Emu XL-7 & MP-7 User's Group

how do you know that given there is no means to access the samples 
directly, without being gain-staged by
the engine itself?

>>>It's all speculation of course.  I recorded the digital outs of the 
XL-7 using a drum ROM preset, one layer, with all of the synth gain 
parameters set to their nominal levels.  Note trigger velocity was set to 
max, as was the sequencer volume control for that track, and the master 
volume was set to full as well. <<

my guess is the samples are at full scale (or close to it), and that 
-14dbFS you are experiencing is a function
of the scaled back digital headroom.

>>>Not disagreeing with you (since we really don't know), but here's my 
thoughts.  We all know that the output of the XX-7's has to be attenuated 
in order to account for summing of multiple parts, not to mention 
resonance peaks.  By normalizing the samples to -14dBFS, you're only 
losing 2 bits of resolution, which is a far easier trade off than coding 
for attenuation in teh sound engine itself.  Just my guess though.  Maybe 
Aaron knows?  I asked Sean about this specifically a few weeks ago, and he 
forwarded my email on to the Emu sound department.  Haven't heard back yet 
though. <<<

rEalm









The information transmitted is intended only for the person(s)or entity 
to which it is addressed and may contain confidential and/or legally 
privileged material. Delivery of this message to any person other than 
the intended recipient(s) is not intended in any way to waive privilege 
or confidentiality. Any review, retransmission, dissemination or other 
use of , or taking of any action in reliance upon, this information by 
entities other than the intended recipient is prohibited. If you 
receive this in error, please contact the sender and delete the 
material from any computer.

For Translation:

http://www.baxter.com/email_disclaimer


[Non-text portions of this message have been removed]

Some of the samples are normalized and some are gain scaled to fit in
with other sounds.  I would guess that the sine and square waves in
each ROM are normalized, so they'd probably be the ones to test.

-Aaron

> how do you know that given there is no means to access the samples 
> directly, without being gain-staged by
> the engine itself?

__________________________________
Do you Yahoo!?
Yahoo! Hotjobs: Enter the "Signing Bonus" Sweepstakes
http://hotjobs.sweepstakes.yahoo.com/signingbonus

on 1/13/04 1:32 PM, Aaron Eppolito wrote:

> Some of the samples are normalized and some are gain scaled to fit in
> with other sounds.  I would guess that the sine and square waves in
> each ROM are normalized, so they'd probably be the ones to test.

Aaron:

I have tested just those samples (among many others)
by sampling my XL-7 via the S/PDIF out with all
engine components disabled or at nominal levels...

the basic analog waveshapes are far below 0dbFS, so
that is why I assumed there is a large degree of
gain scaling occurring in the engine itself...even
when boosting a layer via +10 gain, it is still below
full scale.


cheers,
aeon

Right, remember the samples are 16 bit while the output is somewhere
between 20 and 24 bits (can't remember).  Playing a single 16 bit
waveform should theoretically be much below 0dbFS, yet still retain its
full 16 bit resolution.

--- aeon <aeonlux@...> wrote:
> I have tested just those samples (among many others)
> by sampling my XL-7 via the S/PDIF out with all
> engine components disabled or at nominal levels...
> 
> the basic analog waveshapes are far below 0dbFS, so
> that is why I assumed there is a large degree of
> gain scaling occurring in the engine itself...even
> when boosting a layer via +10 gain, it is still below
> full scale.


__________________________________
Do you Yahoo!?
Yahoo! Hotjobs: Enter the "Signing Bonus" Sweepstakes
http://hotjobs.sweepstakes.yahoo.com/signingbonus

>> I have tested just those samples (among many others)
>> by sampling my XL-7 via the S/PDIF out with all
>> engine components disabled or at nominal levels...
>> 
>> the basic analog waveshapes are far below 0dbFS, so
>> that is why I assumed there is a large degree of
>> gain scaling occurring in the engine itself...even
>> when boosting a layer via +10 gain, it is still below
>> full scale.on 1/13/04 4:32 PM, Aaron Eppolito wrote:

> Right, remember the samples are 16 bit while the output
> is somewhere between 20 and 24 bits (can't remember).
> Playing a single 16 bit waveform should theoretically
> be much below 0dbFS, yet still retain its full 16 bit
> resolution.

nope...

0dbFS is the baseline, and it goes down from there...

theoretically 96dB for the 16-bit, and 144dB for the
24-bit, but in either case, the first bit of the
sample corresponds to the first 6dB of dynamic range...

so regardless of the fact the samples are 16-bit,
if they are at full scale and the digital output
is not, the E-mu engine is doing gain scaling
that cannot be overcome/compensated for by
adjusting engine parameters.


cheers,
aeon

The entire signal path of the XL-7 is 24 bit.  Envision the 16 bit
samples LSB justified in 24 bit space.  The high 8 bits are headroom. 
When you add 128 of these samples, you get a 23 bit number.  When you
add in gain of up to +10dB, you can get 25 bits.  When you add in
resonance you can get even more bits.  (This is why the filters can
clip, by the way)

At no point do you lose the 16 bit precision of the original samples. 
Just because they're down at the LSB side instead of the MSB side
doesn't mean that it's in any way less quality.

That being said, if you record digitally at 16 bit, yes, you will lose
important data.  You should record at 24 bit then normalize.  This
gives the best of all worlds (maximum headroom with the full original
sample precision).

|23|22|21|20|19|18|17|16|15|14|13|12|11|10|9|8|7|6|5|4|3|2|1|0|
  0  0  0  0  0  0  0  0  1  1  1  1  1  1 1 1 1 1 1 1 1 1 1 1
 \__these are headroom__/ \___16 bit sample LSB justified____/

  1  1  1  1  1  1  1  1  1  1  1  1  1  1 1 1 0 0 0 0 0 0 0 0
 \_______16 bit sample MSB justified_________/ \__USELESS!!__/

If you were to left (MSB) justify the samples, the low 8 bits would be
useless, as they'd always be zero.  All that aside, a headroom control
would still be useful for those using the S/PDIF without benefit of a
normalizer.

-Aaron

--- aeon <aeonlux@...> wrote:
> so regardless of the fact the samples are 16-bit,
> if they are at full scale and the digital output
> is not, the E-mu engine is doing gain scaling
> that cannot be overcome/compensated for by
> adjusting engine parameters.


__________________________________
Do you Yahoo!?
Yahoo! Hotjobs: Enter the "Signing Bonus" Sweepstakes
http://hotjobs.sweepstakes.yahoo.com/signingbonus

on 1/13/04 8:34 PM, Aaron Eppolito wrote:

> The entire signal path of the XL-7 is 24 bit.

OK. Does the CS engine (or modern Proteii engine
in general) use integer DSP with some larger
accumulator for multiplies and adds?

> Envision the 16 bit samples LSB justified in
> 24 bit space.  The high 8 bits are headroom.

but why would it be done this way?

I ask because regardless of the bit depth used,
0dBFS should be the same signal peak voltage on the
analogue output...24-bit would just give you the
ability to resolve dynamic range below -96dBFS...

> When you add 128 of these samples, you get a
> 23 bit number.  When you add in gain of up to
> +10dB, you can get 25 bits.  When you add in
> resonance you can get even more bits.  (This
> is why the filters can clip, by the way)

OK, now I understand why it is be done that way
(LSB justified).

> At no point do you lose the 16 bit precision of
> the original samples. Just because they're down
> at the LSB side instead of the MSB side doesn't
> mean that it's in any way less quality.

in the engine, this is true...but as it hits the
D/A converter, it is no longer the case, as the
chipset used in these machines (along with the
analog path) will not resolve a number of bits on
the LSB side, because they are below the noise
floor of the system.

> That being said, if you record digitally at 16 bit,
> yes, you will lose important data.  You should
> record at 24 bit then normalize.  This
> gives the best of all worlds (maximum headroom with
> the full original sample precision).

OK, understood...but still, it would be nice to have
a param by which we could control the gain scaling
on the voice level, so as it have the best control
of that digital out.

> ...a headroom control would still be useful for
> those using the S/PDIF without benefit of a
> normalizer.

oops, just said that! :)

Aaron, thanks so much for the explanation!


cheers,
aeon