Re: [LAD] [ot] rme fireface: weird balanced output measurements

On Saturday, February 18, 2012 07:02:22 AM Fons Adriaensen did opine:

> On Fri, Feb 17, 2012 at 07:23:12PM -0500, gene heskett wrote:
> > > The concept of cable impedance makes sense only if the lenght
> > > becomes a non-trivial fraction of wavelength. For audio that
> > > means that for anything shorter than a few hundred meters it's
> > > only capacitance that matters.
> >
> > But at a 60 ohm impedance, that capacitance is considerable. I could
> > put 20 volts P-P into a cable headed for our news dept, at say 15
> > kilohertz, and 200' away, I could only see 6 or 7 volts.
>
> Sure, no discussion about that. But at that lenght and frequency, this
> is not due to the cable's characteristic impedance but only the result
> of the line amp's output impedance (600 ohm) and the the cable
> capacitance forming a simple RC lowpass circuit. The C in this case is
> proportional to length, while a cable's characteristic impedance is
> not.

Something I didn't consider in my first generation design since I was still
learning myself, which although it had an output impedance at the TLO84 pin
of under an ohm, then used 300 ohm per leg build outs. The next generation
card, the main design change was to change that 300 ohm to 30 ohms. This
then worked much better out in the newsroom, but was even more susceptible
to the emp spikes the longer runs picked up. Since the card had to fit the
cage, it never did grow the schotkey power diodes to the supply rails that
would have absorbed those spikes, no room left on the the card for diodes
that at the time (1984-85) were about 1/2" long & 1/4" in diameter. It
would have taken 12 of them to protect the inputs as well as the outputs.
These cards all had their own rail regulators (78-7915's) for +-15 volts as
the cage supply was about 22 volts +-, filtered some but not regulated,
good for about 10 amps a rail. 22 cards with 44 pin edge connectors in one
3 space high rack, it did have noticeable but tolerable heat output. The
only place with enough farads available to dump the emp was the main rails,
and a quick test of that idea using a bench supply showed that wasn't a
cure as the output stage of the TLO84 was still destroyed when the output
was pulled above or below the cards internal +-15 volts by about 2 volts,
long before the schotkeys to the +-22 volt rail would turn on.

Since they were very transparent & distortion measurements were difficult
because it was well under .01% at 28v p-p output, and even with the emp
problems, were 10x as dependable as another rack of Ramco DA's we paid 5
grand for before I came in the door, I made sure we had sticks of the
chips, and bags of the capacitors on hand in case mother nature heard
somebody call that stuff butter, and ran with it. At least the TLO84's
were socketed, those quad 741's (4432's?) in the Ramco weren't. Those
eventually were pulled out, they weren't as good sounding by a long ways.
Those 4432's were so slew rated they weren't really any good above about a
volt p-p out. Pure crap to my ears. With that slew rate clipping, the
intermod turned your teeth on edge worse that fingernails on a slate
blackboard. Downright fugly was how I described it at the time.

> At audio freqencies the cable's series resistance dominates the
> impedance of its inductance, which means that the classic equation
> for impedance, sqrt(L/C) is no longer valid. It still has an impedance
> which turns out to be something like sqrt (R/wC). In theory you could
> try and match that at both ends but since it depends on frequency that
> would be very complicated.

Sounds like. In those cases I always figured the cable ohmage if otherwise
impedance matched at the src, was a std factor of the load currents
flowing, and with a nominally 10k load on the far end, a generally
ignorable. Effectively un-terminated, it of course dependent on the src
terms to absorb any 'vswr'. In any event, by then I had managed to obtain
a function generator, and a 10 to 60 khz sweep actually looked pretty good
at the far end, easily within about 5% of flat on the scope. Technically
incorrect?, yes. Was better by magnitudes than what was there when I
walked in the door?, hell yes.
 
> The practical solution for long lines, and what Ma Bell does, is to
> increase the cable's apparent inductance by inserting series inductors
> at regular intervals. These combine with the cable's capacitance to
> produce a purely resistive impedance even at audio frequencies. The
> distance between the inductors determines the bandwidth. Lines used
> for full bandwidth audio (e.g. broadcasting) require more inductors
> per unit lenght.

Which also had the effect of making a long 15 kilohertz rated line into one
of the best 'brick wall' filters you've ever measured. Maybe a db down at
15 khz when they were done, a twenty mile circuit was down 90 db by 17 khz.
And they never did get the S/SN below 48 db, so we eventually bought a
subcarrier radio, which made 75db rather nicely. ;-)

> The practical solution for audio connections up to a few hundred
> meters is to ignore cable impedance completely, accept the capacitive
> load and just drive it with a very low impedance. Anything longer than
> that will be digital these days.

Yep, sure is today Fons, even for a 10 foot run between racks in our new
digital control room. So now the problem is levels, digital doesn't come
with a knob. The std, if there even is one, is ignored, which explains the
commercials that are 20db louder than the program, nobody cares and we
catch hell from the listeners.

I don't believe that will be solved until the commish puts some real teeth
into it, traces the violating material back to the production facility &
then figures out a way to make them legally responsible so they can fine
their ass & make it stick in court. Sure, we can re-encode if we have to,
but there goes the lip sync unless we bucket brigade the video too.

That also takes truly competent manpower most stations don't have near
enough of. In many cases, it isn't even available to hire! This is one
business where we often have to grow our own technical people. Then,
hanging onto them becomes the next problem because its an industry wide
problem and the head hunter business is alive and well.

We are often caught in the "damned if we do, and damned if we don't" world,
so we wind up playing it by ear, letting the ratings tell us how we are
doing. IMO, that isn't always optimum. When I first started back in the
'60's, where I worked often helped the 'other guy' stay on the air, based
on the premise that if he went down and somebody else got that license,
they might come in & spend some real money on physical plant and wipe us
out of the market. A case of preserving the status quo was good business
sense for us. Obviously to protect since this list covers the planet, and
its largely historical anyway, that small market won't be named. :)

> Ciao,

Cheers Fons, Gene

-- 
"There are four boxes to be used in defense of liberty:
 soap, ballot, jury, and ammo. Please use in that order."
-Ed Howdershelt (Author)
My web page: <http://coyoteden.dyndns-free.com:85/gene>
You mean you didn't *know* she was off making lots of little phone 
companies?
_______________________________________________
Linux-audio-dev mailing list
Linux-audio-dev@email-addr-hidden
http://lists.linuxaudio.org/listinfo/linux-audio-dev