On this tug everything goes round the deckhouse because it always seems to get in the way.
I have added a few cm to its wires to ease things a bit.
The replacement pair of H bridges has been constructed, fitted ( or to be honest, squeezed in) and seems to work. So I am back to where I was 2 days ago with a working boat waiting for a bath test before refitting the deck.
Another blog has been talking about RFI problems.
According to my scope I have one too, but its not me. I observed one mode of operation where there seemed to be a little imbalance in the extremes of speed.
This was rapidly discovered to be an error in the first op amp of the circuit.
This has a huge amount of 100MHz on its pins and I reckon that is causing the error.
Unlikely to be the opamp as that is a cheap device capable of just 1MHz.
Perversely, even if I turn off all power supplies its still there! If I remove the opamp its still picking up
So I guess its due to the fact I live very near to a major transmitter and any bit of metal picks up signal of the FM band 88 to 108 MHz.

Well I had a short video showing the Tug wiggling its bottom in the bath. The bath is smaller than the Tug so it could not do a circle.
But although I got it to less than 5MByte it looks like the site won't accept the video. Sorry. Can anyone tell me what to do?
More seriously, it was able to wiggle port and starboard while I used no throttle, just rudder.
However, it did move forward slowly as well, which I guess is because its more efficient going forward than back. No problem, one just gives a twitch back on the throttle, but that does require momentarily reversing the rudder.
I think it would be easier to drive if right rudder on the controller always asked for a clockwise turn rather than the theoretically correct opposite rudder in reverse.
But it is at least all on one controller.
I am not too happy about the fairly low maximum prop speed chosen under rudder control only and will adjust this up. Presently, the max speed trying to turn on the spot is about 45% of full speed, and I think more like 60% would be better
More seriously and completely unrelated, why does one boat use two different hatch locking systems? The front hatch is a compact joy, very flat.
The rear hatch has a complex slide mechanism that sticks down a long way. Its also slightly offset so as to miss the rudder servo. My new battery location at the rear only just works for a flat 6x1 battery, and I think it might be fouling the hatch thus lifting it off the seal.
The 3x2 version will need to be rotated or moved forward a bit.
This is why we do testing, most of these things will be quick to adjust.
The real fix is to do what I always intended, and use 2 lithium cells that are AA size. I have 4 on order but the Chinese Mail appears to have hired a slow snail. I may have to shift the battery location forward a little.

Not about the build....Just to confirm that I double checked re video; definitely uploaded an mp4 file and that is not what I subsequently downloaded. As suggested, once I tracked it down on the tablet, I had to change the extension back to mp4 before I could view it. So something in the loop is altering the file extension. I do hope its not my tablet....

Attached is the start of the circuitry and description. It was a WORD document, then RTF, then plain text but the website would not accept any of these. So I extracted the picture and just cut and pasted into here. PDF here we come? If anything is not clear, ask me or Google the phrase that is not understood. even my PAINT picture seemed to have problems and was originally reported as an invalid file, though that has altered even as I type. You can see it on a Windows computer at least by clicking and selecting one of the options such as Background image. I have now tried an Android tablet and clicking away managed to view the diagram.

Now that the boat has floated, I'm properly starting this design blog. I'm intending to describe the dual ESC and mixer with the aim of teaching some relevant electronics as we go.

The first thing in any design is to define what we are going to do, Then consider what sources we have to manipulate in order to achieve the objective.
In this case we have a Tug that manoeuvres worse than a slug. Windage is a disaster with little keel and a high non central superstructure.

The obvious thing to do is to fit a rudder mixer so rudder operation also changes the speed of the motors to help out, especially in reverse, or even stopped. These are commercially available, but this version of the boat cannot use such commercial gear because the required motor signal is not available. All we have is the present motor drive signal direct on the motors and the servo signal to the rudder, which is conventional. We could scrap everything and start again, but…..

The requirements are:-
When stopped, rudder only will turn the boat by sending the motors in opposite directions, with no throttle applied.
When going backwards, if the rudder is deflected more than about 25% the motors will change speed (be modulated) so as to help create a turn.
When going forward slowly, large rudder signal will change motor speed significantly. However, at full speed, even full rudder will not change the motors much, if at all, because the boat might tip over. In fact, full rudder at full speed can only slow one motor, because full speed cannot be exceeded.
With less than 25% rudder, with throttle applied the boat responds to rudder only.
With throttle applied, no motor may be reversed nor may more than full speed be demanded.

So now let's look at the inputs required.
We need to know Throttle setting (speed) and direction, forward or back. That is not too bad. The motor on this boat version is driven by an H bridge with a forward signal and a backward signal. The repetition period is 8.6 msec and the mark/space ratio determines the speed. So we have to blend these two to get speed and record which of the two is active so as to obtain a forward and back direction signal. If neither is present then speed must be zero and we can derive a Stopped signal.
We also need Rudder deflection and direction. The servo signal containing this information for the rudders is nearly conventional.
A short pulse is repeated every 8.6msec. 1.05msec is full left rudder and 1.95 msec is full right rudder. 1.5 msec is straight. There are several methods of extracting this; I chose a simple filtering and gain method as being a low component count.

The other thing to consider is the battery. These have a distressing habit of having a changing voltage as they discharge. One can work at just less than a minimum voltage, using a voltage regulator. Or take the more aggressive approach and do a design that uses every bit of battery voltage available. I choose the latter because it has less component requirements. This is called a ratiometric technique; all circuit design is done using ratios of the battery voltage at that instant. With one exception we don't generally worry about volts absolutely, at any point we have a proportion of the battery voltage.

So let's look at the input circuit FIG1 that gets these signals for us. It’s actually quite simple. The picture is attached to this post. I hope its readable.

Abbreviations.
Below, for instance SPeeD would be written SPD. I interpret once only, but abbreviations like this usually go in groups of 3 characters.
If you see, for instance, B/3 that means 1/3 or 33% of Battery Voltage
All inputs are changed (limited) to be either zeros or full battery volts. So voltages beforehand do not matter. Protection resistors also make it difficult for a mistake to blow up the existing PCB.
WHITE and YELLOW are the wires presently going to the motors. These are removed and go to this circuit. There is no need to go into the deckhouse PCB.
I just describe the WHITE forward channel. As we have disconnected the normal load, I feel better giving the old circuit a little load, that’s R1. D1, R3 and C1 form a fast attack, slow decay filter; a short pulse will easily keep this charged enough until the next one arrives.
Going down and left, Q1 is an ORed inverter. WHITE OR YELLOW will turn on Q1 so SPeeDPulseWidthModulation is the signal that digitally represents speed in either direction.
R8C3 are used later to tell another circuit about SPeeD in ForWarD only.
U1,2 and 5 are Schmitt trigger inverters. These are funny beasts, but handy. If the input reaches 2B/3 the output goes to zero. It will stay there until the input drops to B/3, when it will go Hi, to full 100% of B. So it gets rid of noise without using lots of filters.
The RUDDER servo signal is extracted by breaking into the white servo wire and grabbing the signal from that. Its generally at only 2.5-3V or so, quite low. Q2 doesn’t care. Anything above 1V causes a LOW on its output, R10, and hence a HI on U5. So there we have the 1.05 to 1.95 msec pulse at a known rate, representing rudder position, now “slaved” as a proportion of our B voltage. With a bit of filtering, we are ready to go.
Finally, AND gate U4 detects that both FWD and BCK are STOPPED. All required signals are present and correct.
It wasn’t that bad was it?

😋Tried the Tug on the pond today. I had fairly windy conditions. Results on the dual ESC generally pleasing. Going forward, as expected, slow turning much better, turns going fast about the same.
Backwards gave best results at medium speeds. This is because a fast motor cannot be made to go faster still and a slow motor is not reversed in that mode; the design precluded that. I still have a list on the boat ( my fault due to being too aggressive with battery position) and going backwards this seems to have an effect as one side being deeper than the other has more water resistance at the stern.
With no throttle and using rudder only to control both motors in opposite sense the Tug generally turns on the spot. It moves maybe half a length as the turn starts and then turns almost on the spot, though large wind gusts slowed things down a little. That deckhouse is huge!
So although this mod does not give 100% motor vector control the improvement to manoeuverability is very significant, allowing decent control of the boat with the original controller. As with all manual control one soon learns the actions which work well.
I previously mentioned the rear hatch. I was a bit discombobulated when at full rudder I saw the hatch lift in the middle.... I had not removed enough of the servo rotor arm. But I had brought my cutters and that soon got sorted.

Before getting into the logic of what we do with the inputs there are two sections to clear up; this is the first. I restrict myself to three main parts at once, as far as interfacing is concerned.

So far, we only have a limited rudder signal , RUDFLT.
This just represents the input pulse length with an analogue signal.
We need to reproduce a similar signal to that inside a rudder servo, where a potentiometer is moved by the rudder servo rotation. However, we need to turn this into two signals, direction and deviation.
The circuit attached has a lower set of Voltage against rudder pulse length diagrams to illustrate what is going on.

I mentioned ratiometric techniques earlier; this circuit uses that technique. B/2 can be thought of as a local ground, so we can have plus and minus around that.

Although RUDFLT has been referred to our battery voltage, B, we still need to amplify and further filter this. The reason for the exact numbers will be explained later, but for now,

rudders centred = 1.5msec input pulse

has to give us B/2. This needs a precise gain, set by R25 and R257. These are actually made up of two or three standard resistors so as to achieve the values.

The gain is quite low, around 2.858, which is near enough to the theoretical value calculated next.

The "perfect" input voltage at the 1.5msec point is B*1.5/8.6
We want B/2, so the required gain is (B/2)/(B*1.5/8.6).
The B drops out so the required gain is 8.6/(2*1.5) =2.866
Nothing is perfect, especially the input limiter previously described, and in practice I had to make a very small adjustment. That is the same as adjusting the rudder position on a controller.

Worse is to come.... When the input signal varies away from the centre B/2 we have to extract the deviation, which is the difference between RUDAMP and B/2 in either direction. We have to add a rectifier. Its actually a standard circuit. The trick is in U2 where we amplify any input less than B/2 by 2; when more than B/2 the amplifier just stops where it is at the point considered. ( I tell the truth in principle, R258 and its friend at 6.8K should really be 10K. But reducing both alters nothing, and enables the amplifier to behave better at low battery voltages). So the actual voltage o/p of this stage is less than indicated in the sketches, but the matching 6.8K puts the correct current into the summing next stage, and that is what matters.

When we add together and invert RUDAMP and D46/R253, the required output is calculated as shown. So this starts near zero, goes up to B/2 and then back down to zero.

Rudder direction is extracted from RUDAMP; when greater than B/2, direction is right rudder, when less than B/2, it's left rudder. I'll show the comparator that does this later, which is the fourth Triangle shape of this bit of the circuit. In practice, there is always a deadband where rudder is central, just as throttle also has deadband and is zero. A nice consequence of this is that actually none of this circuit need be very precise and we can use cheap amplifiers.

It looks a lot of bits, but can actually be one quad amplifier, an LM324 costing about 40pence. I actually use LM328 duals, just to make a bit of space.
Many servos also still use one chip, plus a potentiometer. I can't buy those chips for less than about £6 for 5 of them. And they have quite a few bits round them too.

Next we'll look at the PulseWidthModulator circuit that interfaces this analogue to the digital logic. That logic uses all the inputs that we have extracted from the raw signals of the Tug, which actually gets us near to actually driving the motors.