That will work

I do mention that in my opening remarks and it’s a very important point you make, as the maximum value of ADC input is quoted as Vdd - 0.3v according to the data sheet or 3v typically.

I mentioned the ESP32 impedance because it affects the calculations, it varies from device to device, but on the whole is relatively high but enough to skew the calculations. Yes power rating of the resistors is a consideration but for all but the highest voltages >50v should be well within a 1/4 watt resistor rating. p=V^2/R

When measuring AC voltages there are a couple of approaches that can be used, the first is to downscale the AC voltage using a voltage divider and apply this via a capacitor to a voltage offset usually two equal resistors across the 3.3v rail, this is to allow the ADC to measure the positive and negative swings, then the incoming waveform needs to have a zero detection and then sample the whole waveform to then compute the RMS value from the peak to peak values. It gets complicated! Alternatively, apply the AC of suitable voltage probably downsized by a voltage divider to a diode and capacitor to ground, this will then charge to the peak value from which the RMS value can be found by measuring the now DC by the ADC like any other DC measurements. I’ll try to summarise all this in another technical note, it’s really quite easy to do, but safety is a concern I need to consider.

@@G6EJD Wow! Thank you so much for the detailed overview. I shall try this while I wait for you to prepare a video on this. Really appreciate the effort :) Cheers!!

Subrata Dey, all you need to do is feed AC to a diode eg IN4001 and then a capacitor to ground of about 10uF then measure the DC with the ADC

The ESP32 ADC reference voltage is already nominally set at the factory to 1.1v although in some earlier devices it can range from 1.0 to 1.2v. You can route the internal reference input to GPIO25 and apply your own, which you’d need to measure accurately and then recalibrate the ADC but the advantages in doing this aren’t that great. What is the problem your trying to solve?

Well even at 1/4 watt a simple resistor voltage divider is perfectly safe up to 100v, but augmented with adequate terminal protection. Even up to 220v it’s still quite safe again with appropriate power ratings, what factors were you considering.

​@@G6EJD I need to take stable measurements over a wide measuring range of maximum 500V rms. "4kW" PF = 0.99/ 0.97 induction cooker and measuring device required for white goods testing

Pleased to hear I’m helping you out. The non-linearity your seeing is not typical, the most linear region of the ADC is between 0.5v and 2.8v and outside of this range you will see non-linearity. Any variation in input impedance is more likely to be the cause, but just less than 100K should provide enough input current to operate the ADC correctly, I assume there are no further attenuation resistors in the circuit. From the parameters you have provided everything should work as expected. You can improve the linearity if the ADC with some mathematics and I produced a function to do this in one of my other TN’s improving ADC linearity.

@@G6EJD Will be sure to read and or watch / learn more about ADC linearity, thanks for the quick reply.

Let's go fry some esp32s... ;)

Piotr Ojdana, the input range will be 0-4.2v and divided down to 0-3.3v. So if the input is 0v the reading will be 0v or if 2.1v the voltage divider output will be 1.65v, it’s just a method of changing the default 0-3.3v to whatever range is required. The ADC reading needs to be scaled to a 4.2v maximum then you will get an accurate voltage reading just like a DVM would do.

Bill, there are a number of reasons, component tolerance associated with the voltage divider, or production variations in the ESP32 and it’s ADC reference voltage and measuring equipment tolerances. Your observations could be due to a combination of all three. The Espressif data sheet shows a maximum ADC reading of Vcc which is 3.3v, I have never seen an ESP32 that can’t read to 3.3v, so assuming yours is a genuine ESP32 then next I would check for DVM accuracy and next that the voltage divider values used are correct. With so many variables it’s near impossible to produce a design solution that compensates for all the tolerance bands. The only solution is to vary the divider as you have already surmised.

​@@G6EJD Yes, many variables, and it's not a genuine ESP32 - the cheapest I could find to experiment with . . . Anyway I want to use it to monitor a Solar Cell output and a 3.7v Battery Level, so I've mapped the range in the linear portion, customised the V-Divider, and have a pretty accurate scale for the narrow range I need. Thnx again for your great series of informative videos. billd

Well it will be a function of your ESP32 supply noise, do you have sufficient and good enough smoothing and filtering. You should have a largish smoothing capacitor say 47uF or maybe 10uF and at least 0.1uF ceramic capacitor. The ceramic type is important as unlike larger tantalum capacitors are not inductive, therefore it can filter out transients on the supply. Also in theory the battery should act as a capacitor and its impedance is so low it should not allow any transients, so it proves almost conclusively it must be supply noise. It will still be there when measuring AC but less noticeable. Adding a small 0.1uF capacitor across the ADC to ground will help too. What value series resistor are you using? You can also counteract quantisation noise from the ADC by using a simple averaging technique of taking say 10 readings in succession then dividing the result by 10, which is a common method. The ADC is so fast in most applications the method will be indiscernible in practice.

Try adcvalue * 3.3 / 4095 * 0.5/0.3;

You can measure DC easily, but an AC voltage needs signal processing to convert it to what’s called a Root Mean Square value that is the equivalent of a DC measurement. To do this a full cycle of the AC waveform would need to be sampled over say 256 samples then the RMS calculation performed to give the final reading. Measurement of an AC waveform is complex especially if the waveform is distorted.

Marc, wired as Vin===320K===ADC input===100K===Gnd gives you a voltage input range of 4.2v or Vin at ADC = Vin * 100/(320+100) in code terms: float voltage = readanalog(A0) / 1023 * 420 / 100; If your using a Wemos D1 mini it already has a 220k-100k voltage divider on-board, so would need some extra series resistance say 100k to make it 320k-100k

So your reading of 328 is giving the correct result for the voltage divider, but I suspect your using a Wemos D1 Mini with an on-board 220k-100k voltage divider which for a 4.2v input AND the addition of a parallel network yields a reading of 342 but exact values depend on the individual ADC reference that do vary. So if using a Wemos D1 Mini just insert a 100K series resistor to the A0 pin where it adds to the on-board voltage divider.

I think to really understand what’s goi g on I need to know the exact board type your using. Also did you measure the cell voltage by independent means like a DVM?

@@G6EJD = 1,3466 v which is far away from 4,20v. Im really messes up... :(

What board type?

Yes it’s all relative but the ADC uses an internal reference so device supply variations have minimal effect on the readings, then there’s the ADC linearity, which is not that good, but can easily be improved to

It depends on the sensor your using, what type is it? You need to know it’s output characteristics

@@G6EJD Its output voltage 0~3.0 and I’m using it with esp 32 30pin.I really need this information,because I’m doing my project related to pH sensor and esp32.

@@G6EJD May I have your Facebook or Instagram account?

Ok so if the output voltage is 0-3.0 then that represents a pH range of ? 0-7 or 5-7 or?

@@G6EJD Represent range 0-14 sir.

Konstantinos Kostis, it is because 0 is a number and there are 4096 values that the ADC can output, it’s a common misconception with binary mathematics.

@@G6EJD Yes, there are 2^12=4096 values but they start at 0 (not 1!) and end with 4095 (not 4096!). So much about common misconception with binary mathematics. A 12-bit value can hold 4096 different values but not the value 4096.

Konstantinos Kostis, the mathematical solution is to get the fractional results the denominator of which has to be 4096 as there are 4096 possible readings. The solution is to make your denominator 4095. In every ADC example I’ve ever seen the denominator is always the ADC range like in this case 4096. You could also reduce the 3.3v reference by 3.3-3.3x1/4096 to give an answer that more accords with your thinking, but I doubt you’d ever notice the 0.8mV error as the quantisation errors will be much greater.

When there’s an ADC built in why bother it just adds cost and complexity of design

@@G6EJD because accuracy sir.

But you can get within 1% accuracy with the in-built ADC, why do you need more, it’s never going to be a scientific instrument? I think you could usefully read into ADC’s and what are the important aspects of them, speed, resolution are important and to a lesser extent linearity, what aspect would a ads1115 improve?

​@@G6EJD I will give you test report with code sir. voltage ---- LCD ---- adc 11.00. 11.00. 3007 10.00. 9.90. 2710 9.00. 8.81. 2416 8.00. 7.78. 2120 7.00. 6.72. 1840 6.00. 5.63. 1550 5.00. 4.61. 1263 4.00. 3.55. 972 3.00. 2.50. 683 2.00. 1.45. 400 1.00. .425. 117 This is a voltage comparison chart. I think esp32 adc is not compatible for linear reading.

The result is integer unless you specify or imply a floating result with 4096.0 or 7500.00

@@G6EJD I've used integer, without floating. So do I need to multiply by 1000 ? Comma's in comment was from mobile calc :( In sketch all parts were int. only analogpin was float

You can obtain the ESP32 internal reference voltage set at manufacture to improve reading accuracy. In general the ESP32 extra resolution should give you better results, but never approaching those of a DVM. To get better results you’d need to use an external ADC. You can also improve the linearity of the ESP32 using g some mathematics and what’s called a polynomial, I have covered both the ADC reference voltage and improving liberation other videos.

@@G6EJD Sorry, I called it the precision when I really meant to say the repeatability of the measurements. 10x better with the Nano.

Your voltage divider is loading the circuit

@@G6EJD I will check and give you feedback. With ADC value. Sir one more thing any simulation software available for ESP32 for voltage measurement.

@@0124akash none that I’m aware of

Yes, but be careful anything over 30v DC is considered dangerous

Just finished watching your tech note 069 and found my answer: adjust attenuation values, I guess to 11db.

Hal, the default ADC range is 0 - 3.2v or actually 0 to Vdd -0.3 according to the data sheet. If you switch on the attenuation then the input range is further constrained to 0 - 1v the only solution is to use two external resistors as voltage dividers.

I also provided an improved accuracy function.

github.com/G6EJD/ESP32-ADC-Accuracy-Improvement-function

@@G6EJD Thank you. I tried your formula but was unable to fix the described behavior. I think this is a physical problem. Here's what happens: I'm using one of the axes of a joystick which I've measured goes from about 260Ω - 4.8kΩ at each end, however, when I connect its vcc pin to the 3.3V pin of the ESP32, the GND pin to the GND of the ESP32, and the signal pin to ADC on pin 34, I see the voltage on the ADC go between 0 and 3.3V, but for about a quarter of the range of the potentiometer, the voltage is stuck at either 3.3V or 0V, similar to the graph in randomnerdtutorials.com/esp32-adc-analog-read-arduino-ide/ "The ADC is non-linear". However, I also read, as you said, that setting the attenuation should eliminate this value clipping at the end of the ranges, but it doesn't seem to be working for me. Are we limited to using the center ranges of potentiometers when using them with the ADC of the ESP32?

Yes, but I’ve never seen an LCD display you can buy

G6EJD I ment was a lcd display connected to the esp modile to display the voltage. For a simple setup.

Ok, I know what you mean an OLED or TFT screen often known (incorrectly) as LCD. I can’t do that as I’m in hospital until at least November ‘19

No bad questions, I'm pleased you asked. The calculations might say a value of 13.6K is required, but resistors can only be purchased in 'preferred' values, like 10K, 12K, 15K. I put the so-called E-Number range in the video, so if you calculate for 13.6K, then you pick the nearest value in my example here 15K is nearest then you have to go back and recalculate using the new value to know the actual range.

@@G6EJDunderstood, so if i use 3 resistor for voltage divider, the calculation and coding still same?

Yes, and it enables you to get must closer to the required values my examples was 13.6K so we could use 3 resistors in series to make up one resistor using 12K in series with 1K and 680R giving 13.68K

@@G6EJD nice, thnks

Your right, what I tried to do is simplify the impedance measurement as actually the ESP32 requires a minimum input current to get the ADC to operate. Taken for instance a voltage divider comprised of two 470k resistors the voltage at the junction will be Vin / 2 but the ADC can’t make a reliable measurement, now repeat the test with say 220k values and all works as expected because there is sufficient input current but not enough to skew the voltage divider values. I was going to give an equivalent circuit of the ADC input but the data sheet gives insufficient design disclosure. So perhaps the best way to describe the input impedance is still to use the word impedance rather than resistance and that the impedance is dynamic, certainly not fixed. Therefore in conclusion I’ve estimated the impedance at 200k to reflect the minimum input current requirement for correct operation.

@@G6EJD many thanks for the clarification.

A very insightful observation. I will try to be more upbeat when I produce another one, Clearly my current mood is in my voice.

@@G6EJD Man, I hope your life gets as good as you make us feel with your very interesting channel. You really help. Thanks and best of luck from Spain.

Well the ESP32 can measure a voltage of any value inside the chosen range, so design for 24v and apply 13.8v and that’s what’s gong to be measured. I don’t understand what you mean by fluctuation as the ADC reading is instantaneous.

A d once the R1 R2 voltage divider is added the ESP32 can measure an unknown voltage applied to the input , there is no knowledge of an initial value other than to calculate the range maximum that the ESP32 can measure.

And the point your making?

@@G6EJD that I wish for another solution. In my case I need to measure up to 3.4v I don't know what to to. Does esp32 Internal software calibration take care of voltage divider variance? Yes I'm lost

@@solidfuel0 the ESP32 can measure up to 3.3v so you will always need a voltage divider, the solution is sine you have to use an external meter to ensure you have a reference, then measure the source violative with the meter and then with the ESP32 and compare, then adjust your ESP32 measured voltage by the ratio of Vmeasured / VESP32 so voltage = analogRead(35) * constant * Vmeasured/VESP32 or simply adjust your source code so that the voltage line is multiplied by either 0.999 if too high or 1.0”1 if too low, repeat until the correct value is obtained, you may need to use eg 0.95 or 1.005 or whatever.

@@G6EJD i could never get it to measure up to 3.3v even if the vdd is 3.3v.

@@solidfuel0 you can but it becomes very inaccurate as the input voltage goes over 3.0v see my video about improving the accuracy using a polynomial

But you have no concept in your mind of the effect of the ESP32 ADC input impedance on the calculations, so you need to brush up on your electrical engineering and think a little bit wider and deeper, because unfortunately you will find your wrong!

@@G6EJD Thank you for the quick response and important advice! Added ESP32 input impedance ~200K to calculations Your calculation formula for 100K and 27K divider: float voltage = (float)analogRead(36) / 4096 * 15 * 28205 / 27000; Check for 15V, 100K and 27K divider, ~200K ESP32 input impedance Voltage Source (VS) 15V Resistance 1 (R1) 100K Resistance 2 (R2) 27K!!200K=23K789 Output Voltage (VOUT) 2.883V (Voltage Divider Calculator) In this case analogRead=2.883V/3.3V*4095=3578 float voltage = 3578 / 4096 * 15 * 28205 / 27000 = 13.688V (not 15V!!!) ---------------------------------------------------------------------- My calculation formula for 100K and 27K divider: float voltage = (float)analogRead(36) / 4095 * 15 * 3300 /2883; Check #1 for 15V, 100K and 27K divider, ~200K ESP32 input impedance Voltage Source (VS) 15V Resistance 1 (R1) 100K Resistance 2 (R2) 27K!!200K=23K789 Output Voltage (VOUT) 2.883V (Voltage Divider Calculator) In this case analogRead=2,883V/3.3V*4095=3578 float voltage = 3578 / 4095 * 15 * 3300 / 2883 = 15.001V (~15V!!!) Check #2 for 12V, 100K and 27K divider, ~200K ESP32 input impedance Voltage Source (VS) 12V Resistance 1 (R1) 100K Resistance 2 (R2) 27K!!200K=23K789 Output Voltage (VOUT) 2.306V (Voltage Divider Calculator) In this case analogRead=2.306V/3.3V*4095=2862 float voltage = 2862 / 4095 * 15 * 3300 / 2883 = 11.999V (~12V!!!)

@@gora6045 but the input impedance is both non-linear and its equivalent circuit is not a parallel impedance but series.