Perhaps sample-rate conversion is something you do rarely. On the other hand, if you're like me, you have to juggle multiple projects that need preparation for different destinations and sample-rates. Saracon is one of the premier SRC applications on the market. It works with PCM digital files, but an expanded version is available should you need to work with Direct Stream Digital (DSD). When you have clients that need output for CD manufacturing, video placement, web streaming, or a variety of other distribution channels, a quality SRC package is not a luxury, it is a necessity.
Some SRC titles are good at some operations but not at others, so it's important to realize not all are created equal. A visual way to compare SRC performance can be found at src.infinitewave.ca, which is maintained by Infinite Wave Mastering (Calgary, Canada). As of this writing, their database includes test results from more than 100 different SRC products, including software and hardware. I have found that the Infinite Waves website is good for identifying terrible SRC programs, but decisions about the "best" sounding applications must be done by ear. In fact, some mastering engineers claim they audition multiple applications for every conversion and choose the best sounding one on a case-by-case basis.
Over the years, I've used most of the usual suspects. Having good and bad experiences, I've come up with my list of critical features for SRC software. My gauges are quality of sound, processing speed, batch capability, file-type support, and cost. It is a plus if the title works on both Mac OS and Windows, but not a necessity. I developed these criteria for two reasons. First, my obsessive side likes to have concrete areas to evaluate. Second, my creative side finds the theory behind SRC algorithms to be as interesting as the US tax code - especially when I need to make a decision and get to work.
Normally, to down-sample from 96 or 88.2 kHz to CD 44.1 kHz or video 48 kHz, I will reprocess the audio through my analog chain. I capture the needed file through an A/D converter running at the desired sample rate. My belief: the best sample-rate conversion is no sample-rate conversion. But sometimes, I need to do a reference or create a version that will be bit-rate-compressed down the line, and using software to do the job is acceptable.
Not all converters down-sample gracefully. When moving down, the SRC software must remove the frequencies that cannot be reproduced at the destination rate; otherwise, frequencies above the upper bound can cause aliasing. Aliasing can be heard when content from an inaudible range is shifted into an audible range, leading to distortion and noise. Aliasing might not be a terrible problem if it happens during one of the last operations on a file. But that is rarely the case. Audio is edited, reconverted, exported, and altered. Be advised - once aliasing distortion is folded back into the audio, future operations can lead to some of the worst digital-sounding harshness and garble you'll encounter.
The way to avoid this problem is to create a filter to keep aliasing noise out of the desired signal. Some designers use a very steep filter while others take a gentle approach. Here is the catch: when it comes to anti-aliasing filters, there is no single "right" way to process the audio. An overly linear approach can cause pre-ringing, while an overly minimal-phase approach will have post-smearing in the output file. Both filter types are acceptable and efficient; and they are typically used in less expensive or integrated applications. But at the best levels, there is an art to balancing the adherence to theoretical principles versus listening for euphonic results. Weiss will confirm that Saracon permits a small amount of aliasing in the 18-22 kHz range in conjunction with a shallower roll-off in order to reduce perceptible ringing. In my tests, I prefer Saracon over other brands - most likely because of this balanced approach. Of course, I'm going by ear and not a spec sheet. But our clients don't care about specifications; they care about the best sound.
Another point of difficulty in evaluating SRC software is that D/A converters alter the sound depending on their source rate. (I just report the news, people.) Put simply - your DAC sounds different depending on the sample-rate. This is due to their internal filters and architecture. For example, to my ears, the Crane Song HEDD 192 [Tape Op #26] sounds more open and clear playing back 96 kHz files than 44.1 kHz ones. Yes, it is the same hardware "unit," but there are different internal operations going on. So, we have two moving variables - the SRC software and the DAC - in any evaluation we do. For my tests, I ran every output file through the same D/A monitor, so at least that variable stayed the same for software versus software. Comparing source versus target still suffered from the different rate issue.
The speed of different SRC applications varies. And I do mean varies. Consider the following up-sampling benchmark. I loaded twelve songs rendered at 24bit, 44.1 kHz and set the SRC packages to convert to 24bit, 96 kHz. Saracon completed the job in 1 minute 49 seconds. The software I had been using previously took 7 minutes 33 seconds to complete. Saracon is roughly four times faster than the competition. What are a few minutes? That's a fair question. If you consider doing five albums a week, the timesaving is about 30 minutes. Over an entire year, it ends up being 25 hours - more than an entire day - of doing conversions, and this is only one of the SRC routines I complete in a week. Also, I will stay put and wait for Saracon to finish. With the other application, I would get a cup of coffee, visit the interwebs, and become less efficient. For me, time is money in this regard.
Another area where Saracon speeds workflow is in its smart interleaving feature. I get many Pro Tools mixes. A fair share are non-interleaved, meaning the right and left channels are not connected. My other software treated these mixes as two separate files, which they are, but humans know they should be a pair. Once the SRC was done, I would load the files into Sequoia, tell the DAW to link the channels, and wait for it to bounce the new stereo mix. Only then I could get to work. If filenames follow standard conventions (e.g., Pro Tools adds an L to the end of the left channel's filename), Saracon will automatically pair the separate channels into a stereo mix. Again, multiply this out over a year, and it's a good deal of time saved.
I have a few concerns with Saracon. First, the copy-protection scheme uses a dongle, which means another USB port consumed and drivers to maintain. I was surprised to learn that it does not recognize or convert files with the M4A extension, which I receive several times a year. Finally, speed and quality of this level has a cost. The adage "good, fast, cheap - pick two" comes to mind. Saracon is very good and very fast, but it is not freeware.
I find Saracon to be the most natural-sounding SRC software I've tried. For down-sampling, the results sound more "normal" than with some other applications. In fact, compared to my longtime favorite Mac OS converter, Saracon seemed to be more airy, open, and detailed on the high frequencies. For up-sampling, Saracon was simply faster than my standard PC-based SRC application. Further tilting the decision in Saracon's favor, my former application did linear or minimal-phase filters, and often one sounded much better than the other. It depended on source material, and I had to audition each. This took way too long. In the end, I could use the three SRC applications I have already, making a grid to remember to use brand X when I want to go from 96 to 44.1 kHz, but use brand Y for 48 to 44.1 kHz, and so on. And don't forget that amazing command-line application that down-samples, but doesn't up-sample. Or I could fire up Saracon and know I'm going to get very good, very fast results every time. And while other titles are very good, in terms of remaining true to the input audio, I prefer Saracon. Faster and better - those are two good choices.
Increasingly, studios must deliver audio across multiple digital formats. From CD production, to sound for picture, to high-resolution releases - sample-rate and bit-depth conversion is a common function. Having used most of the commercially available programs, I've found that not all sample-rate conversion software is the same. If you want one of the best sounding, fastest, batch-capable options, my recommendation is Weiss Saracon. Do some head-to-head tests in your studio; I'll bet you agree.