Last weekend, I recorded two of the three days at the Austin City Limits Music Festival, using the various components in Core Sound's PDAudio system. I'd earlier this year written up PDAudio-CF and Mic2496 in EDN Magazine, based only on target product specifications and interviews with owner Len Moskowitz, and I appreciate the opportunity to supplement my prior coverage with this hands-on analysis.
Sound savings: Portable audio recorder takes on tape: Part 1
Sound savings: Portable audio recorder takes on tape: Part 2
I've been working with Len and Gordon Gidluck (the owner of Gidluck Mastering and developer of the Live2496 recording software for PCs and Pocket PCs) through many iterations of both hardware and software since the early part of this year, along with Pocco Software (developers of the Wichita recording software for the Pocket PC) for a shorter time. I first used PDAudio-CF a few weeks ago at the Trinity Tribal Stomp, but a variety of factors conspired to produce sub-optimal results. Mic2496 wasn't yet available, a typo I'd put in the Live2496 configuration file (coupled with the fact that I'd left the configuration file documentation at home) left me unable to record with it, and my marcSound 'stealth' binaural mics clipped to the sides of my hat (note: chosen so that my setup could be compact and portable and wouldn't block the view of others....recording was legal at this festival!) were in retrospect an inferior alternative to my Oktava mics on a tall stand!
I've provided both Len and Gordon with copies of my recordings, which are in sum total over 3 GBytes, FLAC-compressed, and nearly 7 GBytes uncompressed. Check their websites, as I've encouraged them both to post (unedited by them, except perhaps MP3-compressed to reduce size) portions or entire recordings of each band's performance that I captured. In anticipation of this posting, you'll find per-recording comments below as appropriate.
Len had first provided me with a PDAudio-CF in mid-June. Pocket PC drivers followed two months later, along with a revised PDAudio-CF with updated FPGA configuration. I received a functional Mic2496 the day before I flew out for the ACL Festival, and that same night Gordon sent me his latest version of Live2496 with functional recording meter display and robust 24-bit support enabled for the first time. Keep the 'freshness' of both hardware and software in mind as you read the journal that follows.
Hardware setup:
Rig #1: Oktava MC012 phantom-powered
microphones with windscreens, cardioid capules and -10 dB pads were arranged in an X-Y
pattern via a Shure A27M stereo adapter and Audio-Technica
AT8415 shockmounts, attached to a 13' Bogen/Manfrotto
air-cushioned stand. The microphone cables direct-connected to a Core Sound
Mic2496 configured for 24-bit/48 kHz digital conversion, whose S/PDIF coax
output was direct-connected to a Digigram
VxPocket v2 PCMCIA sound card installed in my Fujitsu
Lifebook P-2040. The recordings were captured via Sonic Foundry (now Sony)
Sound Forge 6.0e. The notebook PC had both high capacity main and modular (i.e.
in optical drive bay) batteries installed, and I've upgraded the factory
configuration to comprehend Microsoft Windows XP
Professional, 368 MBytes of DRAM and a 40 GByte Seagate
Momentus 2.5" hard drive.
Rig #2: An Audio-Technica AT825 self-powered (i.e. with AA battery installed)
dual-channel cardioid microphone with windscreen was rubber-banded and
foam-cushioned at the top of the Shure A27M (the description reads far uglier than
the setup looked
and performed in practice). The microphone cable was direct-connected to a Denecke
AD-20 microphone preamp and ADC, with 20-bit/44.1 kHz output. The AD-20 was
optical-connected to the PDAudio-CF sound card installed in a Hewlett
Packard Dual PC Card Expansion Pack for the iPAQ 3835 Pocket PC. The
expansion pack's other slot held a Toshiba
5 GByte 1.8" PCMCIA hard drive. I ran Live2496 for the first two recordings
and Wichita for last one, all at 16-bit, 44.1 kHz.
Location:
I was right in front of, and just to the left of center of, the sound stage (as
seen from audio engineer's position behind the sound board).
Liz Phair 3-4PM (16/44.1, Rig #2):
The sound quality was overall quite poor; the
audio engineer at the sound board must have been drunk, asleep or
who-knows-what. Liz's
voice completely faded away on several occasions and was generally lost in the 'boomy'
mix. Recording levels according to Live2496 never exceeded -10 dBFS, but when I
played back the file afterward in Wichita I saw meter levels go into the 'red' and heard
distortion, so I surmised (correctly, as I later determined when I copied the
WAV to my notebook PC....which by the way takes a VERY long time with the 3900
RPM Toshiba 5 GByte hard drive) that the levels displayed in Live2496 were lower than actual. You'll
see and hear that I ended up with lots of clipping on this recording.
But when it's not clipping....it sounds ok!
Liz Phair (24/48, Rig #1):
Aside from the fact that I missed the first few seconds of
the first song with this rig, I'm generally pleased with this recording. There's a several-dB level
variation between the two channels (left lower than right) which may have
reflected either inexact microphone arrangement or imbalanced mic preamp gain setting on my part, or
imbalanced right-vs-left side levels coming off the stage. Looking at this waveform and
listening to the recording, versus the earlier one, you'll clearly see and hear that the
AT825 microphone 'compresses' the sound much more than the wider dynamic range
Oktavas (which makes the nominal level of the Oktava recording 'quieter',
although its peaks are still near 0 dBFS). This contrast will also be very evident on subsequent
recordings.
Galactic 5-6PM (16/44.1):
I backed off on the mic preamp levels this time, but
still got some clipping. Unfortunately the Denecke AD-20 doesn't have a
'clipping' LED so in the absence of accurate metering reported by Live2496, I
had no other means of determining if I was overdriving the ADC. Somewhere in both
this and the 24/48 recording (I heard it the first time I listened, but can't
visually see it in the waveform now and don't have time to listen to it again to
re-find it) you'll hear some momentary rumble caused by a
drunk woman grabbing my mic stand as she walked by. Sigh ;-(
Galactic (24/48):
Partway through this set, I adjusted the mic preamp levels to
equalize the right and left channels. There's a very weird
speedup-and-distortion at the very end of this recording, reflective of loss of
clock lock somewhere in the digital chain, that Len and I are still trying to
figure out. I didn't encounter this on any other recordings I made.
Steve Winwood 7-8PM (16/44.1):
For this recording, I switched to Wichita in
search of reliable metering. I ended up with a recording that I was quite
pleased with, although the Wichita beta is not without flaws. See the 'future
plans' section below for more on this.
Steve Winwood (24/48):
Again, I was pretty pleased with this, although for
some strange reason the levels came out really low (but balanced, reflective of
my earlier adjustment mid-Galactic). One possible explanation for the glitch at the end
of the earlier Galactic recording is a battery that was nearing the end of its
life, and although I can't explain why that same battery would subsequently make
it though a 1 hour Winwood set, maybe it was on its way out (see below for
next-day Nickel Creek comments, which give further support to this theory). I'm sure that I didn't see an illuminated low
battery indicator on Mic2496 during this recording, and am pretty sure (though not
absolutely sure) that Sound Forge was giving me higher level readouts (ie near-0
dbFS peaks) on its display while recording was in progress.
Hardware setup:
One and only one rig: Oktava MC012 phantom-powered microphones with windscreens,
cardioid capules and -10 dB pads were arranged in an X-Y pattern via a Shure A27M stereo
adapter and Audio-Technica AT8415 shockmounts, attached to a 13' Bogen/Manfrotto
air-cushioned stand. The microphone cables direct-connected to a Core Sound
Mic2496 configured for 24-bit/48 kHz digital conversion, whose S/PDIF optical
output was direct-connected to the PDAudio-CF sound card installed in a Hewlett
Packard dual- slot PC Card expansion pack for the iPAQ 3835 Pocket PC. The
expansion pack's other slot held a Toshiba 5 GByte 1.8" PCMCIA hard drive.
I ran Live2496 for both recordings, at 16-bit, 44.1 kHz.
Location:
For Nickel Creek, I was right in front of, and
just to the left of center of, the sound stage (as seen from audio engineer's
position behind the sound board). For SCI, I was on the left side of the sound stage, at the front corner
(ditto).
Nickel Creek 7-8PM:
It was a pretty wet day, and I'd need to sprint a
quarter-mile or so between this set and SCI, so I decided to be more mobile this
time and leave the notebook PC and second set of microphone-and-other-gear
behind. Anyway, I wanted to test the unified PDAudio system, and since Windows
XP drivers weren't yet available for PDAudio-CF, the Pocket PC was the only way I was going to
do it. Pretty cool; everything fit in my mic stand bag, with lots of room to
spare! I set up in the rain (no, I didn't have an umbrella for the gear, but I
'did' have lots of plastic grocery bags!). The rain stopped just before Nickel Creek
hit the stage. You'll hear a brief glitch at 1:20, caused by me frantically
swapping out 9V batteries when I realized the 'low battery' LED in Mic2496 was
glowing red! Otherwise, I'm really happy with how this turned out. My original
intent was to record 24/48, but in testing the setup prior to Nickel Creek
taking the stage I noticed that Live2496's meters were frozen. In and of itself
this wasn't a problem; based on the prior day's experience, I was going to rely
exclusively on the dual-channel 'clip' LEDs on Mic2496 instead. But since I
couldn't directly play back 24-bit files in the Pocket PC, and since I didn't have my
notebook PC along to test them on, I couldn't verify that recording was even
taking place. So I throttled back to 16-bit, which from past experience I knew would work.
SCI Set 2 8:40PM-10PM:
As I mentioned above, I head to quickly tear down after
Nickel Creek and sprint a quarter mile across the length of Zilker Park to where
SCI was playing. I was set up again 10 minutes after I arrived.....one of the
other people recording, who'd already been there for set 1, made a special point
of coming over and commenting on how portable I was and how quickly I was ready
to go! It was a good thing too...the band fired up 20 minutes earlier than
scheduled. They began with a recording of Pink Floyd's 'Pigs on the Wing',
followed by a few-minute montage of appropriate-themed Simpsons clips (as you'll
hear, it took
me a few seconds to realize what was going on and hit the record button!), then
the band launched into 'Another Brick in the Wall' and the authentic,
resurrected inflatable pig from the old Pink Floyd tours appeared overhead where
it bobbed about for a while. Great fun. Since I wasn't there for set 1, my
location was not ideal, though given the crush of the crowd I'm delighted I got what I did! I was set up on a
narrow wooden board, just a bit wider than the tripod spread of my mic stand,
that'd rock back and forth if someone stepped on it. Given how much the stand
was bobbing about the whole set, I'm amazed how well the recording turned out (shockmounts
are a GOOD thing). Again, another quite good capture of an excellent SCI set.
It was still quite wet outside (my gear had survived the previous day, and I was reluctant to tempt fate again), we wanted to see four consecutive bands on four separate stages (Zilker Park is huge!) and my wife nicely asked if she could have all of my attention this day, instead of me fussing with my recording equipment the whole time, so I left the gear at home ;-) Some outstanding sets (we arrived mid-afternoon so missed other bands I'd hoped to see).....Jack Johnson, Ben Harper and the Inncocent Criminals, Yonder Mountan String Band and R.E.M.
The portability of the PDAudio system is very very cool. I'm especially impressed with the combination of functional integration and quality results delivered by Mic2496. One little box, powered by a single 9V battery (I went through about 1.5 batteries in five hours of recording; note that I was phantom-powering the microphones through Mic2496, and power consumption would have likely been even lower had this not been the case), does a lot and does it well. After three hours' worth of recording to the Pocket PC, I still had 80% remaining battery life in the iPAQ 3835 (even though I'd forgotten to turn off its backlight), and 50% remaining life in the expansion pack (which has its own battery). The dual-battery setup on my Lifebook P notebook computer, coupled with judicious use of its display (i.e. shutting it off when not using it) left me with 75% battery life left after three hours of recording, too!
Bottom line: I'm doing something I've been contemplating for some time. I'm selling all my DAT-related gear beginning this weekend (see here for my Ebay auctions, which I hope to post tomorrow). I can frankly think of no better testimonial to the capability of both the hardware I tested at ACL and the other HDD-based recording hardware and software I've used in the past, such as Creative Labs' Nomad Jukebox 3. I've got audio-capable DDS drives in my PCs, which'll enable me to listen to my DAT archives, copy them to CD-Rs and DVD-Rs, etc. But going forward, I've now got solutions that are far more reliable, and deliver much higher quality results, than DAT.
Gordon reports that he has discovered the reason why Live2496 under-represented the recording levels (he saw the same thing in his recording sessions last weekend), and has also figured out to get an accurate recording level display working for 24-bit recording. He's also figured out how to deliver real-time recording meter performance at 16-bit settings. I anticipate receiving an updated version of Live2496 this weekend that encompasses all these improvements (he's already sent me a nearly completed version, but I haven't yet tried it), which I'll test asap.
Here's some feedback I already provided
Pocco Software on the beta version of Wichita (and which they've promised to
address in the final version of the program), based on my experiences at the Trinity
Tribal Stomp:
1) You MUST alter the final version of Wichita such that it
overrides any existing power management settings, so the Pocket PC doesn't shut
down after 3 minutes, etc past the last screen tap....as mine did until I
manually altered the power management! Of course, when the user exits Wichita,
you should also return the power management settings to their prior values
2)
The response time on the signal strength indicators was so sluggish as to be
virtually useless. This, coupled with the lack of robust signal indicators on the
Sony SBM-1 I was using (nothing but an 'over' LED), caused frequent, long-duration clipping
3)
Several times, the display froze (i.e. the recording time, time remaining, signal
strength, etc. indicators no longer moved). When I hit 'save' the application
would eventually save the file, so it hadn't completely locked up, but that's
how it appeared. FYI I ran Wichita after a fresh soft-reboot of the Pocket PC
and with no other applications running in foreground or background. Under these
conditions, I've got around 27 MBytes of onboard memory free
4) In attempting to
solve the problem above, which appeared to be program sluggishness due to lack
of CPU overhead, I ran Jimmy Software's JSOverclock
freeware program (on my system I'm usually
able to run stably at 221 MHz and often can run for long durations at 236 MHz).
This didn't see to help
5) When I attempted to rename a file within Wichita, I'd
end up with TWO files...my renamed one, and a 0 byte sized 'Recording1' file.
Very strange
6) Oh one other thing...I also thought that the sluggishness
mentioned in #3 above might be due to Microdrive problems, so I set it to not be
high speed storage. I subsequently realized that then the recording was cached
to system RAM. Only being able to do a 1-or-so minute recording at 16-bit, 44.1
kHz isn't very useful imho! But I acknowledge that it might be useful with
lower-quality recording settings
Both Gidluck Mastering and Pocco Software, as well as the other Pocket PC software developers, should add the ability to not only disable the backlight but to completely blank the screen (and reawaken it via key press or other non-invasive method), to maximize battery life. The screen is a major 'juice' consumer.
Len is trying to figure out what happened at the end of the Galactic set...whether he has a circuit problem in Mic2496 or, more likely, I experienced a battery power or connectivity glitch. Yeseterday morning, he sent me an alpha set of PDAudio-CF Windows XP drivers, right now supporting only 16-bit resolution and 44.1 kHz sample rate, and I plan to test them as soon as I can free up some spare time.
I've got a 2.5" HDD enclosure on the way from Addonics (from whom I already own a PCMCIA cabling system and AA battery pack, previously acquired for an external optical drive for a prior notebook computer), which I'll pair with a 40 GB Toshiba HDD in preparation for testing 24-bit/96 kHz recording on my Pocket PC. For more, see here.
Any comments or questions? I welcome your emails! Check back for updates in coming days/weeks/months....
A combination of bad weather, uncooperative concert organizers and other, more pressing personal and professional commitments has kept me from doing any more festival recording this past ~2 months, but I have been able to do some additional testing at the office. I'm happy to report that the latest iterations of both Glidluck Mastering's Live2496 and Pocco Software's Wichita (now out of beta) will reliably record long sessions of dual-channel audio at 24-bit sample sizes and 96 kHz sampling rates on my iPaq 3835, as long as the storage media can keep up. Their capabilities include fast-responding, accurate level meter displays at all but the highest sample size and sampling rate settings. I've also tested v2.52 of Resco's Audio Recorder, which is the first iteration of the program with claimed PDAudio support. At the moment it's too buggy to receive my recommendation, but I'm sure the developers will improve their product in time.
It's funny; originally Gordon Gidluck (whose software I've tested many many iterations of) and I feared that the processing 'muscle' of the iPaq 3835's 206 MHz StrongARM CPU would be insufficient for high resolution audio capture. Extensive optimization both of the high-level applications and low-level PDAudio-CF sound card driver have eliminated these concerns (the 400 MHz PXA255 XScale CPUs in latest-generation Pocket PCs should have even more processing headroom!), and have moved the potential bottleneck elsewhere. Semiconductor and rotating magnetic storage media with slow write speeds, or initialized with a non-optimum format type or cluster size, will result in (best case) dropped samples or (worst case) premature termination of the recording session. Gordon's website has extensive data detailing his test results and recommendations.
Speaking of storage media, I have successfully gotten an Addonics 2.5" drive enclosure, containing a 40 GByte IBM HDD, working with my setup. If you have an Addonics CardBus adapter, as I do, make sure you flip the switch to put the adapter in PC Card mode. This minor 'gotcha' caused me several days' worth of head-scratching until I figured it out. For some reason, the website containing the driver and installation instructions seems to be down, so I've mirrored v1.05 of the Storage Brick driver here (you're on your own if you try it out; no guarantees, and as always a full backup of your Pocket PC is recommended prior to installation).
Finally, the participants in the laptop tapers forum ('laptop' and 'tape' make a strange combination, don't they?) and others are anxiously awaiting the release of Windows 2000 and XP drivers for PDAudio-CF. Several beta users (including myself) have been stress-testing a series of incremental driver revs over the past few months, and I feel confident stating that at this point the developer is very close to a finished product. Keep an eye out on the Core Sounds website, and on this 'blog', for updates in this and other PDAudio-CF support topics.
Core Sounds has released the first public beta of its WDM (Windows Driver Model) drivers for the PDAudio-CF card, intended for use with Windows 2000 and Windows XP (and also potentially usable with Windows 98SE and ME as well, in conjunction with an up-to-date DirectX installation). You can download it here (note, this is a ZIP archive of a self-extracting RAR-compressed EXE). Please email feedback to developer Wade Dawson, Len Moskowitz and myself.
I've also discovered another compelling use for the Mic2496; as a phantom power, mic pre-amp and analog-to-digital converter front-end for my Creative Labs Nomad Jukebox 3. With this combination, I've reliably made numerous dual-channel, 16-bit 48 kHz recordings of lectures to both MP3 and WAV formats, in combination with a short mic stand and my Audio-Technica AT825 stereo cardioid mic. Since the AT825 can optionally run self-powered via a AA battery and since I'm not doing high resolution audio recording, I could instead use my Denecke AD-20, but the Mic2496's larger, dual-channel knob simplifies the task of setting and adjusting recording levels in comparison to the AD-20's separate, small recording level controls.