Asked this question on the Raspberry Pi Stack exchange and was redirected here!

I'm designing a low-noise audio gadget that uses an MCP3008 ADC (for sensors) together with a Raspberry Pi and a PCM5102a high quality DAC. The PCM5102a is powered by a 3.3V LDO. I'm wondering if I should share that same LDO to also power the MCP3008 or should I just use the pi's onboard 3.3V?

The MCP3008 is already using a precision 1V reference voltage IC. Is it necessary to use a low noise 3.3v power supply as well or will it make no difference?

If I use the LDO to power both chips, will the MCP3008 generate noise that will disrupt my (very very low noise) PCM5012a?

The PSRR on the MCP3008 is about -68db whereas on the PCM5102a it's around -100db!

The alternatives are to a) just power the MCP3008 from the Pi's onboard 3.3V supply, or b) use a separate 3.3V LDO, but that seems like overkill.

  • 1
    \$\begingroup\$ So what level of performance are you looking for? We can't advise you without a definite specification. \$\endgroup\$
    – Dave Tweed
    Jul 14, 2019 at 14:03

3 Answers 3


The DAC uses a 1.7V LDO internal for the digital core and so you may share 3.3V and keep return paths isolated to avoid ground shift just in case of audio filter coupling. But it creates its own V- charge pump for differential DAC output with 112 dB dynamic range.

Follow the layout and part selections and read all the fine print.

The 10 bit MCP ADC is unsuitable for audio recording but OK for simple sensors . You need a sigma-delta type 24bit ADC for Audio, but for simple 60 dB dynamic range sensors, no problem.

Without design criteria, specs, I can only give hand-waving recommendations if you follow their design.


I assume that the MPC3008 is for voice activation of your device so is not being used for high quality audio digitization.

I would advise that the 3.3V generated to drive both your ADC and DAC is generated from 5V and uses a shunt regulator (TLV431) rather than a series regulator. This allows the best possible PSRR from the 5V supply, and makes the supply current more constant for the ADC/DAC portions of your design. Since the devices are low cost you could even consider separate shunt regulators for the analog devices.
If you are doing modeling of the analog front end, here is a pointer to a quite good Spice model of the TL431 I've used.
One point to watch is that the TL431/TLV431 series can be sensitive to the capacitance on the output. Make sure you have sufficiently large capacitors (I always use 4.7 uF) or sufficiently large Ik to keep out of trouble (see Fgiure 18 in the datasheet).

From there it's just standard rules for implementing low noise analog peripherals:

  1. Bypass capacitors close to the devices
  2. Star layout for power and ground (separate analog ground pour)
  3. Short tracks
  4. Low pass filtering to your range
  5. Hi pass filtering for hum (50-120 Hz)
  • \$\begingroup\$ Has TL431 really better PSRR in higher frequencies than some modern LDO? I actually cannot find PSRR chart in the datasheet of TL431. \$\endgroup\$ Jul 16, 2019 at 8:39
  • \$\begingroup\$ Not sure what you are getting at? What does high frequency mean in PSRR? The specified dynamic impedance allows you to calculate a baseline PSRR or look at the equivalent noise (ti.com/lit/ds/symlink/tl431.pdf) and the test circuit (Figure 7 and Figure 9) to see the best configuration. You can split the series resistance and create a low pass filter node for total noise in the 250 nV range. I don't see you achieving this level of noise output with say an LM317 which is in the 8-10 uV range. \$\endgroup\$ Jul 16, 2019 at 10:43
  • \$\begingroup\$ Probably the best LDO available is the LT3042, which is about equal to a TL431. Great writeup here: analog.com/en/technical-articles/… You could certainly use this, but at close to $6 one off, it's a far cry from the TL431 at 50 cents. \$\endgroup\$ Jul 16, 2019 at 10:57
  • \$\begingroup\$ High frequency in LDOs mean something like 100kHz and above. Simply frequencies that basically pass through LDO input to output almost untouched. Imagine noisy switching power source with 1MHz (and above) noise. LDOs (linear regulators I should say) are powerless against that. I am not talking about inherent noise. So figure 7 you mentioned has nothing to do with PSRR. \$\endgroup\$ Jul 16, 2019 at 13:31
  • \$\begingroup\$ @Chupacabras I do know what high frequency means …..but since most of the devices are manufactured with GBW must less than 1Mhz it's always difficult to ascertain what is being implied by PSRR. You need to see the freq range being claimed. And yes the inherent noise measured at the output (just as in the LT3042) is a measure of input to output ripple ratios, since you are measuring a broad spectrum. \$\endgroup\$ Jul 16, 2019 at 15:57

Insert 100 ohm resistor in the VDD to the ADC. Place 100uF right at the ADC's VDD pin. Use a Ground Plane under all ICs.

If you find the 100uF is overkill, then use a smaller capacitor.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.