Do you really want that scope?

2022-05-29 music

In-rack oscilloscope Eurorack modules come up pretty often on wish lists and ModularGrid fantasy racks. Many people, and especially beginners, seem to think oscilloscope modules are useful, oscilloscope modules may even be must-have modules, and they want those oscilloscope modules. If you are one of those people, I hope you will think carefully before shelling out for an oscilloscope module. They are less valuable for musical performance than you may imagine, and here are some examples pointing at why.

Let me start by giving some disclaimers that the haters will probably ignore. I am talking here specifically about Eurorack oscilloscope modules, not oscilloscopes in any other context, and about using them for musical purposes. I'm a big fan of oscilloscopes for electronic debugging; I think a scope is probably second only to a multimeter as a general-purpose troubleshooting tool for SDIY and other hobby electronics. For troubleshooting purposes you would usually want a "real" oscilloscope: a separate piece of equipment with bandwidth significantly higher than audio, larger display than can fit in a module, well characterized high input impedance, a probe that connects to BNC connectors on the front, and so on.

The typical Eurorack scope module is billed as primarily a musical tool: people who add it to their wish lists seem to think they will often have cause to look at waveforms while building patches and making music. These modules seldom have bandwidth much beyond the 20kHz audio range, and they are designed to accept the same Eurorack signals as other modules.

I can think of several reasons someone might want an in-rack oscilloscope module:

• Oscilloscopes are fun in general
• It could provide an interesting display for spectators in a live performance situation
• You often want to create a specific waveform by patching, and being able to see the waveform as you work on it helps with that
• You think it is important to know objective parameters of a modular signal, especially the voltage range, when planning patches
• An in-rack oscilloscope could also serve as a general-purpose oscilloscope for debugging SDIY builds

I can't argue with the "fun" rationale. If your reason for buying an oscilloscope module is just that you like it and you want to play with it, and you can afford to do so, then go ahead! But don't fool yourself with claims that it will be useful if fun is your real reason for buying it.

The idea of providing a visual display for spectators is an interesting one but in-rack oscilloscope modules are seldom well-optimized for this purpose. If the spectators are not right at your elbow (and only one or two can be there), they will have a hard time watching the scope display unless you sort of amplify it, by pointing a camera at it and putting the result through a projector or similar. You may be doing that with your rack anyway, to make your patching process more visible. But any visually interesting module could work as well, and given how starved for HP most wigglers seem to be, devoting a whole module to just being interesting for spectators to look at seems wasteful. You could use the rack space for a module that actually makes music and maybe also has a light show aspect (like a Fixed Sine Bank); you could point your camera at a separate oscilloscope; or you could use something else, even a laptop running WinAmp-style visualizations, to feed your projector.

As I've written before, I think any oscilloscope is better than none for circuit debugging. But if your real goal is to have a piece of test equipment, then an in-rack oscilloscope module is a disproportionately expensive and inconvenient way to get that. Everything costs much more and is subject to compromise when it's built into a Eurorack module, because of the small market and the physical and power limitations of Eurorack. (This effect is also part of why stereo mixers in the rack are a bad idea.) So you will be getting a less good piece of test equipment for more money than you could have just by buying a regular oscilloscope. Even just the physical inconveniences of needing to bring your synthesizer rack and your device under test close to each other; adapt a probe cable into something that can plug into the module; and deal with a small screen and limited number of buttons, can be significant. With a standalone oscilloscope designed as test equipment instead of as a module, most of those issues would vanish.

The remaining points, about an oscilloscope being musically useful, are the most interesting for me because they are based on an important assumption: the idea that the voltage waveforms displayed by an oscilloscope are important for how things sound, and useful for making musical decisions. I think they usually are not. Waveforms are confusing and usually irrelevant to music, and I've prepared a few examples showing how they can lead you astray. A tool that displays waveforms is not very useful in day-to-day patching of a modular synthesizer.

In these examples, I have randomized the order of the waveforms, spectrograms, and audio samples. For each example there are two waveforms, two spectrograms, and two audio samples; but it is not consistent which spectrogram (left or right) goes with which waveform (top or bottom); nor which audio sample goes with which waveform and spectrogram. You can try to guess, but part of the point is that guessing accurately is seldom possible. Also, be warned that most of the audio samples are pretty loud, and the volume is inconsistent from one example to the next.

First, here is a sawtooth wave (somewhat wiggly because it's band-limited, synthesized by adding up harmonics only through the 40th), and another waveform that contains exactly the same frequencies and amplitudes as the sawtooth but with random phases. How similar or different would you expect these to sound just from the oscilloscope display?

There are many different waveforms that sound the same or nearly the same, so if you're trying to achieve a particular sound, then there may be no one right answer to how it should look, and the scope is not much help in knowing whether you are close.

Here are two sine-on-sine FM waveforms, with the same carrier frequency and deviation but different modulation frequencies. Again, how easy is it to tell them apart by looking at the oscilloscope display, and how easy by listening to them?

Two very different sounds can have waveforms that look similar on the scope, so the problem cuts in the other direction as well: you cannot necessarily start with just a mental or physical image of a waveform and use that to describe a sound. The fact that waveforms and sounds do not map cleanly in either direction is related to the fact that our ears operate primarily in the frequency domain. Oscilloscopes, showing voltage in the time domain, are displaying something only indirectly related to sound, and it is difficult to translate mentally between these two domains. That limits the usefulness of "draw the waveform and then play it" oscillators, the Buchla 132, and similar. That link talks a bit about Don Buchla's dissatisfaction with the idea.

Starting with a desired waveform and then trying to patch to create that waveform just isn't how playing a synthesizer works. People who play modular synthesizers are usually thinking about and trying to create spectra, not waveforms; at most, they may end up translating their spectral ideas into a waveform as an intermediate stage, because of technical requirements.

Here is another sine-on-sine FM signal, and a signal that attempts to imitate the same spectrum using AM instead - that is, it's balanced amplitude modulation with a sum of sine wave modulators matching all the main sideband frequencies calculated for the FM signal. As with the random-phase sawtooth, the oscilloscope display shows a clear difference between the two signals, but how important is that difference for musical purposes, and can you tell from the sound which one is which?

The oscilloscope can be misleading even when it comes to very simple attributes of a signal, like loudness. Here are two white noise signals, one generated using a Bernoulli distribution of voltages and the other using a Gaussian distribution. Which one is louder?

The Bernoulli noise signal is in theory always at either its maximum or minimum voltage, instantaneously switching between the two. The Gaussian noise signal usually stays near zero, only rarely spiking out to its peak voltage in either direction. On the other hand, the peaks of the Gaussian signal are much further from zero when they occur. From the oscilloscope display it seems that the Gaussian signal covers almost three times as wide a range of voltages as the Bernoulli signal. Measurements of the peaks in the audio files indicate a difference of about 9.3dB in peak-to-peak voltage.

Listen again: is one signal really 9.3dB louder than the other? In fact, these signals are scaled to have the same RMS voltage, which is a good proxy for loudness given they have identically shaped spectra, and they sound equally loud to me. A skilled user looking at the oscilloscope display would notice the very different voltage distributions and would recognize that the waveform will not make clear whether there is a loudness difference, nor in which direction it might be. The peak-to-peak voltage is actively misleading here, representing one of the signals as covering a much wider voltage range when, really, any difference between them is barely perceptible. That is one reason I try to discourage people from focusing on "voltage range" in modular synthesis. "Voltage range" is usually the wrong way to think about signals in a modular context.

Some oscilloscopes (and some plain old voltmeters) are capable of measuring RMS voltage and displaying that as a number. Measuring devices tend to call the feature "true RMS" if it is meant to display RMS voltage correctly even on non-sinusoidal waveforms; an RMS feature not described as "true" may be designed only for single sine waves and may give inaccurate readings on other spectra. The RMS voltage is of some value for estimating loudness, and certainly works better for that than the peak-to-peak voltage does, but when comparing two signals with different spectra, the RMS voltage is not the whole story either. Really balancing loudness properly for audio signals in general, is difficult.

Here are two chords (C major and A minor) performed on a digital hardware synthesizer. Can you tell them apart by looking at the voltage waveform?

Although these examples suggest that the voltage waveform is not a good way of understanding how signals will sound, there is a silver lining in the spectrograms, shown below the waveforms in my graphics. When the signals sound the same, they usually have similar or identical spectra; and when they don't, they don't. So why not have a module that displays a frequency-domain spectrum instead of a time-domain waveform?

The standalone piece of test equipment for that would be a spectrum analyser, and most Eurorack "oscilloscope" modules actually do offer a spectrum analyser mode too. There are technical limitations resulting both from the compromises necessary to operate as a Eurorack module, and the inherent sampling time needed for recognizing a spectrum. Even with these limitations, an instrument that displays the spectrum of an audio signal may be of some use in understanding the signal, especially for offline sound-design projects.

But when playing a modular synthesizer in real time, I don't think any visual display of signals, whether time-domain, frequency-domain, or other, is really a high priority and a good reason to buy a module. Patching is best experienced as a flow state - add, change, and adjust parts of the patch while listening to the results. If you're intent on the sound of the patch, that right there tells how best to analyze the signals involved.

And he who rashly grabs the shears
Will find too late, with bitter tears
That there's no substitute for ears.

- Walter R. Brooks, The Collected Poems of Freddy the Pig

UPDATE (September 2023): I hate to cast shade on another manufacturer's product by name, but I've now had two significant tech-support cases where my customers were trying to use a Mordax Data in particular for troubleshooting a DIY build, and they had serious trouble because of the Data's limitations. In one case, the customer was trying to adjust an MSK 008 Octave Switch by plugging the octave switch's outputs into the Data first and then measuring with a voltmeter on the Data's "buffered" outputs. It turned out that the "buffered" outputs did not accurately reproduce their input voltages, so voltage readings taken that way were incorrect and gave inconsistent results compared to other measurements taken on the Octave Switch directly.

In the other case, a low-ultrasonic signal (probably about 26.3kHz) from a malfunctioning DIY Middle Path VCO appeared on the Data's "oscilloscope" display as if it were 2.3kHz because of temporal aliasing: the Data was sampling it at what I guess was something like 24kHz, with no anti-alias filter, so the signal which was really a little above the sampling rate, appeared on the screen as if it were the same amount above zero frequency. It took me quite a while to figure out why the customer wasn't able to hear a signal at 2.3kHz. A real oscilloscope wouldn't do that. It would use a higher sampling rate in the first place, and would have anti-alias filtering to prevent incorrect display of signals too high to accurately represent. And, at the very least, an oscilloscope designed for serious use as test equipment would say in the manual what the sampling rate actually was, whereas (at this writing) Mordax's manual just leaves us to guess.

Someone who had a lot of experience with oscilloscopes would be able to figure out and work around the Data's limitations when used as a substitute for an oscilloscope; but someone capable of doing that would probably already own a standalone oscilloscope anyway. Beginners end up trusting the Data and becoming confused. What I'm seeing in practice is someone buying a Data as their first "oscilloscope," planning to use it not only for musical patching but also as test equipment. It really isn't suited for use as test equipment. A basic standalone benchtop oscilloscope would serve you better for DIY, at half the price of the Data.

◀ PREV A fractal sequencer toy || Equivalent circuits NEXT ▶

MSK 007 Leapfrog VCF SDIY Kit

US$234.32 including shipping

Comments