FM Bandwidth Blues
 

Just to let people know that I have now put up a webpage on the effects of
finite bandwidth on the distortion levels for FM audio broadcasts. Can be
found on my audiomisc site.

The page combines the content from an earlier HFN article with some extra
material I produced a while ago when discussing this issue on uk.rec.audio.

Slainte,

Jim

--
Electronics http://www.st-and.ac.uk/~www_pa/Scot...o/electron.htm
Audio Misc http://www.audiomisc.co.uk/index.html
Armstrong Audio http://www.audiomisc.co.uk/Armstrong/armstrong.html

FM Bandwidth Blues
 

On Wed, 21 Nov 2007 10:13:26 +0000 (GMT), Jim Lesurf
wrote:

Just to let people know that I have now put up a webpage on the effects of
finite bandwidth on the distortion levels for FM audio broadcasts. Can be
found on my audiomisc site.

The page combines the content from an earlier HFN article with some extra
material I produced a while ago when discussing this issue on uk.rec.audio.

Slainte,

Jim

Just made me think back to my "Ultimate tuner" I built back in the
late seventies when FM stations were over 2MHz apart. I built it with
a very wide, linear phase IF filter to prevent exactly this kind of
distortion. I don't have it any more, but if I did, it would no longer
work because of in-band interference.

Actually, now I think of it, the phase response of the IF filter is
also important here. Square, steep sided filters would seem to be the
best choice, but they tend to have a huge group delay wiggle as you
approach the corners, so you do need to compromise.

d

--
Pearce Consulting
http://www.pearce.uk.com

FM Bandwidth Blues
 

In article , Don Pearce
wrote:


Just made me think back to my "Ultimate tuner" I built back in the late
seventies when FM stations were over 2MHz apart. I built it with a very
wide, linear phase IF filter to prevent exactly this kind of distortion.
I don't have it any more, but if I did, it would no longer work because
of in-band interference.

Alas, the transmission may already be processed/filtered, so even without
the filtering in the RX you might get some distortion. Hard to determine
the details, though, as it is difficult to determine quite what is being
done by some of the boxes in the TX chain. Particular problem here is that
with some units like the Orban ones it isn't at all clear exactly what they
are up to.

Actually, now I think of it, the phase response of the IF filter is also
important here. Square, steep sided filters would seem to be the best
choice, but they tend to have a huge group delay wiggle as you approach
the corners, so you do need to compromise.

Yes, indeed. I first twigged all this some decades ago. My colleague at
Armstrong, Ted Rule, had been working on a new tuner to accompany the 700
range amps I was developing. At that time RF was his end of the work. But
over a period of time it became apparent that there were a set of 'target
specs' where he could meet one or two at any time, but not all
symltaneously. Things like distortion for various modulations, capture
ratio, adjacent channel rejection, etc.

After he'd left I was asked to look at tuners. Found that you could 'tweak'
the IF/demodulator to suit any specific spec, but generally by degrading
something else. Sat down and did some analysis.

One result is as you describe. In practice the combination of RF, IF, and
demodulator tend to create a group delay value that varies across the
required IF band, with sharper edged filters giving more gross waggles in
the phase responses, These upset the distortion properties. Turns out that
this is all a fundamental set of trade-offs which stem from the properties
of FM signals when real filters and finite bandwidths are part of the
equation. Unavoidable in reality, therefore.

You can therefore tweak a set to give amazingly low apparent distortion
for, say, 30 percent mod mono 300Hz, but at the expense of, say, the few
kHz R-only distortion going sky-high, or ruining the adjacent channel
rejection, or...

I also twigged that the bench FM generator had no output filtering, so
would have no such effects applied at source and was sending a signal of
arbitrarily wide bandwidth to the tuner. Unlike a real-world TX.

Since experimenting then I have become quite cautious of reviews where a
tuner gives a remarkably good value for some specs on magazine bench tests.
Not surprised if such a tuner might have been 'lined up' for them, or -
perhaps worse - if the production tuners have rather poorer performance
that the magazine values indicates.

Knowing this, if you look at reviews where they give a range of values you
can soon spot the trade-offs involved affecting the results.

Also impressed by some reviews where the comments made about the
'subjective' results were clearly based on the measurements, and not what
would have really been happening if the published values were doing what
simple theory and experience led me to expect would be going on. :-)

In addition to which, in many cases with narrow IFs even tiny FM offsets in
tuning have a marked effect on all the figures. And in some tuners the AFC
pulls the tuning to a place which gives worse results than if hand-tuned.
So it is just as well that most people don't notice these levels of
distortion. ;-

Hence my decision to do the above article. Really to warn people to take
the idea that FM has distortions as low as the magazine values with some
salt. For modest signals the distortion can indeed be low enough to deliver
good results. But that isn't all the story - particularly when listening to
stations which are level compressed to max modulation.

Plan to tackle multipath and impulse interference as well in due course.
Compare the effects on FM with those on DAB and DTTV.

Alas, never did finish off Ted's tuner. This used helical resonators for
the IF and was quite a nifty design. But never made it to manufacture as
the company decided to shut up shop before it reached that point. Shame as
his ideas were quite good even if he'd fallen into a trap which I suspect
has caught many people.

Slainte,

Jim

--
Electronics http://www.st-and.ac.uk/~www_pa/Scot...o/electron.htm
Audio Misc http://www.audiomisc.co.uk/index.html
Armstrong Audio http://www.audiomisc.co.uk/Armstrong/armstrong.html

FM Bandwidth Blues
 

Hence my decision to do the above article. Really to warn people to take
the idea that FM has distortions as low as the magazine values with some
salt. For modest signals the distortion can indeed be low enough to deliver
good results. But that isn't all the story - particularly when listening to
stations which are level compressed to max modulation.

Plan to tackle multipath and impulse interference as well in due course.
Compare the effects on FM with those on DAB and DTTV.

Alas, never did finish off Ted's tuner. This used helical resonators for
the IF and was quite a nifty design. But never made it to manufacture as
the company decided to shut up shop before it reached that point. Shame as
his ideas were quite good even if he'd fallen into a trap which I suspect
has caught many people.

Slainte,

Jim

When I get a moment!....

I'll send you an off-air CD of a really good local FM station around
here and see what you think of the quality;)...
--
Tony Sayer

FM Bandwidth Blues
 

On Wed, 21 Nov 2007 13:53:09 +0000 (GMT), Jim Lesurf
wrote:

In article , Don Pearce
wrote:


Just made me think back to my "Ultimate tuner" I built back in the late
seventies when FM stations were over 2MHz apart. I built it with a very
wide, linear phase IF filter to prevent exactly this kind of distortion.
I don't have it any more, but if I did, it would no longer work because
of in-band interference.

Alas, the transmission may already be processed/filtered, so even without
the filtering in the RX you might get some distortion. Hard to determine
the details, though, as it is difficult to determine quite what is being
done by some of the boxes in the TX chain. Particular problem here is that
with some units like the Orban ones it isn't at all clear exactly what they
are up to.

Actually, now I think of it, the phase response of the IF filter is also
important here. Square, steep sided filters would seem to be the best
choice, but they tend to have a huge group delay wiggle as you approach
the corners, so you do need to compromise.

Yes, indeed. I first twigged all this some decades ago. My colleague at
Armstrong, Ted Rule, had been working on a new tuner to accompany the 700
range amps I was developing. At that time RF was his end of the work. But
over a period of time it became apparent that there were a set of 'target
specs' where he could meet one or two at any time, but not all
symltaneously. Things like distortion for various modulations, capture
ratio, adjacent channel rejection, etc.

After he'd left I was asked to look at tuners. Found that you could 'tweak'
the IF/demodulator to suit any specific spec, but generally by degrading
something else. Sat down and did some analysis.

One result is as you describe. In practice the combination of RF, IF, and
demodulator tend to create a group delay value that varies across the
required IF band, with sharper edged filters giving more gross waggles in
the phase responses, These upset the distortion properties. Turns out that
this is all a fundamental set of trade-offs which stem from the properties
of FM signals when real filters and finite bandwidths are part of the
equation. Unavoidable in reality, therefore.

You can therefore tweak a set to give amazingly low apparent distortion
for, say, 30 percent mod mono 300Hz, but at the expense of, say, the few
kHz R-only distortion going sky-high, or ruining the adjacent channel
rejection, or...

I also twigged that the bench FM generator had no output filtering, so
would have no such effects applied at source and was sending a signal of
arbitrarily wide bandwidth to the tuner. Unlike a real-world TX.

Since experimenting then I have become quite cautious of reviews where a
tuner gives a remarkably good value for some specs on magazine bench tests.
Not surprised if such a tuner might have been 'lined up' for them, or -
perhaps worse - if the production tuners have rather poorer performance
that the magazine values indicates.

Knowing this, if you look at reviews where they give a range of values you
can soon spot the trade-offs involved affecting the results.

Also impressed by some reviews where the comments made about the
'subjective' results were clearly based on the measurements, and not what
would have really been happening if the published values were doing what
simple theory and experience led me to expect would be going on. :-)

In addition to which, in many cases with narrow IFs even tiny FM offsets in
tuning have a marked effect on all the figures. And in some tuners the AFC
pulls the tuning to a place which gives worse results than if hand-tuned.
So it is just as well that most people don't notice these levels of
distortion. ;-

Hence my decision to do the above article. Really to warn people to take
the idea that FM has distortions as low as the magazine values with some
salt. For modest signals the distortion can indeed be low enough to deliver
good results. But that isn't all the story - particularly when listening to
stations which are level compressed to max modulation.

Plan to tackle multipath and impulse interference as well in due course.
Compare the effects on FM with those on DAB and DTTV.

Alas, never did finish off Ted's tuner. This used helical resonators for
the IF and was quite a nifty design. But never made it to manufacture as
the company decided to shut up shop before it reached that point. Shame as
his ideas were quite good even if he'd fallen into a trap which I suspect
has caught many people.

Slainte,

Jim

Do you use Mathcad? I wrote a worksheet to implement the helical
filter design method from Zverev, plus a few bits from elsewhere. You
just put in all the details of frequency, BW, number of sections etc,
plus the k and q values from the appropriate table, and it gives you
all the coil dimensions, tap points, shield heights etc. I used a
helical IF in my super tuner - had to work it all out longhand back
then, of course.

Anyway, you are welcome to a copy if you can make use of it.

d

--
Pearce Consulting
http://www.pearce.uk.com

FM Bandwidth Blues
 

"Jim Lesurf"

Just to let people know that I have now put up a webpage on the effects of
finite bandwidth on the distortion levels for FM audio broadcasts. Can be
found on my audiomisc site.

The page combines the content from an earlier HFN article with some extra
material I produced a while ago when discussing this issue on
uk.rec.audio.

** See: http://www.audiomisc.co.uk/HFN/BandwidthBlues/page.html

1. The claim made that THD figures for FM receivers in mono are speced at
30% modulation is false.

It is usually speced at full +/- 75 kHz deviation. Even the long obsolete
LM3089 ( FM IF system and detector IC ) has a published THD figure of 0.1%
with full 75kHz deviation (with double tuned quadrature detection).

The figure is 100% for real and verified in many commercial tuners.

http://www.national.com/ds/LM/LM3089.pdf


2. There is another, much simpler way to think about how FM stereo operates
than the one presented.

Essentially, FM stereo is transmitted in time division multiplexing - that
is the left and right channels of the original audio are transmitted as
alternate *analogue* samples frequency modulating the carrier.

There are 38,000 *analogue* samples of each channel, per second. The low
level, 19kHz pilot tone simply allows the receiver to know when each sample
is arriving and which channel it is.

With mono audio input, the L and R samples all follow exactly the same
curve, so the derived frequency modulations all match up and there is full
compatibility.

However, when only one channel is sent, the detected audio signal has 38,000
gaps per second - there being no modulation during the silent channel's
allocated sample times.

A direct result of this is the recovered audio level at the detector is only
half that of the mono case, given the same FM deviation.

3. The effect of pre-emphasis.

FM radio uses high frequency pre-emphasis - with a 50uS time constant in
Europe and Australia - so, all FM receivers have a corresponding 50uS
de-emphasis filter installed in the audio path. This helps to reduce HF
noise.

The effect of this filter is such that at 4kHz, the recovered audio is
attenuated by just over 4dB or by a factor of 0.62

Only 50% of the mono signal level is available when a single stereo channel
is transmitted - then the receiver attenuates that by 0.62 if the
frequency is 4kHz. 0.5 times 0.62 = 0.3

So, if one bench tests an FM stereo receiver with 4kHz tone modulation in
only the L or R channels, the recovered audio level will be 30% of the mono
level at 400Hz. Trying to get more will only lead to gross distortion.

This is nothing to worry about here, as the nature of music and speech
signals is such that full level at 4 kHz is completely un-necessary.

Something every tweeter maker has long known and thoroughly exploited.



....... Phil

FM Bandwidth Blues
 

In article , Don Pearce
wrote:
On Wed, 21 Nov 2007 13:53:09 +0000 (GMT), Jim Lesurf
wrote:

In article , Don Pearce
wrote:
[big snip]

Do you use Mathcad?

Just as well I wasn't sipping my orange juice. For a moment I thought you
meant back when I was working on Ted's tuner in c1980. That was in the days
of pen and paper. I did have an HP calculator. The 41 IIRC. ;-

No. I write all my programmes in ANSI 'C'. So all of the results on the
page come from that approach. Means I can define how the computation is
done. I am too old fashioned to take to Mathcad. :-)

I wrote a worksheet to implement the helical filter
design method from Zverev, plus a few bits from elsewhere. You just put
in all the details of frequency, BW, number of sections etc, plus the k
and q values from the appropriate table, and it gives you all the coil
dimensions, tap points, shield heights etc. I used a helical IF in my
super tuner - had to work it all out longhand back then, of course.

Indeed. As above. Alas the drawings for Ted's design were all lost long
ago. When I left the company I took my work on amps, but not any of the
stuff on tuners.

Anyway, you are welcome to a copy if you can make use of it.

Not likely to do any tuner design these days, I'm afraid. Thanks for the
offer, though.

Slainte,

Jim

--
Electronics http://www.st-and.ac.uk/~www_pa/Scot...o/electron.htm
Audio Misc http://www.audiomisc.co.uk/index.html
Armstrong Audio http://www.audiomisc.co.uk/Armstrong/armstrong.html