XLR's and AES/EBU impedance
Tall Paul
In article <32776F...@biddeford.com>, etn...@biddeford.com wrote:
> In a recent issue of Studio Sound (or what's left of it) a couple of
> articles discussed issues concerning digital interconnection and
> interfacing.
>
> Mention was made of XLR connectors' impedance characteristics having an
> effect on the signal path, but no mfr's were singled out for making good
> or bad units in this regard.
>
> Has anyone experience/knowledge/whatever about this issue? I'd sure hate
> to think I've got the wrong XLRs in my installation!
> --
Have been using Neutrik and Switchcraft XLR's for years.
Why, do your's sound bad?
Paul
--
"One day I want an exciting and challenging career in the music and/or entertainment industry", he says as he makes a pot of coffee, empties the trash, sweeps the floor, cleans the toilet, fills the candy machine, drives for endless hours looking for some exotic beer for the artist in studio A because he can't relate to life without it. "Would I trade this for anything..... nope!"
John Etnier
In a recent issue of Studio Sound (or what's left of it) a couple of
articles discussed issues concerning digital interconnection and
interfacing.
Mention was made of XLR connectors' impedance characteristics having an
effect on the signal path, but no mfr's were singled out for making good
or bad units in this regard.
Has anyone experience/knowledge/whatever about this issue? I'd sure hate
to think I've got the wrong XLRs in my installation!
--
===========================================================
John Etnier
Studio Dual
===========================================================
TEL (207) 799-8711
FAX (207) 799-4139
-------------------------------------
http://www.biddeford.com/~etnier
-------------------------------------
etn...@biddeford.com
etn...@aol.com
===========================================================
Jos Van Dyck
tall...@erols.com (Tall Paul) wrote:
>In article <32776F...@biddeford.com>, etn...@biddeford.com wrote:
>
>> articles discussed issues concerning digital interconnection and
>> interfacing.
>>
>> Mention was made of XLR connectors' impedance characteristics having an
XLR connectors were never designed for digital interfaces.
Digital signals contain much higher frequencies (several MHz), and HF
design techniques have to be applied, such as constant impedance.
This is also true for connectors. Therefore, the new 75 ohm coax AES/EBU
standard is more appropirate than the XLR 110 Ohms, derived from analog
experience.
A good 75 ohm RF coax (e.g. RG59) and 75 ohm BNC connectors are better
for digital interfaces than any XLR-based solution.
Telecommunication companies, who are using digital technology since many
years at bit rates ranging from 1.5 Mbit/s up to 2.5 Gbit/s, rely on 75
ohm coax and appropriate 75 connectors. This has been internationally
standardised (ITU-T G.703).
Jos
Scott Dorsey
In article <55a4r1$6...@newstoo.ericsson.se> Jos Van Dyck <j...@seq1.ebr.ericsson.se> writes:
>tall...@erols.com (Tall Paul) wrote:
>>In article <32776F...@biddeford.com>, etn...@biddeford.com wrote:
>>
>>> In a recent issue of Studio Sound (or what's left of it) a couple of
>>> articles discussed issues concerning digital interconnection and
>>> interfacing.
>>>
>>> Mention was made of XLR connectors' impedance characteristics having an
>>> effect on the signal path.
>
>XLR connectors were never designed for digital interfaces.
>Digital signals contain much higher frequencies (several MHz), and HF
>design techniques have to be applied, such as constant impedance.
>
>This is also true for connectors. Therefore, the new 75 ohm coax AES/EBU
>standard is more appropirate than the XLR 110 Ohms, derived from analog
>experience.
>A good 75 ohm RF coax (e.g. RG59) and 75 ohm BNC connectors are better
>for digital interfaces than any XLR-based solution.
This is true. Unfortunately, though, most folks are using those godawful
RCA connectors for 75 ohm digital, which is anything but constant impedance.
The truth is, though, that the XLR connectors were never designed to be
a constant 110 ohms, and therefore I would suspect that they do produce
a pretty heavy impedance discontinuity. I don't know how much it differs
from one manufacturer to another, but I would bet that it's not too significant
a difference as the geometry is pretty similar.
Now that you bring the subject up, I am pretty curious, and I will probably
spend my lunch break with a scope and a pulse generator and any XLRs that
happen to be kicking around my office. Now that the chef at the Thai
restaurant down the street is on vacation I don't have anything else to do.
>Telecommunication companies, who are using digital technology since many
>years at bit rates ranging from 1.5 Mbit/s up to 2.5 Gbit/s, rely on 75
>ohm coax and appropriate 75 connectors. This has been internationally
>standardised (ITU-T G.703).
I have noticed, though, that they aren't throwing out much of that neat
DS-1 coax stuff. Where can I get some of those nifty bundled coaxes in
short lengths? It looks like fun stuff for digital audio.
--scott
--
"C'est un Nagra. C'est suisse, et tres, tres precis."
Jos Van Dyck
klu...@netcom.com (Scott Dorsey) wrote:
>In article <55a4r1$6...@newstoo.ericsson.se> Jos Van Dyck <j...@seq1.ebr.ericsson.se> writes:
>>XLR connectors were never designed for digital interfaces.
>>Digital signals contain much higher frequencies (several MHz), and HF
>>design techniques have to be applied, such as constant impedance.
>>
>>This is also true for connectors. Therefore, the new 75 ohm coax AES/EBU
>>standard is more appropirate than the XLR 110 Ohms, derived from analog
>>experience.
>>A good 75 ohm RF coax (e.g. RG59) and 75 ohm BNC connectors are better
>>for digital interfaces than any XLR-based solution.
>
>This is true. Unfortunately, though, most folks are using those godawful
>RCA connectors for 75 ohm digital, which is anything but constant impedance.
>
>The truth is, though, that the XLR connectors were never designed to be
>a constant 110 ohms, and therefore I would suspect that they do produce
>a pretty heavy impedance discontinuity.
>
>>Telecommunication companies, who are using digital technology since many
>>years at bit rates ranging from 1.5 Mbit/s up to 2.5 Gbit/s, rely on 75
>>ohm coax and appropriate 75 connectors. This has been internationally
>>standardised (ITU-T G.703).
>
>I have noticed, though, that they aren't throwing out much of that neat
>DS-1 coax stuff. Where can I get some of those nifty bundled coaxes in
>short lengths? It looks like fun stuff for digital audio.
I recommend replacing the RCA/Cinch plugs and receptacles by good BNC
(take care to select the 75 ohm version, BNCs also exist in 50 ohms!),
and using propoer 75 ohm RF coax (such as RG59). You can buy these items
from Radio Shack or other radio component suppliers. Cable and connector
have to be matched.
For their digital distribution frames, TELCOs tend to use even
higher-spec cables (silver-plated, double screen, impedance tolerance
under 0.1%) and connectors (screw-lock, gold plated) to reduce
reflections and other disturbances, such as micro-interruptions due to
mechanical vibrations.
It's a pity that Pro-Audio folks (AES, EBU et al.) didn't look into TELCO
standards (CCITT, ITU-T). Many problems related to digital interfaces
have been experienced before, and solutions have been found and
standardized. Many interface and compatibility problems could have been
avoided.
I remember having seen an advertisement for a 110/75 impedance
transformer for AES/EBU interfaces, with an XLR plug at one side and BNC
receptacle at the other, but I don't remember the manufacturer's name.
Have fun!
Jos
Mike Rivers
In article <32776F...@biddeford.com> etn...@biddeford.com writes:
> Mention was made of XLR connectors' impedance characteristics having an
> effect on the signal path, but no mfr's were singled out for making good
> or bad units in this regard.
That's because the XLR wasn't designed with any characteristic
impedance in mind. I'm surprised that nobody's yet measured and
reported it yet, though, if for nothing other than academic interest.
I don't have any idea what it's impedance is, but I suspect, from
looking at connectors that I know are 50 and 75 ohms, that it's not
even close to 110 ohms.
What's at work here is that you have a transmission line that starts
with the transformer on the AES/EBU output, goes through a pair of
connectors, a piece of cable, perhaps a patch panel, and then to the
input transformer of the AES/EBU input on the other end. Each piece
of the line has a characteristic impedance which is a function of it's
physical size, construction, and materials. Each time there's a
change in impedance along the line, some of the signal gets reflected
back to the source. In a properly matched system, these impedance
"bumps" are minimal.
In an RF system, where all of this technology was developed, the
coaxial cables and connectors have been well researched and built in
different sizes to accommodate the popular impedances used. It's easy
to set up a well matched S/PDIF system since it uses standard 75 ohm
components. There are better and worse connectors and coax cables,
but they're all pretty close to what they're supposed to be, at least
over the typically short lengths that we encounter in the studio. The
110 ohm balanced (twinaxial) standard used for AES/EBU, however, isn't
one of the well developed RF standards used by the carload, so
nobody's developed an XLR-style connector that's actually 110 ohms.
The industry now offers us 110 ohm cable (Apogee for one), but it
will take a change in the AES/EBU standard to give us a proper
connector. Roger Nichols proposes that the RJ (telephone style)
modular connectors used in 10BaseT LAN wiring is a pretty good match.
And speaking of cables and impedances for AES/EBU, does anyone know
how to display the "eye pattern" used in cable impedance and phase
testing without buying an Audio Precision System 2? What's it a
measurement of? Apogee advertises their "Wide Eye" cables with
oscilloscope (or maybe System 2) photos, saying that the wider the
opening in the "eye", the better the cable. I originally thought this
was advertising hype based on some measurement technique they came up
with that made a cool scope trace, but I've seen other references to
the "eye pattern" outside the world of ad copy. It's obviously
something that's real. I have some RF measurement equipment around
when I'm out in the field and I'd be curious to play with it.
------------
I'm really mri...@d-and-d.com (Mike Rivers) On the road in Memphis
Ladies and gentlemen, Elvis HAS left the airport!
John Schmidt
Jos Van Dyck <j...@seq1.ebr.ericsson.se> wrote:
>tall...@erols.com (Tall Paul) wrote:
>>In article <32776F...@biddeford.com>, etn...@biddeford.com wrote:
>>
>>> In a recent issue of Studio Sound (or what's left of it) a couple of
>>> articles discussed issues concerning digital interconnection and
>>> interfacing.
>>>
>>> Mention was made of XLR connectors' impedance characteristics having an
>>> effect on the signal path.
>XLR connectors were never designed for digital interfaces.
>Digital signals contain much higher frequencies (several MHz), and HF
>design techniques have to be applied, such as constant impedance.
>This is also true for connectors. Therefore, the new 75 ohm coax AES/EBU
>standard is more appropirate than the XLR 110 Ohms, derived from analog
>experience.
>A good 75 ohm RF coax (e.g. RG59) and 75 ohm BNC connectors are better
>for digital interfaces than any XLR-based solution.
>Telecommunication companies, who are using digital technology since many
>years at bit rates ranging from 1.5 Mbit/s up to 2.5 Gbit/s, rely on 75
>ohm coax and appropriate 75 connectors. This has been internationally
>standardised (ITU-T G.703).
>Jos
Yes, but the most common BNC connectors are 50 ohm, not 75. The
television industry has used the 50 ohm BNC connectors with the
'precision' 75 ohm video cable for over 50 years on analog video
circuits where frequencies of up to about 10 MHz are involved. Only
in the last few years where digital serial video with bit rates of
hundreds of megabits per second are used have true 75 ohm BNC's become
common, together with new foam dielectric low loss cables. I suspect
that at AES audio bit rates, the 50 ohm BNC's make a negligable
difference. We have a lot of them used with AES audio at ABC TV (my
employer) on 75 ohm coax, and have no difficulty. You'll find that 6
MHz is pretty forgiving as long as you make any semblance of impedance
matching. One wavelength (in a vacuum, yes, I know its different in
coax, but not THAT different) at 6 Mhz is something like 49 meters, so
a one or two inch connector doesn't make too much difference.
John
*********************************************************************
John H. Schmidt, P.E. |Internet: SCH...@ix.netcom.com
Broadcast Engineer |Phone--Days (212)456-4218
Hempstead, New York | Evenings (516)483-2915
|Fax:------------(212)456-2424
Join the Friends Of WBAU
Make Adelphi University put WBAU back on the air!
*********************************************************************
Jonathan Edelson
In article <znr846763985k@trad>, Mike Rivers <mri...@d-and-d.com> wrote:
>And speaking of cables and impedances for AES/EBU, does anyone know
>how to display the "eye pattern" used in cable impedance and phase
>testing without buying an Audio Precision System 2? What's it a
>measurement of? Apogee advertises their "Wide Eye" cables with
>oscilloscope (or maybe System 2) photos, saying that the wider the
>opening in the "eye", the better the cable. I originally thought this
>was advertising hype based on some measurement technique they came up
>with that made a cool scope trace, but I've seen other references to
>the "eye pattern" outside the world of ad copy.
When digital information is transmitted using any sort of electrical
signal, the actual signal itself is an analog beastie, subject to things
like attenuation in the signal path, noise, frequency limitations of the
signal path, etc. In order to be able to decode a digital signal, you
have to be able to differentiate the various levels associated with
different transmitted values; in some cases you need to extract timing
from the signal as well.
An eye pattern is generated when you feed a digital signal into a scope,
and trigger the scope with the timing of the signal. With a 'perfect'
digital signal, you will see horizontal lines associated with the various
signal levels, and vertical lines associated with the transition between
these levels. The 'eye' is the opening between the transitions and
between the signal levels. Anything which attenuates the signal will
bring the signal levels closer together; too close and the noise level is
bigger than the difference between adjacent signal levels; you can no
longer tell 1s from 0s. Clock irregularity will close the opening in the
other direction; you may be able to tell the '1' level from the '0' level,
but you don't know which bit the 1 or 0 should fall into. Any of these
will cause the 'eye' to get smaller.
-Jon
--
win...@teleport.COM Public Access User --- Not affiliated with Teleport
Public Access UNIX and Internet at (503) 220-1016 (2400-14400, N81)
Del Winiecki
On Thu, 31 Oct 1996 12:13:05 GMT, mri...@d-and-d.com (Mike Rivers)
wrote:
>
>In article <32776F...@biddeford.com> etn...@biddeford.com writes:
>
>> Mention was made of XLR connectors' impedance characteristics having an
Look at a typical line using a Time Domain Reflectometer. I've had a
chance to play with squeezing the lines together, separating the
conductors, bending them, etc while doing the TDR software for
Microtest (cable testers), and at the frequencies used in digital
audio gear, it takes quite a severe change in conductor spacing or
bending to cause an appreciable amplitude reflected wave. Its the
amplitude of the reflection that is of concern. If the amplitude is
high enough, and the input circuit uses low-threshold (standard)
chips, you can get false bits and jittering. Proper receiver design
would use high-threshold chips, or even better, trigger a given amount
below the peaks.
As sample rates climb, and data rates increase, I think we will see a
move to different connectors.
The XLR's don't present a really severe blip on the TDR, when they are
wired carefully, so you can probably not worry too much about it. The
trouble is that its too dependent on the expertise of the guy who
wires the connector. With a BNC on good coax, it forces you to do it
right.
Del Winiecki
WinSyst Productions
Mike Rivers
In article <55cg72$6...@nadine.teleport.com> win...@teleport.com writes:
> When digital information is transmitted using any sort of electrical
> signal, the actual signal itself is an analog beastie, subject to things
> like attenuation in the signal path, noise, frequency limitations of the
> signal path, etc. In order to be able to decode a digital signal, you
> have to be able to differentiate the various levels associated with
> different transmitted values; in some cases you need to extract timing
> from the signal as well.
The various levels associated with different transmitted values? Like
what level represents a one and what level represents a zero? I
suspect that this eye pattern business has more to do with timing - or
more specificially, when the level crosses the threshold between a
zero and a one, as compared to when it's supposed to cross that
threshold. The difference is a function of capacitance causing the
edges of the "ideal" square wave not to be perfectly vertical, and the
ambiguity caused by clock jitter telling the system to "read" at a
time that's not quite correct.
> An eye pattern is generated when you feed a digital signal into a scope,
> and trigger the scope with the timing of the signal.
OK - and what do you look at?
> With a 'perfect'
> digital signal, you will see horizontal lines associated with the various
> signal levels, and vertical lines associated with the transition between
> these levels.
Are you describing a pulse train here, the simplest case being a
square wave? Or are there really more than two levels to a "perfect"
version of the signal you're describing?
> The 'eye' is the opening between the transitions and
> between the signal levels.
Then why does it rise and fall in the classic "eye" pattern? Let me
repeat my question - what is the pattern a plot of? A voltage vs.
time (as in a conventional oscilloscope trace)? An X-Y plot of an
input vs. an output?
> Anything which attenuates the signal will
> bring the signal levels closer together
That just sounds like changing the gain on the scope would "open" the
eye again. How is this calibrated?
> ; too close and the noise level is
> bigger than the difference between adjacent signal levels; you can no
> longer tell 1s from 0s. Clock irregularity will close the opening in the
> other direction; you may be able to tell the '1' level from the '0' level,
> but you don't know which bit the 1 or 0 should fall into. Any of these
> will cause the 'eye' to get smaller.
Now we're getting somewhere - it sounds like the trace gets "fuzzier"
with signal degradation - that is, it becomes thicker, so the outside
of the "eye" remains at it's original size, but the "hole" inside
becomes smaller. Is that what's happening here?
A block diagram or a test setup would be useful. Or a reference to an
article or book with such a test setup that I might be able to find?
--
I'm really Mike Rivers (mri...@d-and-d.com)
Mikhail B. Matusov
I am not completely sure in what I am going to say but I guess the following
steps to get the eye diagram on a 2-ray scope:
1. Make sure that both traces on the scope with no input signal combine into
one line;
2. Feed two inputs of the scope with the investigated signal and its phase
reversed copy (or reverse phase on the one of scope's inputs);
3. You'll see a plenty of eyes like these: /~\ /~\ ;
\_/ \_/
4. AES/EBU standard says that a receiver has to correctly sense the data
when a random input signal produces the eye diagram characterized by
some Vmin and Tmin. The eye is OK if you can insert into it a rectangle
with the width of Tmin and height of Vmin.
Mikhail Matusov
mat...@mmb.spb.ru
J. Russell Lemon
pattern. With the AES/EBU standard, there is only one vertical eye
because it is manchester encoded. The eye is formed at the bit decision
point. Between the eyes is the transition area. With wide frequency
response, linear phase, no noise and no phase jitter, a well defined eye
can be seen when the data is shown and the scope is synced to the
separate clock. (The clock is at the data rate.) The upper and lower
lines represent the two different values the digital data can take.
Noise make the lines fuzzy and tends to close the eye from the vertical
dimention. Phase Jitter also makes the lines fuzzy and tends to close
the eye from the horizontal dimention. Loss of frequency response and/or
unequalized group delay closes the eye in both dimentions.
Other digital systems that have more then one bit/baud (multi-level) the
scope sync must be at the baud rate (i.e. symbol rate). With multi-level
systems, more then one vertical eye can be seen.
--- Russ
J. Russell Lemon
Mikhail B. Matusov <mat...@mmb.spb.ru> wrote:
>I am not completely sure in what I am going to say but I guess the following
>steps to get the eye diagram on a 2-ray scope:
>1. Make sure that both traces on the scope with no input signal combine into
> one line;
>2. Feed two inputs of the scope with the investigated signal and its phase
> reversed copy (or reverse phase on the one of scope's inputs);
>3. You'll see a plenty of eyes like these: /~\ /~\ ;
> \_/ \_/
>4. AES/EBU standard says that a receiver has to correctly sense the data
> when a random input signal produces the eye diagram characterized by
> some Vmin and Tmin. The eye is OK if you can insert into it a rectangle
> with the width of Tmin and height of Vmin.
>
>Mikhail Matusov
>mat...@mmb.spb.ru
>
The trick is to sync the scope on the data's clock so that data
transistions are stationary on the display. At the data decision point,
the upper and lower lines represent the two different states the data can
take. between the data decision points is the transition area. Noise
makes the lines fuzzy in the vertical dimention. Phase Jitter (timing
error) makes the lines fuzzy in the horizontal dimention. Limited
frequency response and/or unequalized group delay distortion makes the
lines fuzzy in both dimentions and is data dependent. For this reason
long length pseudo-random generators are necessary for an accurate eye
pattern. (A square wave does not work.) Typical pattern lengths are:
63, 127, 511, and 1023. Longer patterns have been used (as 2^20-1) but
are very difficult to sync to.
--- Russ
J. Russell Lemon
Mikhail B. Matusov <mat...@mmb.spb.ru> wrote:
>I am not completely sure in what I am going to say but I guess the following
>steps to get the eye diagram on a 2-ray scope:
>1. Make sure that both traces on the scope with no input signal combine into
> one line;
>2. Feed two inputs of the scope with the investigated signal and its phase
> reversed copy (or reverse phase on the one of scope's inputs);
>3. You'll see a plenty of eyes like these: /~\ /~\ ;
> \_/ \_/
>4. AES/EBU standard says that a receiver has to correctly sense the data
> when a random input signal produces the eye diagram characterized by
> some Vmin and Tmin. The eye is OK if you can insert into it a rectangle
> with the width of Tmin and height of Vmin.
>
>Mikhail Matusov
>mat...@mmb.spb.ru
>
pattern. With the AES/EBU standard, there is only one vertical eye
because it is manchester encoded. The eye is formed at the bit decision
point. Between the eyes is the transition area. With wide frequency
response, linear phase, no noise and no phase jitter, a well defined eye
can be seen when the data is shown and the scope is synced to the
lines represent the two different valuses the digital data can take.
J. Russell Lemon
Scott Dorsey
In article <znr846763985k@trad> mri...@d-and-d.com writes:
>
>In article <32776F...@biddeford.com> etn...@biddeford.com writes:
>
>> Mention was made of XLR connectors' impedance characteristics having an
>> effect on the signal path, but no mfr's were singled out for making good
>> or bad units in this regard.
>
>That's because the XLR wasn't designed with any characteristic
>impedance in mind. I'm surprised that nobody's yet measured and
>reported it yet, though, if for nothing other than academic interest.
>I don't have any idea what it's impedance is, but I suspect, from
>looking at connectors that I know are 50 and 75 ohms, that it's not
>even close to 110 ohms.
I actually just did do a quick and dirty test this weekend... the answer
is that all of the XLR connectors I tried had a pretty significant mismatch
and they were all about the same. Can't give you any real firm numbers
because I was basically running a 110 ohm cable into a 110 ohm load, looking
at the signal source (from a pulse generator) with a boatanchor Tek scope.
I'd say it was probably something in the 150-175 ohm range mostly, but I
doubt it's even constant over its length.
I tried Switchcraft, Neutrik, Radio Shack, and the right angle Calrad
connectors, because those are what I had in the junk bin at work. I
did not try the quick-connect Neutriks or any other right-angle connectors.
>The industry now offers us 110 ohm cable (Apogee for one), but it
>will take a change in the AES/EBU standard to give us a proper
>connector. Roger Nichols proposes that the RJ (telephone style)
>modular connectors used in 10BaseT LAN wiring is a pretty good match.
Dunno, I bet they are also in the ballpark but they're awfully flimsy. I
mean, you can use an XLR connector to tow the sound truck if you have to.
Chris Caudle
Scott Dorsey <klu...@netcom.com> wrote in article
<kludgeE0...@netcom.com>...
> I'd say it was probably something in the 150-175 ohm range mostly, but I
> doubt it's even constant over its length.
Did anyone else see that article in JAES a couple of years back? Somebody
actually tried this, and I don't remember if the actual characteristic
impedance of an XLR was measured, but the author performed TDR measurements
of XLR connections, and basically concluded it wasn't a big deal in most
situations. I seem to remember he connected about 10 or 12 XLR connectors
in series to simulate worst case conditions, and did not notice serious
degradation of signal quality.
The same author also pointed out than an alarmingly high number of devices
he tested did not have anywhere near 110 ohm input and output impedances,
either. I would hope that has improved in the intervening few years, but I
don't have a large number of AES/EBU devices to compare. Seems like
several DAT machines, and the Yamaha DMP1000 digital mixer had really
screwed up output impedances which could cause problems over more than a
few meters.
Del Winiecki
On Mon, 4 Nov 1996 16:55:32 GMT, klu...@netcom.com (Scott Dorsey)
wrote:
>I actually just did do a quick and dirty test this weekend... the answer
>is that all of the XLR connectors I tried had a pretty significant mismatch
>and they were all about the same. Can't give you any real firm numbers
>because I was basically running a 110 ohm cable into a 110 ohm load, looking
>at the signal source (from a pulse generator) with a boatanchor Tek scope.
>doubt it's even constant over its length.
>
>connectors, because those are what I had in the junk bin at work. I
>did not try the quick-connect Neutriks or any other right-angle connectors.
Thats about what I'd expect to see, Scott. Its a significant mismatch,
but the reflected spike still should not be up in the critical
switching range voltage-wise. Is that what you observe?
I'd expect right-angle connectors might add enough extra mismatch that
you'd get into trouble.
Some companies want you to use only their connectors (tascam etc.)
because their connectors have precise lead dress to keep the Z more
constant.
I'm a little hesitant about using the telephone J type connectors in
the field, but -maybe- in a fixed installation, though I sure prefer
XLR toughness. Actually, the 3 prong XLR seems a poor choice. It
promotes people picking up the wrong cable and using a mic cable
instead of a digital cable. 4 prong XLR might have been better. I
think Ampex used to use these for something in old reel-to-reel's?
-Del Winiecki
WinSyst Productions
Iain A B Lindsay
John Etnier <etn...@biddeford.com> wrote:
>In a recent issue of Studio Sound .....
>
>Mention was made of XLR connectors' impedance characteristics having an
>effect on the signal path, but no mfr's were singled out for making good
>or bad units in this regard.
>
>to think I've got the wrong XLRs in my installation!
XLRs are not designed to have a controlled value of characteristic impedance,
so the short answer to your question is: "Don't worry". Now for the long
answer.
Whether or not an impedance mismatch is a problem depends on the degree of
mismatch, the distance over which the mismatch occurs and the variation of
your signal with time (frequency/pulse width/rise-fall time).
Consider the situation where you have 110 ohm transmission line on either side
of an XLR connector pair. Now let's assume that the XLR section has an
impedance of 220 ohms (twice that which we would prefer).
A B
----->---110 ohm cable--->-----|x 220 ohm XLR x|--->---110 ohm cable--->---
as long as you like <- 3 inches -> as long as you like
For the sake of argument, let's take the length of the XLR connector pair to be
3 inches (i.e. a much longer/worse discontinuity than is really the case as
most
of the XLR body has our 110 ohm cable inside it). Then, assuming a velocity
factor for the dielectric holding the XLR pins of about 2, the time taken for a
pulse to pass through the mis-matched XLR section is about 0.5ns (speed of
light is 1ft/ns).
A pulse arriving at the junction between 110 ohm cable and 220 ohm connector
will suffer some reflection. The reflection coefficient, which describes how
much of the signal voltage is reflected is given by:
Coeff = (Z2 - Z1) / (Z2 + Z1)
Where Z1 is the impedance of the line on which the pulse arrives at the
junction
and Z2 is the impedance after the junction (the XLR pair in this case). So,
for
the first reflection, at point A, we have a coefficient of:
(220 - 110) / (220 + 110)
i.e. +1/3. Thus, for a 1 volt pulse arriving from the left, a reflected pulse
of 1/3 volt gets sent back to where the original 1 volt pulse came from. Now,
to keep the Universe happy, the voltage on the connector side of junction A
must be the same as that on the cable side, which we have just calculated to
consist of +1 volt of incident pulse and +1/3 volts of reflected pulse. Thus,
a +4/3 volts pulse runs through the (220 ohm) connector pair (Yes, it really is
a bigger voltage than the incident pulse). After 0.5ns, it arrives at point B,
where it meets another discontinuity in impedance. This time, the impedance
changes from 220 ohms to 110 ohms, so the reflection coefficient is now:
(110 - 220) / (110 + 220)
i.e. -1/3. So, this time, the reflection produces a pulse going back towards
the left of -1/3 * +4/3, i.e. -4/9 volts. The incident pulse was +12/9 volts,
so the pulse which is launched onto the right hand piece of cable is +8/9
volts.
Another 0.5ns passes. The -4/9 volts pulse reflected from B now arrives at A.
Here the impedance changes from 220 to 110 (remember that this pulse is going
in the opposite direction to the original 1 volt pulse). So, once again we
have
a reflection coefficient of -1/3. The -4/9 volts leftward travelling pulse
gets
split into a -1/3 * -4/9 => 4/27 volt rightward travelling reflection and a
-8/27 volt pulse which goes chasing off down the left hand piece of cable, in
pursuit of the first reflection we saw of +1/3 volt. As both pulses are
travelling at the same speed (they're on the same piece of cable), the
-8/27 volt pulse will always lag behind the +1/3 volt pulse by the 1ns which it
took to get to the same starting point.
Meanwhile, the +4/27 volts rightward travelling reflection heads back to
junction B. Once again it gets partially reflected, with -4/81 volts heading
to the left towards junction A and +8/81 volts getting through the B junction
to go chasing off after the first +8/9 volts, but being always 1ns behind it.
Of course, the -4/81 volt pulse which resulted from the 4th reflection arrives
at junction A 0.5 ns after leaving junction B, is partially reflected again and
after another 0.5ns a third pulse leaves from junction B travelling right (with
its reflection partner heading back leftwards to A again) 1ns behind the 8/81
volt pulse and 2ns behind the 8/9 volt pulse.
This reflecting business goes on ad infinitum, but the amplitude of the pulses
is decreasing all the time, as the magnitude of the reflection coefficient is
always less than unity (or equal to it for the case of an open or short circuit
where the connector should be).
So, a receiver at the right hand end of the right hand piece of cable sees a
rising pulse which is:
initially 0.88889 = 8/9 volts
1ns later 0.98765 = 8/9 + 8/81 volts
2ns later 0.99863 = 8/9 + 8/81 + 8/729 volts
3ns later 0.99985 = 8/9 + 8/81 + 8/729 + 8/6561 volts
4ns later .......
If you were sending pulses down the line at, say, 10 Mbps, then each pulse is
100ns long and the upshot of the above is that the rising edge gets smeared
over an extra couple of ns beyond its original rise time and you'd be hard
pushed to see the effect on a 'scope, what with all the other noise and
distortion around the edges. Of course, if you chain 20 or 30 such connectors,
the problem will become more severe, but then you should also be worried about
the impedance variation between all those different lengths of "nominally" 110
ohm (or whatever) cable.
(You also get a short spike running back towards the source of the original
pulse. I've omitted that half of the calculation, for brevity. It is left
as an exercise for the reader...)
The above calculation over-estimates the severity of the problem. In reality,
the electrical length of the XLRs will be less than 0.5ns. You can do the
calculations yourself for other impedance values for connector mismatch. For
all reasonable values of impedance (below and above 110 ohms), you'll find that
it's the length of the mismatched section (w.r.t. the pulse duration and
rise/fall times) which is more of a concern than the value of the "odd"
impedance. Thus, cable which has been crushed, had knots tied in it and
otherwise used in place of wedges and gaffer tape, causing changes in its
physical dimensions and hence characteristic impedance, should be of more
immediate concern than the XLR connectors, which were better designed to cope
with these extra-curricular duties ;-)
In fact, characteristic impedance, attenuation and group delay all vary to
some extent as a function of frequency and the extent of this variation
depends,
to a very large extent, on what the cable was designed for. Digital pulse
transmission is one situation where choosing good cable doesn't just
depend on buying the most expensive or the one with most magic smoke in it.
It's also not just a case of getting cable with good high frequency
performance,
the low and high frequency behaviour (over the effective bandwith of your
signal) need to be well equalised in order to avoid distorting the pulses.
Coming soon: why earth current doesn't travel in loops around an earth loop...
(and why RFI/EMC current does, even when you're in an aeroplane and are a long
long way from the "ground", which is quite unnecessary for radio wave
propagation or detection, much to the relief of astronauts, radio astonomers
and
anyone who likes the idea of sunshine getting from there to here).
-Iain Lindsay..
Electrical Engineering
University of Edinburgh
Scotland GB
Mike Rivers
In article <01bbca79$59e82420$71bc12ac@enzo-sys> Chr...@Xgate.Compaq.com writes:
> Did anyone else see that article in JAES a couple of years back? Somebody
> actually tried this, and I don't remember if the actual characteristic
> impedance of an XLR was measured, but the author performed TDR measurements
> of XLR connections, and basically concluded it wasn't a big deal in most
> situations.
That's the bottom line - it really ISN'T a big deal, we just like to
discuss the things that keep us from technical perfection. The proper
impedance cable for longer runs is a bigger deal - people were having
problems getting digital audio between studios on mic cable, and now
there's a solution.
> The same author also pointed out than an alarmingly high number of devices
> he tested did not have anywhere near 110 ohm input and output impedances,
> either. I would hope that has improved in the intervening few years
Me, too. It probably has as more devices come on the market, and
there's more use of standard components in the input and output
circuits. Still, there's a cheezy looking little transformer in most
of them that doesn't look all that trustworthy.