 8 dB or so improvement possible with RF processing. Of
course, I can see how splitting the audio spectrum to enable effective
filtering of harmonics could be as effective, but not more so. BTW,
I've missed the articles on the split band approach due to two
absences from the hobby. If the explanation is complex, I'll try to
round them up.
>Wes N7WS had a nice article in some magazine years ago, it was good
>enough to be copied (hopefully he was compensated) and sold by two
>other companies.
>
>I see Gary's point. If an unfiltered DSB signal is clipped and re-
>mixed to audio, it will have extra unnecessary distortion products in
>the output.
>But I have no feel for the level.
I thought about that too, but I don't see how the distortion products
could fall in the RF passband. For instance, let's say the balanced
modulator is operating at 6 MHz and that it's modulated with a 1 KHz
tone. The output will be two carriers (DSB), one at 5,999 KHz and
the other at 6,001 KHz. What combination of those two carriers
produces a distortion product that falls near the two fundamentals?
Incidentally, there was an RF processor showcased in the October '95
issue of CQ. It's made in by GFS Electronics in Australia and it's
supplied on a small PC board, intended for internal mounting.
According to the short description, it connects in series with the
microphone audio. Looking at the picture, there does not appear to be
a crystal filter on the board. Anyone know anything about this unit?
73,
Jack WB3U
W8JI Tom
<gar...@ohstpy.mps.ohio-state.edu> writes:
>That's a great idea, Jack, but I'm afraid you're not the first to think
>of it. As I recall, there were several audio processors on the market
>during the 70's which used the basic idea: mix an audio-derived DSB
>signal up to an IF (usually 50-500KhZ), clip it, then mix it back to
>audio. ETO sold one of these units, if I remember correctly, as did
>VOMAX. I also recall reading a do-it-yourself article on the concept in
>one of the ham mags -- probably Ham Radio.
>Jim W8ZR
>
>
Hi Jim,
The Vomax was a split band audio processor which is superior to RF
processing. I still use a modified VOMAX with my 751A.
Wes N7WS had a nice article in some magazine years ago, it was good enough
to be copied (hopefully he was compensated) and sold by two other
companies.
I see Gary's point. If an unfiltered DSB signal is clipped and re-mixed to
audio, it will have extra unnecessary distortion products in the output.
But I have no feel for the level.
Splitting into sub-bands, processing, filtering, and re-combining is
always the best approach. As long as correct phase between signal channels
is maintained. Hopefully Wes will show up, since he designed a system.
73 Tom
Murray Kelly
harmonics at the double freq etc which remix with the original in all
the possible combinations. Unfortunately the opposite sidebands of the
fundamental and the first harmonic mix and trash each other before
either a: going thru the side band filter or b: being detected back to
audio.
The result is no better than audio clipping.
What I fail to understand is why designers use such extreme clipping
to actually make the harmonic generation worse. The b-e junction is
often used for this purpose, whereas I would have thought something
a bit 'slusshy' would have generated less clipping/mixing/hash?
Murray Kelly. vk4aok
Gary Coffman
> w8j...@aol.com (W8JI Tom) wrote:
>>
>>I see Gary's point. If an unfiltered DSB signal is clipped and re-
>>mixed to audio, it will have extra unnecessary distortion products in
>>the output.
>>But I have no feel for the level.
>
>could fall in the RF passband. For instance, let's say the balanced
>modulator is operating at 6 MHz and that it's modulated with a 1 KHz
>tone. The output will be two carriers (DSB), one at 5,999 KHz and
>the other at 6,001 KHz. What combination of those two carriers
>produces a distortion product that falls near the two fundamentals?
Ok, here's the problem. When you go to demodulate DSB back to
audio, the output is the vector sum of *both* sidebands with the
reintroduced carrier. You know what AM sounds like under selective
fading, well the same thing is possible here if the clipping of
the upper and lower sidebands isn't precisely symmetrical in both
the amplitude and frequency domains. The latter is what's different
about DSB. The sidebands are *mirror* images in the frequency domain,
so the clipping also has to be mirror imaged in the frequency domain.
That's not necessary with SSB.
Depending on how the clipper is implemented, there can be considerable
differences induced in the phase/amplitude envelopes of one sideband and
its mirror image. Those differences will show up as in-band distortion
products. I don't have a feel for how great a problem this may be, but
I'm sure it is a problem (it is for AM broadcast transmitters using
compression), and there's no way to simply filter out these components.
My gut feel is that it's important to use a carrier as high in frequency
as possible for modulating and demodulating the DSB signal so that
the percentage bandwidth is minimized and possible asymmetries in
the frequency domain are reduced.
Another potential problem is reciprocal mixing in the demodulator.
The clipped sidebands will generate a *pair* of harmonics for each
audio generated product. To take the steady state example you pose,
the LSB will have a harmonic at 11.998 MHz while the USB will have
a harmonic at 12.002 MHz. Those two can *also* mix in the demodulator
yielding a difference product of 4 kHz which will be in the output
audio passband. Now that particular product is high enough to be
audio filtered in the output, but consider if there is a 500 Hz
component present in the input audio, its product will fall on
2 kHz in the output, and that's in the speech band. The obvious
solution to this problem is a lowpass filter between the clipper
and the demodulator. But that lowpass filter has to be pretty
good to eliminate sensible products in the output. The lowpass
in a SSB clipper doesn't have to be quite as good since it doesn't
have to deal with paired products, IE there are only half as many
potential mixing products that can fall in the speech bandwidth
of the output with SSB *and* they are unrelated in phase, so
there's a potential for at least 6 db worse distortion products
out of a DSB clipper than a SSB clipper.
It's a rule of thumb that products more than 20 db below the
average level of an audio signal are not sensible for communications
purposes. So we need a lowpass with at least 20 db of attenuation in
the SSB case, but at least 26 db in the DSB case to reduce this in-band
distortion below sensible levels. I don't think this is a big deal, and
it rapidly gets better as the order of the harmonics increase, but it
is another disadvantage of the DSB clipper approach. And we should
remember that clipping raises the average level of the audio (that's
why we're doing it), so 8 db of compression makes our filter requirement
8 db greater too. A clean DSB clipper thus may need a filter of 34 db.
Wes Stewart
>From: ga...@ke4zv.atl.ga.us (Gary Coffman)
>Subject: Re: Question: RF Speech Processing
[snipped to save BW, or was that clipped?]
>Ok, here's the problem. When you go to demodulate DSB back to
>audio, the output is the vector sum of *both* sidebands with the
>reintroduced carrier. You know what AM sounds like under selective
>fading, well the same thing is possible here if the clipping of
>the upper and lower sidebands isn't precisely symmetrical in both
>the amplitude and frequency domains. The latter is what's different
>about DSB. The sidebands are *mirror* images in the frequency domain,
>so the clipping also has to be mirror imaged in the frequency domain.
>That's not necessary with SSB.
I have this crazy way of looking at things sometimes. I just said to myself,
"Self, if you *wanted* to clip each sideband in this DSB signal independently,
what is the likelyhood that you could do it with a simple clipping
circuit?" And I answered, "Beats me :-)"
I think it's too late to use a low-pass after the clipper. The intermod
products can (will) be generated in the clipper. After all, it's a
highly non-linear device, perfectly suited for mixing. For example:
Two audio tones, 500 and 1000 Hz applied to the balanced mixer along with
a 6 MHz carrier. Output is 6 MHz (suppressed), 5.999, 5.9995, 6.0005 and
6.001 MHz. Assume an USB filter. Output is now 6.0005 and 6.001 MHz.
Clip the hell out of it and get 6.0005, 6.001, 12.001, 12.002, 18.0015,
18.003... Also, get (12.002-6.0005 = 6.0015) and (18.003-12.001 = 6.002)
which both fall in the passband of the following LPF and BPF. It might be
argued that if the clipping is perfectly symmetrical, there won't be a 12.001
and 12.002, but (18.0015-6.0005 = 12.001) and (18.003-6.001 = 12.002) so there
they are. The split-band approach doesn't suffer this particular problem.
One can further argue (correctly) that this example is steady state as in
two-tone testing and not much of a problem with intermittent speech. However,
steady state is certainly the easiest way to compare systems.
[more clipping]
>Gary
Wes -- N7WS
Wes Stewart
>In article <4ck2p5$k...@pacific.mps.ohio-state.edu>, James Garland
><gar...@ohstpy.mps.ohio-state.edu> writes:
>>That's a great idea, Jack, but I'm afraid you're not the first to think
>>of it. As I recall, there were several audio processors on the market
>>during the 70's which used the basic idea: mix an audio-derived DSB
>>signal up to an IF (usually 50-500KhZ), clip it, then mix it back to
>>audio. ETO sold one of these units, if I remember correctly, as did
>>VOMAX. I also recall reading a do-it-yourself article on the concept in
>>one of the ham mags -- probably Ham Radio.
>>Jim W8ZR
>>
>>
>Hi Jim,
>The Vomax was a split band audio processor which is superior to RF
>processing. I still use a modified VOMAX with my 751A.
>Wes N7WS had a nice article in some magazine years ago, it was good enough
>to be copied (hopefully he was compensated) and sold by two other
>companies.
I wish!
>I see Gary's point. If an unfiltered DSB signal is clipped and re-mixed to
>audio, it will have extra unnecessary distortion products in the output.
>But I have no feel for the level.
I posted on this before, but hadn't read Jack' s post
carefully enough to see that he was talking about a *DSB*
signal. The Comdel approach that I mentioned was an *SSB*
processor. As Gary and Tom note, there will probably be a
distortion penalty to pay. You also have to be very careful
to have the clipping occur only in the clipping circuit, not
elsewhere where it is uncontrolled.
>Splitting into sub-bands, processing, filtering, and re-combining is
>always the best approach. As long as correct phase between signal channels
>is maintained. Hopefully Wes will show up, since he designed a system.
I'm here!
I agree, although, you must be very careful to minimize
lower frequency distortion ahead of the first band-splitting
filters. Distortion products are harmonics and if the input
signal contains them, they can pass into one of the higher
frequency filters, where if they are small enough, they
remain unclipped. Meanwhile, the higher amplitude, lower
frequency signal is clipped, ie. its amplitude is reduced.
Coming out of the second set of filters, the two signals are
combined and... oops; compared to the fundamental, the
harmonic amplitude has actually increased. The distortion
has increased by the clipping ratio; exactly what we don't
want. Of course, we usually analyze these things under
steady-state conditions. With speech, this is much less of
a problem.
>73 Tom
73, Wes
Some references:
"R.F. Clippers for S.S.B." by William Sabin, W0IYH, QST,
July 1967.
"Ordinary and Processed Speech In S.S.B. Application" by
Harold Collins, W6JES, QST, Jan 1969.
"Speech Clipping in Single-sideband Equipment" by Walter
Schreuer, K1YZW, ham radio magazine, Feb 1971.
"Performance of RF SPeech Clippers" by Leslie Moxon, G6XN,
ham radio magazine, Nov 1972.
Letter to ham radio magazine, by L. R. Newsome, VK4LR, May
1975, pp 75-76.
"Split-band Speech Processor" by Wes Stewart, N7WS, ham
radio magazine, Sept 1979.
Letter to ham radio magazine, by W. Schreuer, with response
by W. Stewart, N7WS, Feb, 1980,
WB3U
RF processor in my previous rig was very effective and I'm
looking for a way to build an inexpensive outboard replacement.
Criticisms of the technique that I described are exactly what is
needed. I think it would be useful to explore the various points
that have been made in greater depth.
ga...@ke4zv.atl.ga.us (Gary Coffman) wrote:
>You know what AM sounds like under selective fading, well the same
>thing is possible here if the clipping of the upper and lower
>sidebands isn't precisely symmetrical in both the amplitude and
>frequency domains. The latter is what's different about DSB.
<snip>
>My gut feel is that it's important to use a carrier as high in
>frequency as possible for modulating and demodulating the DSB signal
>so that the percentage bandwidth is minimized and possible
>asymmetries in the frequency domain are reduced.
This is the first thing to consider, because it relates specifically
to the decision of whether to use DSB. As in the previous example,
let's say the carrier oscillator is 6 MHz. The DSB passband for audio
frequencies out to 3 KHz will be 6 KHz. Therefore, the passband is
0.1% of the frequency of operation. If there are no tuned circuits
between the balanced modulator and the clipping diodes, will frequency
response irregularities due to non-symmetrical clipping really be a
consideration? To go a step further, let's raise the carrier
oscillator to something around 30 MHz. Now the passband width is only
0.02% of the frequency. Surely that would prevent this type of
problem?
Looking at this somewhat empirically, the RF processor in my previous
rig clipped the SSB signal at an IF frequency of 3.18 MHz. Because it
was SSB, the passband width was approximately 3 KHz. That means the
passband width was 0.094%, only slightly less than the 0.1% for the
DSB processor at 6 MHz. If this percentage bandwidth can be
symmetrically clipped for SSB at 3.18 MHz, would you agree that non-
symmetrical clipping probably won't be a factor with DSB at 6 MHz,
assuming the circuitry is otherwise similar?
>Another potential problem is reciprocal mixing in the demodulator.
<snip>
>The obvious solution to this problem is a lowpass filter between the
>clipper and the demodulator.
<snip>
>A clean DSB clipper thus may need a filter of 34 db.
Assuming the worst-case scenario of 34 dB is correct, the second
harmonic products at 12 MHz are still a good distance away from the
fundamental. Wouldn't something like a 7- or 9-element Chebyshev
low-pass filter provide this degree of rejection?
73,
Jack WB3U
WB3U
> ... you must be very careful to minimize lower frequency distortion
>ahead of the first band-splitting filters. Distortion products are
>harmonics and if the input signal contains them, they can pass into
>one of the higher frequency filters, where if they are small enough,
>they remain unclipped.
w8j...@aol.com (W8JI Tom) wrote:
>Splitting into sub-bands, processing, filtering, and re-combining is
>always the best approach. As long as correct phase between signal
>channels is maintained.
This points out some shortcomings in the split-band scheme that I
hadn't thought of. I do have a few comments.
First, maintaining the correct phase at the output of each filter is
probably impossible because the shift will vary across each filter's
passband. I'm not sure how significant this phase shift will be in
terms of the percieved quality of the recovered audio, but having been
in the audio business a number of years, it's a troublesome concept.
Multiple-band phase shift is one of the characteristics of graphic
equalizers said by many (including yours truly) to degrade their sound
quality. This may not be a significant issue in this application,
but I'd be very interested in a critical listening test of the audio
quality of one of these processors (with the filters active but the
clipping turned off). To clarify this a little, I'm not just referring
to the amplitude cancellation that takes place as a result of phase
shift, but also to the audible effects of the phase shift itself.
Second, the possibility that the split-band processor might increase
the overall percentage of harmonic distortion in the manner described
above can't be overlooked simply by virtue of a low-distortion
preamp/filter circuit. Most harmonic distortion encountered in audio
is produced in microphones and speakers, not in the electronics. In
this application, THD of the microphone will be impacted in the same
manner as harmonic distortion produced in the input circuitry.
Again, I don't know just how significant this factor is. In
double-blind A/B tests, it's been shown that listeners often cannot
identify harmonic distortion in music until it is above 5%. Below
that level, most listeners detect a difference between the pure and
intentionally-distorted programming, but they fail to identify it as
distortion. In many cases, they will describe a signal with 1% or 2%
THD as sounding louder than the original, but not as being distorted.
Given the fact that speech lacks the sustained tones found in music, I
suspect it is affected even less in terms of perceived distortion.
73,
Jack WB3U
Bill Sorsby
ja...@pinetree.microserve.com says...
>
> w8j...@aol.com (W8JI Tom) wrote:
>
>>The Vomax was a split band audio processor which is superior to RF
>>processing. I still use a modified VOMAX with my 751A.
>
>given the 8 dB or so improvement possible with RF processing. Of
>course, I can see how splitting the audio spectrum to enable effective
>filtering of harmonics could be as effective, but not more so.
I've still got a VOMAX which I used for years. In the tests I ran with it I
was never convinced that it accomplished nearly as much as was claimed for
it. I used it for whatever marginal benefit it might provide and because it
didn't distort my signal too badly (although some people complained!)
The concept of speech compression is a fascinating subject. It is easy to
compress a signal so that the average power is increased, but improving
comprehension (the desired end result) is another matter; the two don't
necessarily go together. In some instances compression can be
counterproductive to comprehension or intelligibility. What happens is that
compression of mid-band speech frequencies (generally not critical to
comprehension) can drown out some of the higher speech frequencies (which are
critical to comprehension). I believe this to be one of the shortcomings with
the VOMAX and the reason why the on-air tests I ran never gave a decisive
advantage to the VOMAX.
I've never had the occasion to use RF processing (other than simple ALC) so I
can't provide a measure of comparison here.
For what it's worth.
Regards,
Bill Sorsby, N5BU
Wes Stewart
>From: bill....@dlep1.itg.ti.com (Bill Sorsby)
>Subject: Re: Question: RF Speech Processing
[snip]
>I've still got a VOMAX which I used for years. In the tests I ran with it I
>was never convinced that it accomplished nearly as much as was claimed for
>it. I used it for whatever marginal benefit it might provide and because it
>didn't distort my signal too badly (although some people complained!)
As the designer of a unit similar to the VOMAX, I feel compelled to respond.
The literature I have on VOMAX claims 10-12 dB improvement. One might ask
what is meant by "improvement", to which I would respond, based on Schreuer's
February 1971, Ham Radio Magazine article: intelligibility gain. OK, what does
this mean, you ask. Good question. Difficult to answer.
>The concept of speech compression is a fascinating subject. It is easy to
>compress a signal so that the average power is increased, but improving
>comprehension (the desired end result) is another matter; the two don't
>necessarily go together.
Well, it isn't easy to do it correctly:-) Increased average power and improved
readability are not necessarily mutually exclusive though. I normally dislike
anecdotal testimonials, but I don't know how to avoid them in this case, so a
couple of personal examples.
When I first tried my processor, I was using a 100W transceiver. I didn't own
a linear amplifier, but I had the loan of one. I ran a number of on-the-air
tests whereby I would ask the other op to compare signal strengths and
intelligibility. I would first use the 100W with no processing. I would then
turn on the amplifier (SB220) and get a report just to verify how much
signal strength change he reported. I would then tell the other guy I was
trying out this new audio filter design and I would like to know what he
thought. I would turn on the processor and TURN OFF the amp and ask for a
report. Almost invariably, the signal strength report was the same and the
comments would be something like "The audio has more punch" or "You sound a
little bassier" or "You're not quite so natural sounding", etc. I never got a
report that suggested that I had dropped power by 10 dB.
I serendipitously chanced upon another fellow, who was using my design, on one
of those old ten meter paths where the signal is S1 but Q5. I asked him to
turn off the processor so I could get a feel for how it sounded in the other
direction. When he did, he was GONE. I mean gone. There was nothing. He had to
turn it back on to tell me to go ahead. We did this several times in both
directions with the same result.
Again the disclaimer. I do mostly weak signal VHF work and I have no lab data
to back this up, but I'm convinced, that a good processor, intelligently
operated, is equivalent to adding a linear amplifier.
> In some instances compression can be
>counterproductive to comprehension or intelligibility. What happens is that
>compression of mid-band speech frequencies (generally not critical to
>comprehension) can drown out some of the higher speech frequencies (which are
>critical to comprehension). I believe this to be one of the shortcomings with
>the VOMAX and the reason why the on-air tests I ran never gave a decisive
>advantage to the VOMAX.
I'm not clear on what you mean by this. Advantage over what? There was one
letter to Ham Radio Magazine from K3ND which discussed comparative tests
between the VOMAX and my design. Gale reported a preference for my design in
90% of the cases. Most of his respondents believed that the VOMAX frequency
response was too narrow. Perhaps this is the problem in your case. There is a
trend to narrower and narrower filters in modern radios. Coupling these with
further BW limiting in an outboard processor might cause problems.
>I've never had the occasion to use RF processing (other than simple ALC) so I
>can't provide a measure of comparison here.
>For what it's worth.
>Regards,
>Bill Sorsby, N5BU
73, Wes -- N7WS
W8JI Tom
ja...@pinetree.microserve.com (WB3U) writes:
>>The Vomax was a split band audio processor which is superior to RF
>>processing. I still use a modified VOMAX with my 751A.
>
>Tom, could you elaborate on this? I was surprised by your statement,
>given the 8 dB or so improvement possible with RF processing. Of
>course, I can see how splitting the audio spectrum to enable effective
>filtering of harmonics could be as effective, but not more so.
I'll defer to Wes on this, but briefly when clipping is done at audio, you
can process small bands that are are less than one octive wide. Harmonics
from clipping not only fall outside the passband of channels, but mixing
is reduced because there is little likelyhood two tones will appear in one
clipping channel at any one time.
RF clipping does not do this, unless the processor has several very narrow
filters feeding separate clippers.
73 Tom
W8JI Tom
ja...@pinetree.microserve.com (WB3U) writes:
>
>First, maintaining the correct phase at the output of each filter is
>probably impossible because the shift will vary across each filter's
>passband. I'm not sure how significant this phase shift will be in
>terms of the percieved quality of the recovered audio, but having been
>in the audio business a number of years, it's a troublesome concept.
>Multiple-band phase shift is one of the characteristics of graphic
>equalizers said by many (including yours truly) to degrade their sound
>quality. This may not be a significant issue in this application,
>but I'd be very interested in a critical listening test of the audio
>quality of one of these processors (with the filters active but the
>clipping turned off). To clarify this a little, I'm not just referring
>to the amplitude cancellation that takes place as a result of phase
>shift, but also to the audible effects of the phase shift itself.
>
>
Hi Jack,
Vomax did this by having the "after clipping" filter shift the phase back
at the same slope as the pre-clipping filter. At RF phase control would be
easier, but the filter cost would be pretty high.
I don't see how phase shift would be a problem with DSB as long as a
narrow RF filter is avoided, but then the worry would become the lack of
signal filtering. There must be some BW control before clipping. Perhaps
an audio bandpass filter before the balanced modulator would help, since
as Wes said, the lows should be rolled off a bit before clipping and we
all know clipping out of band audio just adds needless distortion.
I'm working on a DSP based processor now, in between 999 other things.
73 Tom
Gary Coffman
>In article <1996Jan7.2...@ke4zv.atl.ga.us> ga...@ke4zv.atl.ga.us (Gary Coffman) writes:
>>Ok, here's the problem. When you go to demodulate DSB back to
>>audio, the output is the vector sum of *both* sidebands with the
>>reintroduced carrier. You know what AM sounds like under selective
>>fading, well the same thing is possible here if the clipping of
>>the upper and lower sidebands isn't precisely symmetrical in both
>>the amplitude and frequency domains. The latter is what's different
>>about DSB. The sidebands are *mirror* images in the frequency domain,
>>so the clipping also has to be mirror imaged in the frequency domain.
>>That's not necessary with SSB.
>
>I have this crazy way of looking at things sometimes. I just said to myself,
>"Self, if you *wanted* to clip each sideband in this DSB signal independently,
>what is the likelyhood that you could do it with a simple clipping
>circuit?" And I answered, "Beats me :-)"
Well, I don't know of any deliberate simple controlled way to do it either,
but what I'm talking about is unintentional and undesired behavior. In
broadcast we call it incidental phase modulation, and it's the result of
sending the signal through a somewhat non-linear system. The phase shift
induced in the two sidebands is different, because they are on different
frequencies, and varies with signal level. When they recombine in the
demodulator, their vector sum is different from the input signal too.
Through a system that deliberately operates non-linearly, the effect should
be much greater.
I agree with this, of course, but I think this high order multiple
mix will be of considerably lower amplitude than the straight products
the filter would remove. You'd need some sort of circulating tank to
keep the base products around to remix with each other to form tertiary
products sensible to the demodulator. I don't think most clipper circuits
have enough tank to do that to any significant degree, but perhaps the
very filter I'm suggesting might supply that tank.
Kein{nen Paul
ja...@pinetree.microserve.com (WB3U) wrote:
> w8j...@aol.com (W8JI Tom) wrote:
>>The Vomax was a split band audio processor which is superior to RF
>>processing. I still use a modified VOMAX with my 751A.
>
> Tom, could you elaborate on this? I was surprised by your statement,
> given the 8 dB or so improvement possible with RF processing. Of
> course, I can see how splitting the audio spectrum to enable effective
> filtering of harmonics could be as effective, but not more so.
I have wondered about this too. I know that multi-band processors
are used to reduce the dynamic range (eg. in Dolby A and in more modern
variants), but usually it is about reducing the dynamic range from, say
60 dB to 40 dB for storage on tape etc.
However, if you are going to use aggressive clipping to reduce the
peak to average ratio, I do not see how tis would improve the situation
a lot. If you split the signal into multiple subbands, clip each of them,
filter out any out of band harmonics and combine the subbands into a
single stream for transmission, you can still get some nasty high peaks,
if the phase relationship is unfavorable.
Think about a simple waveform consisting of frequencies f and 2f and both
have the positive peak at the same time (this resemples the speech
waveform). aAssume that f and 2f fall into different subbands as clean
stable sinus signals with no need for clipping. Summing these signals
and you get a positive peak twice as high as the individual signal. This
corresponds to four time peak power in an SSB transmitter compared
to a single tone. You have gained nothing with a multiband clipper.
[About DSB clipping]
> I thought about that too, but I don't see how the distortion products
> could fall in the RF passband. For instance, let's say the balanced
> modulator is operating at 6 MHz and that it's modulated with a 1 KHz
> tone. The output will be two carriers (DSB), one at 5,999 KHz and
> the other at 6,001 KHz. What combination of those two carriers
> produces a distortion product that falls near the two fundamentals?
How about third (and 5th) order intermodulation products ?
2*6001-5999 = 6003 kHz and 2*5999-6001=5997 kHz and after detection
you have two signals at 3 kHz.
However, a single tone is not a realistic test signal for a clipper.
Assuming a simple two tone signal with 400 Hz and 1000 Hz tones, you get
in addition to the 3rd order products for the 1000 Hz tones (as above)
also the 3rd order products of the 400 Hz tones.
Calculating all the intermodulation products up to the 5th order,
will for the DSB case products produce within +/- 4 kHz of the DSB carrier:
<deleted 41 intermodulation product below 6.000,
which are images of those above 6.000>
3rd order 6.0002, 6.0010, 6.0012, 6.0016, 6.0018, 6.0024, 6.0030 MHz
5th order 6.0002, 6.0004, 6.0006, 6.0008, 6.0008, 6.0010, 6.0012,
6.0016, 6.0018, 6.0020, 6.0022, 6.0024, 6.0026, 6.0030,
6.0032, 6.0032, 6.0036, 6.0038 MHz
After the product detector you get intermodulation products practically
every 200 Hz.
In SSB clipping the only intermodulation products that fall within
4/- 4 kHz of the carrier are:
3rd order 5.9998, 6.0016 MHz
5th order 5.9992, 6.0022 MHz
if there is an USB filter after the clipper, the only audible IM products
are 1.6 kHz and 2.2 kHz.
If you run the test with more tones that are not harmonically related
to each other, the difference is even more evident.
Paul OH3LWR
--
Phone : +358-31-213 3657 Mail: Hameenpuisto 42 A 26
Internet: Paul.K...@cc.tut.fi FIN-33200 TAMPERE
Telex : 58-100 1825 (ATTN: Keinanen Paul) FINLAND
X.400 : G=Paul S=Keinanen O=Kotiposti A=ELISA C=FI
Gary Coffman
> ga...@ke4zv.atl.ga.us (Gary Coffman) wrote:
>
>>You know what AM sounds like under selective fading, well the same
>>thing is possible here if the clipping of the upper and lower
>>sidebands isn't precisely symmetrical in both the amplitude and
>>frequency domains. The latter is what's different about DSB.
>
>Looking at this somewhat empirically, the RF processor in my previous
>rig clipped the SSB signal at an IF frequency of 3.18 MHz. Because it
>was SSB, the passband width was approximately 3 KHz. That means the
>passband width was 0.094%, only slightly less than the 0.1% for the
>DSB processor at 6 MHz. If this percentage bandwidth can be
>symmetrically clipped for SSB at 3.18 MHz, would you agree that non-
>symmetrical clipping probably won't be a factor with DSB at 6 MHz,
>assuming the circuitry is otherwise similar?
I couldn't say, because one case is SSB and the other is DSB. There's
more than just bandwidth at stake here. SSB demodulation is just a
down-conversion mixing process. If there are phase shifts between
the highest and lowest frequencies, you just get a little time
distortion, IE the old click-boom effect familiar to users of horn
speakers. But DSB demodulation is different. There, the output
audio is the vector sum of the two sidebands with respect to the
reinserted carrier. If one sideband is phase delayed with respect
to the other, amplitude distortions and waveshape changes occur
in the output.
Any shaping done after modulation and before demodulation may
result in a phase delay between the matching components of the
sidebands, and that will in turn lead to amplitude and waveshape
distortion in the output audio for DSB. I believe the correct
factor we should be concerned about is called group delay.
The classic example of this occurs in a VSB TV transmitter.
You'll get a nasty spike on the tip of sync when you demodulate
the signal because one sideband is delayed differently than the
other through the transmitter. The cure requires *predistorting*
the modulating waveform so that the incidental phase delay
introduced by the system is cancelled out. The bandwidth for
a channel 11 transmitter and H sync is 0.015%, but that's not
the important measure. The important measure is the differential
delay through the system. That only needs to be a few degrees
at 200 MHz, a very small time difference indeed.
Claude Frantz
>I'm working on a DSP based processor now, in between 999 other things.
Can you give us more information about this work ?
Thanks es 73,
Claude
G.Mo...@massey.ac.nz
Interesting discussion ... I built one of Wes' Split band speech processors and I'm
still using it - works great - thanks Wes.
An idle thought regarding comprehension. Has anyone tried using formant filters
rather than just 300-3000 speechband filters? I'm mainly thinking about receive
filtering but possibly in a processor as well. I vaguely remember a simple RC filter
years ago (from NASA) that was supposed to put notches into the transmitted
audio where there shouldn't be any speech info.
Ring any bells?
Cheers
Giovanni ZL2BOI - currently discovering 40M after years with an 80/20 QRP SSB rig.
W8JI Tom
You wrote:
>Typically a MOV will be in hard conduction at 150% to 200% of its
>rated working voltage, IE for a V130 that would be 195 to 260 volts.
>It'll reach hard conduction a few nanoseconds after a transient reaches
>that value, and remain so until the transient decays (up to the peak>
>joule rating of the device).
Could you please give me an example of what the "hard conduction"
resistance is?
What is the "on" resistance at say...300 volts peak for a MOV that will
safely opeate on a normal 120 volt power line (say 180 volt peaks). Do you
have the exact specs?
The last time I looked at the data from manufacturers of MOV's, they were
about worthless! Transformers would have been into saturation way before
the MOV conducted, and the transformer's saturated impedance would have
been several times lower than even a very large MOV's "on" resistance.
Maybe that has changed in the past few years.
Do you have data that indicates a MOV that can protect a transformer
operated supply? I understand MOV's help on line operated devices without
a transformer (they are better than nothing), but what about if a
transforer is used?
73 Tom
Now assuming the PS being protected has
a dV/dt longer than a few nanoseconds, the transient will be safely
clamped before damaging voltage can pass through.
Remember a MOV isn't intended to be an overvoltage protector,
it's a *transient* surge protector. It won't protect a PS if
the line voltage rises to 140 volts for a sustained time, but
it will keep transient peaks below values that could punch
a junction or an electrolytic cap in the supply. (It is assumed
that a good PS design will use parts rated for peaks at least
150% to 200% of their planned sustained voltage.)
Gary
--
Murray Kelly
with the upper SB of the 1st harmonic. Add em together and it looks
like a clipped audio signal - the AF harmonics mix. The only way round this
is to use side band (single) and/or very soft clipping.
If one clips SSB the harmonics reflect away from the base band rapidly
and the mixing products are removed by the very broad 2nd SSB filter
after the clipper.
No probs.
Murray Kelly vk4aok.
W8JI Tom
(Gary Coffman) writes:
>
>I think you just analyzed it wrong. The dV/dt for a transformer is
>going to be such that it won't even begin to saturate in the 2 nS
>it takes the MOV to clamp. So it will never see the transient.
>
>Gary
Hi Gary,
I don't think I analyzed it wrong.
The rise time of the transformer actually helps prevent the coupling of
high frequency transients to the secondary (at least in the designs I am
familiar with) much better than a MOV ever could. You aren't trying to say
a typical 60 HZ transformer requires time to saturate, and responds very
well to transients?
Most transformers I'm aware of are operated *reasonably* close to core
saturation. If they are not operated that way, efficiency, regulation, and
space is wasted. No manufacturer likes to do that. Core saturation makes
it virtually impossible to couple transients to the secondary through
magnetic coupling. If the transient is large enough to do damage, the MOV
would likely be useless anyway since it's shunt resistance is so high
compared to the saturated winding impedance.
The on resistance between 6 and 20 ohms is about worthless across a power
line, unless the transient is pretty puny or comes from a high impedance
source. If the transcient is that frail, the transformer can handle it
fine. If the transient comes from a "stiff" or low impedance (high
current) source, the MOV has too high an impedance. The equipment is
history anyway!
I'm not saying MOV's are usless in all applications. They are great in
some applications, such as telephone lines (where the goal is to protect
the customer or house). They can certainly help protect line operated
transformerless supplies, and can reduce the chance of arcing from
extremely high common mode voltages between the line and chassis. MOV's
just don't do much to help things on the secondary side of a transformer.
A six ohm clamp isn't what I would consider particularly "hard" compared
to the impedance of a 20 amp 120 volt supply line.
73 Tom
Rich_S
once did a lot of his own research on power conditioning techniques.
Using a transient generator and scope, he had good luck using MOV's if
they were across all three inputs (line, neutral, and ground) and
preferably followed by a low-pass filter network. The LPF helped "slow
down" the pulse while the MOV went into conduction. All of this
preceded the transformer primary.
By the way, semiconductor devices such as "Transorbs" are much faster
but are also less robust and more expensive.
In <1996Jan13....@ke4zv.atl.ga.us> ga...@ke4zv.atl.ga.us (Gary
Coffman) writes:
>
>In article <4d7lvt$c...@newsbf02.news.aol.com> w8j...@aol.com (W8JI
Tom) writes:
>>
>>The on resistance between 6 and 20 ohms is about worthless across a
power
>>line, unless the transient is pretty puny or comes from a high
impedance
>>source. If the transcient is that frail, the transformer can handle
it
>>fine. If the transient comes from a "stiff" or low impedance (high
>>current) source, the MOV has too high an impedance. The equipment is
>>history anyway!
>>
>>I'm not saying MOV's are usless in all applications. They are great
in
>>some applications, such as telephone lines (where the goal is to
protect
>>the customer or house). They can certainly help protect line operated
>>transformerless supplies, and can reduce the chance of arcing from
>>extremely high common mode voltages between the line and chassis.
MOV's
>>just don't do much to help things on the secondary side of a
transformer.
>>A six ohm clamp isn't what I would consider particularly "hard"
compared
>>to the impedance of a 20 amp 120 volt supply line.
>
>You didn't read what I wrote, Tom. During the IEEE 8/20 transient,
>the MOV has an effective "on" resistance of 6.15e-6 ohm, that's
>*not* 6 ohms, that's 6 *millionths* of an ohm. This is dynamic
>behavior, of course, and only lasts for the 20 microseconds of
>the transient waveform. If the transient persists, steady state
>"on" resistance is around 20 ohms, but the device will self-
>destruct under that condition.
>
>MOVs certainly do work to protect transformer powered equipment,
>as the blown MOVs in my Astron supply attest. They gave their
>lives clamping the lightning transient, and the power supply
>survived. On reflection, MOVs of greater than 60 joule surge
>rating should have been in there. They wouldn't have died while
>doing their duty.
>
>You might think that the transformer would swallow the transient,
>but it won't. The core saturation has nothing to do with it. The
>transient will couple through the transformer via the inter-winding
>capacitance and puncture electrolytic capacitor junctions and
>solid state device junctions if the MOV is not present. The transient
>is very high frequency RF.
>
>A MOV can't be modeled as just a voltage dependent resistor. Its
>dynamic behavior is much more complex than that. It's made up of
>a sintered mix of electron donor and electron receptor materials.
>The 60 joule device for which I quoted specs has the ability to
>"eat" 3/4 of a coulomb of charge, and leak it away gradually
>and safely after the event. The best model for a MOV is that of
>a paralleled RC snubber, but that doesn't accurately portray the
>complete device physics either. Once the charge is "swallowed"
>it can't get out either way without going through the high "off"
>resistance of the device.
>
>Let me try a crude description of how a MOV works. Inside the
>MOV, picture a bunch of capacitor plates separated by insulating
>material. Initially, the "capacitors" have a neutral charge, and
>the insulation value is high enough to prevent sensible current
>flow. During a transient, the insulation breaks down, a
non-destructive
>(if within device joule ratings) arc forms which rapidly charges
>the capacitor plates. This distributes charge throughout the device.
>The potential between one "capacitor" and another is small enough
>that the insulation doesn't break down, and the current can't get
>back out except by very slow leakage. It "eats" charge, up to
>3/4 coulomb of it for the device I described. It can't do that
>indefinitely, of course, which is why it can only handle a transient
>of limited duration. Longer duration overvoltages will make the
>insulation arcing destructive rather than non-destructive and
>the device becomes conductive and destroys itself.
Gary Coffman
Kein{nen Paul
Wes Stewart (n7...@azstarnet.com) wrote:
> In article <4d0kr1$1...@proffa.cc.tut.fi> k23...@proffa.cc.tut.fi
> (Kein{nen Paul) writes:
>>Think about a simple waveform consisting of frequencies f and 2f and both
>>have the positive peak at the same time (this resemples the speech
>>waveform). aAssume that f and 2f fall into different subbands as clean
>>stable sinus signals with no need for clipping. Summing these signals
>>and you get a positive peak twice as high as the individual signal. This
>>corresponds to four time peak power in an SSB transmitter compared
>>to a single tone.
>
>This is what we call an amplifier with a gain of two. If this
>overmodulates the transmitter, we can reduce the audio gain control
>accordingly.
>
>> You have gained nothing with a multiband clipper.
>
>With your example, we can't call it a clipper because you specified no
>clipping is occurring. Obviously, there is nothing gained until we *do* allow
>clipping. That's what it's all about.
Allowing clipping does not change the situation.
If we set the clipping at +/- 1/2 the peak amplitude in each channel,
we get square waves with half the amplitude of the pure tones (f and 2 f).
Bandlimiting the channels, the harmonics of the square wave are filtered
out and we get sine waves at 1/2 the original amplitude. Summing these
signals, we get the original signal envelope which is attenuated by 6 dB.
Since the envolope is still the same, there is not an improvement in
peak-to-average ratio compared to the case of no clipping.
I admit that assuming that the f and 2f tones of the original signal are
a bit unrealistic. For typical voice signals the harmonics are weaker
and the high frequency channels will not limit as easily as the low
frequency channels. The net effect is a change in the spectral contents
of the signal favoring the high frequencies.
This is just a high-pass filter with the corner frequency controlled
by the signal level :-)
While it is interesting to debate the peak-to-average, readability is a
much more complex subject. Other factors, such as the quality of the voice
(male/female) the spectrum of the noise+interference at the receiving
site affect the readability.
Paul OH3LWR
W8JI Tom
want 2 years of work to be stolen by another company and their code
out to use in another "like product". >>
That statement indicates data was withheld just to protect "two years of
work" that went into the software.
So which reason was it really? Was data accidently left out by QST as a
mere oversight or space saving gesture, or did the author request it be
left out as a favor to protect "two years" of work?
This is begining to sound like someone danced around the truth a bit in
the earlier replies.
Tom
W8JI Tom
(Kein{nen Paul) writes:
>If we set the clipping at +/- 1/2 the peak amplitude in each channel,
>we get square waves with half the amplitude of the pure tones (f and 2
f).
>Bandlimiting the channels, the harmonics of the square wave are filtered
>out and we get sine waves at 1/2 the original amplitude. Summing these
>signals, we get the original signal envelope which is attenuated by 6 dB.
>Since the envolope is still the same, there is not an improvement in
>peak-to-average ratio compared to the case of no clipping.
>
>
No, what we get is a signal that has all the levels at different
frequencies limited at the same threshold, exactly what we try to do in
any processor!
The result is instantaneous "compression" or limiting of each frequency
sub-band to a predetermined level. It can then be re-mixed at the level we
desire without the addition of unwanted intermodulation products or
harmonics of the original input. We have simply removed all amplitude
variations in individual sub-bands and re-combined them at the ratios that
improve communications without adding unwanted byproducts.
The improvement is very real on speech, but not on a two tone test.
73 Tom
Gary Coffman
>Gary,
>Oppps. Your right, I misread your response.
>>Lets look at an IS-MPT-200 MOV.
>Who's component is that?
That's a Polyphaser part number, I don't know who makes it for them.
I expect it's GE.
>>It has a turn on voltage of 200 volts, a clamping time of 2 nS for a 1
>kV/nS IEEE >820/20 waveform, and has an effective "on" resistance of
>6.15e-6 ohm for the duration >of the IEEE 820/20 waveform (2 uS peak, 20
>uS tail).
>
>What is the peak voltage of the IEEE waveform? A simple low pass filter
>could do a great job on a sharp pulse like that. That's why I am mainly
>concerned with longer duration clamping ability. The engineering
>guidelines I must follow require not duplicating functions without cost
>justified benefits.
The IEEE 820/20 waveform looks like a 2 uS sinc^2 pulse with an asymtotic
decay tail over 20 uS. The pulse rises at 1 kV/nS for 1 uS, so it peaks
at 1 Megavolt (no load). Of course depending on the energy of the pulse
in joules, and the shunt impedance, it won't necessarily make it to 1 MV
into a load. This pulse is designed to mimic a nuclear EMP. The IEEE 587
pulse is designed to mimic a lightning surge transient and has a somewhat
slower risetime and a negative "ringing" component at 80% of peak positive
amplitude following the initial 2 uS sinc^2 pulse, and trails off in
typical damped wave fashion over a longer interval. The IEEE 820/20 pulse
is the more stringent test, and is what's typically used to qualify surge
suppressors. If the suppressor can handle it, the IEEE 587 pulse should
be child's play (but that's not always true, that negative "ringing"
component can stymie some suppressors).
And, there's yet another IEEE pulse shape, the IEEE 8/20 pulse, which
has a 8 uS rise and a 20 uS decay in essentially a triangular form.
(I may have inadvertently mixed this up with the IEEE 820/20 pulse
in an earlier post.) This pulse may be seen downstream of certain
types of transformers or entrance networks. Typically it will have
a peak voltage of only 6 kV, but will have essentially the same energy
as the IEEE 587 pulse.
>Without MOV's in equipment, and only using simple low pass filters (that
>also function as RFI filters), I have never seen a transient induced
>rectifier or component transient induced failure. But then all of my line
>operated equipment designs use transformers. I'm sure if they were line
>operated switching supplies a zener or MOV clamp would be a worthwhile
>addition.
I can show you the power supply choke out of a Harris 100 kW TV transmitter
that's charcoal thanks to a lightning transient that jumped the insulation
and established an ionized path. The bulk of the damage was done by stepped
up line current, of course, following the path established by the transient,
but the transient was the "ignition" source. If there had been adequate
suppression on the 480 line, the transient would never have been high
enough to start the ionization path that led to the failure.
Now you're probably right that a lowpass filter would have snubbed
the peak voltage, but it wouldn't have damped the energy at all,
just spread it out into a "ringing" voltage at the lowpass cutoff.
As I'm sure you know, that can have unexpected and undesired
results in a high energy circuit too.
>>The steady state "on" resistance is around 20 ohms, but if the overload
>is sustained >so that the steady state is achieved, the MOV will
>self-destruct in about
>>3/4 second. It only has a transient capability of 160 joules.
>
>The 50 / 60 Hz power transformers I work with do not pass pulses (in the
>differential mode) of 1 mS very well (if at all). They generally reach
>core saturation at around 1-1/2 to 2 times the design voltage. Even if I
>tried to use them as a pulse transformer with a stiffly sourced 1 kV 20 mS
>pulse, the transformer would be almost non-responsive to the pulse. The
>only danger is the common mode voltage causing the transformer to arc, but
>the line bypasses and terminal strips would fail first.
As I noted, core saturation plays no part with 1 kV/nS pulses. They
*capacitively* couple between the transformer windings, unless you
use a ferroresonant or Faraday shielded transformer of course. Note
that's *nanosecond* not *millisecond*. Transients almost by definition
are *very* fast pulses. They usually don't pack a lot of energy (which
is good or little bitty MOVs would be of no value), and don't do the
damage themselves in line operated equipment. What they do is serve as
junction puncturers or "ignition" sources. It's the ordinary line current
that then does the damage by following the path created by the transient.
>In article <4d96v0$r...@ixnews8.ix.netcom.com>, ion...@ix.netcom.com(Rich_S
>) writes:
>>Using a transient generator and scope, he had good luck using MOV's if
>>they were across all three inputs (line, neutral, and ground) and
>>preferably followed by a low-pass filter network. The LPF helped "slow
>>down" the pulse while the MOV went into conduction. All of this
>>preceded the transformer primary.
>>"Transorbs" are much faster but are also less robust and more expensive.
>
>I'm not familiar with Transorbs, but I am always looking for ways to
>improve protection of high power transformer-operated equipment. Who is
>the manufacturer of transorbs?
Transorbs are silicon devices, think of them as *very* fast zeners.
An ordinary zener doesn't have nearly fast enough response to be of
any use for transient suppression (turn on is typically in the 40 uS
range, the transient is gone by then). Transorb is someone's trademark
name (I don't recall whose at the moment). You'll find generic examples
of these devices in the 1N62XX number range. For example, the 1N6296A
is a 110 working volt protector with a 10A surge rating (1.5 joule),
and a turn on threshold of 152 volts. Turn on is in the 400-500
*picosecond* range. (Fast buggers.)
Transorbs have the fastest response, MOVs are next, and gas tubes
last. But the order of robustness and energy handling is the reverse.
Gas tubes can handle surge of very high energy while MOVs and Transorbs
can handle much less. For very very high energy surge, silicon carbide
suppressors can handle the most, but they are one-shot devices and
self-destruct (explosively, it's quite a sight) when they protect.
They're the best thing to use right at the facility entrance (and
the cheapest too). For most transients, they won't fire, and you still
need gas tubes, MOVs, or Transorbs downstream, but if the granddaddy
direct strike occurs, the silicon carbide protector will fire, and
probably save your plant.
Tim Mathis
cumulative and the primary failure mode is short circuiting of the the
MOV.
Is this true?
Tim
Glenn Thorn
<snip>
>So which reason was it really? Was data accidently left out by QST as a
>mere oversight or space saving gesture, or did the author request it be
>left out as a favor to protect "two years" of work?
The source code wasn't submitted to QST, the software flow chart was,
and they chose not to publish that. There was no accident. QST, being
the fine journal that it is, wouldn't practice the ethics implied.
They bench tested the unit to ensure it did what was claimed. They
aren't going to stake their reputation on something as trivial.
>This is begining to sound like someone danced around the truth a bit in
>the earlier replies.
I fail to understand the statement. Dwayne and Gene did a lot of
pondering on the release of the source before they submitted the
article. They (I can't speak for QST) didn't know peoples feelings on
the source code release. That is why they are releasing the code...
because people want it. All the comments posted and E-Mailed had a
voice and it was listened to. I never know if I did something right
or wrong unless there is feedback. Feedback is a good thing.
There is no dancing, right, or wrong to the situation. What ever
their reasons, they listen to comments. If someone doesn't like the
article, turn the page. Any future articles will have source code
(probley downloadable from the ARRL board).
Thank you and everyone for the great comments!
73,
Glenn
W8JI Tom
In article <1f7cc$12f11.27a@NEWS>, ke...@atc.ameritel.net (Glenn Thorn)
writes:
I wrote:
>>This is begining to sound like someone danced around the truth a bit in
>>the earlier replies.
>I fail to understand the statement.
Let me explain it then. Reading this thread, I first received the
impression there was no reason the code was omited other than space and a
lack of understanding people would be interested in it. I received the
impression QST omitted the flow chart basically on it's own.
As I understood your comments, you said the author was concerned about
"giving away two years of work to a manufacturer". That leaves me with the
impression the author's desire was to have the data left out.
If that was even part of the reason for the omission, then everyone should
have just said so instead of implying the omission was due to
forgetfulness or ignorance.
In my opinion, "protecting work" is a very poor reason to omit anything
from an friendly helpful educational article.
Tom
Booth, Thomas G
upper and lower case characters) is held by General Semiconductor
Industries, Inc. An alternate source of silicon transient voltage
suppressors is Microsemi Corporation - the trademark for their product line
is TAZ.
TGB
\\ The opinions expressed herein are my own. //
W8JI Tom
>As I noted, core saturation plays no part with 1 kV/nS pulses. They
>*capacitively* couple between the transformer windings, unless you
>use a ferroresonant or Faraday shielded transformer of course. Note
>that's *nanosecond* not *millisecond*. Transients almost by definition
>are *very* fast pulses. They usually don't pack a lot of energy (which
>is good or little bitty MOVs would be of no value), and don't do the
>damage themselves in line operated equipment. What they do is serve as
>junction puncturers or "ignition" sources. It's the ordinary line current
>that then does the damage by following the path created by the transient.
>
If the pulse is capacitively coupled, how does the turns ratio step it up?
73 Tom
Gary Coffman
The normal failure mode is shorted. That's good. If they failed
open, you might have no protection and not know it. A catastropic
failure usually results in the MOV *vaporizing*, so that sort of
failure is definitely an *open*. :-)
Damage can be cumulative, but if the transient is within the joule
limits of the device, it should cause no damage at all.
Roy Lewallen
; FWIW the trademark for TransZorb (this is the correct spelling and use of
; upper and lower case characters) is held by General Semiconductor
; Industries, Inc. An alternate source of silicon transient voltage
; suppressors is Microsemi Corporation - the trademark for their product line
; is TAZ.
Motorola also has a line of zeners characterized for transient protection.
They call theirs TVS (Transient Voltage Suppressors).
Roy Lewallen, W7EL
Gary Coffman
It doesn't, nor does it need to, a megavolt transient is quite sufficient
in itself.
W8JI Tom
(Gary Coffman) writes:
>It doesn't, nor does it need to, a megavolt transient is quite sufficient
>in itself.
>
>
How does a megavolt ultra short rise time transient make it past the line
bypass caps, terminal blocks outlet, fuse box and so on?
73 Tom
Gary Coffman
It often doesn't. The impedances across which it is expressed will determine
whether it will flash over there or not. The transient, for the case of
lightning, is a current pulse of up to 18,000 amperes, but of very short
duration (so energy is low). It doesn't take much of an impedance for that
pulse to translate to a very high voltage (a megavolt is not uncommon),
which can then flash across ordinary insulation or air gaps. The purpose
of transient suppressors is to give the pulse a defined and controlled
place to discharge rather than depending on chance to allow it to flash
somewhere unpredicted.
W8JI Tom
(Gary Coffman) writes:
>It doesn't take much of an impedance for that
>pulse to translate to a very high voltage (a megavolt is not uncommon),
>which can then flash across ordinary insulation or air gaps. The purpose
>of transient suppressors is to give the pulse a defined and controlled
>place to discharge rather than depending on chance to allow it to flash
>somewhere unpredicted.
That's a lot of current, it almost sounds like a direct hit on the outlet!
Are you saying the MOV you were describing would protect equipment from
damage with that type of hit? Perhaps they are worthwhile!
Can you run a model of the MOV you described in a typical line input
circuit to see what happens?
73 Tom
Jeff Anderson
writes:
>
>Motorola also has a line of zeners characterized for transient
protection.
>They call theirs TVS (Transient Voltage Suppressors).
>
>Roy Lewallen, W7EL
>
..and at the back of Motorola's book (Motorola TVS/Zener Device Data)
are some app notes regarding transients and their suppression. I
happened to stumble across them several days ago while looking up a
zener's spec's.
- Jeff, WA6AHL
Peter Shintani
A company called Mitsubishi Material makes an awesome surge protector.
It's fast and once it's clamping voltage is exceede it's holding
voltage drops to nearly zero. This may provide better protection than a MOV
for a radio's input and also for phone lines, and modems. The device incidentally
has less than 1 pF of capacitance so it's easier to use than a MOV. It can take
repeated hits and it's characteristics stay constant were as MOV can deteriorate.
PS
James P. Meyer
>While we're on the subject, can anybody identify this MOV (at least I think
>it's an MOV):
>
><triangle symbol>ZNR
>14K201U
><backwards R>U 46
Check your DigiKey catalog. They're nominal 130 VAC "ZNR"s. 4500
Amp surge rating. They'd make ideal surge suppressors for 120 Volt AC
applications like turning a regular outlet strip into a "protected" strip.
Jim
Shawn Ruttledge
it's an MOV):
<triangle symbol>ZNR
14K201U
<backwards R>U 46
It's black, shiny, looks like a disc cap but is a bit too thick to be one of
those, and I can see through the coating that the leads bend and go to the
center of the material. I'd like to find out what the joule rating is for these
- maybe they'd be useful. I got a bunch of them in a Circuit Specialists
mystery box. My meter says the capacitance is 878 pF.
--
_______ KB7PWD @ KC7Y.AZ.US.NOAM ecl...@goodnet.com
(_ | |_) html: http://www.goodnet.com/~ecloud
__) | | \__________________________________________________________________
* OO * Khoros * Linux * robotics * techno * Gravis Ultrasound * X window *
Glenn E. Little
>
>While we're on the subject, can anybody identify this MOV (at least I think
>it's an MOV):
>
><triangle symbol>ZNR
>14K201U
><backwards R>U 46
>
>It's black, shiny, looks like a disc cap but is a bit too thick to be one of
>those, and I can see through the coating that the leads bend and go to the
>center of the material. I'd like to find out what the joule rating is for these
>- maybe they'd be useful. I got a bunch of them in a Circuit Specialists
>mystery box. My meter says the capacitance is 878 pF.
>
Digi-Key sells Panasonic MOVs.
Their catalog should shead some light on the ratings.
201 gives voltage at 200 volts +- 10%
14 is the series surge current is 4500 Amps.
Max allowable voltage ia 130 VAC, 270 VDC
Hope this helps
73
Glenn
WB4UIV
Wanda M Desmond
cell phone into a transceiver. I picked up a couple of Motorola cell
phones for next to nothing at a "junk" store the other day thinking
they were transceivers (the handset was missing so I didn't know
exactly what they were <g>).
Anyone know if this is possible? And how do you go about it?
Thanks for any help you can give me!
John
KE6TGN
M Simon
>While we're on the subject, can anybody identify this MOV (at least I think
>it's an MOV):
><triangle symbol>ZNR
>14K201U
><backwards R>U 46
>It's black, shiny, looks like a disc cap but is a bit too thick to be one of
>those, and I can see through the coating that the leads bend and go to the
>center of the material. I'd like to find out what the joule rating is for these
>- maybe they'd be useful. I got a bunch of them in a Circuit Specialists
>mystery box. My meter says the capacitance is 878 pF.
>--
> _______ KB7PWD @ KC7Y.AZ.US.NOAM ecl...@goodnet.com
> (_ | |_) html: http://www.goodnet.com/~ecloud
> __) | | \__________________________________________________________________
>* OO * Khoros * Linux * robotics * techno * Gravis Ultrasound * X window *
Check the Digikey catalog. Look for equivalent sizes. Also do a
leakage check to find out at what voltage they 'zener.' 1mA of
current vs voltage.
Simon