PC Notes
Technique : PC Notes
All USB ports are not created equal, as PC Notes finds out this month...
Further to my PC Musician feature concerning Windows XP Tweaks in last month's SOS,
you may remember that, as an experiment, I carefully disabled a total
of 46 Windows Services on my music partition, according to a list
provided by a well-known commercial builder of music PCs. This effort
resulted in absolutely no change in CPU overhead, a fairly insignificant
release of just 8MB of system RAM and no measurable improvement in song
performance, confirming my opinion that tweaking Services is a
pointless exercise.
I also warned of possible instability as a result of
tweaking Services. Well, I left my super-clean music partition with
these tweaks in place, and during the next week found time for several
long sessions working on some of my own songs. While my music
applications all booted up correctly and seemed to run as before, during
each session I experienced a number of infuriating crashes,
particularly with Gigastudio 3, and I eventually found a sequence of actions that would crash the Gigapulse reverb window every time, with an application error that required a reboot. Cubase SX 3 seemed less affected, although on the second session it also crashed badly (previously a rare occurrence on my PC).
Fortunately, I'd created a hard drive image
immediately before the Service tweaks, so I could restore my partition
to its previous state. As soon as I did so, my application crashes
disappeared, and I was once more able to work with Cubase SX rewired to Gigastudio over an eight-hour session, without any bad behaviour at all from my PC.
So let me reiterate my previous advice from last
month's feature: "Disabling various Windows Services in an effort to
further streamline your PC's audio performance is, in my opinion,
pointless. Moreover, if you don't know exactly what you're doing, your
system can become unstable or even refuse to boot up afterwards." Let me
now add to this that even if you do know what you're doing,
and carefully follow a list published on the Internet, you can still end
up with an unstable PC. Just say no to Service tweaks!
USB Shenanigans
This month I also had an incredibly frustrating
experience while attempting to install a new USB peripheral, so just in
case it ever happens to you, here are the gory details. Interested in
on-body printing of CD-R disks, I ordered a Canon PIXMA iP4200 USB 2.0
printer — a bargain at £60 including VAT (although printer cartridges
for it are expensive).
I fitted the print head and a set of ink tanks, and
then followed the driver installation instructions to the letter, only
plugging in and then switching on the printer when asked (to bypass
Window's own 'New Hardware Found' regime). However, as soon as I did so,
up popped the Windows message 'USB Device Not Recognised (one of the
USB devices attached to this computer has malfunctioned, and Windows
does not recognise it'.
If you ever see this error message when you plug in
a USB peripheral, don't assume it's faulty — it may be your choice of
USB port that's causing a problem.
|
I replaced the USB cable, just in case it was
faulty, but there was no change. I then plugged the printer into another
spare USB port, still with no success. Just in case both these USB
ports were faulty, as a test I plugged a USB memory stick into them,
which worked perfectly. I followed all the fault-finding advice in the
Canon PDF manual, but the printer stubbornly refused to be recognised.
At this point I was almost convinced that the
printer must be faulty, but I persevered, and tried installing its
drivers on my laptop. Success! The printer was recognised first time and
printed test pages perfectly. Back on my desktop PC, I reorganised my
other USB peripherals so that I could try the printer in yet another
port, and this time it worked perfectly, but still refused to work in
the first two ports I'd tried. The only difference was that the port
that worked was a captive one mounted on the motherboard, while the two
that didn't were on a backplate attached to one of the motherboard's
headers via a short internal cable. Bizarre!
At this point I remembered some similar behaviour a
year or so ago, when a different USB memory stick wasn't recognised at
all when plugged into my desktop PC's backplate USB ports, but worked
perfectly on my laptop. Armed with this new knowledge, I plugged the
stick into a captive desktop USB port and it too was recognised!
Given that this USB backplate assembly was bundled
with my Asus motherboard and consists of just a few inches of ribbon and
no active components, I found it very surprising that its ports don't
provide performance identical to the other USB ports. However, you have
been warned: if you ever encounter problems with a new USB device not
being recognised at all, or being recognised and declared faulty, make
sure you try a captive port before throwing in the towel.
Unfortunately, my relief was short-lived: my new USB
printer still wouldn't print CDs properly, so it had to be sent back
for a replacement. The only reassuring thing to come out of the whole
sorry affair is that I can recommend the company I bought the printer
from, for their professional tech support and their prompt and efficient
returns policy. Take a bow Scan (www.scan.co.uk), whose employees had no idea who I was, or that I'd reviewed one of their music PC systems in SOS.
I notice plenty of musicians claiming that their
audio interfaces manage a latency of 1.5ms, when the true figure is
closer to 3ms. The reason for this confusion is that the 'real world'
latency not only includes the interface buffer but also the delays due
to A-D and D-A converters, plus any overhead caused by extra DSP
functions or buffers on the interface.
If you send a short audio pulse from your PC
sequencer to your audio interface output, through an audio loopback
cable connected between its analogue output and input, and then back in
through the audio interface inputs, you can measure the round trip, as I
described in SOS September 2002. However, a new utility from
Centrance automates the whole process, making it far quicker and less
prone to user error. This handy ASIO Latency Test Utility (LTU) is a free download from their web site (www.centrance.com/products/ltu).
Download this handy utility from Centrance and you
can measure the 'real-world' latency of your audio interface. Here's a
typical example, where the buffer size/latency of my Echo Mia soundcard
is reported at 5.8ms inside software applications but the expected
11.6ms output/input round trip actually measures 14.42ms, due to
converter and DSP extras.
|
Centrance claim accuracy within 0.5ms, but I found
that my audio interfaces provided very repeatable results often to
single-sample accuracy. Once you've chosen a buffer size and sample
rate, and clicked on the Measure button, you can subtract twice the
buffer size from the result to reveal the 'extras', which should be
independent of buffer size. For instance, with a buffer size of 264
samples (6ms at 44.1kHz) my Emu 1820M measured a path length of 628
samples (14.24ms). After subtracting twice the buffer size from this
total, to remove the input and output buffer contributions, you end up
with a fairly typical 100 samples (2.2ms at 44.1kHz) for 'extras'. This,
to all intents and purposes, is your 'zero'-latency delay (zero-latency
monitoring connects the output of the A-D converter internally to the
input of the D-A converter).
So far, such results may seem largely academic, but
other interfaces may reveal more interesting facts. For example, the
extras on my Echo Mia measure 125 samples (2.83ms at 44.1kHz), which, as
I explained in SOS October 2002, comprises 31 samples for the
A-D converter, 30 samples for the D-A converter, and a hidden 32-sample
DSP buffer on both outward and inward journeys, for PCI-buss efficiency.
Some interface manufacturers are quite open about
such hidden buffers. RME, for instance, employ a 64-sample 'safety'
buffer (which would add just 1.45ms at 44.1kHz) on their Fireface, to
ensure smooth playback. This is because, in contrast with PCI
soundcards, Firewire interfaces lack DMA (Direct Memory Access) support.
However, other manufacturers don't reveal such information, so it's
quite possible that an interface that musicians find they can run
successfully with a 64-sample buffer size does so because it has extra
DSP buffers on the interface itself, that more than double the true
latency.
Of course, the most important thing is to have a rock-solid interface, but such techniques do mislead the average musician, which is why I intend to use the Centrance LTU to reveal 'hidden extras' in my future interface reviews.
No comments:
Post a Comment