CS-320 is the Water family's creeper: a cast elastomeric pulsed bell where a propeller would be, a small dry core beneath it and a 2 kg / 2 L cradle below that. No screw, no cavitation, no blade-rate line, no wake. This page is the bell, and everything that follows from it.
Three bodies, one of them dry. A free-flooded elastomeric bell over a rigid polymer chassis, a small dry canister inside it and a cradle underneath: 16.34 kg and 15.95 L wet, rated to 30 m.
A cast elastomer dome contracts on a slow stroke, expels a ring of water through a Ø40 mm nozzle and refills on the recovery, 0.5 to 2 Hz, for days. Propulsor, fairing and contact-compliance layer are one part.
A glass-filled nylon canister with 6 mm domes, holding the 0.2 kWh pack, the compute and the avionics. It is the only volume aboard at one atmosphere.
A mechanical mount, a wet-mate connector and a release, defined once, with the ballast module and the drop weight below it. Every cartridge is a sensor, sampler, relay or inspection tool.
There is no shaft to seal, no gearbox to whine and no blade tip to cavitate, because there is no screw. Everything the water touches is free-flooded moulded polymer off the fleet's own tooling, rPET-eligible where it carries no load: the most-plastic platform in the fleet. The signature posture is the construction, an all-plastic non-metallic body with no screw, no cavitation and a near-ambient thermal plume.
A cast bell and a moulded chassis from the fleet's own tooling make the vehicle cheap enough to field in numbers, and the same mouldings leave little of it to find. Distance is another machine's job: CS-310 runs the long transit on its screw and carries this one in its bay. Both hulls sit side by side on the Water family, and the buyer's view is who it is for.
Two architectures were open at this scale: a small dry canister inside a flooded body, or no air volume anywhere, with potted electronics and pouch cells in oil at ambient pressure. The second was taken seriously and not selected.
Going pressure-tolerant deletes the collapse problem, the penetrator problem and most leak-failure modes in one move. What it also deletes is the air, and on a 16 kg vehicle the air inside the canister is most of the buoyancy budget.
| Architecture | Displaced | Buoyancy | Mass | Net |
|---|---|---|---|---|
| Dry canister with hemispherical ends | 11.19 L | 11.47 kg | 3.62 kg structure + 2.23 kg contents | +5.62 kg |
| Potted and oil-compensated, no air anywhere | 2.30 L | 2.36 kg | 3.40 kg | −1.04 kg |
Delete the air and 6.66 kg of lift has to be bought back in buoyancy material.
| Buoyancy material | Litres needed | Flooded-route total mass | Against the dry canister |
|---|---|---|---|
| Deep syntactic foam, 0.40 g/cm³ | 10.7 L | 7.66 kg | +1.8 L / +1.8 kg |
| 100 m-class syntactic, 0.25 g/cm³ | 8.6 L | 5.55 kg | −0.3 L / −0.3 kg |
| Closed-cell PVC, 0.15 g/cm³ | 7.6 L | 4.54 kg | −1.3 L / −1.3 kg |
At a grade appropriate to 30 m the flooded route is the smaller vehicle, by about a litre and a kilogram.
So the flooded route is not disqualified, and it is carried as a quantified upgrade. What the arithmetic settles is that it is not free: it costs displacement, the one currency this vehicle has least of, and at 30 m it spends that currency on the cheapest problem on the platform. Two millimetres of extra wall already solve it.
The dry canister is selected on serviceability. Potted assemblies are hard to rework and constrain component choice to parts with no air gaps and no hard-case cells; being able to open the box and change a component is worth more than the litre.
External pressure at 30 m is 0.302 MPa. The critical length for a Ø160 mm × 8 mm cylinder is 816 mm and the canister is 450 mm, so it sits in the short-cylinder regime, where the end restraint carries load.
The regime decides the formula and the formula decides the wall. A short cylinder resists external pressure roughly 2.5 times better than the long-cylinder relation predicts, and creep and ovality take all of that back: a polymer under sustained external pressure loses stiffness with time, and real tubes collapse below the perfect-geometry prediction. The closure below is written on the Windenburg–Trilling short-cylinder relation over an effective unsupported length that includes head depth, 450 + 2 × 80/3 = 503 mm, with creep at ×0.50 on modulus and ovality at ×0.70 on pressure.
| Wall | Long cylinder | Short cylinder | After creep and ovality | SF at 30 m | Verdict |
|---|---|---|---|---|---|
| 4 mm | 0.21 MPa | 0.51 MPa | 0.18 MPa | 0.60 | Fails |
| 6 mm | 0.72 MPa | 1.43 MPa | 0.50 MPa | 1.65 | Below a safety factor of 2 |
| 7 mm | 1.15 MPa | 2.10 MPa | 0.74 MPa | 2.44 | Passes thinly |
| 8 mm, selected | 1.71 MPa | 2.94 MPa | 1.03 MPa | 3.41 | Passes |
Ø160 mm canister at 30 m; both knock-downs applied to every row.
Going from 7 mm to 8 mm costs 0.28 kg and takes the safety factor from 2.44 to 3.41, an obvious purchase on a platform whose case is being cheap enough to field in numbers. The 6 mm domes clear at 5.20 MPa after the ×0.25 knock-down a spherical shell earns, so on this machine, unlike on CS-310, the closures do not govern. Every canister is proof-tested before its first wet use.
Take a credible stroke: 0.5 L expelled in 0.30 s through a Ø40 mm nozzle. That is a jet velocity of 1.33 m/s and 2.27 N of momentum thrust while the stroke runs. Mean thrust is that figure multiplied by duty cycle.
Efficiency comes off it twice. Froude efficiency on this jet is 0.203 at 0.15 m/s, and drive and electrical losses roughly halve it again, to 0.10 overall. Against that, the drag the vehicle has to beat is 0.35 N at 0.15 m/s, 0.96 N at 0.25 m/s and 3.84 N at half a metre a second.
| Pulse rate | Duty | Mean thrust | 0.35 N at 0.15 m/s | 0.96 N at 0.25 m/s | 3.84 N at 0.50 m/s |
|---|---|---|---|---|---|
| 1.0 Hz | 30% | 0.68 N | Met | Not met | Not met |
| 1.5 Hz | 45% | 1.02 N | Met | Met | Not met |
| 2.0 Hz | 60% | 1.36 N | Met | Met | Not met |
The stroke above, at an overall propulsive efficiency of 0.10.
The top of the band is 0.25 m/s at 1.5 Hz, and that is the whole speed the machine has. Half a metre a second would need 3.8 times the mean thrust a bell stroke of this size delivers, and the shortfall survives a generous drag figure: at a drag coefficient of 0.30 and a 2 Hz, 60% duty cycle it is still 1.41 times. A small screw would buy the sprint and forfeit the identity the platform exists for, so the mission set is sized around a creeper that has none.
A 0.5 L stroke from a ~2 L bell is a 25% volume change, about 7.7% linear strain on an isotropic conversion, and a 59-hour mission at 1.5 Hz is 0.32 million cycles.
That arithmetic sits comfortably inside the band where elastomers live, and it is the average rather than the number the material answers to. A real bell contracts anisotropically, in a bending and folding mode, and local strain at the crown and root runs several times higher, so the elastomer is specified against the local strain and the moulding's section is shaped around it. Three further properties of a pulsed jet are designed in rather than discovered:
Cold water changes the pulse directly: elastomer stiffens near 0 °C, which moves the stroke and the rate together. The material is chosen against cold water rather than around it, because inland and northern water is where this machine works.
A small submersible is bought twice: once in material, and once in the buoyancy the material consumes. Every line carries a mass and a displaced volume, so displacement is a sum rather than an estimate.
| Item | Mass | Displaced |
|---|---|---|
| Canister, Ø160 × 8 mm × 450 mm, with 6 mm domes | 3.62 kg | 11.19 L |
| Rigid chassis, cradle and fairings | 2.02 kg | 1.50 L |
| Elastomeric bell: propulsor, fairing and contact layer | 1.38 kg | 1.20 L |
| Battery, 0.2 kWh at 140 Wh/kg packaged | 1.43 kg | 0 |
| Avionics, compute and the optional micro-modem | 0.80 kg | 0 |
| Pulser actuator, magnetic coupling and drive | 0.90 kg | 0 |
| Variable ballast, dry and water | 1.00 kg | 0 |
| Moving-mass trim | 0.30 kg | 0 |
| Penetrators and harness | 0.40 kg | 0 |
| Drop-weight abort, cast iron | 0.80 kg | 0.10 L |
| Total | 12.66 kg | 13.99 L |
| Displacement in seawater | 14.34 kg | — |
| Fixed trim and reserve, 12% of displacement | 1.69 kg | — |
| Payload cartridge, neutral by rule | 2.00 kg | — |
| Vehicle wet displacement | 16.34 kg | 15.95 L |
The selected baseline's ledger; items inside the canister displace nothing of their own.
The fresh-to-salt swing on this displacement is 0.35 kg, inside the ±0.5 kg of routine ballast authority, and a separate compensation of up to ~2 kg on payload release is what sizes the system. Handling follows from the same 16 kg: one person launches it by hand from a small boat, a pier or the bank, and a slow soft body is easy to net at the end.
Every wire that crosses a pressure boundary is a leak path, a pressure-test item and a cost line. The mothership needs five electrical crossings. The creeper needs two, plus one water port.
| Crossing | Type | Carries |
|---|---|---|
| J1 · Shore umbilical | Electrical | Charge and data, pre-dive only |
| J2 · Cradle wet-mate | Electrical | Power, data and the release command |
| W1 · Ballast port | Hydraulic | Water in and out |
| — · Pulser drive | Magnetic coupling | Torque only: no wire, no seal, not a penetration |
The circuit and penetrator schedules and the harness itself are a Wired Industries deliverable.
Both the pulser actuator and the ballast pump live dry inside the canister and reach the water through a magnetic coupling and a hydraulic port. A water port is a far easier seal than a multi-conductor electrical penetrator, and the depth cell reads through a diaphragm and needs no penetration at all. This is the small platform's one structural advantage over the larger hull, and it comes from scale.
The abort follows the fleet's inverted doctrine. A normally-energized hold coil retains the drop weight, so cutting power releases it: a flat battery, a hung computer or a severed harness all end the same way, with the weight gone and the vehicle rising. Every other circuit aboard fails safe by being switched off, and so does this one, wired the other way round.
The weight is 0.80 kg of cast iron rather than lead, because every abort leaves it on the bottom, and a vehicle abandoned in a harbour is an environmental failure before it is an economic one. A fault surfaces the machine and brings the core, the battery and the data home.
At creep the propulsion draw is under a watt, so the hotel load governs everything. Drag is taken at a 0.6 drag coefficient over 0.05 m² of frontal area at an overall propulsive efficiency of 0.10, on a 200 Wh pack.
| Silent-running hotel load | Best-range speed | Total draw | Endurance | Still-water range |
|---|---|---|---|---|
| 0.5 W | 0.118 m/s | 0.75 W | 266.7 h | 113 km |
| 1.0 W | 0.148 m/s | 1.50 W | 133.3 h | 71 km |
| 2.0 W | 0.187 m/s | 3.00 W | 66.7 h | 45 km |
| 3.0 W | 0.214 m/s | 4.50 W | 44.4 h | 34 km |
Still water. Each row carries the speed it assumes.
Range scales as hotel load to the power −2/3, so doubling the silent-running draw from 1.0 W to 2.0 W costs 37% of the vehicle's range. Underneath that is a small exact result: at the best-range speed the propulsion draw is always half the hotel draw, whatever the propulsive efficiency happens to be, so efficiency moves the best speed and barely moves the best range.
The consequence orders the engineering. A watt taken out of the sleep state is worth more here than any refinement of hull or propulsor. On CS-310 the cube law on speed dominates; on this machine it barely appears.
At the 0.25 m/s creep the vehicle draws 3.40 W for 59 h and 53 km in still water; at 0.15 m/s it runs 133 h and 71 km. Froude efficiency is itself speed-dependent, at 0.166, 0.203 and 0.317 for 0.12, 0.15 and 0.25 m/s, so the true optimum sits a little above the table's.
The vehicle adds at most 0.25 m/s to whatever the water is already doing. The reachable set after time T is a disc of radius 0.25·T centred U·T downstream, and the bounding cone from the release point is tangent to that disc: the half-angle is arcsin(v/U), not arctan.
| Current set | Reachable cone half-angle | 6 h along-track | 6 h cross-track |
|---|---|---|---|
| 0.20 m/s | Unbounded | 4.3 km | 5.4 km |
| 0.25 m/s | Unbounded | 5.4 km | 5.4 km |
| 0.50 m/s | 30.0° | 10.8 km | 5.4 km |
| 0.75 m/s | 19.5° | 16.2 km | 5.4 km |
| 1.00 m/s | 14.5° | 21.6 km | 5.4 km |
| 1.50 m/s | 9.6° | 32.4 km | 5.4 km |
At or below its own speed the vehicle holds station or works upstream and the set is unbounded in angle: a different regime, not a wider cone.
The arrival error is the other half, and the vehicle does not own it. On a six-hour transit in a 0.75 m/s set, a 10% error in the predicted current speed puts it ±1.62 km off along-track and a 20% error ±3.24 km; a 10° error in predicted direction adds ±2.81 km across. They combine, so kilometre-class arrival error is a floor, and it sits in the forecast rather than in the navigator.
Then the sharper number. At 0.25 m/s in the 0.75 m/s set the design assumes, ground speed is −0.50 m/s. Drift works for getting there; it cannot work for arriving. Three things follow, and all three are design constraints.
Tidal currents are depth-stratified, so the way to steer in a drift is to change depth and select a different stream. That makes the variable-ballast system, rather than the pulser, the primary flight control.
The energy cost of steering that way is close to nothing. Moving 0.1 L against 3.02 bar at a 30% pump efficiency is 101 J per stroke, and 24 strokes over a six-hour drift is 0.67 Wh, or 0.34% of the battery. A drift leg is effectively free, and the bell is held for cross-set corrections and final positioning. That is the drift doctrine in one line: the current is transit energy, and the window comes off the tide table instead of being fought.
The design consequence is a duty requirement rather than an authority one. The pump is sized for cycle count and duty life; one adequate for the ±0.5 kg trim swing but rated for a handful of cycles per mission is the wrong pump for a vehicle that flies on it. Moving-mass trim sets static pitch alongside it.
The cradle takes one module at a time, about 2 kg and 2 L, drawn from six non-kinetic cartridges. Anything larger is the mothership's cargo, not this machine's.
Emplaces a small acoustic or environmental listener for persistent harbour and approach monitoring, at the node scale below the mothership's seabed packages.
Delivers a small marker, transponder or instrumentation tag to the seabed or to fixed civil infrastructure, for sensing.
Drops a small acoustic-network relay below the mothership's moored relay class. The modem hardware itself is a partner's scope.
A water-quality sonde or a small sediment sampler: the civil work the line does, on the same cradle as everything else.
A compact camera, lighting or a small imaging sonar. The creep that makes the machine hard to find also lets it approach a structure without disturbing it.
Low-signature delivery or recovery where a surface presence is unwanted. Treated as export-sensitive in its own right.
The interface is defined once, as a mechanical mount, a connector, retention and release, so a new cartridge is tooling and not redesign; it is a scaled member of the mothership's bay interface family rather than an independent specification. Every cartridge is neutrally buoyant to ±0.2 kg in seawater, which is what makes releasing one trim-neutral.
The list ends there by rule rather than by omission. No warhead, no fuze, no energetic material, no terminal-effect payload and no mechanism whose purpose is harm, in any cartridge and for any customer. Arming the platform is a counsel-first corporate question and never a product decision.
The creeper closes at 16.34 kg and 15.95 L wet, inside the CS-310's 20 kg / 20 L wet bay with 18% margin on mass and 20% on volume.
Carriage is a cartridge fit, not a dock. The creeper rides in a cradle and leaves on a command the bay interface already carries. It is one-way: the mothership transits and releases, and the creeper does the last, quietest leg and surfaces at a planned point for its own recovery.
The neutral-cartridge rule is what makes it cheap. Releasing 16 kg from a 113 kg vehicle would otherwise be a 14% buoyancy transient at the exact moment the larger hull is trying to hold station; because every cartridge is neutral by rule, it is nothing at all.
The reason it matters is the arithmetic above. For much of the mission set, delivery to a surveyed release point is the only thing that makes the delivered-precision tier load-bearing rather than decorative, and it settles the mothership question on engineering grounds rather than taste. The expensive navigation stays on the larger hull. Each vehicle is independently launchable, and this one keeps its other door: one person, by hand, from a small boat or the shore. The mothership's own page is CS-310.
Seawater conductivity kills RF outright, the way rock kills it underground. The fleet's no-radio, no-GPS doctrine costs this platform nothing, because the water enforces it for every side equally.
Pre-programmed legs on magnetic heading, a depth cell, time and a current prediction, with the drift planned against it. Nothing to jam and nothing to spoof; the current term dominates the error budget, so the drift plan is the mission plan.
Guidance over optical fibre from a payout spool, the same mechanism family as the fibre tether on CS-410, and kindest here: no launch shock, walking-pace speeds. Spool volume competes with the 2 L cradle, so this is the inspection tier.
The mothership carries the navigation stack, transits and releases at a surveyed point for a short, bounded final leg: precision by delivery rather than by a navigator a 16 kg vehicle cannot carry.
A micro acoustic modem rides as an option rather than a baseline, and it is never a control channel: every transmission spends the signature budget the platform exists to conserve. Silent running is the default, and human authority is exercised in the mission plan and the abort criteria.
Where the doctrine removes the radio, the design removes the screw, the wake and most of the metal. Human safety runs first here as everywhere: one person launches the machine by hand rather than working over the side of a vessel, the abort is mechanical, and a critical fault surfaces it. The four-domain doctrine is on the doctrine page.
It suits a buyer whose problem is the last kilometre and not the transit, and who wants numbers in the water instead of one vehicle too valuable to risk. It suits nobody looking for a weapon.
If the problem is the close work, a pier or an intake or a listener that has to go somewhere exact, the conversation starts with the machine that does it slowly.
Nothing here is quoted; a screened enquiry is the way in. Bring the water, the tide, the depth band and the thing that has to be placed or looked at. What comes back covers the bell, the cradle and the interface between them. Counsel comes first on any international transfer and permits are taken per shipment, and the emplace-and-retrieve cartridge is treated as export-sensitive in its own right. Every design in the family is patent pending.
The Water family is the two swimming hulls side by side, CS-310 is the mothership that does the distance, and KMM-01 is the uncrewed clearance workboat that works the surface. Who it is for is the buyer's view.
All Water family designs are patent pending; international transfer is export-controlled and subject to Canadian government permits taken per shipment, and nothing on this page is an offer.
