Rev 7 noise floor vs Rev 4

I have both of these and the Rev 7 is significantly noisier than the Rev 4. I read that the earlier rev used an AKM codec that isn’t available anymore, but Rev 7 uses the PCM3060, which is supposed to be an improvement over the WM8731 used in Rev 5.

I’m measuring 40uVrms noise from Rev 4, and 225uVrms for Rev 7, plugged into a Terrarium with the digital audio out set to 0. That’s 15dB difference and too much for a guitar level input. The WM8731 was claimed in another thread to be 11dB max above the AKM, but it appears that the PCM is even worse!

Is this normal? I’m going to make my own dev board with a preamp, which should help, but for now this seems abnormal.

1 Like

It seems that Rev7 is way noisier than Rev5, at least in the Terrarium. I just received today a Rev5 module and will do a detailed comparison tomorrow. I will share the information here.

Thanks for the reply! Sort of surprising that the new codec would make things worse…

Hi,
there are some reasons already known for noise here:

  1. PCM3060 has a higher analogue voltage 3.6Vpp versus 2,1Vpp of AK4556, so absolute Vrms is higher. This alone gives 4.5dB.
  2. PCM3060 has out of band noise, which cannot be heard but can be measured. See datasheet of PCM3060, Page 23. “An external low-pass filter is used if further out-of-band noise rejection in required.”
  3. In my opinion the design of Terrarium could be improved. Suggestion:

@masd curious question: How do you measure such low voltages as 40uVrms? When I try to do such things, my oscilloscope always picks up lots of noise not coming from the intended source… A directly shorted probe or no cable at all in the input alone brings about 300µVrms. (Rigol DS1074)

Hi, @cebersp and @masd
I don´t have a Rev4 board, but a Rev5 with the WM8731 witch in theory has the same 3,6 Vpp output level as the PCM3060, so the difference between Rev5 and Rev7 should be 0,0 dB
As for the recommended external LP filter for the PCM3060, it should be applied at the output, not the input. I modified my Terrarium by adding a capacitor after the 100R resistor at the output of the opamp to get a 15.9 kHz cut-off. It would be better to have this LP before the opamp, but I was looking for a solution that didn’t involve disassembling the pedal. I haven’t really noticed any practical difference to the ear, I’ll have to do the measurements at some point. The current input circuit of the Terrarium has an LP filter, perhaps unintentionally, with a 159 kHz cutoff, which can easily be lowered to 15.9 kHz by changing the capacitor to 10 nF. This provides aliasing protection in the event of high frequency signals or interference coming into the input (someone mentioned in this forum that they had problems when connecting two Terrariums in series, it is possible that this is due to a lack of filtering at the output).
Attached is the Terrarium schematic to illustrate the above:


I agree with @cebersp in his suggestion to create a virtual half-voltage source to reference the analogue signals.
I assume that @masd has measured the output signals using an analogue interface and a suitable program like Audacity, as I did. Even the best oscilloscope is of little help in these cases. More info in the next post.

Hi again, I am going to post the results of my noise comparison measurements.

About a month ago I bought a DS Rev7 directly from Electrosmith to install in a Terrarium kit to evaluate its capabilities. I was surprised at the amount of noise output I was getting, so I started researching and finally bought a DS Rev5 from a dealer who had it in stock to make comparisons and try to reach some conclusions.

My test bench has consisted of the Terrarium kit modified to add the capacitor at the output mentioned in the previous post and a program that alternatively generates zeros at the output or bypasses the input (shorted to ground) to the output, changing with the press of a footswitch.
To capture and measure the output signal I used the microphone jack of my laptop connected to the pedal output, together with the Audacity program. The results seem coherent to me, but I am going to repeat the test with a sound interface to be more sure that I am doing it right.

The first thing I must say is that the difference in noise is very clear. The best case is the Rev5 with zero at the output, then the Rev5 in bypass, in which a hiss is already beginning to be noticed in various tones, and finally the Rev7, both with zeros and in bypass with hardly any difference, with a very loud noise. annoying and high that is a mixture of many frequencies without being able to determine any dominant one.

The RMS levels obtained in each case (Analyze - Measure RMS) are as follows, first is the level measured directly by the interface and in parentheses is the level referred to 1 Vrms, and then the supposed level in uV, to compare with the levels obtained by @masd

Rev5_Zero => -58,6 dB (-94,6 dB) 18,6 uV
Rev5_Bypass => -49,3 dB (-85,3 dB) 54,3 uV
Rev7_Zero => -36,6 dB (-72,6 dB) 234,4 uV
Rev7_Bypass => -35,9 dB (-71,9 dB) 254,1 uV

The difference between the Rev5 and Rev7 is 22 dB with the output at zero and 13.4 dB with the pedal in bypass. Interestingly, @masd measurements in uV are quite similar to mine when in bypass…

I have also analyzed the frequency spectrum of the signals (Analyze - Plot Spectrum), these are the results:


Rev5_Zero


Rev5_Bypass


Rev7_Zero


Rev7_Bypass

Here are also the MP3 sound files shortened to 5s, I could also upload the original 15s WAV files if anyone is interested.

Rev5_Zero

Rev5_Bypass

Rev7_Zero

Rev7_Bypass

Sometimes Rev7_Bypass generates strange hissing noises, not always but from time to time.

Rev7_Bypass_noises

My first impression is that the culprit of the noise is not the codec but the layout and/or decoupling of the DS board. Rev5 is quite good in its output and not so good but acceptable in its input, but Rev7 has a clear problem and, in my opinion, is not suitable for a sound system that aspires to be of any quality.

I have examined as much as I can the layout of the components in the new codec but it is not possible to determine if there are any design mistake, for this I would have to have the PCB design, but not even the schematic is available, so I think that Is that asking too much…

If I were Electrosmith, I’d be urgently designing a Rev8 before its reputation is lost. Even a Rev7 board recall could be a good idea…

1 Like

What I meant to say is, that PCM3060 has a 4.5V analog power supply, while the others both have 3.3V. PCM3060 scales its output to the higher voltage range which gets an output range of 3,6Vpp, while AK4556 has 2,1Vpp and the datasheet of WM8731 says 3.3/3.3 * 2 * 1,41=2,82Vpp, but they recommend to use -3dB for low distortion, while PCM3060 gives distortion values for -1dB.

Of course, the differences, which you show, are much bigger than this difference.

===========
Other topic:
As far as I understand, the Petal board was sold with a battery to have better S/N ratio.

I have access to an audio analyzer and used it for the noise measurements.

I’ll be designing my own stereo dev board so that gives me an opportunity to try to improve SNR, thanks for the suggestions on improving the Terrarium circuit. Probably will reach out to customer support to see if a Rev8 is in progress to fix this. If not, I might look into an external DAC for now.

Have you tried adding a simple 20kHz RC low pass filter on the outputs of the Rev7 seed before any opamps? That’s what I ended up doing on my guitar pedal design and it seemed to solve the noise issues with the Rev7 for me and now it’s much less noisy than the Rev4 in my experience.

OK, after I had learned, that power supply had a certain influence, I revised my setup, and TATAH!

Power supply to Daisy Seed Rev7 via USB coming from my laptop. (A 9V battery was not better).
Use differential amp to cut ground loop.
Essential: Take output ground directly from Analog_GND Pin!!!

2kHz 100% output => set Zoom R8 input sensivity
2kHz, 50% output => take measurement,

As Audacity does not display <-90dB, I set output to 1%=-40dB.

Comparing relative rms values 620 for 100% amplitude and 0.024 for 0% gives S/N=88dB and ENOB= 15.7bits.

Nevertheless, when I had a look at the Datasheets of LP2985, TI recommends an input voltage more than 1V over the output voltage (4.5V), and there is a diode before this chip. So if you have a weak and noisy usb supply, there might be problems?

USB connection via a USB-HUB:

If the 9V Battery at Pin 39 is used in parallel to the bad USB:

The influence of output level to distortion is:
Level= 1%: THD=0,05%
Level= 50%: THD=0,02%
Level= 80%: THD=0,03%
Level= 90%: THD=0,1%
Level=100%: THD=2,6%

Lessons learned:
Beware of ground loops! Use analog GND very closely to the PIN! If you have a bad USB supply, then an additional 9V Battery supply helps! Reduce output level to 90% or so for low distortion.

Is it the rev7 schematic you were looking for ? I found it available on product page if this helps.

Thanks, lemsto-SBO, but this schematic is very incomplete, it only shows some external connections, and even that is only partially shown…
They published complete schematics until Rev5, but Rev7 may have something obscure…

1 Like

Thanks for all the information, it is very valuable. I have somewhat abandoned this project, partly because of lack of time and partly because of disillusionment with the Daisy_Seed. When I have done more research I will post my findings here.

1 Like