LOOKING FOR YOUR SEAT IN THE
When comparing hard and soft processing,
the common misconception is that using the "exact same settings" should
match the hardware. At best, that gets you in the ballpark because
analog component tolerance - especially of vintage gear - is too wide to
be do hardware comparisons. D2C will show how to roll your own comparison.
First you must consider how many jobs the
plug is being asked to do and what degree of resolution can be expected.
We have the luxury of running 24-plus tracks all the time with plugs on
every channel. Obviously one of the programmer's goals is code efficiency,
which is not necessarily in the best interest of absolute emulation.
If and when true emulation becomes possible, it might be so processor intensive
that, for example, you might only get to run one instance of a Fairchild
670 on the mix buss.
Simply put, it's easier to make the graphics
look like the original than to make them sound like some of the originals.
But that doesn't mean we should toss all our plugs in the trash.
Evren when they don't emulate exactly as claimed, many plugs are useful
tools once we figure out what they are good for - just like the analog
That said, the processing that should be
happening falls into two categories - emulation of the feature set (EQ
or Dynamics Processing) and of the artifacts (desirable distortion).
I would persoally trade all the retro grapics for more sonic options -
being able to isolate and run throug the LA-2's tube amplifier or the 1176's
input transformer, for example. But, no one is doing that.
There are, IMHO, four reasons for this
- (1) the marketing / programmers think the users are not that intelligent,
(2) emulating gear-specific circuitry is not really what's behind the facade
(3) it's not possible at this time, (4) why work harder than you have to?
I will admit that even I tend to use stuff that works the first time -
most of us do need to work quickly in order to do more tan break even.
That said, the D2C project put me in Tenacious Geek mode - I got sucked
into hours of fine tuning plugs to get as close an emulation as possible.
Once you know what to listen for - the
trained ear is an excellent evalaution device - you will realize how complex
the analysis process is.. Distortion analyzers in the analog and
digital doman typically provide one static evalustion facet - sound is
obviously so much more complex than that.
Thanks to Zach Ridgeway for pointing out
an article by Bob Katz - about oversampling plugins, in his book,
Audio - The Art And Science" pages 206 - 208 - as well as an Ohmforce
link. Their plugs all have a "high-quality" button that puts them
in oversampling mode. The Weiss DS1-MK2 also oversamples. Even
if your plug-ins do not have this capability, you can always manually achieve
higher resolution by first upsampling - from 44.1 to 88.2, for example
- then processing before making a CD master or equivalent.
Thanks also to Dave
Hill at Crane Song. Of the many conversations we'd had, the most
recent touched on all of these points. I can honestly say that those
who successfully Dare to Compare will find themselves less affected by
hype and hearsay more empowered to do their own reviews. Please feel
free to share your results by sendig files or providing links.
WHAT ARE WE LISTENING FOR?
At the heart of any sonic comparison is
the ability to NULL - subtracting one track from another yields the differences.
Mid-Side (also known as Sum + Difference) is another process that takes
advantgae of the difference between two channels.
Your job is to tweak until the differences
are minimized and the signals come as cose to complete emulation, and therefore
cancellation, as possible. This is accomplished by starting with
a RAW, unprocessed track, processing with and capturing the Device Under
Test (DUT) and then opening the appropriate plug-in (PLUG). The DUT
can be hardware or software, whatever does the job for you.
Reverse the PLUG polarity and mix with
the DUT At first, there will only be partial cancellation -
that's the difference between the original signal and what's being done
to it. In this application, the null technique allows our ears to
measure "distortion," or the devation from the exact copy or "linear" amplification.
Take the following step to maximize cancellation
TERMS OF SONIC ENDEARMENT
Start with an unprocessed (RAW) reference
track such as drums, vocal, piano, etc. - process as needed and capture
the Device Under Test (DUT).
Visually line up the two tracks to be sample
Route the RAW (reference), DUT and process/
plug tracks to the mix buss.
Insert the short sample delay / invert plug
on all channels
Reverse polarity and match levels to find
the NULL. Make as much audio disappear as possible.
Experiment with sample delay to confirm that
timing is not the obstacle. "Best" is when the high frequencies are
Copy the RAW track and create a new channel
strip for it.
Mute the original RAW / reference track.
Insert sample delay, polarity and processing
plugs and again find the NULL
Tweak the controls - attack, release and ratio
(for a dynamics processor) or the various EQ parameters to further improve
Phase Shift NULL: as caused by intentional
and non-intentional frequency altering components (capacitors in analog
and latency in the digital domain). Adjust for max treble cancellation.
Amplitude NULL: When judging
by ear, it is essential that the levels be perfectly matched. The null
technique is the best way to achieve this.
EQ NULL: Digital Domain equalization
often requires more than one EQ PLUG to emulate in a "simple" analog equalizer
DYNAMICS NULL: This is the trickiest
of all because every tweak requires a new GAIN OFFSET
Take a moment to check out the audio samples
(below) and then try to roll your own. Your audio snippets are also
welcome. You can either point me to your web post or email - please
note that I only need about 30 seconds of audio. Music should preferably
have an obvious transistion (verse to chorus, etc.) or obvious pattern
change because this will reveal the challenges / flaws.
1064 EQ, API 550B EQ, 1176, Alan Smart
Samples provided by Tom Garneau and Adam
Samples provided by Peter Bregman and Dusty
Samples provided by Colt Leeb
Samples provided by David Hedding
Samples provided by Colin McArdel