Tutorial – Calibrating Your Subtractive Synth
If you have a synth that just has knob markers for filter cutoff, etc. and you’d like to know what those values correspond to, here’s how to discover what’s going on under the hood.
by David Baer, Sept. 2014
This is a discussion of what it takes to calibrate a synth. The first obvious question is “what do you mean by ‘Synth Calibration’”? Once answered, a second obvious question is “why would you ever want/need to do that”? Let me answer both.
Although many synths have digital readouts for parameters like filter cutoff frequency, envelope attack time, etc., many do not, especially those synths that in some way model old analog equipment. So, it can be useful to do a little digging and discover those things yourself. A filter-cutoff knob, for example, might have eleven position markers between the 7:00 and 5:00 clock positions. But to what frequency does, say, 1:00 correspond? Finding such correspondences is what is meant by “calibrating a synth”.
Why go to the bother? Accomplished sound designers will just use their ears in the first place. But many of us are, at best, neophytes when it comes to sound design. A highly effective way to learn synth sound design is to recreate a patch from one synth on another. Since most subtractive synths have similar architectures (filters, envelopes, et al), this is often eminently doable. But first, it helps to know the from-patch and the to-patch parameter values. So let’s get started.
Introducing the Patient
In all the examples that follow, we’ll be using NI’s sadly discontinued, but delightful, Pro53 synth. This has no readouts of any kind, so it’s a perfect specimen for our probing.
It has one LFO, two envelopes (dedicated to amp and filter cutoff respectively), one filter (normally configured as low-pass) that has a 24 dB per octave slope, and a delay. What we’ll be determining is filter cutoff, LFO rate, envelope attack/delay/release times and delay times.
Filter Cutoff – the Most Important Thing
The filters are often modelled on classic designs, and filters on different synths will often have different audio characteristics. There’s nothing we can do about that, but we can at least get two filters in the same ballpark in terms of cutoff frequency. And it’s extremely easy to discover what cutoff values are.
When it comes to measuring values in this and all the other calibration chores ahead, you may be comfortable with getting approximate values – two significant digits is plenty of accuracy. Don’t sweat differentiating between, for example, cutoff values of 508 Hz Vs 514 Hz. Just call it 510 Hz and move on.
To find filter cutoff, we need a frequency spectrum display. Voxengo’s invaluable (and free!) SPAN effect will get the job done nicely, but feel free to use whatever you have at hand. There may be one included in your DAW software.
To find the cutoff frequencies, there are a few things you need to do to set up for measurement: it is important to turn off all modulation of filter cutoff: velocity, envelope, LFO, and most importantly key-tracking. Have an oscillator (one will be sufficient) set to saw wave. Turn filter resonance all the way up. Turn the cutoff control to the desired measurement position. Play a handful of adjacent notes low in the keyboard range (e.g., the five notes between C1 and E1, C3 being middle C). The reason is that this cluster of notes playing saw waves will be guaranteed to produce a lot of partial frequencies in the cutoff area of the frequency spectrum no matter where that happens to be. Observe the frequency spectrum and the cutoff will be obvious, as shown below. The resonance spike at the cutoff frequency is hard to miss.
Repeat this for a reasonable number of control positions. If you have markers at the “o’clock” positions, do it for all of those. You don’t need to go overboard. Eight or ten readings will be sufficient. Also, you should expect the readings at the lowest cutoff frequencies to be a little vague. Don’t worry about it. Make your best guess and leave it at that.
Now, we’re almost done, but not quite. The other important thing that needs to be determined is the key-tracking pivot point (assuming the subject synth has that feature, of course). Sadly, synth documentation rarely includes this important piece of information.
This is also quite easy. With the resonance knob all the way up, play notes with no key-tracking for cutoff and then with full key-tracking. When you play the note at the pivot point, it will sound identical without and with key-tracking. If in doubt and you cannot decide between several adjacent notes, use the spectrum display to identify your final choice. But with resonance on full, you will be unlikely to need it – it’s probably going to be quite audible. It turns out that on the Pro53, the pivot note is the B natural at MIDI note 35. Who can fathom why that odd choice was made?
OK, now we have this information (for the Pro53, this shown to the right). How do we use it? If a patch has no key-tracking modulation of cutoff, then just duplicate the cutoff reading of the from-patch into the to-patch. Easy peasy.
If there is key-tracking used in a patch, then we have to take the pivot point into account. If the from-synth has a pivot point of, say, middle C (MIDI note 60) and the to-patch has one an octave lower (MIDI note 48), you can’t just set the cutoff to the same value on both. The to-synth patch will be noticeably brighter. To compensate, you need to set the to-synth cutoff one octave down. If the from-synth has a cutoff of 800 Hz, then make the to-synth cutoff 400 Hz.
If this is confusing, just consider this. On the from-synth, when we play MIDI note 60 (the pivot note), the key-track-influenced cutoff will be 800 Hz. If we set the cutoff on the to-synth to 800 Hz, when we play MIDI note 60, the key-track-influenced cutoff will be 1600 Hz since the note is now one octave above the pivot. But if we set the to-cutoff to 400 Hz, then playing MIDI note 60 will get us an octave higher and we’re back at the desired 800 Hz courtesy of key-tracking.
Of course, you probably won’t be so lucky as to have them one octave apart where the math is just a matter of multiplying or dividing by 2. It’s still not very difficult to calculate to-cutoff values for the general case, but you will need a calculator that can do exponentiation. If N is the number of notes between the from- and to-pivot notes (i.e., MIDI note number of higher minus MIDI note number of lower), then calculate:
To-cutoff = From-cutoff * (1.06 ** N), if the to-pivot is higher
To-cutoff = From-cutoff / (1.06 ** N), if the to-pivot is lower
In the above octave-apart example, To-cutoff = 800 / 1.06 ** 12, which equals 397.58 Hz – plenty close enough!
If the 1.06-to-some-power calculation puts you off, there’s an alternative that just requires plain old division. Find the frequencies of the two pivot notes and calculate their ratio. For MIDI notes 48 and 60, these are respectively 261.6 Hz and 130.8 Hz for a ratio of 2. MIDI note frequency charts are easily found on the Internet. This second way is easier and you can do without a fancy calculator, but you do need the frequency chart.
The Rest is Easy
If what’s come so far hasn’t scared you off, the remainder of calibration measurements is child’s play – although it now gets a little tedious. We are interested in timing calibrations for envelopes and LFOs. It will help if your DAW has a feature found in SONAR. When you freeze an instrument track in SONAR, the project window track shows the waveform. Cubase, doesn’t do this, so there will be an extra step or two for each measurement. When you freeze a track, you’ll need to pull up the wave form in a separate audio file viewer or editor.
The most important timings you need are the envelope times. But let’s start with LFO rate because it’s the easiest to describe. All we need to do is set up a patch on the subject synth that has no envelope modulation and just one LFO. That LFO can modulate amplitude (if available) or filter cutoff. The Pro53 cannot modulate amplitude, so we’ll use filter cutoff.
In this and following measurements, we’ll need a simple MIDI track containing one note of sufficient length to allow measurement of the longest possible envelope attack and decay times and/or longest LFO cycle time. Use a high note in this case to avoid confusion between the waveform itself and the modulation.
Set the filter cutoff low, set the LFO to a square wave or saw wave – anything that makes a cycle start is easy to see. Modulate filter cutoff with the LFO. Set the LFO to the measurement rate you want to capture and freeze the track. Look at the resultant wave form at the appropriate zoom level and note the time difference between the start of two adjacent cycles. Divide 1.0 by that and you’ve got the LFO rate. In the example below, the rate is 1.1 Hz.
If we want to capture the timing information for the delay, it’s much the same process (although the Pro53 delay, which is intended to serve as a phaser and chorus unit as well as conventional delay, has complexity we don’t need to go into here). Set up the simplest of delays, with a brief pulse as the initial signal. Set dry/wet levels to 50/50 and freeze. The resultant wave form should look like something in the image to the right. The time distance between the two peaks is the delay time.
Attack, Decay and Release
By now you can probably figure out how to proceed with envelope measurements. I won’t bother walking you through the process – it’s much the same as what we’ve just been looking at: make the envelope settings, freeze, examine the waveform and determine the time between the start and the end of the envelope segment. Attack is easy. Delay is also straightforward, but there is one nuance. Does the delay setting designate a time or a slope? If the former, the delay is independent of the sustain level. We see that with the Pro53, the time to get to a 0% sustain level and a 50% level is about the same.
If decay governed slope, then we’d need to take that into account and modify settings accordingly. In other words, if we wanted a two-second decay and sustain was at 50%, we’d dial in a four-second decay and the end result would be two seconds. But other than that, this is all very straightforward. At the right are the measurements for the Pro53.
The Filter – Not Again!
One final point should be mentioned. You synth documentation will likely tell you what kind of low-pass filter is present (24 dB/octave, 12 dB/octave, maybe there’s a selection). If it does not, you need to find out. If you are duplicating a patch that uses a 12 dB/octave filter and you only have a 24 dB/octave filter on the to-synth, you’ll need to nudge the filter cutoff higher to make up for the more aggressive attenuation on the to-synth. Now, we know the Pro-53 has a 24 dB/octave filter because it’s in the documentation. But if we did not know that, look at the two images below. Both show the results of a cluster of saw waves played in the lower range of the keyboard. In the first, the filter is wide open. In the second, we’ve closed the filter down quite a bit. The first image shows us that between 1 KHz and 2 KHz, the level drops about 6 dB with the filter open. In the second, with the filter in effect, we see a drop of about 30 dB. Subtract the 6 dB drop that’s in the unfiltered signal and we’ve got a 24 dB drop in one octave caused by the filtering.
In closing, let me express my gratitude to Voxengo for making the wonderful, indispensable SPAN VST freely available for download. Computer sound geeks worldwide appreciate this generosity!