The Offshore Voyaging Reference Site

How To Tune an Autopilot to Steer Better—Part 1

In the last article on the subject I explored autopilot options and particularly the difference between truly smart autopilots and the vast majority of units that, despite the extravagant claims of their makers, are not a patch on a half-way decent human helmsperson or, for that matter, when under sail, a vane gear.

I also looked at why good steering matters to us cruisers, particularly offshore in waves, and rather brutally explained why, for most of us, our own steering skills are not a reliable benchmark when assessing how good a job our autopilot is doing.

Of course, one way to fix poor autopilot steering is to buy a truly smart autopilot but, as we saw in that article, these things are silly-expensive, to the point that anyone with budget limitations (most of us) would be better off spending the extra money on more important things like, for example: new sails, standing and/or running rigging, etc.

And, further, many of us have older, deeply stupid autopilots that have years of life left in them—our old Simrad went nearly 30 years and well over 100,000 miles—so it would be good to make these pilots steer better, rather than throw them away.

The good news is that, once we have recognized the crappy steering of most autopilots, there’s a lot we can do to improve it, and the best news is that so doing does not cost a penny.

That’s what this two-part article is about, based on Phyllis and my experience of making deeply stupid autopilots steer reasonably well for well over 100,000 mostly offshore miles.

How Autopilots Work

Before we improve autopilot steering, though, we need to understand how autopilots work.

The good news is that an autopilot only makes four easy-to-understand changes to its steering responses (along with some secondary ones):

  • Gain: The amount and how fast the pilot changes the rudder angle when the boat is off course.
  • Counter-rudder: The amount and how fast the pilot changes the rudder back to neutral and a little past it when the boat has responded to the initial steering correction.
  • Deadband: The amount the autopilot ignores small deviations in course before stepping in with a rudder adjustment.
  • Trim: The amount the pilot biases the rudder angle neutral point to take care of weather or (heaven forbid) lee helm.

Different vendors will have different names but that’s all there is.

Initial Settings

Most (probably all these days) autopilots have some sort of initial sea-trial algorithm that requires us to motor in smooth water at cruising speed while the pilot experiments with ever-larger rudder-angle changes, while monitoring the heading sensor to see how the boat reacts, and then sets values for the first two out of the four variables above.

After that, really smart autopilots use a separate computer module (as we discussed in the last article) to vary all four parameters continuously as conditions change and taking into account the boat’s characteristics.

These complex systems also require days of calibration and tuning to reach their full potential, another reason that many of us should just say no.

That’s the last time I will refer to smart autopilots in this article since this is about making stupid autopilots steer better, not spending great gobs of your money.

Why We Need To Step In

However, most autopilots never change how they respond to an off-course situation after initial sea trials, and even if they do they generally suck at it.

Just think about this for a moment.

If we don’t do something about it, the two most important settings on most autopilots will stay as they were set in smooth water inshore while motoring…forever!

Now think about the last time you steered in big waves offshore, particularly on a broad reach. Was that anything like steering the same boat on a sheltered bay under power? Clearly not.

But not to worry, there’s a lot we can do to give these dim bulbs a helping hand.

This Works

And if you are wondering if reading further is worth it, never mind actually tuning your autopilot, I can assure you it is:

A buddy of mine and I won our class in the Newport Bermuda Race twice, and had the best corrected time in fleet once—beat all the fully crewed race boats—using a deeply stupid autopilot, but with one huge benefit: it had knobs to control…you guessed it: gain, trim and deadband, which we constantly tweaked as the conditions changed—it would have been even better if it had a counter-rudder control knob.

We were, in effect, taking the place of the extra brain and inputs that super-expensive race autopilots have these days.

Ok, I admit it, that was not all there was to winning: We also trimmed sails like fiends and worked out the optimal true-wind angles to maximize VMG to Bermuda from the boat’s polar curves twice an hour (when applicable)…graphically by hand…while constantly monitoring water temperature and current to optimize our Gulf Stream crossing. Double-handed racing is a lot of work! And also the most fun I have ever had offshore racing, but that’s another article.

And while I’m certainly not suggesting that cruisers need to go that far, a lot of the speed we enjoyed, as well as electricity savings and increased comfort, can be had in return for a fraction of that workload.

And so doing also has the advantage of keeping us awake and making long watches more fun and much less interminable, without being anywhere near as exhausting as hand steering.

That said, the bad news is that I’m pretty sure most cruisers don’t adjust their autopilots these days.


Because, in the interest of making autopilot control heads smaller and cheaper to make, the manufacturers removed the knobs and buried the adjustment in the menus, and further exacerbated the problem by convincing us with all that marketing BS that our autopilots are taking care of all that for us.

But even without knobs we can make most autopilots steer way better using the menu settings. Let’s do it.

Once again, I’m going to use the B&G NAC-2/3 autopilots as an example, but the fundamentals will apply to tuning most autopilots.

Those of you who want to follow along in detail can download the manual here.

Make It Fun

To make this interesting, let’s pretend we are setting off on a passage (not racing) from Newport to Bermuda.

And to make it even more useful, I will work in some sail-trim tips that will get us to Bermuda faster and make the trip more comfortable.

Starting Point

Before we leave the dock, and after doing the sea trials as specified in the manual, we need to write down the settings that the pilot came up with for both the high- and low-speed settings (if present). That way, if we totally screw things up, we can get back to an at least halfway useable set of parameters.

We will get into what the high- and low-speed settings do later. Hint, ignore the B&G manual on this one, at least when under sail.

Note that for the purposes of this article I’m assuming that the autopilot on our pretend boat is properly installed with an adequately powerful and fast-enough-acting actuator, and that all dock and sea-trial settings have been done as detailed in the manual.

Broad Reaching

OK, with all that blather out of the way, let’s go to sea. That super-chicken and ginger ale (Run & Ginger) is awaiting us in Bermuda.¹

¹Most Bermudians don’t actually drink Dark and Stormies, we just feed ’em to the tourists, just like few Newfoundlanders actually drink Screech.

If we are smart about weather windows, we have probably waited in Newport for a cold front to go through and immediately departed on a broad reach in a brisk and gusty west/northwest wind.

We get the sails up and get clear of the land, where the sea starts to build and, before we know it, the autopilot is steering erratically and burning battery capacity like crazy.

Worse still, the resulting motion is making everyone aboard rather uncertain about their breakfast choices as well as making it dangerous to move around the boat.

Not to worry, we can make this better:

Login to continue reading (scroll down)

Inline Feedbacks
View all comments
James Chase

This is fantastic! I just wish it had been available to read BEFORE my family and I set off on our TransAtlantic Loop last year,… I definitely made all the wrong assumptions about my NAC-3, that you have pointed out in these articles, including the idea that it would continue to “learn” my boat as it was used.

Will you be including any information on setting up the autopilot, if using it together WITH a self-steering vane? We primarily used our Hydrovane to do most of the steering on our trip, usually with the autopilot off- but there were times that I had both engaged at the same time- (eg. when one of my teenaged daughters were on solo nightwatch in gusty/wavy conditions, and I didn’t want to risk having the boat round up on them),but I never really optimized that setup.

Ben Logsdon

I just installed a new Raymarine Evolution system on my Sabre 34. I immediately opened the manual to make sure the settings listed were actually available. They are! “Deadband” = “Rudder Damping” in Raymarine-speak. Looking forward to tuning it this season. Thank you for the timely and informative article series.

Rob Gill

Hi Ben, we may be on a similar journey. Couldn’t find how to enable rudder damping, it wasn’t there in my controller manual or on the control head. But on the official Raymarine support forum in comments by a Raymarine tech, I found this …

“Using the p70r control unit, access SET-UP/ AUTOPILOT CALIBRATION/ VESSEL SETTINGS/ CALIBRATION LOCK, turn Calibration lock OFF and then return to AUTOPILOT CALIBRATION/ DRIVE SETTINGS and you will find RUDDER DAMPING is now accessible.”

Elsewhere, from the same forum and tech guy, our dockside setup should have a helm, hardover -> hardover response time of around 10 seconds, to react fast enough with sailing downhill in large waves, but not being too twitchy and over-working. So my next job is to time this and figure out how to adjust it if needed.

But individually controlling gain and counter rudder, I haven’t found. From what I have now learned, they seem to be packaged together by Raymarine in a combination of helm sensitivity (Leisure, Cruise, Performance), and in set-up for vessel Type (Sail & Slow-turn sail).

If you do find out where to adjust gain and counter-rudder separately, please share!

Best regards, Rob

Michael Jack

Hi, Rob. I am also on the same journey. Just had the Evo installed yesterday and thanks to John’s great article I am playing with it (much to the dismay of my installer who is reliable but might believe the automagic BS from Raymarine a bit too much). Please keep us updated on anything you find out.

Eric Nordquist

Thanks so much for posting this, couldn’t find ‘deadband’ on my SIMRAD AP44 and NAC-3 Computer. This did it, even though mine is slightly different.

Rob Gill

Such a good article John. Great information, explained in your unique style. Looking forward to Part 2, which I hope will help me understand how to optimise what we can control with the Raymarine setup.

Look forward to experimenting with our newly found variable, “rudder damping AKA dead-band.” Many thanks. Rob

Rob Thompson

Thanks John, the timing of this couldn’t be better for us as we are heading down the home stretch on the launch of our new custom cat. We have purposely left the final decisions on some electronics until the last stage, mostly to not get gear that is not up to latest improvements, with autopilot among them.
I guess if the B&G 9 axis electronic compass adds no value to a NAC 3 installation, we could pass on that and substitute another brand or model of compass/ heading sensor? Would that be advisable, or not?

Rob Thompson

OK thanks for the feedback on the compass.

Eric Meury

hello. We have dual Raymarine 6000 Auto pilot Controls. I do not believe there is a way to change the deadband much other than the change the “Response” to one of three settings. i believe they call it Auto Sea State. Level 1-3

Eric Meury

im sorry it is the Raymarine 6002 that we have with dual ACU300.(copy and pasted) The response level controls the relationship between the autopilot’s course keeping accuracy and the amount of helm/drive activity. • Response Level 1, AutoSeastate (Automatic Deadband), causes the autopilot to gradually ignore repetitive movements of the vessel and only react to true variations in course. This provides the best compromise between power consumption and course keeping accuracy, and is the default calibration setting. • Response Level 2 (Minimum Deadband) provides the tightest course keeping possible. However, tighter course keeping results in increased power consumption and drive unit activity. • Response Level 3 (also Minimum Deadband) provides the tightest course keeping possible by introducing yaw damping.

Eric Meury

yes, there are no other settings.

Drew Frye

They’ve been using the same basic mathamatical parameters to tune refinery (and other) controls for 100 years, but with different names. They’re called PID controllers.

  • Proportional
  • Integral
  • Derivative
  • Offset

And they still work.

Good article.

Drew Frye

Originally, they were not digital or even electric. Air pressure and springs (proportional or gain), accumulators or clockworks (integral), screws (offset or trim), pressure relief valves and clutches (derivative or deadband), combining to open valves. When I first started I was often repairing these mechanisms from the 50s. Tedious and you could never find parts. But curiously, the math was roughly the same. Digital is better!

William Murdoch

Ah, PID controllers… Foxboro and Taylor…

Proportional – How far are we off our heading?
Integral – How long and far have we been off our heading?
Derivative – How fast is the heading changing?
Deadband – How much error is too little to worry about?

First there was air, orifices, flappers, and bellows. Then there was electricity and simple electronics. Now there are programs that learn. They change their constants as they “learn” what works and what does not. They can monitor more than one one variable, determine the influence (if any) of each variable on the output, and provide an ever changing set of constants for each variable.

It was a big change over my career.

John Visser

I just subscribed on the basis of a recommendation to read this article series. I’m a bit obsessed with autopilots, but unfortunately don’t know as much as you! I would like to know, what old Simrad pilot are you using as the basis of this series? I have the AC42 with AP24 heads, and Zeus2 MFDs. Thanks, jv