Instantaneous (analogue) compression of speech signals

[John Woodgate]

Does anyone here have any experience of instantaneous (analogue)
compression (aka soft clipping) of speech signals? I've been doing a
little work on it but I'm unable to judge the resulting sound quality.
Why do treble boost controls no longer have any audible effect for me?
(;-)

 

[Jim Thompson]

Does anyone here have any experience of instantaneous (analogue)
compression (aka soft clipping) of speech signals? I've been doing a
little work on it but I'm unable to judge the resulting sound quality.
Why do treble boost controls no longer have any audible effect for me?
(;-)

Huh?

...Jim Thompson

 

[Rich Grise]

Does anyone here have any experience of instantaneous (analogue)
compression (aka soft clipping) of speech signals?

Intentionally? ;-)

I've been doing a
little work on it but I'm unable to judge the resulting sound quality.

Well, isn't that a pretty good indication that your clipper isn't
introducing objectionable distortion? ;-) With speech, AIUI, you can do a
dramatic amount of clipping (10 dB? 20 dB? I read it in some Ham
magazine article mumble years ago) and still be intelligible. And the only
thing soft about it is an ordinary silicon knee; that's about all the
softness you really need, with speech.

Why do treble boost controls no longer have any audible effect for me?
(;-)

Too much Twisted Sister at volume level 11 in your misspent youth? ;-P

Cheers!
Rich

 

[Adrian Jansen]

Does anyone here have any experience of instantaneous (analogue)
compression (aka soft clipping) of speech signals? I've been doing a
little work on it but I'm unable to judge the resulting sound quality.
Why do treble boost controls no longer have any audible effect for me?
(;-)

Back in the ancient days of 27MHz CB radios there was heaps of work done
on compression and clipping to get the most "voice power" onto the
carrier. Is that the sort of compression you want ?

Your ears, like mine, probably date from this period too, if you cant
hear treble anymore

--
Regards,

Adrian Jansen adrianjansen at internode dot on dot net
Design Engineer J & K Micro Systems
Microcomputer solutions for industrial control
Note reply address is invalid, convert address above to machine form.

 

[John S. Dyson]

Intentionally? ;-)


Well, isn't that a pretty good indication that your clipper isn't
introducing objectionable distortion? ;-) With speech, AIUI, you can do a

One thing about audio 'clipping' is that more clipping can be done if
intermod is controlled. So, one trick has been to modulate the audio
onto an SSB type carrier, and then clip it there, and demod it (that is
a long way to do it.) Another possibility is to chop up the spectrum,
and apply soft clipping to each chunk.

I did something similar by doing an fft of audio (music, in fact), and
by using a carefully chosen window function, I was able to do evil, nonlinear
processing of the FFT'd signal, and then to reverse FFT the result. With
the proper window, the result is difficult to distingush from the original,
except it is more level compressed. If you choose the wrong window function,
you'll get a buzz effect (because of the overlap of the FFTs isn't optimal.)

With the right method, you can really make audio (even music) much more
dense, yet it is still 'musical.' If you let too much intermod occur, then
it can get ugly and distorted sounding.

John

 

[gwhite]

Does anyone here have any experience of instantaneous (analogue)
compression (aka soft clipping) of speech signals?

What the hell is "soft clipping?" It sounds like "golden ears" audiophile
mumbo-jumbo.

I've been doing a
little work on it but I'm unable to judge the resulting sound quality.
Why do treble boost controls no longer have any audible effect for me?
(;-)

Hey, the bright side is you don't have to spend any money on components to boost
the bright side.

 

[gwhite]

Yes. It is standard practice in voice-grade two-way radio. In that application
distortion is less important than long range intelligibility -- so the trade-off
is made.

One thing about audio 'clipping' is that more clipping can be done if
intermod is controlled. So, one trick has been to modulate the audio
onto an SSB type carrier, and then clip it there, and demod it (that is
a long way to do it.)

I've built one of these and it does not eliminate the odd-order intermod. In
fact, that's the one thing it can't do. However, it totally eliminates both
even- and odd-order harmonics. For example if you do the dual tone test with
800 and 1000 Hz into the RF clipper, you'll get 600, 800, 1000, and 1200 Hz out
(due to third-order effects, higher order effects certainly likely at some
amplitude). However, 1600, 2400, 2000, and 3000 Hz harmonics will be absent.

For SSB radio, conventional clipping in the baseband requires extraordinary PEP
in the transmitter. Of course this is highly undesirable. (Do the Hilbert
transform of a square wave.) While increased intelligibility claims are true,
the best reason for the RF clipper is due to the high PEP required if
conventional and otherwise effective baseband clipping is used.

Craiglow and Werth patented a baseband version of the RF clipper for
Rockwell-Collins. It requires a Hilbert transformer. I wonder if the whole
thing would be better implemented in DSP these days. At any rate, "RF clippers"
eliminate harmonic distortion in instantaneous clippers.

Craiglow-Werth patent (expired):
http://www.freepatentsonline.com/patents/us/441/4410764/4410764.pdf

 

[Al Clark]

Yes. It is standard practice in voice-grade two-way radio. In that
application distortion is less important than long range
intelligibility -- so the trade-off is made.


I've built one of these and it does not eliminate the odd-order
intermod. In fact, that's the one thing it can't do. However, it
totally eliminates both even- and odd-order harmonics. For example if
you do the dual tone test with 800 and 1000 Hz into the RF clipper,
you'll get 600, 800, 1000, and 1200 Hz out (due to third-order
effects, higher order effects certainly likely at some amplitude).
However, 1600, 2400, 2000, and 3000 Hz harmonics will be absent.

For SSB radio, conventional clipping in the baseband requires
extraordinary PEP in the transmitter. Of course this is highly
undesirable. (Do the Hilbert transform of a square wave.) While
increased intelligibility claims are true, the best reason for the RF
clipper is due to the high PEP required if conventional and otherwise
effective baseband clipping is used.

Craiglow and Werth patented a baseband version of the RF clipper for
Rockwell-Collins. It requires a Hilbert transformer. I wonder if the
whole thing would be better implemented in DSP these days. At any
rate, "RF clippers" eliminate harmonic distortion in instantaneous
clippers.

Craiglow-Werth patent (expired):
http://www.freepatentsonline.com/patents/us/441/4410764/4410764.pdf

I believe the DSP version is described in Marvin Frerking's book:
Digital Signal Processing in Communication Systems. Frerking was/is an
engineer at Rockwell-Collins. His DSP book is one of my favorites and
well worth the money if you are interested in DSP (regardless of whether
you have an interest in communications). It might be hard to find.

 

[Jim Thompson]

What the hell is "soft clipping?" It sounds like "golden ears" audiophile
mumbo-jumbo.

[snip]

As in, "run it thru a tooob" ;-)

...Jim Thompson

 

[John Woodgate]

I read in sci.electronics.design that Rich Grise <[email protected]>
wrote (in <[email protected]>) about
'Instantaneous (analogue) compression of speech signals', on Mon, 3 Jan
2005:

Too much Twisted Sister at volume level 11 in your misspent youth? ;-P

You have determined the problem, but the cause is idiopathic and
familial. My misspent youth dates from just after Greensleeves was top-
of-the-pops. (;-)

 

[John Woodgate]

I read in sci.electronics.design that Adrian Jansen <[email protected]>

Back in the ancient days of 27MHz CB radios there was heaps of work done
on compression and clipping to get the most "voice power" onto the
carrier. Is that the sort of compression you want ?

Yes, and low-cost. FFT and DSP solutions are unlikely to be justifiable
on a cost/benefit basis, but a couple of diodes and a quad op-amp is a
different matter.

Your ears, like mine, probably date from this period too, if you cant
hear treble anymore

Indeed.

 

[John Woodgate]

What the hell is "soft clipping?" It sounds like "golden ears"
audiophile mumbo-jumbo.

No, it's quite legitimate. You know what hard clipping is; flat tops on
the sine waves. Soft clipping gives rounded tops; in principle produced
by:

o----R------+----+-----o
| _|_
_V_ /\
In | | Out
R' R'
| |
o-----------+----+-----o

 

[Rich Grise]

One thing about audio 'clipping' is that more clipping can be done if
intermod is controlled. So, one trick has been to modulate the audio
onto an SSB type carrier, and then clip it there, and demod it (that is
a long way to do it.) Another possibility is to chop up the spectrum,
and apply soft clipping to each chunk.

Thank you for bringing this up. I now remember, the context of "'Deep'
clipping" was modulated RF, although my mental dredge is coming up AM, as
opposed to SSB, but the point is entirely the same.

I did something similar by doing an fft of audio (music, in fact), and
by using a carefully chosen window function, I was able to do evil,
nonlinear processing of the FFT'd signal, and then to reverse FFT the
result.

I'm afraid you're out of my league here, although I do want to say that
it's only the implementation I'm ignorant about - I get the _point_ of
what you're saying about transforming signals quite clearly, thanks.

In my seven-dimensional universe model, the equivalent is simply, turn
the signal sideways. ;-)

Thanks!
Rich

 

[Anthony C Smith]

John- a very good soft clipper can be formed by placing a pair of back to
back zeners in parallel with the feedback resistor on an opamp- the leakage
from the zeners brakes the signal slowly before the Vz+0.7 point- this
increases the THD on even harmonics only and sounds OK in practice - if you
want a better limiter a fet and rectifier would be the way to go but more
complex.
regards
Anthony

 

[John Woodgate]

I read in sci.electronics.design that Anthony C Smith
et.com>) about 'Instantaneous (analogue) compression of speech signals',

John- a very good soft clipper can be formed by placing a pair of back to
back zeners in parallel with the feedback resistor on an opamp- the leakage
from the zeners brakes the signal slowly before the Vz+0.7 point- this
increases the THD on even harmonics only and sounds OK in practice - if you
want a better limiter a fet and rectifier would be the way to go but more
complex.

Thanks for that. Are you sure it's even harmonics? If the clipping is
precisely symmetrical the harmonics are all odd order.

Anything using a rectifier involves a time constant, and I want to avoid
that because it introduces an extra variable - the time constant.

 

[martin griffith]

I read in sci.electronics.design that Adrian Jansen <[email protected]>


Yes, and low-cost. FFT and DSP solutions are unlikely to be justifiable
on a cost/benefit basis, but a couple of diodes and a quad op-amp is a
different matter.

Indeed.

Hi John,
have a look at alesis semiconductor/aka wavefront semi available from
profusion. They do a very cheep DSP, AL3101 IIRC, less than a fiver,
that is designed for this sort of thing. There is a Perl based
complier for it somewhere as well, freeware


martin

Serious error.
All shortcuts have disappeared.
Screen. Mind. Both are blank.

 

[Keith Wootten]

Does anyone here have any experience of instantaneous (analogue)
compression (aka soft clipping) of speech signals? I've been doing a
little work on it but I'm unable to judge the resulting sound quality.
Why do treble boost controls no longer have any audible effect for me?
(;-)

Howzabout a potential divider with the top leg being a small
incandescent lamp. Also acts as an emergency beacon where the louder
you call for help, the brighter is the lamp.

Wayne Bridges is the expert with this method of course.

Cheers

 

[John Woodgate]

I read in sci.electronics.design that Keith Wootten
'Instantaneous (analogue) compression of speech signals', on Tue, 4 Jan
2005:

Howzabout a potential divider with the top leg being a small
incandescent lamp. Also acts as an emergency beacon where the louder
you call for help, the brighter is the lamp.

Interesting, because it has potentially less distortion than diode
clipping. It's not quite 'instantaneous', though, and it's not so easy
to find suitable lamps.

Wayne Bridges is the expert with this method of course.

Does he have a web site?

 

[Jim Thompson]

Does anyone here have any experience of instantaneous (analogue)
compression (aka soft clipping) of speech signals? I've been doing a
little work on it but I'm unable to judge the resulting sound quality.
Why do treble boost controls no longer have any audible effect for me?
(;-)

As a start, at the very least, don't you need to define the level at
which ONSET of amplitude reduction begins, and then an absolute
MAXIMUM output level?

...Jim Thompson

 

[Anthony C Smith]

_snip-

Thanks for that. Are you sure it's even harmonics? If the clipping is
precisely symmetrical the harmonics are all odd order.

Good point now i look at it- I used this method with a pair of 3v3 zeners in
parallel with a 10K feedback resistor on a TL072 (2k2 Rin) and the even
harmonics increased IIRC when viewed with FFT analyser on my AP- still the
ears test is the best and it sounds OK when used on a mic preamp- limiting
starts around 2.6V and still has round corners when you hit the limit at 4V
(peak) ie no sharp transistion from the rising of the sine wave and the clip
point line.
simplest is to try it and see- its simple enough-
regards
Anthony