In Part 1 of this article we set off for Bermuda on a broad reach in a gusty wind and biggish seas, reefed and properly trimmed the sails, and set the autopilot deadband for the conditions.
So now the autopilot is steering much better than when we first turned it on, the crew are looking less green, and the cook has arisen from his bunk, dumped the stew he puked in earlier over the side, and is bustling around the galley heating up a new batch.
Life aboard is way better, but it could be better still if we take the trouble to properly tune the gain and counter rudder (see Part 1) and this is exactly what the skipper orders, only to have the crew whine,
Why do we have to bother, surely a modern autopilot will sense the conditions and automatically adjust these settings for us?
The lazy crew
Did you learn nothing from how much better things got after we changed the deadband, and have you forgotten what we learned from that great article over at AAC? (A pity that lazy lout Harries didn’t publish Part 2 before we left for Bermuda.)
Anyway, have you already forgotten that auto learning only works with super-smart autopilots with separate computers that cost way more than any cruiser would sensibly spend and take weeks of calibration and tuning to work right?
Remember, most autopilots, including many so-called smart ones, only learn the boat’s steering characteristics when we do the sea trial in flat water under power and then never change it.
Does this booming broad reach in big waves look like motoring in flat water to you?
Skipper, while going red in the face and looking around for a handy rope’s end
No Adjustment on Your Pilot?
Several members pointed out in the comments to the last article that some autopilots do not have gain and counter-rudder adjustments—thanks for the heads-up.
But before you give up on this, dig deep into the manuals to make sure these settings are not buried under some advanced settings menu and/or called something else.
If not, it’s still worth reading this article to get more understanding of how pilots steer, particularly under sail offshore.
By the way, while I would not suggest junking a working pilot because it lacks these settings, I would recommend that those with sailboats buy new or replacement pilots that do have them.
And If your pilot only has a coarser version of these adjustments that combines gain and counter rudder —might be called Leisure, Cruise and Performance mode, or some such (Thanks, Rob)—reading on will clarify what the settings do (the names are deceptive) and when to make changes.
More Adjustment
OK, now that the cranky skipper has reminded us of why we are doing this, let’s get the gain and counter-rudder set right…we all know what those are…since we learned that in Part 1…which we all read…right?
Seriously, if you are even slightly hazy on this, or the other stuff the cranky skipper was ranting on about above, please re-read Part 1.
Also, remember that our heroes (of the story) have already reefed and trimmed the sails for the conditions.
Gain
Here are the steps to setting gain right for this booming broad reach:
John,
On a previous boat, a rather twitchy and tender Farr 1220, I had to constantly tweak the AP parameters for different sea conditions. The original fluxgate compass was an electromechanical device with its own lagging response to boat motion so it was impossible to tune it properly for dynamic conditions. After swapping out the old compass for a solid-state compass, tuning became more effective in more dynamic conditions, but still not ideal. For the past seven years, I have been sailing my Outbound 46 fitted with the Raymarine Evo autopilot system and, once having selected the correct settings: Boat type = Sail, Slow Turning, Response = Cruising, I make no further adjustments and the boat steers very well in almost all conditions, even downwind in big overtaking seas using wind-vane steering mode. My days of tweaking are over. The OB46 is a larger and heavier boat, better suited to offshore sailing than my previous boat so the dramatic improvement in AP performance can probably be attributed to both boat characteristics and AP technology. I am not aware if any detailed description of the control algorithms built into the Evo AP have been published.
Chris.
Hi Chris,
Thanks for the report on the Evo and your decision to leave it at that is certainly a valid one.
That said, if it were me, I would want to at least look at what improvements could be made through experimentation, tuning and calibration, particularly given that I find it hard to credit that Raymarine are providing all of the steering intelligence of the B&G H5000 or NKE at a quarter the price. https://www.morganscloud.com/2023/03/26/autopilot-buyers-guide/
The point being that, as far as I know, and also based on input from an industry insider, those are the only pilots that actually auto-learn and vary gain and counter rudder for conditions. So at the very least I would experimenting with the three aggressiveness settings and deadband.
But then I have never used an EVO, so I could be wrong and also as a long term performance sailor I’m much less tolerant of poor, or even not quite good, steering than most.
I can confirm what Chris has found out about the Evolution series autopilots, these are quite smart (or adaptive) autopilots.
While I have not found a good description of what goes on, this is a entry for the Raymarine user forum (https://forum.raymarine.com/showthread.php?tid=7837):
“The Evolution range don’t have a fixed gain. Older or less intelligent pilots had a fixed rule that applied a certain amount of drive output in order to achieve a certain change in course without regard for whether that course change is upwind or down, up a wave or down into a hollow, and of course that means that sometimes they’ll apply too little helm and sometimes too much. Instead, the Evo pilots steer like a human helmsperson does: they use their accelerometers and yaw-rate sensors to feel how the boat is moving in response to the helm and apply more helm or less from one moment to the next as required.”
My experience is that this description is quite correct, very robust performance.
Maybe the Evolution range provide most of the advantage (for normal cruising boats) that H5000 or NKE provide for fast/planing boats racing boats?
Very helpful comments from Chris & Svein,
This is what the product manager for Raymarine told me also. Svein, please would you share what model and size boat you have, and what setting you have your Autopilot on; SAIL / SLOW-TURN or SAIL?
We have a Beneteau 473 (14.5 m) and have the setting on SAIL as we have a large semi-balanced spade rudder and she responds well to wheel movement. But I find we oversteer in large waves downwind. I’m going to be moving my EVO sensor from up forward to near the centre of the boat, and then re-calibrate.
Wondering which settings to use?
Many thanks, Rob
I had the oversteering problem that you have described when I initially set up the Evo AP and selected the SAIL setting, which seemed like the obvious setting. The oversteering is corrected with the SLOW SAIL setting. The Raymarine setup information suggests trying all settings to see which one suits your particular boat. It would appear that the SAIL setting might be more suitable for smaller light weight sail boats and larger heavier sail boats with more turning inertia require the SLOW SAIL setting. Maybe that is obvious now in hindsight.
Hi Chris, thanks for the input. Will definitely try this.
Hi Svein,
Possibly, but then again, B&G make the same sort of claims for the NAC 2/3, so we need to be careful of vendor claims.
That said, my guess would be that since Raymarine don’t have a more expensive offering they may be adding more intelligence at a lower price.
That said, I still think it will be worth experimenting with the three aggressiveness settings and deadband.
To my knowledge, the three aggressiveness settings (response levels of leisure, cruising and performance) controls the deadband (i.e. how much deviation of course is accepted).
The rudder-damping setting has a different purpose/function (at least in Raymarine terms):
https://raymarine.custhelp.com/app/answers/detail/a_id/5479/~/helm-is-oscillating-back-and-forth-when-evolution-pilot-is-in-auto,
“(…) RUDDER DAMPING in the drive calibration menu may need to be increased, especially on larger boats to compensate for the ‘flexing’ of the steering system and system inertia, (this is the correct purpose for rudder damping). Note that RUDDER DAMPING should be kept as low as possible, but should be increased to the appropriate level to improve the drive motor stability. ”
(The Raymarine documentation is a bit thin in places, it has to be admitted.)
I think there are two-three factors that make the Evolution autopilot perform better than older/legacy autopilots:
-the combined heading sensor, accelerometers and rate gyros is a step change in instrumentation
-the control algorithms is more advanced, including higher order terms (think PIDD2-regulator, ‘Proportional’, ‘Integral’, ‘Derivative’ and ‘Second Derivative’ ), and probably coupling between the axes.
-a very robust autolearning algorithm that continuously adapts the parameters in control algorithms.
The only missing feature I would like to see is an easy way of switching between AWA and TWA sailing modes (now it is buried deep in the menus).
I might come across as a Raymarine fanboy .
I am not, I am only very pleased with the performance of a piece of technology that just works, leaving me with time to adjust and trim sails, navigate and keep a sharp lookout.
Hi Svein,
On rudder damping. This is a good example of how careful we have to be when reading autopilot documentation because most of it was written for motorboat autopilots before sailing features were added and by people who have probably never sailed. In fact the B&G manual says almost the same thing.
So with that in mind I’m near certain that “rudder damping” is indeed deadband and can be used as such when sailing in bigger waves and the three modes are a combination of gain and counter rudder.
One question, when you say this just works without any tuning have you tried it offshore in big waves and breeze?
For example earlier in this thread Rob shared that he must go to Performance mode to get the boat sailing safely which pretty much proves that this is a combo of gain and counter rudder.
Hi,
I think it the description of rudder damping makes sense in terms of the mechanics: a rudder connected to a big/heavy cockpit-spanning wheel will have much more inertia (spinning angular momentum) than a lightweight tiller, and the friction (i.e. damping) in the steering system might be low. Accurate rudder positioning is then going to be much more difficult to achieve with a big wheel.
I have indeed tried it offshore with waves, swell and breeze, with robust and predictive performance. Based on this I known it is working as well as I claim (otherwise I would not go on and on about it).
I have not yet tried it with a spinnaker in a breeze, but I will do that this summer, both inshore and offshore. I will report back 🙂 It will be very interesting to see how the AWA vs TWA modes behave (the TWA mode might be more dependent on accurate sensor data). Somewhere this system also has to be overwhelmed, it might be interesting to see when that happens.
I have (briefly) reviewed a couple of the recent papers on small sailboat autopilot development, based on these I am not surprised that it is possible to develop a robust auto tuning algorithm. But by now we are deep into a another rabbit-hole (and there are not enough time for all rabbit-holes).
Hi Svein,
I am pretty sure the Raymarine TWA and TWS rely on your manual impeller log to provide boat speed through the water (rather than GWA which uses GPS positioning for SOG).
Should the impeller get jammed up on passage, or even broken by a log or alike, then I would worry about the effect on the autopilot steering and possible gybe?
We keep our pilot on Auto mode, where we set the course and manually adjust the sails. Or we use wind vane mode & AWA, where we set the sails and the autopilot adjusts the course.
May I ask what benefits you see for using TWA, since sails are normally set to match the apparent wind they are experiencing, not the true wind? Now wondering if there is a Wind Vane – GWA option?
Thanks, Rob
Hi Rob,
You are absolutely right that true wind requirs dead nuts accurate wind and boat speed data. More in the next article.
And all the experts I have been in communications with (including one deeply experienced with the software) have agreed that true wind is the only way to sail on a broad reach in big waves safely. Apparent does not work safely because it changes so much as the boat accelerates and slows on each wave.
Hi.
I just checked the settings for the autopilot, and there is an option to select which speed source to use, either speed through water (the impeller speed sensor), speed over ground (from GPS), or cruise speed (manual input).
If the impeller stops working, then something surely will happen. The question is if it fails gracefully or not.
There is also another setting in the autopilot for preventing gybes (“Gybe Inhibit”) which might reduce the chance of unintended gybes if the impeller stops working, provided the “Gybe Inhibit” only relies on apparent wind (which might be interesting to test, in calm conditions by disabling the impeller speed sensor).
The main reason for trying out the different modes (auto/compass, AWA, TWA) is basically envelope expansion, finding out in which conditions the modes work best, and when does it all break down. As John mentioned, in reaching maybe TWA is preferred, but what is the difference from Auto-mode if the wind is constant?
Hi Svein
Great you are digging into this. A couple of thoughts for you.
Thanks, I will keep that in mind, and maybe report back. My thinking is more along the line that when reaching, there is likely enough power to attain “maximum” speed anyway, and the sails can be trimmed for a wide groove, so heading in the right direction is maybe more important.
By the way, I have found (after extensive searching) some more information on the development of the Raymarine Evolution autopilot.
There is a professor at University of Illinois Naira Hovakimyan who has developed a technology called L1 adaptive controllers:
https://aerospace.illinois.edu/directory/profile/nhovakim
It seems this is the theoretical framework for the autopilot. There is a lot of papers on how these adaptive controllers work, not enough time 🙂
Hi John, feel free to move this information to a more appropriate place.
I think it useful information for many, and there is a few engineers on this site that might appreciate it (I know I would 🙂 ).
Hi Svein,
The more I think about that, the more I think we will have to agree to disagree. With the autopilot ram directly connected to rudder shaft and installed right with no play (as on most sailboats) the only difference a big wheel will make, to my mind at least, is in the power required of the ram to change rudder movement direction, which has nothing to do with deadband. The only time deadband would come into this is if there was play in the system.
Hi.
I do not think there is much to disagree about, after all this is mostly physics.
The description you give of the effect of inertia is that more power is required to achieve a give rudder movement for a system with high inertia is also valid for when the rudder movement should be stopped/braked. More energy needs to be feed into the system and then removed from the system.
More power is the same as applying a force for a longer time period, during which the rudder still moves until coming to rest. If the rudder moves to far it overshoots, and the the autopilot will soon correct back when the heading deviation becomes to large. Preventing the overshoot in rudder positioning is one way of reducing the tendency to “Snaking” (if it the correct term?).
This article from Jefa might give a slightly better description: https://www.jefa.com/steering/products/drives/qanda.htm
I do not know how old the Jefa article is, maybe all modern autopilots have the features described (power electronics has advanced in the later years, after all).
Hi Stein,
I do understand the momentum issue and that deadband can be used to reduce that problem, but that’s different from the use of deadband to reduce unnecessary steering, as I explain in the part one of the article above—same adjustment, different use case.
Hi, I think I have tracked down a patent that covers the Evolution AP, issued to to FLIR Systems (with inventors from Raymarine UK), which describe some of the details. Searching for WO2016109832A2 should find it.
I have not read it (yet), but it might give some more insight into the what the different settings actually do.
Hi John,
As you mention, many will think that this stuff is too complicated or too much hassle, compared to the benefits, which is wrong. It’s certainly worth it. Same goes with improving and maintaining our own helming skills. I think you’re right that coming from racing means an increased sensitivity to useful details and decreased tolerance for poor boat behavior, but that it’s very easy to learn enough to use the knowledge in these articles well.
Since I’m mostly a multihull sailor, I’d like to mention that they have some significant differences in their autopilot behavior. Some points make it easier and some clearly the opposite. The lack of roll, no “pendulum” is an important benefit, often but not always . The bigger variation of speed, especially on performance boats, giving larger changes of apparent wind angle is an important problem, even potentially dangerous. I guess you’ll discuss this in the next article in context to using the wind rather than compass as the steering reference.
Hi Stein,
Very good point on performance multihulls, that said I’m not sure their needs are much different than performance mono hulls. The key point on both is that only a very smart autopilot can steer these boat safely, particularly with the wind aft, (as you say) and then only if the inputs are really well calibrated so true wind is correct.
That said, I will be sharing a hack to get around that limitation in the next article.
Hi John,
I suspect you won’t cover “Trim” in future articles, so I share the following while we are still on topic.
“Trim” is indeed often left to its default value both because the effects of sub-optimal settings are not as easily noticed as incorrect Deadband, Gain, Counter Rudder, and because well, even this article doesn’t tell us how we could go about it 😉
Getting Trim wrong causes the autopilot to operate around the wrong weather helm: we will be asking Gain and Counter Rudder to deal with an offset they are not designed to manage. Often, the default Trim settings are deliberately set too high by the manufacturer: the advantage is that weather helm is always right in stable conditions. But when things change as they do, the autopilot may not find the new weather helm as quickly as it could, the other parameters will kick in, and there will be more action in the galley. Worse, we could be misled in trying to improve things with less Deadband, or more Gain and Counter Rudder.
So how could we go about it?
This is for autopilots where Trim is automatically calculated over a time interval that is set by the user as you mention in the article. Let’s call this interval “Autotrim”, which is a name used by B&G.
1. Start at a low Autotrim value, well below the default, low enough that you will see the weather helm (on the helm itself or possibly as a value that can be displayed on an instrument) is jerky and unstable.
2. Increase Autotrim until you see stability and consistency in weather helm for the prevailing conditions (you can use bracketing).
3. At this point, to check whether you are not over-averaging, create extra weather helm by over-sheeting the main, and see whether you like the autopilot’s reaction time in terms of settling on the new weather helm. Likewise by easing the main sheet. If it is too slow to adjust, reduce Autotrim, but keep the system stable.
I can’t comment on the NAC-3, but on other autopilots from B&G, Autotrim is not an outcome of the sea trial: it is a default value that doesn’t change unless modified by the user.
The main benefit I have found in tuning Autotrim is that it makes tuning Deadband, Gain and Counter Rudder easier and more accurate. And it’s satisfying to see the autopilot adjust weather helm reliably so that it can perform as designed.
Hope this helps.
Jean-Louis
Hi Jean-Louis,
Sounds good to me, and probably worth trying on a NAC-3. I had not thought of the idea that the manufacturer would default too high, but that makes sense now you point it out.
For others: while I think Jean-Louis’ method will be very good on a NAC-3 be aware that he has an H5000, which is a far more sophisticated pilot, so we have not tested this on the NAC-3 or pilots like it.
Not intending to turn this into a Raymarine discussion, but I have come to a dead-end with the Evo when it comes to downwind sailing in almost any size wave (at least the .5 meter to 3 meter waves I have sailed in since I installed it). Given that the Evo pretty much only has the Rudder Damping to adjust now, I have tried it in all positions (1-10) and none have had any discernable effect. The Evo is just working non-stop all the time. I have nothing to complain about when reaching and close-hauled using the settings you suggest John. Any other suggestions from you John (or the other folks who have the new Evo) would be most welcome.
Hi Michael,
Frustrating! First off, are you using it off the wind steering to wind or compass? And does which of the two modes make a difference to the problem?
Hi, John. I haven’t used Wind Vane mode (if that is what you are referring to) yet. So compass steering.
Hi Michael,
OK, are you using the compass that is specified as part of the system, or different source of heading data. Maybe you have more than one source of heading on your network? Also have you experimented with the modes (cruise, sport, or whatever Raymarine call them)?
Hi, John. So I have played with the Leisure, Cruise, Performance settings a bit. I generally leave it on Leisure which is the least responsive setting. None of them make a difference though to this issue. In the interests of you not having to waste time guessing at my settings, below are the screenshots of all the settings I have available. I have have only played with Vessel Hull Type (Sail is the only other appropriate option) and Rudder Damping. In regards to your question on compass, I don’t know the answer and upon reading parts of the manual again to try and find and and looking at some Raymarine forums, I realize that I don’t really know the system well enough. I think I have to do a lot more study before wasting your time. Will get back to you.
Hi Michael,
This may be just the way it is, but others have had good luck with your pilot, so I would indeed look further and problems with the heading data are the first place to look.