1kHz noise with Terrarium board

Through implementing a few of the tricks people have mentioned here and on the Electro-Smith forum (small blocksize, distributed ProcessControls) I think I finally have a new version of the Rhythmic Delay without any 1kHz whine (at least with my Terrarium builds). The same tricks also seem to work with a harmonic tremolo I am almost finished with. There is a bit of finesse to keep things working when using these tricks, but so far it has not really held back my programs.

I am still tweaking some of the features, but I hope to post the new whine-free version of the Rhythmic Delay soon. I have had one other person test the program and they did not hear any whine, but if someone else is interested in testing it to confirm the lack of whine send me a message and I can pass the current version of the .bin along.
 
Through implementing a few of the tricks people have mentioned here and on the Electro-Smith forum (small blocksize, distributed ProcessControls) I think I finally have a new version of the Rhythmic Delay without any 1kHz whine (at least with my Terrarium builds). The same tricks also seem to work with a harmonic tremolo I am almost finished with. There is a bit of finesse to keep things working when using these tricks, but so far it has not really held back my programs.

I am still tweaking some of the features, but I hope to post the new whine-free version of the Rhythmic Delay soon. I have had one other person test the program and they did not hear any whine, but if someone else is interested in testing it to confirm the lack of whine send me a message and I can pass the current version of the .bin along.
I'm happy to give it a go if you need another tester.
 
I won't get into a big discussion about the "quality" of the PCB design but it's at least worth mentioning that the noise issue is just as bad with the official Petal. This is not a problem exclusive to the Terrarium and was reported on the Electro-Smith forums before the Terrarium PCB was created.

Your point about the separation of analog / digital grounds is certainly valid, but I don't believe that alone is going to be the root of the problem.

You can hear the noise in this clip, which is an algorithm running on the official Petal before the Terrarium PCB existed.
It's like yeah, but I have the Daisy Patch with no such noise. It's possible to design it to work. And it's possible for that design to completely eliminate any noise from the signal.
 
If he could read too.

The title is 1khz noise WITH Terrarium board, which is a Pedalpcb board, sold by Pedalpcb.

And the post is asking in the terrarium board specificaly.
I'm confused by this comment. I gave a picture of the issues with the terrarium board and only talked about the PedalPCB terrarium board in my comment.
 
It's like yeah, but I have the Daisy Patch with no such noise. It's possible to design it to work. And it's possible for that design to completely eliminate any noise from the signal.

I've had the Petal since the first wave of shipments from the Kickstarter.

The fact that the noise occurs with the official Petal hardware, other PCBs, vero layouts, and breadboards somewhat disqualifies the implication that the Terrarium is the cause of the problem.

Like I said, your points are valid, no one is arguing with that, but unfortunately they aren't the cause of the noise. I wish improved trace routing and decoupling would clear everything up and give us a nice clean noise floor, unfortunately I don't believe that is going to be the case.

If I thought the Terrarium PCB was the cause of the problems discussed here I would remove the project until the necessary fixes were made.... but I can say first hand that it is not the source of the noise.


I'd also like to add that I don't think Electro-Smith nor the Daisy is getting a "bad name"... It's an excellent piece of hardware and everyone here is working to get the most out of it.... but it is relatively new, and there are quirks that need improvements / workarounds. Electro-Smith (and the community) is constantly fixing bugs and improving the platform.

Likewise, the Terrarium is a prototyping board, and a first revision at that. No one here is expecting to drop it into a mass-produced production device. There are plenty of people here making contributions who otherwise wouldn't be onboard at the entry price of $319 with no prior experience with the hardware.
 
Last edited by a moderator:
Through implementing a few of the tricks people have mentioned here and on the Electro-Smith forum (small blocksize, distributed ProcessControls) I think I finally have a new version of the Rhythmic Delay without any 1kHz whine (at least with my Terrarium builds). The same tricks also seem to work with a harmonic tremolo I am almost finished with. There is a bit of finesse to keep things working when using these tricks, but so far it has not really held back my programs.

I am still tweaking some of the features, but I hope to post the new whine-free version of the Rhythmic Delay soon. I have had one other person test the program and they did not hear any whine, but if someone else is interested in testing it to confirm the lack of whine send me a message and I can pass the current version of the .bin along.
I'd also be interested in testing the improved delay. Also really excited to take a look into your code and what exactly you did to minimize the noise since I am getting quite a bit with my reverb and also the ladder filter.
 
I'm confused by this comment. I gave a picture of the issues with the terrarium board and only talked about the PedalPCB terrarium board in my comment.

Then that's even worse than your initial tone. You simply don't understand what you wrote.

Pedalpcb f'd up and it's giving daisy a bad name. By the real shit from Electrosmith.

Still confused or need more in depth explanation ?

How the hell, a thread named NOISE ON THE TERRARIUM PCB BOARD (Sold by Pedalpcb has you mentionned) is giving daisy a bad name ?

Because there's indeed a cause effect in your sentences regarding Pedalpcb missing ground point, and Daisy getting a Bad name.

Which is untrue by the way, since the noise problems has been discuted on other forum with other system, and some official Daisy stuff itself ?

You see, some miss ground point, some miss point when they wan't to take a rant, we all got to be f'ing somewhere 🙂
 
Last edited:
I'd also like to add that I don't think Electro-Smith nor the Daisy is getting a "bad name"...

Every participant I've encountered on this board and the Daisy Slack has realistic expectations with regard to first steps. We've all got to learn to walk before we can run.

This is a very collaborative, supportive community and there are only solutions and good times ahead, as far as I'm concerned.
 
Last edited:
Honestly, I'm comfortable with the comments across the board. This thing is a proto board and the results and input I'm seeing from several users seem to address the issues and push this thing into really cool territory. Folks like sonic_explorer are doing really amazing things with this and with respect to him I won't comment on his latest dev that he's let me work with unless he asks me to.

<soapbox>
Speaking totally from my perspective...I appreciate the fact that everyone has been pretty civil about an invasion by the typical "ass-hat" assault that seems to be the norm on the web these days and I also understand the misunderstanding that this place might be part of that mentality. It isn't, we're a respectful community here and all we ask is that you respect that. If that doesn't sit well, then please feel free to not be a part of it. We will always welcome input by knowledgeable peeps and are grateful for it.
</soapbox>

Carry on.
 
I am following this with interest (but no judgement) as an FV-1 veteran but Daisy newcomer. As a retired analog and digital EE, now music effects hobbyist, I offer this perspective: The FV-1 was very smartly conceived and carefully designed such that it could be treated as and analog part, with internal codec, low frequency crystal, and well-behaved power characteristics. I think that E-S, and all of us potential Daisy users, want to treat the Seed module in the same way, trusting it to encapsulate all of the scary complexity of high-speed mixed-signal circuit design. I know I do.

OK, maybe this amounts to a judgement after all. To realize this "virtual analog" model, the burden needs to fall back on E-S to deal with processor noise leaking into the audio, through better grounding, bypassing, or whatever within the Seed module itself. At the very least, there should be clear guidelines for how to ground, bypass, or otherwise treat the module externally to avoid any noise issues.
 
It's been a while. Was the issue with the 1k noise ever addressed other than tweaking the sampling in code?
 
It's been a while. Was the issue with the 1k noise ever addressed other than tweaking the sampling in code?
I got my first brand new Seed in and wired up to Terrarium specs on the Madbean ProtoRig breadboard and powered with a Onespot. I put the lastest Rhythmic delay on it and it is hella noisey including the very distinct 1k noise.

I've tried a couple of different placements and values/types of bypass/decoupling caps and it is likely not optimal. Are there any known, optimal values/types?

It seems like low esr/esl electro caps are needed but what about mlcc types? I mostly have X7R types from tayda. Would others perform better?

I've had a look at the Petal to see if I could glean any values or power related suggestions from it and they seem to be doing some extreme isolation with the 5v-5v dc converter and some kind of audio bias buffering with the MCP6004 that I don't really understand.

Any insights are appreciated.
 
Last edited:
  • Like
Reactions: fig
Based on the suggestion from the DaisySeed documentation, you can mitigate some (but not all) of the 1khz whine by keeping the STM32 "busy" with sin/cos functions outside your audio callback. This prevents the chip from having to ramp up and down its current draw 1000 (or whatever) times per second. I imagine jumping from a few microamps all the way to 400mA on the power rail creates a ton of noise.


C++:
volatile float dummy_trig_value = 0.0f;  // volatile keyword prevents compiler from optimizing it

void main() { 
    while(1) {
        // loop code here

        // do some math to keep the stm32 from entering a low-power state
        dummy_trig_value = sinf(0.1234f) * cosf(0.54321f);
    }

}

Also if you're using a "digital" bypass (like the terrarium build) you may also want to write your audio callback in such a manner where you create two copies of your input buffer: one thats always processed regardless of bypass state and one that stays "dry" and gets applied to the output if the pedal is in a bypass mode. This way the same work is done whether the pedal is bypassed or active.
 
Last edited:
I was affected by the 1kHz callback noise, and I found a solution that works great, at least for my application, so I wanted to share it here (as this thread helped me a lot to understand what was going on).

I designed a custom board based on the Daisy Seed, similar to the Terrarium, as in it has analog input and output buffers. I use it to run a NAM A2 model, an impulse response for cabinet simulation, and a reverb. The 1kHz whine was really annoying, no matter how I tried to ensure there were no ground loops, digital and analog ground were connected with a single low-impedance path, and the analog input power was stabilized. The load stabilization suggested by @Newcastle in the previous comment (and in the DaisySeed docs too) was unfortunately not helping in my case. All the computation I am doing in the audio callback causes spikes in the power consumption once per callback (so, 1kHz), clearly visible with an oscilloscope on the Daisy Seed VIN pin, and unfortunately propagating to the analog part through the input buffer and possibly through the 4.5V audio reference.

Reducing the block size from 48 down to 8 or less would have sent the frequency in a range that is mostly filtered by guitar cabinets, and going down to 2 would go outside of the audible range, but that is not feasible in my application (or, at least, I could not make the callback efficient enough to allow for it).

What worked for me was digitally filtering the noise away. Here's the thing, this noise has three very important properties, at least in my case:
1. It mostly affects the circuit through its analog input
2. It aligns perfectly with the audio block (by definition, because it is caused by the audio callback itself)
3. It is constant across blocks: as long as I don't change the analog input gain, the noise always has the same pattern on every block.

This provides an opportunity to filter it: I keep 48 variables, each holding the running average of each one of the 48 samples in the block (across different invocations of the audio callback, trailing off over around 1 second). Then, I subtract such average from the input block. Effectively, this removes from the input signal any component that perfectly aligns with the block and is stable over time.

The noise is completely gone, one can only hear a small transient noise over 1 second when I tweak the analog input gain, that can be changed with a potentiometer in my board, but that's fine as this is not changed while playing. The tone is unaffected. Theoretically, an input signal very close to 1kHz or its multiples and relatively constant over time could be affected, but in practice guitar has too much transient and variation in phase/amplitude/pitch that I do not detect any negative impact, even if I play a B at the 19th fret (at around 988Hz it's the closest note to 1kHz), or even if I try to bend it to 1kHz. If anything was audible, a longer moving average would make it even less noticeable, at the (acceptable to me) cost of a slightly longer transient when changing input gain.

I know it might not be a viable option for all applications (for example, if your computation can change a lot while playing), and that fixing it in hardware would be better. If I design a new revision of the board, I might try using a stabilized 5V as audio reference instead of dividing 9V into 4.5V, and to isolate the analog power section even more. That said, this solution did save me from having to design an entirely new board, and for my application it has basically no practical downsides.

I hope it may help others too.
 
Last edited:
Back
Top