Response to the WorldCast/Orban press release

WorldCast and Orban have recently issued a press release about the APTmpX codec. We do not usually respond to something like this, and let our technology speak for itself. However, their press release contains some falsehoods about the current MPX composite codec landscape, that we cannot ignore. Their claims misrepresent the capabilities and performance of our codec. As a company committed to innovation and excellence, we feel it's necessary to set the record straight and address these inaccuracies.

We will start by responding to their claims, and end with briefly comparing real-world compression results.

Their claims

The key benefits over other MPX compression methods in the market is the absolute transparent nature of the codec, being rooted in ADPCM as opposed to a psychoacoustic code base. This gives the immediate benefit of maintaining perfect sonic fidelity and precise peak control of the high-end processor from which the MPX signal has been derived. The other crucial advantage with APTmpX, thanks to the transparent nature of the algorithm, is that it will not degrade in any way the Nielsen watermarking technology - used in the USA market to validate station ratings.

We agree that ADPCM can be superior for transparency, sonic fidelity, peak control and not degrading watermarking. Which is exactly why MicroMPX has always been ADPCM based. It's a benefit of both codecs, even though we see massive differences in fidelity between the two.

The APTmpX codec is the only reduced bandwidth codec technology for distributing analog FM composite today that doesn’t negatively impact watermark encoding.

This is simply not true, as MicroMPX has been showing for years now. 

[APTmpX] audio quality is spectacular and it’s virtually indistinguishable from the source.

Let’s look at some comparisons of source and compressed audio to get an idea of how one would distinguish the two.

Head to head comparison

Our initial comparisons are based on the example files that are provided by WorldCast. Those samples do not seem to contain any audio that's difficult to handle for ADPCM-based codecs, unfortunately. More in-depth comparisons will have to follow after we return from the NAB Show in Las Vegas (come visit our booth, W3322!).

Despite the simple nature of the source material, there are very clear differences. Listening to the decoded version of their recording of Tom's Diner by Suzanne Vega, we can very clearly hear a metallic sound, most audible at 300 kbit.

To test what MicroMPX would do, we ran the provided source file through Stereo Tool, adding a pilot and RDS and performing MicroMPX encoding. We then decoded it with a MicroMPX decoder and demodulated the MPX signal, giving us our own decoded recording for direct comparison. There was no metallic effect at all.

The source spectrum clearly shows that Tom's Diner would be a very difficult track to encode for MP3 - its developers even used it to calibrate their algorithms on: since MP3 only encodes louder sounds, you can easily get empty parts in the spectrum where audio has been completely cut out, which give MP3 its distinct and annoying “watery” sound. This shouldn’t be a problem for ADPCM-based codecs, as unlike MP3, they don’t have a psychoacoustic basis.

As the second image shows, the lower volume around 6-8 kHz is indeed no problem at all for our MicroMPX codec. The area is basically as it was before. Any difference is inaudible (to us, and confirmed by multiple days of listening tests with dozens of volunteers).

The third image shows the spectrum of the decoded APTmpX audio, and the amount of noise added in the 6-8 kHz band (and in some other places as well) is surprising. Let’s focus a bit more on that.

The following image contain a detailed frequency analysis of a small slice of the original and the two decoded files (from 14.00s to 14.15s).

The decoded MicroMPX audio contains some noise, but not much. The peaks are at around -88 dB, which is going to be very hard to notice, especially on FM.

The same chunk of audio, decoded by APTmpX, has its noise level peak to -74 dB. That's 14 dB higher than MicroMPX and would be clearly audible on FM.

In relative numbers, over the whole area, MicroMPX added about 4 dB of noise, APTmpX around 22 dB.

We had to pick a specific spot in the recording for these graphs. but the same effect is visible - and very audible - throughout the entire recording.

This brings us back to one of the statements in their press release:

The APTmpX codec is the only reduced bandwidth codec technology for distributing analog FM composite today that doesn’t negatively impact watermark encoding

As said above, I agree that ADPCM codecs won't affect watermarking. But MicroMPX stays much closer the original signal. Not only does that make it sound better, but if there's any difference for watermarking at all MicroMPX will outperform APTmpX.

Even lower bitrates: MicroMPX+

When we created MicroMPX, we were aiming for complete transparency rather than a specific bit rate. After days of listening tests with dozens of volunteers, we ended up at 320 kbit/s as the minimum bit rate that achieves this.

However, for some users, around 300 kbit/s is still too much. We created MicroMPX+, a more aggressive version of our codec that can go down as deep as 176 kbit/s, albeit at a (in our opinion, acceptable) quality cost.

Our tests show that MicroMPX+ at 176 kbit/s still clearly outperforms APTmpX at 300 kbit/s, delivering better signal quality at close to half the bandwidth, while still sending the full signal (including RDS) which APTmpX does not.