I should have mentioned I did my tests with all inputs disabled except
for a single recording source. This appears to be the way the mixer
software is designed to function, only 1 recording source can be
enabled at a time (I recall other sound cards, particularly SB16
cards, would allow recording from multiple sources, I think).
Enabling "digitize input" subjects the input signal to the processing
of the digital audio circuits/drivers. Enabling "digitize input" may
make the noise better or worse depending upon your equalizer and
"environment" settings.
I can't recall if I have the 4-wire CD audio cables installed in this
box.
The noise floor is not awful, on the contrary, it's decent or lower
than -50 dB which gets the noise floor down where one will not hear it
above typical computer fan noise or room noise. But, the mic boost
clearly raises the noise floor 20 dB, which is to be expected since
it's simply an amplification of the mic input signal.
I did notice tonight while playing with the 3com NIC (which has been
disabled for some time, but I'm using it for a brief span for some
testing purposes) that a network speed test using Sisoft Sandra
resulted in identifiable noise in the headphones I was using during
the brief burst of data, and I never noticed any such noise with the
Nvidia NIC. The Nforce thread you linked above where users suggest
some cures is humorous in a way, b/c each users seemed to find the
magic fix to noise, e.g., paper washers, using only the 3com nic,
using only the Nvidia NIC, disabling a PCI bus service,
enabling/disabling "digitize input", etc.
It is clear the multifunction LSI circuits that include onboard sound
face huge design hurdles in establishing a low noise level. In
addition, the location of the audio signals in the mobo (near the CPU)
puts those signals in a very noise environment. However, I thought to
get and use the Dolby Digital trademarks, certain performance levels
were required in the S/N area, or maybe I am mistaken.
I'll have to have a look at my system with the SBLive 5.1 and see if
it's noise floor is any better, just a sec. . . . (/me fiddles with
sblive 5.1 card in an older K7S5A based 1.4G Tbird system)
No doubt about it, the SBLive card is much better in the noise floor
area, the mic input had lower than -70 dB reading with no mic plugged
in and "mic boost" enabled, and "line in" noise floor was less
than -60 dB level (I do have the "line in" jack connected to a stereo
receiver at the moment as I'm recording old LPs and making MP3s
therefrom, so that might explain the difference in noise floor there).
One observation I note, the SBLive board has a good quantity of very
small electrolytic caps, typically used for noise filtering in audio
circuits.[/QUOTE]
But this is not your typical "noise filtering" problem.
First of all, the AC97 codec does the digitizing. Transmission from the
Realtek chip (ALC650?) on the left of the board to the Southbridge, is
done digitally over the AClink.
The noise problem will exist from the audio connector rack in the I/O
area, to the AC97 codec. Now, I looked at two A7N8X family boards, and
I can see audio signals travelling on the surface of the board. This makes
them "antennas" for noise, and is a definite no-no. They should be routed
stripline, with solid ground plane over top of them, to protect against
E fields. Also, there should be sufficient clearance between them and any
H fields, like the transformers for the Ethernet interface.
The whole Asus nforce2 layout is compromised by the dual channel memory
interface. It pushes the processor socket and Northbridge too far to the
left, and forces other circuits to be placed sub-optimally. If you want
to see a better layout, compare an A7N8X board to the Abit NF7-S (the
first one I grabbed).
http://www.abit-usa.com/images/products/NF7_S_large.jpg
Notice how the DIMMs are much closer to the front of the computer,
the Vcore power circuit is pushed North, away from critical routing
areas near the I/O panel. The area south of PCI5 has been used for
chips/headers. The layout is less crowded where the audio chip lives,
and there is more opportunity to treat the audio right. (I wish I had
a hi res picture, to be able to see more of it).
If you remember the early history of the A7N8X, Asus did lose their
SoundStorm "sticker", and who is to say why. It could be those very
things, like measured noise floor and quality of implementation.
Imagine how good a board like this could have been, if an extra two
routing layers were available. Adding 2 more layers of copper adds
10% to the cost of the PCB board blank, but it would have allowed
that huge copper "rats nest" between the Northbridge and the DIMMs
to be compressed, and allowed the use of ground fill around circuits
that really needed it.
Also, the audio circuit is powered by a small linear regulator. In
theory, this is supposed to clean up power to the circuit, but I've
been taught that in fact a switching power supply can be cleaner
than a linear regulator. The linear regulator rejects noise well at
low frequencies, but not at high frequencies, and the +12V used to
power the regulator, is bound to be dirty after travelling half way
around the board, to get to the audio circuit. Yet, I don't see any
RLC filters feeding the regulator, to improve its output.
So, there are a bunch of things that could be done differently. As the
board is intended for budget AMD applications though, it is hard to
do the job right. The NF7-S layout proves there is room for
improvement, in terms of general layout issues.
A sound card, on the other hand, has no compromises in its design.
Plenty of room on a PCI card, to do an audio function, without
messing it up.
Paul
