AUDIOSIEGE

06/30/2025

Well this is certainly exciting.

06/29/2025

06/24/2025

To finish up on compression, I thought I would add a few graphs of my hardware units, mostly in the way I use them. As you can see in these-especially with the Rocksonics MB-3X (which is also modified a bit), analog behavior can look sort of funny on a graph when you’re used to seeing the predictability of plugins, and that’s one thing that makes it fun. But predictability is still key, and that comes with knowing your gear.

I appreciate the interest in the graphs and information I’ve shared, and will be happy to go further if anyone has any questions. I’d originally planned on adding a section on how to choose a compressor with a bit on FET style comps (which I don’t use), feed-forward vs feedback, etc., and I may do that in the future, but I’m going to be moving on to equalization soon.

06/20/2025

I save every single record that I’m sent, and I keep the thank you notes in them too. I’m currently building a new shelf here to house them all, and “The Endless Hour”, the new 7” by St. Louis’s Worn Down will be the newest addition. This record rips, and sounds great, thanks to a great recording and mix by Brandon Hoffman. Much appreciation to y’all for sending this to me!

06/19/2025

Last week, “Adult Child” by Buick Audra was released, and I’m so thankful for being entrusted with the mastering. Hearing it again, without wearing my work hat, it’s such a beautiful and inspirational record, and it sounds fantastic, thanks to the team on this one. Recorded by Justin Francis and Buick Audra, mixed by Kurt Ballou, and mastered here at Audiosiege. Thank you so much for sending a copy Buick!

06/18/2025

When talking about dynamics processing, there are nuances in the terminology that can’t be overlooked. For example, while compression and expansion could be considered opposites, they work in similar ways. But upward compression is often mistaken for the latter, when it’s really not the same at all. While a normal, downward compressor attenuates peaks, and an expander makes them louder, upward compression doesn’t focus on the peaks; lower level signals are made louder, thereby still compressing the audio, as the distance between the peaks and valleys is still squashed. This is still done with thresholds and ratios, albeit differently. For example, as we see in slide 1 (Izotope Ozone Dynamics), signals below the threshold are made louder. Note that here, the threshold is indicated as (X):1 for upward compression. You’ll also see the numbers reversed (e.g. 1:2). So, unlike the threshold behavior on a downward compressor, the signal is more dynamic with lower thresholds. That’s because the threshold is focused on signals below it, and not above. The ratios shown in slide 2 (ADPTR Audio Sculpt) are easier to understand once you get the threshold sorted out (look at the blue line, relative to the others), and in slide 3, using DMG’s Essence, you can see how the knee shapes covered earlier are still applicable.

There are several advantages of upward compression, especially when used in parallel, but it can also bring the noise floor up, or increase sustain of components that make the sound a bit weird when used on a full mix.

06/17/2025

I really enjoyed working on the new LP from Poland’s JAD, and I enjoyed hearing it on vinyl even more-it sounds fantastic. Luckily, the nice neighbor who had it erroneously delivered to her home knocked on our door with it, saying, gently, “it’s from Poland!” Thanks for sending this across the pond y’all!

06/16/2025

Now let’s go back a bit, to the more simple controls of a compressor-the main ones that set where, when, and how a compressor responds to audio. The most basic of these are threshold and ratio. The threshold is, simply, the point at which the compressor reacts to the signal. That reaction triggers gain reduction, which is of a magnitude typically indicated by ratio. For example, as seen in slide 3, a 2:1 ratio will only allow 1 decibel to pass for every 2 decibels of signal, and so on; it’s really just the ratio of output to input above the threshold.

The compressor’s reaction is controlled by a detection circuit, usually fed by a sidechain. The sidechain can be the full spectrum of audio, or it can be filtered. In some cases, it can even be boosted in certain ranges so that the compressor reacts more to signals in those ranges. Most often, you’ll see these filters on the detector circuit indicated as a frequency of a high pass circuit. This is not a cut of the signal, nor is it a boost of the signal in that range (although, as we’ll see, it can have that effect). By nature of physics-with low end waveforms being longer-compressors will react more the lower the frequency (assuming general balance). The HPF or sidechain controls set the rolloff of the detection circuit only, letting the compressor react to higher frequency transients more than lower ones. Keeping kick drums from pumping a mix is a good example of a usage scenario here.

06/12/2025

After a compressor attacks a transient, it has to release it. Just like the attack time is variable-altering the slope of the transient-so is the release time; it’s just the time it takes for the compressor to turn off the gain reduction, letting the signal recover. Think of it like a light switch, or the gas pedal in a car. How quickly does the light turn on? How quickly does your vehicle accelerate? And how quickly are you back to the initial state?

As seen in the first chart, it’s not necessarily a linear process, depending on threshold, ratio, and other characteristics inherent to the compressor. It isn’t always the speed at which the signal reaches its original level. It’s just how quickly the compressor releases the transient. For example, with a 150 ms release time, a deeply compressed signal may still take longer to fully recover, and this is dependent on the attack time as well. Using long release times can actually soften the attack of the transient, resulting in a more evenly compressed signal. As with every tool in audio, this is useful in some applications (like increasing or enhancing sustain) and very much unwanted in others (certain program material).

In the final images, we see captures of a short bit of audio, where the attack time remains the same, but the release times are changed. The overall level gets lower and less dynamic with longer release times in this case, since the transients are happening while the compressor is still in recovery. As evidenced in the last shot, a slow release time on program material can be problematic with too deep a threshold, too large a ratio, and/or too fast an attack time.
Notice how the signal is initially louder on the left, then gradually slopes downward to a flatter, more even level. The sound is very recognizable: you’ll hear the snap of the first transient, then the compressor struggling to recover after the initial loudness, pumping through the rest of the material. This sound can get buried, to an extent, if followed by an aggressive limiter, but that’s one of many ways it’s easily missed.

06/09/2025

If we think about audio in terms of amplitude, time, and frequency, then it helps when using signal processors to attach different parameters to each of those components.

With compression, we’ve established that the compressor’s knee is the slope of the transition into gain reduction, so let’s apply that to amplitude or level. And for time, it’s going to be attack and release. These parameters are very important with compressors, and the interactions between the two of them can be quite complex, but we’re going to simplify it for the sake of visualization.

The attack time is basically just the reaction time of a compressor after a transient has eclipsed the threshold. Slow attack times allow more of the transient through before gain reduction is applied, while fast attack times kick it down sooner. In these graphs, we can see the gain reduction on the right, after the transient spike that happens from the lower signal on the left. We’ll get into release times later, so it’s important to view these signal ramps as levels, since the release on these examples is a constant 50 ms, using the TDR Molotok compressor. Take note here of the distance between the transient and attenuation: with slower attack times, more of the transient gets through, and vice versa.

While this may be rudimentary knowledge to most engineers, I’m building on the aforementioned slopes of amplitude, time, and frequency, so I plan to talk about release times next, then get into the interactions of the attack and release controls.

06/04/2025

With the exception of the handful of laptops I’ve owned over the years, I’ve built every computer used in the studio. The Intel i7-4790K based PC I’m using now has been a workhorse for the past ten years or so, but it’s time to retire it and build another. I decided this after succeeding in trying to get a print slower than realtime… All it took was 4x oversampling at 96k with one very hungry plugin.