Monday, April 10, 2006

Digitizing Casettes with the iMic

So, I picked up a Griffin iMic, an inexpensive device designed to allow stereo audio input and output via USB. The purpose was to digitize some casettes. The marketing copy for the iMic says:

"iMic shines as the essential tool for converting your old LPs and tapes into MP3s and CDs. Griffin's audio recording software Final Vinyl for Mac OS X (provided for free exclusively to iMic owners) makes recording old records and tapes super easy, with several advanced features such as waveform-based cue editing and built-in 10-band EQ. Final Vinyl can also equalize LPs without having to connect a turntable to a pre-amp. Just hook the turntable directly to the iMic, and Final Vinyl will record your LPs perfectly."

So, how well does it actually work (for casettes, at least)?

First off, a word on subjectivity. Human ears are incredibly versatile. We can carry on conversations on the back of a motorcycle, or next to a jackhammer. We are really good at filtering out signal from noise. Our default setting seems to be to enjoy reproduced music, even if it is quite low-fidelity -- the recorded music really only has to _suggest_ the original, and our imagination does the rest. We're very forgiving.

That is, until you start training your ears _not_ to be forgiving. Several of my jobs have required me to work on multimedia production. I've had to try to get the best audio possible out of low-end equipment. I did this as well for a somewhat serious hobby, when I was setting up audio systems for deejaying, or setting up music gear for a church band that played weekly.

Training yourself like this removes some of the fun of listening to music. You stop thinking of it as just music and start thinking of audio as a technical problem to be solved. Hum becomes your mortal enemy. Distortion makes you grit your teeth. You find out that there is such a thing as "better than CD-quality sound." You learn to hear the difference.

With my ears now, the iMic sounds reasonably good -- about as good as I was hoping for, in a $40 interface. However, compared to something like a Mark of the Unicorn A/D device, there is a lot of compromise involved to make the price that low.

There are some basic audio problems, but they are not severe. More importantly, there are some usability problems, most of which have more to do with documentation and software than with the iMic hardware itself.

The near-complete lack of documentation makes this device somewhat inappropriate for a less technically knowledgeable user, even though this is the apparently the target audience. So be aware of that -- if you buy it because you aren't a super-critical listener, but just want to get better sound out of your computer than the standard microphone input, you are going to need to tweak and mess with the device a little. After you do this for a while, you may find that you aren't a technical beginner anymore, and then you may want something that aounds better!

The iMic is a little round "dongle" with a USB cable on one end, two 1/8" stereo jacks (one input, one output), and a switch on it that allows the user to switch between microphone and line levels. it comes with a cable with two female RCA connectors on one end (for phono-plug inputs) and a 1/8" stereo male connector on the other end. I say this because it is NOT clear from either the ad copy or the package what cabling is actually included -- I had to open up the box at CompUSA to figure this out.

The wiring as supplied will connect nicely to a consumer casette deck. If you want to connect it to a turntable, you will probably want Griffin's specialized cable for this purpose, which comes with a separate grounding wire. There are also cables available for XLR microphone input and 1/4" instrument input.

This doesn't mean you can just buy the cable and plug in your microphone. The microphone I own is a low-end professional microphones, like the very popular Shure BG 4.1 (made for recording instruments like acoustic guitar), and provide balanced XLR outputs. A balanced or "low-Z" signal won't work with the iMic; although it has a preamp, it expects a "hi-z" unbalanced signal and is made for consumer-type or "computer" microphones. The iMic also does not provide phantom power, which quite a few microphones need.

The iMic will also supposedly support an electric guitar or other high-impedance instrument, but I don't think I would be satisfied with the results [see end note 1]. I might give it a try later, but for now I am going to stick to using it to transcribe casettes.

My setup was to simply run the output from a Yamaha Natural Sound casette deck into the supplied cable. I should point out that the USB cable is annoyingly short -- only 1.5 feet -- and the supplied patch cable is even shorter. You will need a regular stereo male-to-male RCA patch cable. The cable you normally use to plug the tape deck's line outputs into your stereo receiver should work fine.

I first tested the iMic on my wife's original iBook, running MacOS X 10.3.9, then with my PC, which is a 2.8 GHz box running Windows 2000.

First off, the device comes with almost no information for setting levels, and also doesn't come with software, or a manual, aside from a few notes printed on a piece of cardboard in the packaging. There is a downloadable manual for the Final Vinyl software, but it is pretty light, and I gave up on Final Vinyl, so I was left with pretty much no usable documentation.

On my wife's iBook, Final Vinyl was a disaster. It crashed repeatedly. I kept getting audible drop-outs in the recording. It lagged painfully far behind in drawing the waveform it was recording, and it didn't provide a useful guide to setting levels. It supposedly provided equalization, but the settings were dimmed out. I heard digital clipping all over the place, but the "clip" light did not come on. About this point I was wondering if I could return the device to the store.

One basic problem is that I wanted to be able to choose a reproducible input level. The analog slider on the control panel does not provide any numeric indication of the level you have chosen. MacOS X supports the iMic directly, but the Sound control panel and the built-in Audio MIDI application don't provide very much information about the device. Final Vinyl displayed slightly more meaningful information about the input level, but this wasn't helpful because I just couldn't get it to work reliably.

An iBook, even a very low-end iBook with a relatively slow G3 processor, should have more than enough CPU power available to record a stereo signal. It can, for example, easily handle iTunes with the visualizer on, pulling music from a shared library on our wireless network and sending it out to an Airport Express. Griffin's specs only require a Mac with a USB port capable of running MacOS X, but clearly this is not sufficient.

I next tried Audacity for MacOS X. Although this seemed to perform better, it also crashed constantly, and kept changing its own settings. In addition, each time it crashed, the next time I ran it, Audacity thought that it had never been run before, and the various settings I had chosen were lost.

On either platform, there is no information on what the optimal settings for audio input should be. You have a lot of choices, though, even before you try to set an input level. I wound up selecting 16-bit, stereo, 44.1 KHz, because that represents a common CD-quality sound level supported in most software. However, after reading some of the somewhat cryptic notes on Griffin's web site, I also tried configuring the device for 16-bit, 48 KHz, which may be the device's "native" format (the documentation is unclear on this subject; do the A/D converters digitize at 16-bit, or 24-bit?).

The reason for using 48 KHz is not because I think the source justifies using this much resolution, but because using a non-native setting probably causes sample-rate conversion to happen while recording, in the driver. This could cause performance issues and reduced quality while recording. In theory at least, it is always better to record at a native format and then down-convert the final product _after_ any post-processing, using a higher-quality sample-rate conversion algorithm that doesn't run in real time. In addition, giving the post-processing stage more of the original data to work with should minimize aliasing effects that can be introduced when applying filters.

After the painful experience on the iBook, I decided not to test the iMic with my PowerBook G4/400, although this might have done a little bit better. Instead, I decided to just go ahead and try it on my PC.

Things went considerably better on the PC. Audacity on Windows 2000 is rock-solid. The software that came with my sound card actually allowed me to control the input level and shows me a number for my chosen level. I spent some time listening critically and found that the best balance between dynamic range and the onset of digital clipping occurred at an input level of 50 to 60%. Tapes are mastered at slightly different levels; if you have a really not one, you may want to lower the input level even further. This is a far cry from an Amazon review where a user reports getting great results after setting the input level to 100%.

I should point out that I never was able to get my PC to drive the output of the iMic, but according to some notes on the Griffin support site, this seems to be a known issue with Windows 2000, and may be fixed on XP. In any case, I already have a reasonably good headphone out on my PC, so I just used that. (And yes, it is somewhat noisy).

Let me say a word about distortion. Audible distortion in devices like this tends to be of two types. Distortion in the analog realm tends to make a signal sound fuzzy, or squashed, or grungy, but it is not all that unpleasant (think heavy-metal guitar). A small amount of that is acceptable, even expected, in a low-end device. The difference between "audio quality" is often actually not in the amount, but in the _quality_ of the distortion that is introduced. Some devices with poorer paper specifications actually sound better because they err in ways that sound better -- more musical -- than the device with the better paper specs.

To my ear, the iMic does not sound all that musical: it seems like the frequency response is not quite right by default. It seems to have a slightly unnatural "edginess" on vocals and high-pitched instruments, adding some un-musical hissing distortion, and a similar bulge in the bass. It is common for consumer-grade devices to have an slightly unnatural highs and lows because this makes stereo components with poor frequency response sound a bit better. It isn't unbearable, and I could probably clean it up (the frequency emphasis, not the distortion) with a little bit of equalization in post-processing.

One thing to avoid, though, is digital distortion. While analog distortion can sound bad, if you manage to overdrive the inputs to the point where digital distortion occurs, you will immediately notice. Digital distortion involves waveforms that are completely chopped off, when they exceed the dynamic range of the analog-to-digital converter. They are turned into square waves. On playback, these clipped waveforms sound like painful "glitches" and have a fingernail-on-the-blackboard quality. They can also damage your speakers. If you're looking at the recorded waveform, you can see this chopping-off.

Note that digital clipping does not always happen in the context of a visibly loud signal! That is, you don't get digital it only on signals that are obviously filling up the whole dynamic range. It is not adequate to set levels by just looking to see if the waveform is getting visibly truncated at the top, and backing off a little bit. If your input level is too high, you will get them on transients that occur even in what appear to be relatively quiet passages. When you zoom in on the waveform, you can still see the clipping.

This happend for me on one tape in particular, which was produced from a live recording, and which did not seem to have adequate low-frequency filtering on the microphones. The result is that there were occasional "rumbles" of low-frequency noise that could barely be heard, and which didn't look high in the recorded waveform, but which would "spike" the signal enough to cause digital clipping. So don't rely entirely on your eyes to tell you if the levels are set correctly. You have to listen.

After getting the levels set, and some music recorded, it was time to try to minimize the noise. Casettes don't have the dynamic range of CD. Even with Dolby, tape hiss is audible in very quiet passages. But most of the problems you will hear in practice when listening to a casette, especially an older casette, take the form of worn tape, dropouts, wow, and flutter. Worn tape sounds muffled, since a lot of the high-frequency details have literally been worn away. Dropouts are the loss of some of the magnetic material as the tape becomes worn and manifest as sporadic volume drops or brief loss of part of the frequency response. Wow and flutter are variations in speed that comes from the tape stretching unevenly or from irregular motor speed. "Flutter" is more noticeable; it sounds like unwanted vibrato in vocals and instruments, and can be particularly annoying on sustained piano chords. Keeping your tape deck heads clean and demagnetized is of some help, but nothing will help a stretched tape, and with such a flawed source, the best you can hope for was to get as musical-sounding a picture of what is actually on the tape as possible, without adding too much additional noise in the process.

In addition to the noise inherent in the casette and the player, all components in the audio chain can add noise. The iMic marketing copy boasts: "iMic's audio is superior to most computers' built-in audio because it uses USB for the audio signal. USB isolates the audio signal from the noisy electronics in your computer, giving you higher quality sound when you record, and higher quality sound output for external speakers."

Well, this is true when compared to the built-in 1/8" audio jacks built into your Mac, or perhaps available on your PC's sound card. The environment inside a computer's case is full of high-frequency electrical noise and thus a terrible place to try to run an analog-to-digital or digital-to-analog converter, or an amplifier. Using an outboard device helps a lot. However, there is still a noticeable amount of audible noise that the iMic adds to the recording. You can hear it on playback or play-through, when you turn up the volume on your headphones, and you can see it on Audigy's input meters. It is kind of a periodic "whine" that indicates digital signals are bleeding noise into audio circuits.

You might be able to ameliorate at least some of this. For example, you can make sure that the casette deck is plugged into the same AC circuit as the computer. This can eliminate "ground loop" as a possible source of hum. You will want to have the iMic unit as far away from devices that project an electromagnetic field. This includes the computer and the computer monitor. Unfortunately, since the cord is so short, if you want to experiment with this, you will have to try a USB extension cord. You could try wrapping the little iMic in tinfoil, too, but by this point you may feel like you need a break and go and take a walk until you regain some perspective.

Regarding noise and other matters of signal quality, Griffin's iMic help pages have an amusing answer on this subject:


What are the exact technical specifications on the iMic –
slew rate, transient response, voltage, impedance, etc.?



[Explanation about how they reserve the right to change
components to maintain the low price removed]...

Having said that, we feel if exact specifications are that
crucial to your project, you might be better served with
one of the products costing hundreds of dollars more.

They're right... and I had to keep reminding myself to re-calibrate my expectations. It does indeed sound pretty good for what I paid, which was in fact $39.95 + tax.

Anyway, since I had gone this far, I decided to see if Audacity could help me remove some of that audible noise. It could -- in fact, it could help quite a bit! First, I tried to figure out how to apply a filter to remove a bit of the microphone "rumble" from the tape, but I was not really satisified with the results. I did not put a lot of time into this, though.

To get rid of the iMic's "whine" as well as tape hiss, I decided to try Audacity's noise filter, which has a "training mode." I selected a couple of seconds of the audio track recorded from the casette lead-in, with no music playing, and told Audacity that this was the noise I wanted it to filter out. That noise was tape hiss, a little bit of low-frequency motor noise, and the quiet but audible digital "whine" of the iMic. Then I told it to apply the trained filter to the audio track, at a level two clicks left of center (another case where a digital readout of what I was doing would be a lot more useful than an unmarked analog slider, so I could make notes and reproduce the same result later).

The result was impressive -- the background noise, especially the tape hiss and irritating "whine," dropped to nearly unnoticeable levels. Some of the low-frequency noise seems to have disappeared as well. There is always "collateral damage" done when applying filters like this -- a bit of the music gets filtered out too -- but in my opinion the result was worth it.

I didn't have a lot of time to put into experimentation with other post-processing such as equalization, so I decided this was good enough for now. Using Audacity, I selected ranges from the resulting audio file that represented individual songs, then exported these clips as 16-bit, 44.1 KHz AIFF files. From there, I imported them into iTunes, typed in some meaningful track information, converted them to Apple Lossless files to reduce their size, put them on a playlist, and then burned a CD. The result, with the filtering applied, actually sounded "better" (or at least cleaner) than the original casette.

So what is the final verdict?

After my terrible experience with Final Vinyl, I can't recommend it at all. I gave up, at least for now, on using the Mac. If you have a more modern system with an audio application like GarageBand, it might work considerably better.

I can't vouch for attempting to transcribe records, which Final Vinyl is designed for -- there are some equalization issues unique to turntables that you will need to handle somehow. You also might want to consider Griffin's special cable with an extra ground connection.

If what you want is to digitize casettes, the iMic provides a reasonable and very inexpensive means to do this. Casettes don't sound that good to begin with, so when you scrutinize the results on headphones, you will probably notice the flaws in the source more in the flaws introduced by the iMic. I was also able to improve the noise situation somewhat using Audacity's noise filter. The results don't sound spectacular, but they sound approximately as good as the casette does, and maybe even a little better.

More importantly, although your casettes are deteriorating with the passage of time, once you have them in digital form, you can play the resulting files as often as you want without degrading the audio further. If you plan to experiment with post-processing later, save them in the original resolution as well as the clips you export for listening or burning to CD. And make sure to keep a backup!


End Notes

[1] On the subject of recording electric guitar and other instruments with the iMic, and why I don't want to try it: while doing live performance and home recording of voice, various electric and acoustic guitars, guitar synthesizer, and Chapman Stick (tm), I have spent quite a bit of time with headphones on, scrutinizing the sound of various preamps, direct boxes, and instrument inputs available on equipment in the under-$1,000 price range -- and I've become very critical of the sound.

The preamp circuit in the iMic is not going to be better, for example, than the well-regarded preamplifiers in the Aardvark DirectPro Q10. That preamp provides phantom power, which the iMic does not, and has a "guitar switch" which changes the impedance and frequency response to provide a quite decent-sounding input channel for an electric guitar. It is _designed_ to handle guitar. It sounds musical. You have to spend quite a bit more to get an audibly better solution for directly recording an electric guitar.

Using a microphone input to handle an unbalanced electric guitar signal just does not sound very good, and in fact _cannot_ sound very good. The microphone preamps available on mixers made by Mackie or Carvin will handle instrument inputs, but the results aren't all that musical, even in conjunction with a decent direct box to handle the impedance mismatch. For decent reproduction of an electric guitar, you really need a preamplifier circuit that is actually designed for a guitar's frequency response.

No comments: