The Thin (Actually Pretty Fat!) Red Line
I’m working on a project for a local film composer. He has put together a bunch of 96 kHz/24-bit, 5.1 surround tunes in the “electronica” style (they include acoustic instruments as well). And they’re really good! They have the familiar pulse drum/percussive beats, lots of smooth filtering, and aggressive 5.1 surround panning. Not surprising considering this guy has spent his life in and around music, recording studios, and dubbing stages. His “garage” studio is one of the most incredible places I’ve ever been — and I’ve been to a lot of studios. He has a very large API analog mixing console (and a small one in the iso room), lots of signal processors, a huge patch bay, a 5.1 surround set of ATC speakers, a drum set, grand piano, guitars, and a bunch of microphones. And it’s all set up ready to record anything passing inspiration. When you’re responsible for 25 minutes of new music every week for a popular TV series, you develop a highly optimized workflow.
The API console is a monster. As you know, I’m a digital guy when it comes to large format recording consoles but I’ve worked on large analog mixing desks from SSL, Neve, Harrison, and Yamaha. The only other time that I’ve encountered an API desk was the custom one that my friend Peter Otto purchased for UC San Diego music department. My memory has faded somewhat when it comes to the API console that they ordered — and which was never completed — but I do remember that the university paid something like $600K for this fancy new state-of-the art mixing console and had to sue the owner of the company and API to get them to finish it. API has a reputation for excellent sounding analog consoles but they have flirted with disaster in their business dealings a number of times. But they’re still in business and still selling expensive consoles. People seem to like the sound.
So I get the first tunes from the composer and take a peak at the frequency response and dynamic contour in Adobe Audition. And there is — a huge red horizontal red line hovering in ultrasonic space above the rest of the music. This is definitely not part of the music. Any harmonic components are always down -80 to -100 dB at 40 kHz. The 40 kHz tone represented by the red line was only -40 to -50 dB down. Amazing! I immediately called my collaborator and asked him about it. I wanted to know if he has used some fancy signal processor or plug in that might account for the intrusive tone. He didn’t know where the spurious tone had come from.
So is the 40 kHz tone a problem or not? Obviously, no one is going to “hear” it and only speakers equipped with super tweeters will try to reproduce it. So what’s the problem? The signal path from source to speaker includes a lot of components that won’t want — and might have a problem with — a -40 dB ultrasonic 40 kHz tone! It’s possible to damage electronics with signals that are outside of their normal operating parameters. You wouldn’t plug 220 volts into a device expecting only 120 volts. And any thinking audio engineer wouldn’t record or mix to DSD 64 because it also has obnoxious amounts of unwanted ultrasonic frequencies that have been conveniently shifted above the audio band.
The mystery was solved when JMK called API and asked them about the 40 kHz signal. “Oh, sure,” they replied, “the automation system has been generating the 40 kHz tone for years but we fixed it some time ago.” No they didn’t — at least on this particular console. It makes me wonder how many so-called “high-resolution” capable consoles or devices have similar problems.
So I’ll remaster the tracks for this project and remove or attenuate the 40 kHz tones. But I shouldn’t have to do.
Mark – Thanks so much for this article explaining those crazy “horizontal noise lines” you see with spectrograph software up in the ultrasonic ranges on some recordings. One example I have is the 24/192 download of Scheherzade on Chandos with the Toronto Symphony on the Chandos label, which has more than one: a main one at about 48kHz, a secondary one at about 66kHz, and some tertiary ones at 62kHz, 67kHz and 73kHz. Most Chandos recordings are pretty exemplary in their absence of these kinds of “horizontal noise lines”, but I guess something happened with this Toronto Scheherazade.
I’m probably missing something technical here, but isn’t there also a sum and difference signal when the 40kHz beats with signal frequencies? What effect does that have?
How is it that EVERY producer of “Hi Res” (whatever that means) music does not own and know how to use some type of spectrum analyzer? It’s like being a chef without a spoon. Consumers would soon come to understanding the graphs and perhaps, maybe, hopefully, the snake oil guys would disappear.
It is unlikely that any 40 kHz tones will create audible artifacts.. But signals this loud are not welcome in the signal path of any analog system.
Just think of it as a watermark indicating genuine higher sample rate.
I think the odds of an isolated 40 khz -40 db tone causing problems is rather remote. Too high in frequency to set off metal dome tweeters. Too low in level to be an issue in most other ways. Sloppy yes, and easy enough to get rid of though it shouldn’t have been there in the first place.
Let’s not give anyone ideas about watermarking of “genuine” high sampling rate in any way. 🙂
It just should not be there as it has nothing to do with the music. Period. A flaw in the production process and hardware used.
If you are going to pump power into a tweeter at 30 to 40 kHz, it’s a good idea to have a dedicated tweeter for ultrasonic frequencies. Applying this kind of signal to tweeters crossed over at say 5 to 10 kHz can cause distortion in the audible range.
Some amplifiers, especially early solid state units can be unstable driven at high frequencies. One speaker cable manufacturer had a product that was notorious for blowing up power amplifiers. My theory was that it was due to inadvertently creating an LC oscillator as a load that was resonant at ultrasonic frequencies and that blew the daylights out of the amps.
Is it necessary to be able to record and reproduce these frequencies? Some audiophiles say yes, I think most engineers would say no. Just my opinion.
I would agree that it is not required to record ultrasonic frequencies. But instruments do produced tones or partials in the ultrasonic range. If it can be done accurately, then I say way not?
Why not? Because 1) it adds to the cost, 2) it is of no value to the user since you said yourself in your presentation in LA that mastering engineers could not hear the difference between your high resolution recordings and down converted 44.1 K signals, and 3) because it can potentially create problems in the range that is audible such as causing tweeters to distort. Frankly I don’t see the point anymore than there is a point to putting a 1000 horsepower engine in a car that is to be driven on the street.
Because it’s easy and reproducing what was actually in the room at the time of the original recording.