why does no one care about the totally wrong optical HR-measurments?

It doesn't seem to be the case with my Garmin Instinct. It works reliably all day round, and regardless of the activity. If the HR changes strongly with your arm movement, it is certainly not intentional or designed. The HR is measured optically, monitoring the light reflected from the skin, and measuring the changes of the blood stream. If ambient light enters under the watch (the strap being too loose), it may lead to false HR detection. If the strap is too tight, it may lead to a false detection too, due to a limited blood stream. There are also other factors playing a role - skin tone, tattoos, periferal vasoconstriction, medical issues, and more.

Read up the following document to find out how to improve the HR reading (from the page #6 to the end of the document):

https://images-eu.ssl-images-amazon.com/images/I/A1c0MzlpaBS.pdf

ok, I didnt have the Instinct available, but all Forerunners, Fenix, Vivoactive, Venu and many more.

The rest of your anser is wrong. It is statistically proofen, that there is a direct coupeling between arm-movement and reported HR-frequency. There is nothing that could help. This is true for everyone Ive asking about this issue and true for every watch and type of movement and so on.

The only question remaining for me: what is different with the Instinct watch... if it really does work?

1271255 said:

It is statistically proofen, that there is a direct coupeling between arm-movement and reported HR-frequency.

If it is statistically proven by you, it does not mean it is true. The arm movement will certainly raise your HR to certain extent, and it will also raise the blood stream in the arm, and make it irregular, which can lead to some false detection, especially at persons, already having problems with a reliable HR detection (for whatever reason, whether it is the fit, skin tone, medical condition, or other). But principally, the accelerometer measuring the arm movement is in no way involved in the HR detection, unlike what you seem to be suggesting. If the arm movement causes false HR detection in your case, it means the blood stream varies significantly, and the variations are captured by the HR sensor. As I wrote, read up the suggested document, it will help you understanding better the principles, and the tips may help you improving the detection.

The optical detection on extremities has its limitations. If you need precise HR monitoring, I advise wearing a HRM belt working with electrical signals, instead of the optical sensors.

I only rely on the watch OHR for general 24/7 HR monitoring when I am not exercising or training. When I do activities like cycling or weight training I always use a chest HR strap. Chest HRM using electrodes to measure heart pulse is always going to be more accurate than the optical method on the wrist.

Absolutely, any serious workout requires a proper heart rate strap for good data.  Not to mention a good chest strap like the Wahoo reports HRV which assists in getting VO2 max.

First: 

explain me why the hear-rate even rises when you take off the watch and place something like leather under the sensor and then shake it (so that the leather does not move relative to the watch). There is nothing the sensor could read from dead leather (no light goes trough), but still the watch reports a heart rate...which is rising. Works not every time and depends on the color of the leather or so, but it clearly indicates a connection to the accelerometer, which you cannot deny. And there are many other situations where the blood stream is clearly not affected by the movement, but the readings are still completely off.

For example, take the watch off after relaxing and do some really heavy sport and then wear the watch again and watch the HR.... your blood stream is normalized at this point and the watch should report the accurate HR... but it needs 3-5 minutes at least until the reported HR matches your actual.

How is this possible? Garmin applies a filter to the HR, it is not allowed to rise faster than 100bpm in 10 minutes without movement detection of the accelerometer. But if the watch is moved, this rise time can be shorter.

Second:

every other watch on the market CAN measure the heart-rate on your wrist with high precision, except the Garmins. So it is clearly possible to do it without limitations. For example the Apple watches have a deviation of only a single digit from the actual heart rate, no matter what you do, even if worn in not optimal positions, jumping around or doing pushups aso..

I understand the working pricipals better then most people, as Im a software developer for embedded systems and sensor-fusion (for robotics). So I know a lot of the inner workings of such a sensor as I have done some similar projects already myself. So dont tell me anything about it, please just belief that my findings are very accurate.

But I see fromt the comments, that most people really dont care about this issue, as they are ready to invest a lot of money for a chest strap, including the effort of wearing such a strap. Why should I do this, when the watch is technically able to measure my HR but is only limited from some very strange filters in the firmware?

1271255 said:

explain me why the hear-rate even rises when you take off the watch and place something like leather under the sensor and then shake it (so that the leather does not move relative to the watch).

I tested it - I taped the sensor, and insulated it carefully. It detects exactly 0 HR regardless how long and how strong I shake it. Do not forget the watch does detect the HR even through rather thick opaque material, because the optical sensor uses several wavelengths, including infrared, penetrating through a variety of materials. You actaully need to insulate it with a piece of metal preventing the radiation of IR, and additionally also a heat-insulating layer preventing the heat conduction through the metal, if you still wear it on the wrist for the shake test. To exclude the influence of the heat entirely, shake the watch without wearing it on the wrist. 

Your results, where you see the correlation between the motion and the detected HR, have nothing to do with the accelerometer, but are caused by the inertial mass of the watch and the consequent variable adhesion, hence variances in the heat conduction, in the penetration of ambient light, and in the light-reflection. Of course, heavy watches like the Fenixes will necessarily react much worse on such shaking, than light models like the Instinct.

This is why I recommended reading the document, where you can learn several tips for improving the reading - including sufficient grip, and appropriate position. For some exercises it is also often better attaching the watch to the ankle instead of the wrist.

1271255 said:

every other watch on the market CAN measure the heart-rate on your wrist with high precision, except the Garmins.

Yes, sure, and that is probably why the web is full of complaints of users of all existing brands, about the unreliability of OHR detection. They certainly all mistook their crappy Garmins for Fitbits, Samsungs, Suuntos, and Apple Watches. Just google the phrase "heart rate detection problems" together with the name of your favorite brand, and you will get as many hits as at Garmin.

You dont get it.

I can predict (!) the amount of deviation the watch will show for a random person. Ive tested this with some friends of mine. Gave them a new watch and additionaly a chest strap connected to my watch and let them do some exercises. I was able to exactly tell the amount of deviation the garmin HR will show to the relatively precise measurement of the chest strap. How am I able to do this? This has nothing to do with the crap link you are referring all the time to. This is purely mathematical and predictable. Would the things you say be true, i would not be able to predict the exact error for a random person... Would I?

Please keep on mind, Im an expert for sensor development, nut just a random guy with no clue about whats happening. 

1271255 said:

Please keep on mind, Im an expert for sensor development, nut just a random guy with no clue about whats happening. 

And despite it you continue claiming nonsense and jump to false conclusions. There is no relation between the accelerometer and the HR algorithm as you claim. As I explained, the false reading of the optical sensor is induced by the hand (and watch) motion due to the changing adhesion, and there are ways how you can improve it. But it is simply a limitation of the optical method, and no intentional distortion of the result by the accelerometer data due to a malicious software design, as you claim.