Using a 1 KHz square wave, the L/R program channels of the Delta masters
show something a little odd at the active balanced outputs.

For each output channel, the circuit consists of a predriver op amp and
two line driver amps. The predriver output is Y'd and drives the + input
of one line amp and the - input of the other. The individual feedback
loops on each driver amp goes to the - and + inputs, respectively.

The square wave is clean from the unbalanced output of the pre-driver amp.

But when I look at the + output of the balanced line I see a narrow spike
at the leading edges of the square wave. The amplitude of the spike is
approximately 15% of the waveform amplitude. The width of the spike is
roughly 10 microseconds.

When I look at the - output, I see a slight "rounding off" of the leading
edges of the square wave, with same width as the spike on the + amp.

The spike/round off width does not vary at all over the hz range of
square waves that the system will will reasonably pass.

If I put the + and - output signals on a dual trace scope, invert the
scope input of one of them, then move the trace vertical positions so that
the waveforms just touch top to bottom, the spike of the + line driver
fits neatly into the leading-edge round off area of the minus side. It's
an exact fit, in fact.

I have tried the original 5532s and also the new 2134s and get exactly the
same thing. Both L and R channels show the same thing.

I have adjusted the balanced line symmetry control and only change the
amplitude of the + and - signals. The spike and rolloff do not go
completely away. (This 1K pot, with its wiper to ground and legs fed
through 1K resistors from the + and - output lines, determines the
relative levels of the + and - signals.)

I'm curious as to what might cause a spike on the + side and a rounding on
the - side, and whether it could have audible effect given that it appears
to be supersonic. Is there something inherently problematic about this
type of balanced line driver configuration? Seems like it ought to be
pretty straight-forward.

I have not yet replaced the resistor packs with discrete resistors as Jim
had suggested sometime back. (These resistor packs are used around the
line driver opamps to set gain/feedback and such.)

What am I missing?

Thanks in advance,
Frank Stearns
Mobile Audio
--

On Sat, 04 Jun 2005 17:55:33 -0000, in rec.audio.pro Frank Stearns
wrote:

Using a 1 KHz square wave, the L/R program channels of the Delta masters
show something a little odd at the active balanced outputs.

For each output channel, the circuit consists of a predriver op amp and
two line driver amps. The predriver output is Y'd and drives the + input
of one line amp and the - input of the other. The individual feedback
loops on each driver amp goes to the - and + inputs, respectively.

The square wave is clean from the unbalanced output of the pre-driver amp.

But when I look at the + output of the balanced line I see a narrow spike
at the leading edges of the square wave. The amplitude of the spike is
approximately 15% of the waveform amplitude. The width of the spike is
roughly 10 microseconds.

When I look at the - output, I see a slight "rounding off" of the leading
edges of the square wave, with same width as the spike on the + amp.

The spike/round off width does not vary at all over the hz range of
square waves that the system will will reasonably pass.

If I put the + and - output signals on a dual trace scope, invert the
scope input of one of them, then move the trace vertical positions so that
the waveforms just touch top to bottom, the spike of the + line driver
fits neatly into the leading-edge round off area of the minus side. It's
an exact fit, in fact.

I have tried the original 5532s and also the new 2134s and get exactly the
same thing. Both L and R channels show the same thing.

I have adjusted the balanced line symmetry control and only change the
amplitude of the + and - signals. The spike and rolloff do not go
completely away. (This 1K pot, with its wiper to ground and legs fed
through 1K resistors from the + and - output lines, determines the
relative levels of the + and - signals.)

I'm curious as to what might cause a spike on the + side and a rounding on
the - side, and whether it could have audible effect given that it appears
to be supersonic. Is there something inherently problematic about this
type of balanced line driver configuration? Seems like it ought to be
pretty straight-forward.

I have not yet replaced the resistor packs with discrete resistors as Jim
had suggested sometime back. (These resistor packs are used around the
line driver opamps to set gain/feedback and such.)

What am I missing?

Thanks in advance,
Frank Stearns
Mobile Audio
The way I look at these simulated balanced outputs is that you must
use a diffential receive amp or transformer to look at the signal.

Looking at just one leg is not satisfactory as you are dealing with a
differntial signal.

I think that you are effectvely looking at a common mode sig that will
be more or less eliminated by the common mode rejection of the receive
amp




martin

Frank Stearns wrote:

Using a 1 KHz square wave, the L/R program channels of the Delta masters
show something a little odd at the active balanced outputs.

For each output channel, the circuit consists of a predriver op amp and
two line driver amps. The predriver output is Y'd and drives the + input
of one line amp and the - input of the other. The individual feedback
loops on each driver amp goes to the - and + inputs, respectively.

The square wave is clean from the unbalanced output of the pre-driver amp.

But when I look at the + output of the balanced line I see a narrow spike
at the leading edges of the square wave. The amplitude of the spike is
approximately 15% of the waveform amplitude. The width of the spike is
roughly 10 microseconds.

When I look at the - output, I see a slight "rounding off" of the leading
edges of the square wave, with same width as the spike on the + amp.

The spike/round off width does not vary at all over the hz range of
square waves that the system will will reasonably pass.

If I put the + and - output signals on a dual trace scope, invert the
scope input of one of them, then move the trace vertical positions so that
the waveforms just touch top to bottom, the spike of the + line driver
fits neatly into the leading-edge round off area of the minus side. It's
an exact fit, in fact.

I have tried the original 5532s and also the new 2134s and get exactly the
same thing. Both L and R channels show the same thing.

I have adjusted the balanced line symmetry control and only change the
amplitude of the + and - signals. The spike and rolloff do not go
completely away. (This 1K pot, with its wiper to ground and legs fed
through 1K resistors from the + and - output lines, determines the
relative levels of the + and - signals.)

I'm curious as to what might cause a spike on the + side and a rounding on
the - side, and whether it could have audible effect given that it appears
to be supersonic. Is there something inherently problematic about this
type of balanced line driver configuration? Seems like it ought to be
pretty straight-forward.

I have not yet replaced the resistor packs with discrete resistors as Jim
had suggested sometime back. (These resistor packs are used around the
line driver opamps to set gain/feedback and such.)

What am I missing?

It's a potential characterisitc of the 'cross-coupled' balanced outputs. Might
be parasitic capacitance somewhere. Or even slightly mismatched caps around
the feedback loop.

The important thing as you found is that the two outputs when summed are
accurate.

It shouldn't be the ICs btw. I do use 1% Rs in this configuration myself
though.

Graham

In article writes:

Using a 1 KHz square wave, the L/R program channels of the Delta masters
show something a little odd at the active balanced outputs.

The square wave is clean from the unbalanced output of the pre-driver amp.

But when I look at the + output of the balanced line I see a narrow spike
at the leading edges of the square wave. The amplitude of the spike is
approximately 15% of the waveform amplitude. The width of the spike is
roughly 10 microseconds.

When I look at the - output, I see a slight "rounding off" of the leading
edges of the square wave, with same width as the spike on the + amp.

If I put the + and - output signals on a dual trace scope, invert the
scope input of one of them, then move the trace vertical positions so that
the waveforms just touch top to bottom, the spike of the + line driver
fits neatly into the leading-edge round off area of the minus side. It's
an exact fit, in fact.

That's the way this cross-coupled transformer-emulating circuit works
when everything is at unity gain (which they do to keep the headroom).
The good news is that it appears that you output stage is working
properly - when summed (differenced?) at a differential input, the
nasties exactly cancel. But it's a good lesson in why you shouldn't
use just one side to ground with this kind of output stage if you're
feeding an unbalanced input. You should tie the low side (pin 3) to
ground (pin 1) and take the unbalanced output between pins 1/3 and 2.

Try slamming it so that the output stage runs into clipping and you
may be really horrified at what you see. The Mackie Onyx line (except
the 1220) use an output stage like this. If this works, you'll find a
picture of what the two sides of the balanced output look like when my
1620 clips. Sum them together and you get a nice, symmetrical waveform
that's flat on the top and bottom.

http://www.imagehosting.us/imagehost...jpg/?id=510053


--
I'm really Mike Rivers )
However, until the spam goes away or Hell freezes over,
lots of IP addresses are blocked from this system. If
you e-mail me and it bounces, use your secret decoder ring
and reach me he double-m-eleven-double-zero at yahoo

Mike Rivers wrote:

In article writes:

Using a 1 KHz square wave, the L/R program channels of the Delta masters
show something a little odd at the active balanced outputs.

The square wave is clean from the unbalanced output of the pre-driver amp.

But when I look at the + output of the balanced line I see a narrow spike
at the leading edges of the square wave. The amplitude of the spike is
approximately 15% of the waveform amplitude. The width of the spike is
roughly 10 microseconds.

When I look at the - output, I see a slight "rounding off" of the leading
edges of the square wave, with same width as the spike on the + amp.

If I put the + and - output signals on a dual trace scope, invert the
scope input of one of them, then move the trace vertical positions so that
the waveforms just touch top to bottom, the spike of the + line driver
fits neatly into the leading-edge round off area of the minus side. It's
an exact fit, in fact.

That's the way this cross-coupled transformer-emulating circuit works
when everything is at unity gain (which they do to keep the headroom).
The good news is that it appears that you output stage is working
properly - when summed (differenced?) at a differential input, the
nasties exactly cancel. But it's a good lesson in why you shouldn't
use just one side to ground with this kind of output stage if you're
feeding an unbalanced input. You should tie the low side (pin 3) to
ground (pin 1) and take the unbalanced output between pins 1/3 and 2.

Try slamming it so that the output stage runs into clipping and you
may be really horrified at what you see. The Mackie Onyx line (except
the 1220) use an output stage like this. If this works, you'll find a
picture of what the two sides of the balanced output look like when my
1620 clips. Sum them together and you get a nice, symmetrical waveform
that's flat on the top and bottom.

http://www.imagehosting.us/imagehost...jpg/?id=510053

Curious. Do you mean Mackie use the cross-coupled configuration with unity gain from
unbalanced in ( usually post fader buffer ) to balanced out ?

I always add 6dB to indeed 'preserve headroom' so that the balanced out doesn't clip until
it's at +28dBu ! This means that all the mix bus etc runs 6dB lower than 'external' level
too giving a bit more mix headroom pre output level fader.

Graham

Don't sweat it. When run into an Audio Precision analyzer, distortion
of this stage is around .0005%, differential. If you match your
resistors to.1%, you could remove the trim pot. Use a 4 digit DVM and
use Dale RN55 resistors from Mouser. Replace the 22pf feedback caps
with mono ceramic NPO types. Match these on a Capacitor bridge to get
better high frequency CMRR.

Yes, it never ends with a console of this pedigree.

Solder on, men.

Jim Williams
Audio Upgrades

In article writes:

Curious. Do you mean Mackie use the cross-coupled configuration with unity gain
from
unbalanced in ( usually post fader buffer ) to balanced out ?

Yes.

I always add 6dB to indeed 'preserve headroom' so that the balanced out doesn't
clip until
it's at +28dBu ! This means that all the mix bus etc runs 6dB lower than
'external' level
too giving a bit more mix headroom pre output level fader.

Good policy. The Mackie is a bit skimpy on the headroom, and I think
the maximum output level (differential, between pins 2 and 3) is only
+22 dBu. Curiously, the Onyx 1220, the "baby" of the line, uses the
same output configuration as the earlier series, a unity gain inverter
between pin 2 and pin 3. That one will put out +26 dBu before
clipping.

Since the Onyx already has a single-ended output on a TRS jack, with
the ring tied to ground through a resistor, I questioned why they
bothered with this cross-coupled output configuration at all. I never
got a really good answer, but I suspect that their thinking was that
it would end the "if you use a balanced output unbalanced, you lose
6 dB" discussion - that is, if you ground the low pin (which you DON'T
want to do if it's driven by a straight inverter). However, I think
that it backfired, at least for anyone (like me) who bothered to look
at the individual halves of the differential output.

If you looked at the photo, you'll see that the amplitudes of the two
sides aren't equal. That wasn't a matter of scope gain not being
equal, they're really different voltages. Not only that, but I
measured a wide variation in the amount of "unbalance" between the two
legs of the output among two mixers and the eight outputs of the 800R
stand-alone preamp. A few were very close to equal in amplitude, some
were as far as 50% different, but always adding up to the right
amplitude when driving a differential input. Without a trimmer (like
the Soundcraft has - I also have a Soundcraft 600) you just can't get
resistors close enough in value, even 1% values, to have it completly
symmetrical.

So (and I think I may have said this here before) if you're using the
XLR outputs on an Onyx mixer, either connect them to differential
inputs or tie Pin 3 to Pin 1. Otherwise you can't be sure what level
you'll get out when the meter reads 0 VU.

--
I'm really Mike Rivers )
However, until the spam goes away or Hell freezes over,
lots of IP addresses are blocked from this system. If
you e-mail me and it bounces, use your secret decoder ring
and reach me he double-m-eleven-double-zero at yahoo

wrote:

Don't sweat it. When run into an Audio Precision analyzer, distortion
of this stage is around .0005%, differential. If you match your
resistors to.1%, you could remove the trim pot.

Are you absolutely *sure* about that ? At Studiomaster I improved the R
match to 1% types hoping to be able to remove the trimpot - but it simply
made things a bit better.

I'd expect use of 0.1% types to simply make it a little better still.

I did an analysis of that circuit once and the funny result is that it
works for *any* combination of + and - outputs. In short the equation
has no resolvable answer !

Use a 4 digit DVM and use Dale RN55 resistors from Mouser. Replace the
22pf feedback caps with mono ceramic NPO types. Match these on a
Capacitor bridge to get better high frequency CMRR.

If you use the former Philips ( now Vishay - BC ) ceramic plate Cs ( the
low value ones ) - these are *manufactured* to 2% tolerance btw. They're
rarely more than 1% out in practice.

Regds, Graham

Mike Rivers wrote:

In article writes:

Curious. Do you mean Mackie use the cross-coupled configuration with unity gain from
unbalanced in ( usually post fader buffer ) to balanced out ?

Yes.

I'm astonished !


I always add 6dB to indeed 'preserve headroom' so that the balanced out doesn't clip
until it's at +28dBu ! This means that all the mix bus etc runs 6dB lower than
'external' level too giving a bit more mix headroom pre output level fader.

Good policy.

Seems common sense to me. Ok - it can make the mix bus a fraction noisier overall but not by
much and the headroom increase is great esp for those who like to run a 'hot mix' - not
unusual in these days of digital.

Let's put it his way. it was good enough for Neve V series !

The Mackie is a bit skimpy on the headroom,

Now I understand all those user comments ! Clearly the gain structure isn't optimised for
what I'd call 'real world' use.

and I think
the maximum output level (differential, between pins 2 and 3) is only
+22 dBu. Curiously, the Onyx 1220, the "baby" of the line, uses the
same output configuration as the earlier series, a unity gain inverter
between pin 2 and pin 3. That one will put out +26 dBu before
clipping.

Oh yuk ! That crazy configuration ! Have they not learnt ?


Since the Onyx already has a single-ended output on a TRS jack, with
the ring tied to ground through a resistor, I questioned why they
bothered with this cross-coupled output configuration at all. I never
got a really good answer, but I suspect that their thinking was that
it would end the "if you use a balanced output unbalanced, you lose
6 dB" discussion - that is, if you ground the low pin (which you DON'T
want to do if it's driven by a straight inverter). However, I think
that it backfired, at least for anyone (like me) who bothered to look
at the individual halves of the differential output.

They don't appear by the sounds of it to have a coherent design policy to me.


If you looked at the photo, you'll see that the amplitudes of the two
sides aren't equal. That wasn't a matter of scope gain not being
equal, they're really different voltages.

I did wonder !

Not only that, but I
measured a wide variation in the amount of "unbalance" between the two
legs of the output among two mixers and the eight outputs of the 800R
stand-alone preamp. A few were very close to equal in amplitude, some
were as far as 50% different, but always adding up to the right
amplitude when driving a differential input. Without a trimmer (like
the Soundcraft has - I also have a Soundcraft 600) you just can't get
resistors close enough in value, even 1% values, to have it completly
symmetrical.

I agree. Well.... I know it to be a fact with 1% values to be entirely accurate ! I believe
the trimmer to be a necessary evil. See my comment to Jim Williams.


So (and I think I may have said this here before) if you're using the
XLR outputs on an Onyx mixer, either connect them to differential
inputs or tie Pin 3 to Pin 1. Otherwise you can't be sure what level
you'll get out when the meter reads 0 VU.

That's bad ! Considering that I thought the Onyx to be a major improvement in Mackie's line
I'm truly amazed.

Btw - you get +28dBu out with proper balanced trimming on this little beast.

http://www.studiomaster.com/c1.html


Graham

I did see dolby do it with .1% resistors on the 363 SR outputs, the
symmetry was very close. Love those Chi-com made BC mono ceramics from
Digi Key. I was matching some yesterday and was suprised to find how
close they are compared to Panasonic or AVX. Wasn't hard to find a
couple of 470 pf's to match within 1 picofarad.

If you folks want to get really good matching in this circuit without
the loss of headroom caused by the attenuation of one output via a trim
pot, do this:

Place a small trim pot around 500 ohms in series with one of the
negative feedback resistors. If it's normally 10k ohms, use a 9.76 k
instead so you have some adjustment room to work with. This will alllow
CMRR to be trimmed to -90 db, 20 to 20k hz. Use an Audio Precision to
get the best alignment. Keep in mind the small feedback caps should
also be matched and any output coupling caps will degrade CMRR in the
low frequencies unless they are also matched. It may be better to do
the trim before the output caps are installed.

I do this in the output stage of the Audio Upgrades High Speed Micpre
and get -90 db CMRR throughout in this differential mic pre design.

Jim Williams
Audio Upgrades