Topogram - noisy output

I’ve been meaning to ask for a while. If were to feed modulation CV from my dirty-powered audio rack to a completely clean, barrel-powered GEN 3 system. Would I automatically lose all the noise-reducing benefits of GEN3’s clean power?
Would it be the equivalent of sticking one end of a hose in my dirty aquarium, sucking on the other end and then dropping it in my clean aquarium?

Not quite. In the analogy, “water” is the power supply, not the signal path. So if you generate a noisy signal, then process it with Gen3 modules – the noise is still part of the signal. The Gen3 modules can’t filter out the noise because that would interfere with the bandwidth of the system (low noise and high bandwidth is a tall order!)

Maybe a better analogy in this case is the traditional process of recording music in a studio.

One studio might have a very high end, ultra low noise mixing console with extremely high fidelity EQ or DSP functions.

But the band might still come in and record a track using a crappy fuzz pedal with an audible hum, or mic a chainsaw with a $2 contact mic, or whatever. Even though the signal source has some noise in it doesn’t mean that the noise will be inherited by everything going into the mix – it’s isolated to the input. For example, if you turn down the fader for that channel, the noise disappears.

A wise mixing engineer will of course adjust the EQ/filtering or even apply a noise gate to the noisy channel so that it sits with the frequency spectrum of the mix as a whole. Maybe more “mastering tools” would be nice to have in the video synth environment (notch filter, de-noise DSP, etc).

And of course some bands may just record at home on an old four track and not care too much about noise in the recording – it becomes a creative choice.


Ah that’s great to hear. I was worried that simply patching an LFO from my uZeus-powered modules would somehow propagate a general noise to any GEN3 module in the chain.

1 Like

The issue I’m seeing with Topogram appears to be in the high frequency range. That is, the noise is a semi-regular waveform in the same general frequency as the video signal. Looks like some harmonics of 60 cycle hum.

I haven’t yet had a chance to test the Topogram as a low-frequency amplitude classifier. My assumption is that if I use Topogram to process low frequency signals, the high frequency noise will be present at the output, but its effect will be negligible. And even if it’s an issue, all I’d need to do is send that dirty low frequency output through an audio module of any kind. That would filter out all of the high frequency crap.

Is that reasonable? And wouldn’t that scenario apply to any low frequency signals, regardless of their point of origin or the specific character of the high-frequency noise? So really, it’s not an issue with audio modules, they’re generally not fast enough to resolve anything above ~20 kHz, including video rate power hum.

1 Like

You’re describing the noise created by a switching power supply, which often oscillates around 100KHz to 1MHz. So it’s usually filtered out of an audio path, but that’s right where we don’t want it, for video!

In Topogram, we used a switching power supply into low noise LDOs (a similar architecture to Gen3) but due to nesting it inside the assembly had to add a grounded shield to reduce the inductor’s effect on the rest of the circuit. So you are more than likely seeing the reduced ripple from that circuit. EIther that, or you are seeing the EuroRack noise floor, or both in concert. The noise is overall much higher than other modules because there’s a large amount of gain in the circuit. With that design, we realized if we wanted to do more high gain circuits and expect clean response, starting with a much lower noise floor in the first place was necessary.


Yes, your understanding is good! Think of it like in audio mixing; if you’re doing a traditional dance track you will usually shelve the mix around 150Hz so that the subharmonic of the kick drum gets a nice clean tone, and then layer that with a transient that has its low end scooped out.

So really, it’s not an issue with audio modules, they’re generally not fast enough to resolve anything above ~20 kHz, including video rate power hum.

I’m sure some audiophiles would argue about that with us (the supersonic components of the signal path and how they may interfere with clarity in the mix) but, yes! If you are sampling CD quality audio, a 500KHz switcher ripple is well above the nyquist frequency.

So if you are the circuit designer and want to get rid of the switching ripple entirely, but need to account for modern low voltage switching power use everywhere, what do you do?

You create local switching power supplies with a controlled ripple frequency, and immediately post-process them with LDO regulators with a high PSRR characteristic designed to target the frequencies you are generating. It’s a bit of controlled chaos designed to both introduce and nullify the high frequency ripple in one step.

Many of the early generation of switching power designs for EuroRack were using standard LDOs that were more suited to traditional linear power supply architectures, which explains why they wreaked havoc on video modules (no filtering of those high frequencies at all!). When Visionary series was developed and up through Visual Cortex’s release, EuroRack was mostly linear power supplies, with Mono Rocket, Analogue Solutions, Doepfer being the only case/power manufacturers around.

1 Like

Thanks as always for the education, @creatorlars !

I also realized after posting that in the worst case scenario, high frequency noise can be filtered from any low frequency modulation by merely sending the signal through a sample and hold circuit locked to vertical sync. I do that routinely on near-frame-rate modulation to avoid horizontal tearing and banding. The modulation value is held for the entire field/frame. In this scenario, there will never be a pattern onscreen from the high frequency noise. The worst you’d see is a little bit of flicker among fields/frames, and the high frequency noise would need to be very high to even see that.