Rev5 seed producing ~13dB higher noise floor

[stm]

Has anyone noticed a significant increase in noise floor in rev5 daisy seed compared to rev4? I measured around ~13dB higher peak and ~7dB higher RMS.

The flashed program is just writing the output buffer to 0.

It doesn’t sound like a lot but the difference is very noticeable when amplified (with a guitar amp). I know there is a change in the audio codec in rev5, but the noise floor chart in the data sheet is not updated.

Is it this problem specific to the unit I have or is it a universal problem? Has anyone found a way to reduce the noise floor problem in rev5?

[shensley]

To address your question, I feel some Daisy-History is necessary, but the short-version is that the difference in noise floor is related to the codec installed on the board.

The original release of Daisy Seed (rev4) had the AK4556 codec installed. The AKM factory that manufactured that codec burned down due to a fire in 2020. Unfortunately, the end result was the part being discontinued, with no plans of continuing to manufacture it.

In order to keep Daisy alive despite this, and to be able to continue manufacturing Daisy Seeds throughout the chip-shortage, we redesigned the board to be hardware compatible with the existing Daisy, but with a more readily available codec. This is the WM8731 that is installed on the Rev5 Daisy Seed.

During this period, there weren’t many components that would fit all of the requirements (voltage, size, availability, compatibility, etc.) required for the Daisy Seed. The WM8731 checked those boxes, and we had actually done a revision of the Daisy Seed with this codec prior to the kickstarter.

The relevant specs that differ with respect to the output noise floor are the SNR (signal to noise ratio). The WM8731 has a specified range of 95dB to 100dB, while the AK4556 has a specified range of 98dB to 106dB. So at a minimum there may be a 3dB difference in SNR performance, but at its maximum there can be up to 11dB (spec’d) difference. In practice, there are other factors that have an effect on the noise (for better and worse) of these ICs; things like the actual devices’ sensitivity to external noise, etc.

We are aware of some of its deficiencies, especially when boosting guitar level signals to the ranges needed. The last of the original “Petal” breakout boards shipped with the last of the Rev4 Seeds, and we decided to discontinue it at that point because we felt it didn’t live up to the quality we like to ensure. We are currently working on a few newer pieces of hardware designed specifically for effects pedal development. At least one of these should be available in the next few months.

As would seem to become an ongoing theme, the WM8731 has also been marked end-of-life, and is now an obsolete part. So we are currently working on a revised version of the Daisy Seed that has another different codec: the PCM3060.

If you’ve used the PatchSM, or have had the Seed2 DFM (another upcoming release) on your radar, you will already be familiar with this codec. Spec-for-spec, it is a lot better than the WM8731, and the AK4556 in the configuration that is used in the redesign (e.g single-ended outputs vs. differential on the other dev boards).

[tunagenes]

Thanks for giving us the short version and the details - it’s very useful to have that all in one place!

[kshep]

This is very helpful to understand. I’m pretty sure this is the background “noise” I’m hearing on my seed based guitar pedal designs. Good to know it’s not something specific I’m doing on the PCB. Also, happy to hear that a future iteration of the seed will likely solve this issue too! Thank you for this post!

[PlasticLobster]

Hi,

I just encountered the same issue, a Rev4 Daisy died for an unknown reason, and I replaced it with a Rev5 in a guitar pedal build similar to the Petal and Terrarium designs. The noise floor on the guitar input is now completely unacceptable.

Do you have any news regarding a release date for a new Seed with a better codec and/or a specific hardware ? (even though, as many others I have designed and manufactured my PCBs around the Seed design)

I just ordered 6 extra Daisy Seeds from Modular Square to build a batch of pedals prior to discovering this issue, if it’s the Rev5 I’ll simply have to cancel the order / send them back…

[Takumi_Ogata]

Hello!

The Daisy Seed will still be available in the future so you do not have to redesign your PCBs. That said, I currently don’t have an estimated date as to when the switch to the PCM3060 codec will happen.

And I recommend contacting Modular Square to check if they’re Rev5.

We’ll keep you all posted with the switch. Thank you for your understanding and the wait.

[PlasticLobster]

Thanks for your answer Takumi,

I’ll stay on alert for the next generation of Seeds then, I can still prototype my guitar effects with the Rev5s as long as I know that a fix is coming somewhere down the line, and it might be a good excuse for me to start using the Daisy for Eurorack (which I imagine is much less sensitive to these kind of issues).

Modular Square indeed had Rev5 boards, but I can return them so that’s not too much of an issue.

Thanks!

[ArguZ]

Hi made extensive tests last winter to get to the bottom of this…

And i also found the Seed 2 which was included in the Q-Bit modules way less noisy than the Rev 5 we could buy at that time…
Here a few graphs, i also have audio recordings of like 10 modules that all show the same pattern.
All recording made with UAD Apollo hardware at the line inputs without any effects

Aurora - very clean, the best of the test

Nautilus - still much cleaner than the Patch

Patch - Worst, especially with the OLED running

cardME - this is our module, based on Petal but with significant efforts to reduce noise
4 layer boards, thick hand routing, shielding , sections in the PCB for each module
It took 11 tries to get there.

And the best one…48 db gain on the Seed 2 module compared to Patch running Rev 5

All of these are with the Seed running a 2 in 2 out programm, once it is in boot mode it dropped to 95 db.
Yes, I am VERY much looking forward to the new codec.

[mjkirk12]

Following up on this codec topic:
Is there a spreadsheet, document or web page with a matrix showing which hardware product and revision uses which codec? I think these are the 3 used:
AKM: AK4556 EOL
Wolfson/Cirrus: WM8731 EOL
TI/Burr Brown: PCM3060 active

Here is what I have, please correct any errors.

Product / Codec

Seed V4 / AKM4556
Seed V5 / WM8731
Seed 2 DFM / PCM3060
Patch SM V4 / WM8731 or PCM3060 ?
Patch.Init() / Uses Patch SM codec
Patch / Uses Seed codec + external codec WM8731 or PCM3060 ? (quad audio)

Thanks for reading,
Mike
(first post)

[tele_player]

Welcome Mike -
The libDaisy source code shows which codecs are used for each board.

[mjkirk12]

@tele_player Thanks - I did some review and came up with this:

Product / Codec
Seed V4 / AKM4556
Seed V5 / WM8731
Seed 2 DFM / PCM3060
Patch SM V4 / PCM3060
Patch.Init() / Patch SM codec (PCM3060)
Patch / Seed codec + external codec AK4556 << quad audio >>

-Mike

[kshep]

I’ve been following this thread for a while now. I was experiencing the higher than expected baseline noise floor with the Rev5 seeds in my daisy seed guitar pedal design when simply doing a passthrough of the audio signal from input to output on the seed without doing any processing. It wasn’t terrible, but it was slightly audible.

I was excited to get my Rev7 Seed 1.2 with the new codec to see how it works. A few interesting things to report:

1. The good news, with the same setup, I’m pleased to report the background noise is completely gone when simply passing the signal through. Very exciting!

2. The bad news, if I simply add a second one of my guitar pedals to the signal chain (also rev7 seed doing a simple pass through) there is A LOT of background noise. Worse than when the Rev5 was alone. Stacking multiple pedals with the Rev5 seeds did not produce this same result. In fact, if I have a Rev7 first then a Rev5 the noise is also there, but if I have the Rev5 first and the Rev7 second, it isn’t there. It makes me think something about the new Rev7 seed output is interacting weirdly when another buffered input is behind it.

Anyone having similar issues, or have any ideas why this might be happening? My guitar pedal schematic input / output buffers are the exact same as they were designed for the Daisy Petal schematics if that makes any difference.

[kshep]

Ok, after much troubleshooting, I’ve narrowed down the issue I’m seeing a bit further.

Here is my setup that is producing a different behavior between the Rev5 and Rev7 seeds:

1. Both the Rev5 and Rev7 Seed are programmed with the Sample Seed Bypass example - https://github.com/electro-smith/DaisyExamples/tree/ceadb82adcd6f0cc44fb718fea77b2da14f21b5e/seed/bypass

All it’s doing is passing input to output.

2. I setup the seed on a breadboard as shown in the attached image. The seed is being powered by 9v battery power, DGND and AGND are connected, I’m using a TRS stereo jack with Tip connected to the Audio In 1 pin, Ring connected to Audio In 2 pin, and Sleeve connected to GND. The Stereo TRS jack is connected to the stereo headphone jack on my iPhone. Finally I’ve connected my oscilloscope probe to the GND on the breadboard and to the Audio Out 1 on the seed.

3. If I look at the Oscilloscope with the Audio Paused from the iPhone it looks like this when the Seed Rev7 is connected.
   

   

4. If I look at the Oscilloscope with the Audio paused from the iPhone it looks like this when the Seed Rev5 is connected.

I would have expected the same results with both measuring no signal since the audio is not playing.

Also, if I play audio through the input and connect headphones to the audio outputs on the seed it sounds just fine with out noise.

It really seems like something is behaving different electrically with the Seed Rev7 compared to the Rev5 when a High Z connection (like my oscilloscope or the input buffer on my next guitar pedal) is attached to the outputs, which seems to be causing the noise I’m hearing when stacking pedals with a Rev7 before another pedal.

[Takumi_Ogata]

Some of that is probably down to the codec changes that we were discussing earlier in the thread. Over the Seed generations, some of the passives on the audio IO have also been tweaked a bit here and there. That was mostly done to keep the Seeds as backwards compatible as possible, but there still may be some small differences.
That being said, it’s not recommended, especially when dealing with guitars, to directly connect to the Seed’s audio IO. Appropriate op-amp buffer circuits on the audio IO should help a lot to improve these issues.

[kshep]

Thanks for the response Takumi. Unfortunately, that’s the issue though. In my guitar pedal, I do have the op amps before and after the seed rev7, and that noise becomes audible. Doing the experiment without the seed was to try to figure out if I could rule out the actual guitar pedals being the issue.

Here is an experiment for you all to try. Take 2 of the Daisy Petal hardware (my guitar pedal uses the exact same op amps and signal chain) and put Rev7 seeds in them. If you connect a guitar to 1 and then to an amp, you will hear no noise in the signal. However, if you take 2 Daisy Petals and daisy chain them. So the signal goes from Guitar → Petal 1 → Petal 2 → Amp, my bet is you will get the same noisy signal I’m seeing. If you were to put Rev5 seeds in both of those Petal’s you would not.

It’s also not a ground loop, as I’ve done this powering all devices from batteries and the issue remains.

My current theory (and I’m no EE) is that this new Audio codec in the Rev7 seed is expecting the outputs to be loaded, and when they are loaded even just a little bit, the noise is gone, but the op-amps at the output and then input of the next pedal are actually preventing the outputs of the first seed in Petal 1 from being loaded enough, and that same noise we saw on just the seed by itself persists but is now amplified.

[kshep]

Ok, I took this a step further on the breadboard. With a single Rev7 seed playing a wav file with the audio output hooked up to the left side of this circuit I get no noise at point A and noise at point B (using either an oscilloscope or audio probe at those two locations on the schematic, the wav file audio is audible at both points just extra noisy at B)

The reason I did this on the breadboard is because the left side of the circuit is the same as the output op amp circuit in my guitar pedal (and the Daisy Petal), and the right side of this circuit is the same as the input op-amp buffer circuit of the second guitar pedal. Basically replicating the interface between the two.

The same exact circuit on the breadboard with a Rev5 seed has clear audio at both point A and B.

The Rev7 seed, clearly just doesn’t like this configuration for some reason. Hope this helps provide some insight into the behavior

[jackcampbell]

I think I’m running into the same/similar issue with the Rev7 (though I don’t have buffered audio I/O set up yet, which is a significant variable I know…) Loud noise floor/white noise which goes away when the output of the Daisy passes through a passive DI connected to a low-z input on an audio interface, powered by battery and running the Seed Bypass example. Seems to have only become an issue in my circuits after upgrading from Rev4 to Rev7

[AsaLPatterson]

I’m also running into this issue. I’m using opamp buffers for inputs and outputs. Switching back to Rev5 has solved my noise floor issue for now, but wondering if we can expect a newer revision to address this? Or some other workaround?

[Takumi_Ogata]

In yesterday’s meeting, I brought up the noise issue that @kshep first reported, and it was added to a list of issues to troubleshoot.

I appreciate all of you for bringing this to our attention and hope to share our findings very soon. Thank you so much for waiting.

[jackcampbell]

@AsaLPatterson While waiting for an official solution, just an FYI, I was able to work around the noise in my case by placing a transformer on the audio output. My testing was fairly limited (tried with my guitar amp, a Headrush, and the line-in on my interface and the noise was down to acceptable levels for those IMO), but mentioning in case you have any laying around and wanted to give that a go for your use case.