Electrical Engineering & Electronics Design

Thanks for this!

It would be nice to include the 3.3V ADC resolution as well.

Where is the graph with parts from Digikey?

It's great to see ADC resolutions increasing on micro controllers too without having to get an extra ADC module.

No this is NOT microcontrollers ! This is for external ADCs.

If you read my post, you will see that going to a lower voltage reduces bandwidth of the ADC, so you definitely do not want to do that !!

The obvious solution is that since all MCUs have core CPU voltages of 3.3V or lower, that you would use the supply rails for everything except the ADC in which you would use the external Analog Reference pins for the higher voltage to get better dynamic range.

This is the chart for MCU ADCs that I finished this afternoon. Notice the big differences and the limitation, so you cannot get high resolution on most MCUs , plus they are also limited below what the MCU rating is for the clock speed; usually only a fraction.

Also this list is just for 32 bit MCUs - no 8b or 16b devices since almost everyone has moved onto 32 bit MCUs.

I will try to see what is available for embedded CPUs and post that later.

I check all of the Embedded CPU parts on Digikey and none of them have ADCs of any type. So in regard to ADC resolution either an external ADC is used, especially for high precision and high speed sampling.

For lower costs, reduced bit resolution, and lower sampling speed, the ADCs in embedded MCUs may be good enough.

Note that any Sigma Delta ADC will always be an oversampling type and therefore will be much lower output speed but will have much higher precision of quantization for every output value.

Also see all of my notes on the 24-bit ADCs in the Radio Frequency and Wireless group.

Yes I totally understand that.

I was just saying since a lot of micro controllers are shifting to 3.3V and lower. It would be nice to see higher bit ADCs integrated into them.

It would be nice to see the if the noise can be reduced some how. I don't have an in depth knowledge of electronics but I wonder if a lower nm node may help with noise issues and increased resolutions.

One of my favorite MCUs is the ESP-32, fast dual core CPU with WiFi.

I also like the new Arduino Nano 33 IoT which has WiFi.

I also like the Teensy 4.1 (600 Mhz! What a beast!).

All these have 12 bit ADC.

Different topic but I thought I should share this:

https://www.microchip.com/en-us/solutions/low-power

"Sleep currents as low as 9 nA with power-efficient wake-up sources and run currents down to 30 μA/MHz"

If they can get really low power consumption, I hope they can get higher integrated ADC as a next feat/goal!

So noise sources are mainly the thermal effects that come from the atoms vibrational patterns that are directly coupled to the thermal effects. This is an internal issue that cannot be avoided with any material.

Reducing size of the atoms in geometry by die and gate size may actually make matters worse, since the physical size is so much smaller, the effects of thermal activity are much more reactive (sensitive) to changes and then are less stable.

Moreover, other complex circuit adjustment compensation circuits may need to be implemented to help stabilize from size reductions, so although the other circuits may shrink, by adding additional compensation circuitry, the net gain of the analog electronics may be a net-sum-zero-gain game.

In fact, if you ever looked at the die of the MCU, you will see that the largest geometry of the die is the ADC and is also becoming the most significant cost factor of the MCU pricing, right after flash size. Its because as the other logic side get smaller, the need to keep the analog side physically the same before the die shrink, so changes in the analog side actually do not scale as fast as the digital side.

You may eventually come to where the ADC may be pushed outside of the MCU in order to lower costs !! I have seen this with the really LOW cost MCUs, so if really don't need an ADC, don't add one just to be safe if you are trying to drive down costs.

Bear in mind that any RF circuits are Analog, but really its the other way around ! ;-) ALL digital circuits are really analog; and all analog circuits are really RF circuits even if the switching speed is really low since ALL circuits are R-L-C components. Even simple resistors are really R-L-C circuits in reality ! Its just that we ignore the significance of the L-C portions relative to the comparative magnitude at frequency.

In regard to the Embedded MCUs that support different wireless protocols and frequency bands, YES, I like them too ! :-) And the Teensey is a top-end cool little' beast'.

< I will continue with the sleep currents on another thread >

Sleep currents are really misleading !!

Why do I say this ? Because they only matter when you are sleeping !! What really matters is the ON duty cycle power consumption to the OFF duty cycle power consumption !!

If you are running 10mA for 1% and 1 uA for 99% it may be lower than a 100mA for 1% and a 1nA for 99% !!

It's both the ACTIVE power ON TIME and the SLEEP power OFF TIME combined to find out how much power you truly save !!

Without ALL the power information, it can be VERY misleading ! Only by having a TOTAL POWER BUDGET can you truly understand your real power savings.

So its best to compare all facets of power consumption in a spreadsheet for all MCUs before jumping to any conclusions because you may be surprised what parts truly save power.

Other factors that tie into power consumption are :

• bit width (8b, 12b, 14b, 16b, 18b, 24b, 32b)
• clock speed (how fast are you on or off)
• power consumption of each and every circuit portion
• circuits that cannot be powered off
• circuits that do not have low power modes
• clock circuits that cannot be turned off
• leakage currents
• core voltages
• external I/O voltages
• external discrete power leakage losses
• external circuits that are not powered off
• parasitic losses
• noise losses
• poor code size
• compiler optimization
• poor coding skills
• excessive RAM usage
• excessive RAM sizing
• RAM speed
• ROM circuits
• internal charge pumps for flash memory
• frequency of storing to flash
• 
  

Here's some additional information on the subject of external vs internal ADCs :

https://www.digikey.com/en/articles/iot-microcontrollers-have-adcs-know-when-apply-an-external-adc

I wanted to mention that in digital systems dynamic range is expressed in Bit Error Rate (BER) used with BER Testers (BERT). But I want to point out that BERTs do not always show true recoverable digital dynamic range since almost all higher speed protocols are not based upon individual bits, but packets ! https://en.wikipedia.org/wiki/Bit_error_rate

If you are sending (or switching) a very long Ethernet packet and you have a single bit error, the entire packet (frame) has to be resent, so 1 bit error in a very long frame (jumbo) and excessively long frames (super jumbo) can be over 80,000 bits ! [10,000 bit jumbo frame x 8 bits).

https://en.wikipedia.org/wiki/Jumbo_frame

And the delay time for even 1 Gbps can be 80,000 x 1us = 80ms which for high speed switching systems can be an eternity, especially when there are random frame errors that are spread out over the bandwidth.

This is why traffic BER for Ethernet on progressively higher traffic have actually improved pushing well past the traditional 1E-12 and now past 1E-15 with 1E-18 over fiber more common to get past all the RF noise on copper wiring.

Wow! Thanks for all that information. I'm a manufacturing engineer, but it's still nice to read and get information in different fields of engineering. Thanks for being a great resource for information. Love your wireless and electrical related models!

Thank you, I try to help out where I can, so don't ask me about political questions ! ;-)