Servo lib has 15ms delay since ESP8266 V3.0.0

[makerobotics]

Basic Infos

• [x ] This issue complies with the issue POLICY doc.
• [x ] I have read the documentation at readthedocs and the issue is not addressed there.
• [x ] I have tested that the issue is present in current master branch (aka latest git).
• [x ] I have searched the issue tracker for a similar issue.
• [x ] I have filled out all fields below.

Platform

• Hardware: [ESP8266]
• Core Version: [3.0.0]
• Development Env: [Arduino IDE]
• Operating System: [Ubuntu]

Settings in IDE

• Module: [Wemos D1 mini r2]
• Flash Size: [4MB]
• lwip Variant: [v2 Lower Memory]
• Reset Method: [nodemcu]
• Flash Frequency: [40Mhz]
• CPU Frequency: [80Mhz]
• Upload Using: [SERIAL]
• Upload Speed: [115200] (serial upload only)

Problem Description

I am programming a useless box with a Wemos D1 mini.
I use a servo : "myservo.write(pos); "
Since the board update to V3.0.0, there is a delay in the servo write instruction (15ms as I could trace).
If I go back to a later board version, this problem is no more present. The servo lib version is not changing anything.

MCVE Sketch

#include <Arduino.h>
#include <Servo.h>

Servo myservo;  // create servo object to control a servo
unsigned long timestamps[4];

void scenarioStepMoveToContact(int delayscenarioStep){
//  timestamps[0] = millis();
  pos--;
  //Serial.println(pos);
//  timestamps[1] = millis();
  myservo.write(pos);                 // tell servo to go to position in variable 'pos'
//  timestamps[2] = millis();
  delay(delayscenarioStep);
  if(!isPushed()){
    Serial.println("Reached contact");
    scenarioStep++;
    btnLock = 0; // unlock next cycle
    Serial.println("unlock");
  }
  /* Safety */
  else if(pos<ENDPOS){
    Serial.println("Reached end position");
    btnLock = 0; // unlock next cycle [DEBUG]
    Serial.println("Unlock");
    scenarioStep++;
  }
//  timestamps[3] = millis();
}

void setup() {
  myservo.attach(SERVO);        // attaches the servo on GIOxx to the servo object
  myservo.write(SERVO_STANDBY); // start servo at STANDBY deg

}

void loop() {
  // this is not the real code, but enough to understand the example...
  scenarioStepMoveToContact(5);
}

Time between the timestamps is 15ms as I use the V3.0.0 board.

[dok-net]

there is a delay in the servo write instruction (15ms as I could trace).

Servo.h:

#define REFRESH_INTERVAL            20000 // classic default period to refresh servos in microseconds

This is according to servo specification, and has been in the code for at least 6 years. If you take some extra time to update your code snippet to an actual MCVE, one that must compile (verifiable), and isn't full of commented lines (minimum), your chances of getting support are much better.

When planning issue reports in the future, please consider:

If I go back to a later board version, this problem is no more present.

Your use of "Later" is probably lost in translation, you mean an earlier version, but please always be specific when reporting issues.

The servo lib version is not changing anything.

This applies to a few other libraries as well, so please check first, then try, then report:
There is no such thing as a "Servo lib version" for ESP8266, as the "library" is part of the ESP8266 Core. Probably it's even safer to uninstall any Servo lib from the Arduino library manager while building for the ESP8266.

[makerobotics]

Hello, thanks for your feedback and sorry for not being so professional. It is my first time... I now created a MCVE (had to google the meaning first). It is directly based on the official servo sketch which I only extended with some trace information: ```cpp #include <Servo.h> #define SERVO 16 Servo myservo; // create servo object to control a servo unsigned long timestamps[4]; int pos = 0; void setup() { Serial.begin(115200); myservo.attach(SERVO); // attaches the servo on GIOxx to the servo object myservo.write(0); delay(500); // wait for servo to be in position } void loop() { for (pos = 0; pos <= 180; pos += 1) { // goes from 0 degrees to 180 degrees // in steps of 1 degree timestamps[0] = millis(); myservo.write(pos); // tell servo to go to position in variable 'pos' timestamps[1] = millis(); delay(5); // waits 5ms for the servo to reach the position timestamps[2] = millis(); Serial.print(timestamps[0]);Serial.print(" "); Serial.print(timestamps[1]-timestamps[0]);Serial.print(" "); Serial.print(timestamps[2]-timestamps[1]);Serial.println(" "); } for (pos = 180; pos >= 0; pos -= 1) { // goes from 180 degrees to 0 degrees myservo.write(pos); // tell servo to go to position in variable 'pos' delay(5); // waits 5ms for the servo to reach the position } } ``` Here is the trace: 571 0 5 576 15 5 596 15 5 616 15 5 636 15 5 656 15 5 676 15 5 696 15 5 716 15 5 736 15 5 756 15 5 776 15 5 796 15 5 816 15 5 836 15 5 856 15 5 876 15 5 896 15 5 916 15 5 936 15 5 956 15 5 976 15 5 996 15 5 1016 15 5 1036 15 5 1056 15 5 You can see the timestamp, the servo write (15ms) and then the delay(5ms). I agree that this timing is out of spec (refresh should be only after 20ms) but it is definitely a different behavior than before. I was sending the write instructions every 5 ms and had a fast servo movement. Now, it is limited by this 20ms. If I use the V3.0.0 of the ESP8266, I am forced to stick to the spec (20ms is included in the servo.write instruction). If I use an older version like V2.7.3, I can also have a faster actuation as I directly leave the servo.write instruction. It is up to you to decide if it is better now (on spec) or before (faster), but for many projects like for mine, it may be a problem as it is changing the system behavior. Thanks again for your time. Le ven. 28 mai 2021 à 11:16, Dirk O. Kaar ***@***.***> a écrit :

…

there is a delay in the servo write instruction (15ms as I could trace). Servo.h: #define REFRESH_INTERVAL 20000 // classic default period to refresh servos in microseconds This is according to servo specification, and has been in the code for at least 6 years. If you take some extra time to update your code snippet to an actual MCVE, one that must compile (verifiable), and isn't full of commented lines (minimum), your chances of getting support are much better. When planning issue reports in the future, please consider: If I go back to a later board version, this problem is no more present. Your use of "Later" is probably lost in translation, you mean an earlier version, but please always be specific when reporting issues. The servo lib version is not changing anything. This applies to a few other libraries as well, so please check first, then try, then report: There is no such thing as a "Servo lib version" for ESP8266, as the "library" is part of the ESP8266 Core. Probably it's even safer to uninstall any Servo lib from the Arduino library manager while building for the ESP8266. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#8081 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/APBLAUYUUJ6IUAIAQXRIAITTP5NNXANCNFSM45UWTNDA> .

[dok-net]

@makerobotics Please fix the code block, we can't copy & paste without editing all the line breaks by hand.

[d-a-v]

I tried to fix as I usually do but in this case: "Email replies do not support Markdown"
…
(so I edit my post and do the copy-paste myself)

#include <Servo.h>

#define SERVO           16

Servo myservo;  // create servo object to control a servo
unsigned long timestamps[4];
int pos = 0;

void setup() {
  Serial.begin(115200);
  myservo.attach(SERVO);        // attaches the servo on GIOxx to the servo object
  myservo.write(0);
  delay(500);                   // wait for servo to be in position
}

void loop() {
  for (pos = 0; pos <= 180; pos += 1) { // goes from 0 degrees to 180 degrees in steps of 1 degree
    timestamps[0] = millis();
    myservo.write(pos);              // tell servo to go to position in variable 'pos'
    timestamps[1] = millis();
    delay(5);                       // waits 5ms for the servo to reach the position
    timestamps[2] = millis();
    Serial.print(timestamps[0]);Serial.print(" ");
    Serial.print(timestamps[1]-timestamps[0]);Serial.print(" ");
    Serial.print(timestamps[2]-timestamps[1]);Serial.println(" ");
  }
  for (pos = 180; pos >= 0; pos -= 1) { // goes from 180 degrees to 0 degrees
    myservo.write(pos);              // tell servo to go to position in variable 'pos'
    delay(5);                       // waits 5ms for the servo to reach the position
  }
}

[dok-net]

@makerobotics You must have hit "the wrong button" :-)
Anyway: For an immediate fix, you can switch over to another waveform generator (full disclosure: I am the author of that alternative implementation).
Like so (adding enablePhaseLockedWaveform()):

void setup() {
   enablePhaseLockedWaveform();
   Serial.begin(115200);
   myservo.attach(SERVO);        // attaches the servo on GIOxx to the servo object
       myservo.write(0);
   delay(500);                   // wait for servo to be in position
}

[dok-net]

@earlephilhower I have only tested that the "phase locked" generator gives instant returns from servo.write, no more, no less, and I don't really know why "pwm locked" apparently delays for up to 20ms (15 = 20 - 5). Please take it from here?

[dok-net]

A bit more analysis for the benefit of my pending PRs, to whom it may concern. This issue remains in PR #8008 ("pwm locked"). See how compatible it is ;-)

[dok-net]

1100 0 5 
1105 1 5 
1111 0 5 
1116 0 5 
1121 0 5 
1126 0 5 
1131 0 5 
1136 0 5 
1141 0 5 
1146 0 5 
1151 0 5 
1156 0 5 
1161 0 5 
1166 0 5 
1171 0 5 
1176 1 5 

Caveat: I haven't attached a servo, don't know if this would go up in flames...

[Tech-TX]

@dok-net The servo doesn't flambé, but it does move about 3-4 times faster with your branch, Dirk.

edit: This may also explain what I saw with my PWM test routines a month ago before I disappeared: the analogWrite() ramp-up ramp-down was running a lot slower than previously.

Minor comment: my servo only goes through about 90 degrees of arc (not 180), typical for an aircraft or hobby servo.

[Tech-TX]

@makerobotics your code isn't doing what you've programmed it for, at least not exactly. You're trying to update the servo every 5mS, but the period of a 50Hz servo signal is 20mS. When it's working 'correctly', the ESP is actually updating the servo position by 4 degrees every 20mS, and not by 1 degree every 5mS. You can't update the servo signal DURING the cycle, only when it starts a new cycle or else it would jitter horribly.

[makerobotics]

As I told in my previous answer, I agree on that now. Honestly, it was not clear to me. But the behaviour was charged. In previous version, my code was working fine even by calling the function every 5 ms. I can change my code accordingly, but I guess it was easier before, for someone like me without knowing the protocol. Thanks for taking care of this.

…

On Mon, 31 May 2021, 02:26 StanJ, ***@***.***> wrote: @makerobotics <https://github.com/makerobotics> your code isn't doing what you've programmed it for, at least *not exactly*. You're trying to update the servo every 5mS, but the period of a 50Hz servo signal is 20mS. When it's working 'correctly', the ESP is actually updating the servo position by 4 degrees every 20mS, and not by 1 degree every 5mS. You can't update the servo signal DURING the cycle, only when it starts a new cycle or else it would jitter horribly. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#8081 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/APBLAUYEZIDWFDCKSG372NDTQLJU3ANCNFSM45UWTNDA> .

[Tech-TX]

No worries, we're looking to see where the strange behavior comes in.

If you're using 'digital' servos (like the tail rotor on a helicopter) you could be running a 5mS refresh rate, but you'd have to change the refresh rate in servo.h,

#define REFRESH_INTERVAL    20000     // minimum time to refresh servos in microseconds 

For modern RC systems, that refresh rate can be anything from 50Hz (ancient analog spec) up to about 333Hz maximum. 5ms is a 200Hz refresh rate, which might work depending on your servos. Some servos will chatter or jitter above ~ 100Hz, some work OK up to 333Hz.

If you add the line in setup that dok-net mentioned above, the servo function doesn't block like the default version does,

enablePhaseLockedWaveform();

so you'll get the expected motion, depending on your refresh rate: at 200Hz it'll step in 1 degree increments, and at 50Hz it'll step in 4 degree increments. The underlying issue should be corrected in the git version of the repo sometime soon, but it could be months before it's rolled into the board manager version. In the meantime, the enablePhaseLockedWaveform(); should fix you up.

[makerobotics]

Thank you all for the kind help.
I could not find the documentation about enablePhaseLockedWaveform(), but I will try it soon.
I also noticed a difference in the servo positions. I mounted a 180° servo, and the positions are not the same with the V3.0.0.
Is it locked to 0-90° ?
Unfortunately, my hardware is not available currently to compare on this level. I will investigate as soon as I have my HW back.

[dok-net]

Find out the extreme min and max and set them in the respective call to to the servo channel. You'll know what I mean if you look at the header file.