PreCharge over 300V...

[methods]

I was perfectly happy with 120V... but we have a bite on BMS dev... so our customers requirements are our requirements.

Call it ~= 300V

This is for Main Contactor
Precharge Contactor
DC-DC
Switching of DC-DC if required

Looking at Relays first... we see some really high end expensive stuff:
http://www.mouser.com/new/TE-Connectivity/te-mini-k-hv-relay/

Those come in at around $50 in volume but $80 - $100 in ones and twos
http://www.mouser.com/ProductDetail/TE-Connectivity/V23700-C0001-A408/?qs=%2fha2pyFadujboUUQ4aP1vA3DobRVN4G7q7jOoknTHQS5pv7%252bUB%2frFSWfpkrYTgD7
http://www.mouser.com/ProductDetail/TE-Connectivity/V23700-F0002-A408/?qs=%2fha2pyFadujboUUQ4aP1vMQIXJg2w2za4vgeqHipro5jbV4WCiNxv%252b9RyGM2dVi1

Those can only switch up to 20A and are only rated up to 450V DC.
Costs about 2.8W to drive them and they are available in N.O.

CLEARLY DOMINATING THAT OPTION>.....

Would be the Gigavac P series... coming in at Under $50 in low quantity
Switching 1500V under load... with continuous loading up to 50A....
Coil can run at 12V, 24V, and 48V... for a range of 7.5V to 60V... with coil power closer to 1.2W on the low end... in a tiny package.

Gigavac clearly winning.

Next to review:

* Solid State Switching options
* The tiny relay used in the Zero Brick

We will start with the Zero Brick.
A customer brought one by for repair and we took it apart. The part number on the relay is (probably custom )
Its a TE part manufactured in Mexico
12V coil
SUL03

CS01X0120CCB markings...
Only thing that matches up with the formfactor (assuming its COTS) is this part:

http://www.te.com/usa-en/product-4-1618391-0.html
LEV100A4ANH=RELAY,SPST-NO,12VDC
Rated at 100A (hmmmmm )
Smaller than the GIgavac... but get ready for the price (I have not looked yet... you can be surprised with me...)
Rated at 900V

Hmm.. the Zero unit is definitely a one-off... as it has additional features packed into it.
Fine - dont need those... but do need pricing:

$128 on Arrow for similar
https://www.arrow.com/en/products/lev100a4anh/te-connectivity?utm_medium=cpc&utm_source=google&utm_term=lev100a4anh&utm_campaign=US%20-%20SKU%20-%20TE%20Connectivity%20Ltd%20-%20Dynamic%20Inventory&gclid=Cj0KCQjwnubLBRC_ARIsAASsNNk6pxk6HDlHKqPGqPfsHkzb8KwAZXb_fwPnCQwOkNHz_W9jDOcShZcaAmC9EALw_wcB&gclsrc=aw.ds&dclid=CJia5Jv_qdUCFZaTfgodQdAEKw

But then... AN OCTOPART SEARCH IS THE ONLY WAY AN INTELLIGENT HAIRLESS MONKEY WOULD SEARCH... SO...
https://octopart.com/search?q=LEV100A4ANH

At a thousand units we see the price drop to $80USD>.. which makes perfect sense if we are paying Amp for Amp for HV switching...

Call it about a dollar an amp.
$50 gets you 50A
$100 gets you 100A
(Mind you thats ultra compact high voltage switching... not normal ol' switching...)



We have a high level requirement to:

* Switch a 3W contactor with coils between 12V and 120V
* Switch a heavy duty precharge with voltages up to 300V (call it 500V... as at that point it does not really matter...)
* Switch a latching relay on mission critical control lines
* Possibly switch our own primary power (for deadman behavior at ultralow LVC with a bootstrap)
* Possibly direct power handling for Ebike level performance... call it 50A continuous... and perhaps we can combine this with PreCharge as an ultra-robust switch)


My engineering experience tells me that at lower voltages... like 120V... mosfet switching can get VERY APPEALING... and will always have some risk and complexity and heat associated with it.
Its really to be avoided where possible... as it does not scale, either in the voltage or current domain. Its sketch basically.
When it gets up toward hundreds of volts the solid state switches get VERY expensive (i.e. not a buck)

Physical switched contactor solutions meet the full range of voltage requirements.
They are simply limited by I^2*R heat... so running a pair of them in parallel is legit for current handling... Which tells me I NEED TO OVERRATE THE COIL DRIVE TO HANDLE AT LEAST 2 COILS!!!
Coils allow the user to put just about whtaever they want out there for switching...

Coils burn 2 watts
Mosfets burn 2uWatts

Sigh...
Here we go again.
New justification

We are handling power systems in the range of 1KW to 100KW
Ratios are:

1% of a KW is 10W
1% of a 100KW is a KW

So... we are at a control power of 1/5th of one percent of ebike load... and three orders of magnitude less than a percent of a Medium Size EV power...

So for 10KW - 100KW its hands down contactor
For 1KW... lets put it in perspective again.

Average rider: 20wh/mile to 60wh/mile (thats milking it to thrashing it)

2wh will take you 1/10th of a mile to 1/30th of a mile
A mile is about 2000 steps
Efficient guy is looking at taking about 200 steps to pay for his contactor. Inefficient guy may see 66 steps

Efficient guys are going to buy LiGo bricks from Justin... so lets drop them off our radar.
Inefficient guys... they eh... blow stuff up all the time equating to MUCH PEDALING.

So... I am over trying to run an ultra efficient BMS. Justin has that nailed.
I want to run a VERSATILE BMS... that scales epic.

Ok - I have made the argument again and we are not revisiting.
We are going with mechanical switching for all controls.
We are spec'ing parts with a minimum of 200% margin in the voltage domain.
We will push margin in the current domain.

Now - with that assumption... to solve PRECHARGE in an affordable way...

Cars, Trucks, Planes, Trains, Bikes, Motorcycles, Boats, enclosed bikes,,,,,,,,

ENTER SOLID STATE SWITCHING FOR 300V or BETTER HERE: No thanks unless I am paid.
ENTER MECHANICAL SWITCHING FOR 500V at low precharge levels here: He... need to dig hard into my memory for these parts. $20 relays. Look like automotive. Can switch 1000V. Sealed. Low current. Ultra special contact materials.

Speaking of ultra special materials... I was at Advanced Laser the other day... and he was cutting material which I found VERY INTERESTING.
Advanced Laser... he is a good guy. Small outfit - I like that. Send business his way.
Watch out on quotes... if you send it in inches he will quote it in mm... you get a $26 quote on a $650 job... your mistake. Know what the job should cost before getting quotes anywhere.

-methods

 

[methods]

Please Post any links to the ability to:

Switch 500V or better at Precharge levels.

I need to define:

Minimum Viable Precharge Current
Maximum Precharge Current

To solve otherwise:

Resistor selection and heat dump
Reliability around the Precharge switch ... and accompanying Boot-Strap switch... to assure it is momentary or timed... such that it can not physically remain ON... where the drive system will overheat the Precharge resistor... otherwise Precharge resistor will need to be 10X the size and cost.

Not addressed:

Full time strapped Precharge resistor across the contactor (VIABLE OPTION>>>)
PWM Precharge with solid state switches and no resistor or limited resistors (I dont want to get into this as it is a black art no matter what anyone things and I am not a pro at high voltage switching)

The trouble with a strapped resistor is that HV is present at the output at all times (even if current limited)
Leakage to death is a viable outcome
Safety and ISO requirements
Not really an option from a "battery only" perspective. Only from a "systems perspective" and we are looking for the "general solution" here, not the specific solution.

Other options are a simple lockout in the BMS that just WAITS for precharge to be truth... letting the USER mechanically boot-strap the Precharge via a momentary button and large resistor.
This is actually viable for Ebikes... but for $20 or $50... I would just "do it right"

We are not going to be able to appease the masses with low cost on this.
The reason one solution is $300 instead of $30 is flexibility, reliability, and an overall ability to scale.

I want to sell a BMS that will grow with a customer... such that they can buy 2 or 3 and use them on anything...
Use it on a 12V Starter Battery
Use it on a 12V-36V Trolling Battery
Use it on a 36V to 120V Ebike battery
Use it on a 80V to 120V Motorcycle battery
Use it on a 120V to 420V full size EV battery
Use it on a 12V to 400V stationary Power battery for Solar, wind, water, etc
Use it in an airplane... where margin is everything

One size can fit all. But just like the kids in rap videos... for those who are at the bottom of the scale... your pants may sag, your crotch may be 8" low, you may need a really good belt, ...

I really need to aim at the 10KW and up folks to make this viable as a solution that we can amortize engineering across without selling 10,000 units

I am looking at starting with 10's, moving into 100's, and if it happens... partnering with an OEM to move into 1,000's

I am building only the Master and we have a qualified set of Slaves already in production.
We have shake-n-bake lined up for both US and Overseas standards
We are consulting with someone who knows more than us about the minutia of standards (I have the general idea... but have no time or energy to read chapter and verse)

And... my 5yo is ready to go to the beach again.
Here we go.

Rent? Who needs to pay rent... its only like $3k/mo

-methods

 

[methods]

On to HV switching relays

Pickering is an expensive place to start.
109 series on Ebay for cheap
http://www.ebay.com/itm/26-PCS-PICKERING-SERIES-109-109P-1-A-5-635-HIGH-VOLTAGE-REED-RELAY-/152483731128?_trksid=p2385738.m2548.l4275

Data
http://www.pickeringrelay.com/pdfs/109-micro-sil-reed-relays.pdf

Not a match for this, but I have used a lot of their gear in Testers. Switch matrix cards for upward of 500V

Figuring now I will have to make precharge an external module.
Have a low cost 120V version and a higher cost 1000V version

Lower cost will be simple enough.

Higher cost would probably be CAN based.

Total budget for the Master is blown by almost any 500V capable precharge system.

-methods

 

[methods]

Picturing the total cost of a system from the perspective of the range of customers.

Gonna need a master and 1, 2, or 4 slaves.
Gonna need a contactor that is suitable
Gonna need a precharge solution which is suitable
Gonna need a DC-DC or other 12V source which is suitable

Later going to need advanced stuff like isolation measurement, moisture or corrosion detection, leakage of any kind detection, self test which is deterministic and complete

So... for the budget folk...

Master, 2 slaves, $50 contactor, manual precharge, cheap DC-DC
Folks can shop around for used higher quality gear.
My job is to put the hooks in to support whatever the customer has or wants to use.
Find some balance of supporting everything... without having a bunch of different models.

Slaves get potted or built in. Never think of them again.

Master ... would like to make it cost about the same as a slave.
Hard to do that and incorporate safety.

Master can be as simple as the cheapest Arduino with an isoSPI adapter, a few voltage dividers, dont even really need a shunt or current meter.. dont even need a contactor for that matter.
Cheapest version can just cut throttle.
Cheapest version can use heuristics to estimate current draw
Cheapest version only has to handle up to 24S + margin
Could just be a shield that plugs into a Mega or something similar...

IF ONLY... all the major manufacturers settled on the SAME SLAVE COMMUNICATIONS STANDARDS...
Well, then... we could mix and match and drive down cost.

A hundred solutions to one problem - all nearly identical.

-methods

 

[methods]

For the Pickering line, highest power is about 40W... so a couple of amps.
These cap out at switching 200VDC but can stand off 500VDC

We are interested in switching ability only... as they will switch ON hot every time and OFF hot at times.
(rather with a potential)

http://www.pickeringrelay.com/pdfs/114-high-power-sil-reed-relays.pdf

The math for precharge seems simple at first but to get the job done fast enough you want to precharge at a higher rate than anticipated... else the final 10% lags.
PWM solves this but introduces other problems.

An amp should be more than enough to precharge
Especially in a system that remains connected at all times...
But we wont remain connected as that would require constant draw on the primary contactor...
So even if a Sevcon is in the system... we probably will have to deal with on/off duty.

Personally... I dont really want to even see Precharge in a BMS. That should be owned by the Controller component or the System owner.
We are still seeing a lot of stuff on the market that overreaches.
Controllers that try to do everthing
BMS packed with all sort of things that belong elsewhere.

BMS really just needs to be able to cut discharge. To do that it needs to be able to connect discharge. To do that it only needs to be able to read charge across the contacts... not actually make that charge match. Push the requirements out to other components.

IF we are in the system with a Sevcon tho... it is not going to do its "precharge thing" until it sees a potential - so its a chicken before the egg.
In that system I would probably just piggyback a switch into the relay drive circuit... but thats a specific systems solution and not a general solution.

General solution should not have two redundant contactors... so the general solution is dependent on proper systems engineering up front.... which is not yet in existence.... unless you want to accept that the battery can only *request* that its discharge be terminated... as opposed to enforcing termination.

We are early enough still that redundancy that can enforce termination of discharge is probably a good idea.
Systems can solve that on a case by case as time goes on.

I do know that you can spoof the Sevcon into not controlling its own contactor.

-methods

 

[methods]

As for mosfets...

600V range high quality have about 2 orders of magnitude more internal resistance than a 4110.
No switching required... but math pans out quick:

5A^2 * 200mOhms = 5W
5W destroys a TO-220 without dealing with the heat.
5W gets things real hot real quick without a large sink, fan, or other means to get the heat out.

https://www.infineon.com/cms/en/product/power/mosfet/20v-650v-automotive-mosfet/600v-and-650v-n-channel-automotive-mosfet/channel.html?channel=db3a30431a571997011a57d0c4eb0002

In comparison to what we know...

3mOhms * 25 = 75mW
Thats what you would see with a 4110

3mOhms * 20^2 = ABOUT A WATT

And thats about the limit of what you can do with one without proper cooling.
SO... 5 of the 600V units does not sound great... but that would be for 5A precharge

Bastards are a couple five bucks each + drive + heat sinking.
Comparable off the shelf switch starts around $80

See... the only way to hit budgetary targets is to pass the buck :wink:
Not my budget buddy... talk to Systems. They either get it or did not think it thru.

This systems approach will work well to whittle down the scope creep.
I already decided how much these will cost.
Now I just have to figure out how much I need to make over what period of time to offset the engineering... and how many features I can include at a given Manufactured cost (including BOM, Assembly, and Test).

None of that includes logistics and customer support.

So if you are wondering why shit is so expensive... when you buy it from niche shops... its the SUPPORT you are paying for.
Ah....
But we are all so spoiled now with an abundance of free information. Who would actually PAY for help these days?

Few actually.
Most choose to wallow instead At least they learn something along the way.

-methods

 

[methods]

From phone.

Kid is watching brainwash on my PC...

So nearly every relay is rated for 240VAC. Higher AC easy to find. 10A or 20A not hard to find.
We could ASSUME a capacitive load and run a McGhetto relay on the assumption that after it closes it won't have to open until the potential across it is equal.

This is totally reasonable for low cost applications.

It won't address:
* Shorted primary outputs
* Shorted fets in a basic controller
* Differential between depleted pack and dumb HVC charger... For 400V this can look like ... Eh... Call it 100S pack... So 300V on the battery and 420V on the charger... 120V dc differential at the charge current! Could be gnar gnar (tho anyone with a 100S pack would be smart charging... I hope...)

I suppose we could perform a sanity check on the output... But it would be at zero volts... So not much help without the ability to pwm blip the precharge.

PWM on the precharge is not insurmountable... Just prone to success on the bench and failure in the field... If we strap it down well... Protect the drive well... Design it to drive right into a short circuit... Have the driver "learn and remember" the load profile such that on first turn-on it takes it's sweet time and on subsequent activations it makes some assumptions...

Back to our assumptions:. Base Case

BMS must be able to terminate charge or discharge of short circuit
BMS must protect contactor to same reliability as battery
It is optional that BMS performs precharge duties

It would be nice.
It will blow the budget.
It must be an external component... It can not be integrated.

It should run 12V to 120V for our stuff... Or 12V to 420V for OEM stuff...
But wait... An OEM should already be managing their contactor and precharge.
We should have only CAN communication with them.
The BMS would not over-reach... Or should not... But could.

I would like to cap this design at 120V (with margin to 150V at high risk)

I will make it modular such that the 120V design can be adapted to work for a 420V design... But not by design.

So now the connectors, interconnects, and communication costs factor in.

Precharge can be CAN controlled
isoSPI controlled
Voltage controlled
PWM controlled
Solid state switched, mechanical switched, or strapped 24/7

CAN is rad... But very expensive!
CAN chip
uC to unpack
Isolation and regulation and clamping
Connectors

On top of driving, shrugging voltage/heat (the actual precharge circuit)
Encapsulation and connectors

Can is sorta out... Unless we can get costs for the above down to around $5 Manufactured... Which is still in the future.

So... It's onboard or dumb.

These things add up...

-methods

 

[methods]

So we need to address every stupid idea and have arguments prepared so that when we sit down to negotiate requirements we don't get sidelined...

(If your wondering why I keep rehashing fail paths forward)

Also... Sometimes stupid ideas in 2005 become obvious paths forward in 2015 as tech develops... So we must always review all previous assumptions.

Budget.

Budget... This is the easiest argument to make for bounding requirements.
"If ya want it to be a hundred bucks... Then... This is what we gotta do"

Lol

For the lowest form of cheap bastard...

"Grrr.. me BMS no workum. Contactor no close. Ok... Called methods... He said pokum big resistor across contactor leads... Makeum voltages match... Then BMS workum"

"Er... Me figure out just leavum big resistor across contacts... Now BMS always working...but controller left on for 2 months now pack dead. Me sad... Methods fault... Lol"

"Er... Put big poke-um button between resistors and terminal. Now pokeum make start. Boot strap... Works good, no dead battery when controller left on... But button need replace every 6 mo"

"Er... Bought PWM option. Works every time at 120V. Now I want to run 180V. Smoke come out..."

"Er... Bought PWM work up to 420V... Now can't feed family. Power company shut off lights. F-merhods, he is a jerk"

Bored with PWM
Gotta push it off board.
I will look only into a few more relay and mosfet options for off board solution.

On that path...
A box that sits with the contactor
Touches coil, both posts, and has a few wires coming in for control
Works stand alone
Programmable
Bypassable
Moves all that logic and high voltage off the master board
Keeps it at the Contactor

Could possibly incorporate the DC to DC to send out 12V
Would require a thick home run to batt neg
So...

We want to offboard dc-dc, precharge...
We want no hv on our pc if we can help it
We need a 12V source which is lvc smart
Maybe value can be had in combination...

The more we group together the lower the bom gets.

If the nugget already has a tiny uc chip on it... Then it can handle basic CAN or isoSPI. Just a few packets and mostly just dumping.

Modular.... But not too modular.

Meh... Meh... Meh...

So the broke customer just skips it.
Uses the BMS as throttle cut only
As pack value goes up, budget for true protection goes up, so they buy nugget.

Easier just to have rich customers only

Hrm
-methods

 

[methods]

Example:

http://www.evwest.com/catalog/product_info.php?cPath=41&products_id=299

Looks about like what I am talking about.
Interesting they came up with about the same range split

-methods

 

[methods]

Ok... like time being up on an 8hr exam... here is my final answer:

BMS Master shall have 3 open collector outputs.

Contactor Control
12V to 120V
Up to 7A
Shall not drive this until condition is met (Contactor posts are within X volts or percent)
Simple low side only switch

Precharge Control
12V to 120V
up to 7A
To be used to drive an external precharge relay coil OR to directly connect contactor terminals with an external precharge resistor.
Requires a bi-directional mosfet totem driven by a dedicated high side driver (to support direct precharge). If we omit direct precharge then a simple low side driver identical to contactor sink.

Throttle Control
12V to 120V
up to 7A
To be used to drive the coil of a latching relay which shall interrupt the throttle before blowing the main contactor.
Can be re-purposed as a PWM open collector output to limit throttle to X percentage (probably will not implement)
To lower cost this can be small signal... but the price difference is about a buck... so best to keep them as similar as possible for swapping if needed.

Master to bring in AT A MINIMUM the downstream (switched) side of the contactor.
Ideally master also brings in B+ side... but this can be calculated from slave data.
Master to take a 12V power source to enable

DC-DC to be external. Powered from either the downstream side (which requires a boot-strap) or if on the upstream side requires either a sleep input, the ability to shut off the HV input with a dead man circuit... so looks like a 3rd low side driver output... same as the above... to drive the coil of a relay or directly sink current.

BMS to shut down upon low system voltage by cutting throttle, opening contactor, then disabling its DC-DC. 12V goes away if we have control over it. If someone else controls the 12V then we stay on but draw only from the 12V

BMS to take CAN input via the isoSPI buss.

BMS to accept either USB communication via FTDI or Bluetooth via the uart programming port.

BMS to see HV in an isolated section of the board, at low resolution, with a high value divider, clamped to hell... only for precharge confirmation.... and possibly to be able to kill the DC-DC at the HV side... Isolation is highly recommended but not a requirement.

BMS to not give a hoot about power use during ON time. BMS to draw no current after an LVC event. BMS to have a sleep mode which draws microwatts.

BMS to turn be controlled via CAN if it has 12V power.

BMS to have auxiliary inputs, 3 minimum, to allow user buttons for features needed.

BMS to broadcast minimally at first, then aggressively, over CAN, its state.

BMS to do all the things we know a BMS does in the software and firmware domain.

BMS to be noise tolerant up to 400V. BMS to work in a 400V system assuming no direct control and only coil control. BMS to operate directly on any voltage below 120V.

BMS to be IPXX rated with locking connectors and in compliance with blah blah

BMS master cost to be on par (within 10%) of a single slave unit cost. Maybe... 150% at most.

BMS to accept at a minimum a standard 5V current sensor input. For a shunt... eh... dont know. BMS to accept current data via CAN if required (like from another node on the bus)

BMS to leave Analog and digital ports exposed for future use.

BMS not to eliminate the future option of doing isolation test, moisture test, temperature reading, LED display, RS-232 or UART input, or other standard stuff that we can do with a uC

Ok... off to fry chicken.

How many times have I finalized the requirements?
Eh... as many times as I have had different customers driving them

-methods

 

[methods]

Dual low side driver in a clean sot8
Very simple wiring with few externals
No low power shutdown ... Few hundred uA qCurrent
Poor match - don't need the speed
Not suitable for driving many trench fets... 1A output
Vin max a bit close to 12V @ 14V seeking greater margin
No regulation of gate voltage (shoot through from 12V must be protected)
http://m.linear.com/product/LTC1693

Another Overkill for our app chip
Single driver 6A output
Has a nice 12uA sleep mode
25V tolerant input is better for 12V
Regulated gate voltage 5V to 8V
Safer chip... More than we need
Too much chip for us
http://www.linear.com/docs/5777

It would be nice to standardize on a driver which can run high or low side.
A SOT package is strongly prefered... Rework and inspection is easier.
Low shutdown capable is a must
Less crap is better
High speed is not needed
Ability to drive up to 4pcs trenchfet preferred
Margin and built in protection preferred... Fewer externals.
Willing to accept a couple externals if it can go high or low
Two to a package preferred
High side has to do at least 150V or bust

Back to Google...

And the reality that I may have to address high speed cutout in the case of short circuit.
I still think this is the job of the breaker or fuse... I mean if we are using contactors... We can't blow open faster than a fuse so...

Over current in general can be much slower.

Back to the 7000 and 7000-1 (with the over current protection)
Or the 7001 with the cleaner package and fewer whistles.

Looking at the LTC7001 this time
Suitable for our voltage range
Bah... Pass again

7000
http://www.linear.com/docs/58153
Yep... That's the juicy datasheet with all the goodness you would want if you intend to hang it's balls between a pair of contactor posts. Gotta have bi-directional control... So back to back high side fets. Such a risk tho... A relay is just so much less risk.

People who have no experience with switching high voltage non-inductive loads around that connect to the outside world just don't get it. All sorts of ways to be sorry in a hurry. Sure the datasheet makes it look trivial. Show me trivial 5 years down the road when magic smoke is in the air... Good luck sorting it.

Relays...
Broke-folk see lower reliability OR slower precharge
Rich folk get to have cake and eat it... What's another $50?!?

Such an important function. Can't put a price on it working and not failing (separate things...)

Ok...
So instead of trying to directly handle precharge... If we fall back to mechanical... Then all our switching can be super simple low side.

Reliability and budget is then in the hands of the customer.
We buy some demo stock... Offer it near cost... A nice tested list to chose from.
Busted RadioShack relay?
Super Overkill spaceship relay?
A 1200V 50A contactor anyone??? That's my sort of precharge device...

Do I get to sleep now?
Where is my ice cream?
Girlfriend looks less than pleased with me... Lol... No billable hours.

But... Prepped to deliver schematics on schedule
Almost ready to estimate parts and labor.

-methods

 

[methods]

OK - system diagram on the whiteboard.

Rough current and pincounts done
Wanted to be all rad and have dedicated connectors...

Took a lesson from Sevcon.
A single gasketed 35 pin Ampseal is a super value
8 amps per channel
Proven to work up to 120V with all the functions I need to do
$15 bones
http://www.mouser.com/ProductDetail/TE-Connectivity/776231-1/?qs=sGAEpiMZZMs7eK6h2EBtKhIfvaF7bp9jcOvCLvOrJqw%3d

Already tooled up for the hardware.
Can go with a different color to use different keying... or save money on single stock part and just use the identical Sevcon connector.

There is a right angle 24 but thats too close to pin count. Margin of 12 with the 35 and thats about right.

Straight PCB mount with silicone seal at the top
Mounts to the back of a flat plastic box face
If I seal the bottom... no need to pot... or can pot if desired.

Makes for a single nasty octopus cable... but I know how to make those.
Seems harder... but just as expensive (or less expensive) that 6 individual cables.
Just needs to be a meter pigtail for finishing off... or finished to spec... or terminated outside with another IP connector.