magic keyword: scalable design

hi everyone,

not new to power stuff - owned a 1000 m2 datacenter - but now designing new modular and endless scalable version on IEEE802.3AT++ (95W @ 48VDC)

powerhouse will be solar/wind and will charge "bricks" of 18650 cells in 13S 6P configuration with smart BMS on I2C, LON, BACNET/IP or equivalent. The GRID will be backup power...

the magic question is what will happen when I add more 13S6P bricks parallel to a power rail which distributes PoE to midspan power units?

In other words: How to balance multiple bricks?

Fred

The BMS will handle any balancing within the pack. Your term brick still applies as a pack, you just have a system voltage pack.

As long as the voltages are the same on the outside ends of the packs as they are connected in parallel, the BMS will work to keep the internal cells balanced with the others.

You don't really need to try to balance one string with another. They are sorta independent of each other, even if they are connected in parallel on the ends. So you really only need to worry about the BMS keeping its own cells balanced and within spec of each other

A pack of cells in parallel don't need balancing - if a cell has a slightly higher voltage than another cell in parallel, it will provide slightly more current, and the parallel pack will stay at the same voltage regardless of load (assuming there's no failed cells). A pack of cells in parallel can be treated as one bigger cell, and a bank of batteries in parallel can be treated as one big battery.

A battery of cells (in series) needs balancing. All cells must allow the same number of electrons to flow through them, so a cell with a lower capacity will go to a lower voltage faster than the other cells in the battery. Let's say you have 2 cells in series, and one has 2000 mAh capacity while the other has 1000 mAh capacity. When the 2S battery has expended close to 1000 mAh, the smaller capacity cell approaches its cutoff voltage (let's say 3.0V), but the other cell has only expended half its capacity, so it's at its nominal voltage (let's say 3.7V). The battery now has a voltage of 3.7V + 3.0V (6.7V) which is above the cutoff voltage for a 2S pack (6.0 V), so the protection circuitry would keep the load connected, and the smaller battery would over-discharge. A BMS would check each cell and disconnect the load if the cell dropped below its cutoff voltage. It would also artificially drain the higher-capacity cell to keep them at the same voltage during discharge.

You need a BMS for each set of series-connected packs. If you have 13 cells in series per pack, and then connect 6 packs in parallel, you would need 6 independent BMSs to keep each pack balanced. The upside is, if one 13S pack is disabled, the other 5 can continue to operate and provide power.

If you instead make packs of 6 parallel cells, and then put 13 of those packs in series, you only need a single BMS. The downside is, if any of the parallel packs dies, the whole battery is disabled.

If you want to add independent 13S6P packs in parallel, no BMS is needed for the bank, but the pack itself will need an internal BMS.

On another note, 802.3bt is the spec you're looking for (4PPoe) and when using type 4, you can only guarantee 71W at the endpoint, and the endpoint will need LLDP to negotiate power above the ghost power provided by default on the line.

FredArendse said:

hi everyone,

not new to power stuff - owned a 1000 m2 datacenter - but now designing new modular and endless scalable version on IEEE802.3AT++ (95W @ 48VDC)

powerhouse will be solar/wind and will charge "bricks" of 18650 cells in 13S 6P configuration with smart BMS on I2C, LON, BACNET/IP or equivalent. The GRID will be backup power...

the magic question is what will happen when I add more 13S6P bricks parallel to a power rail which distributes PoE to midspan power units?

In other words: How to balance multiple bricks?

Fred

Click to expand...

Korishan said:

The BMS will handle any balancing within the pack. Your term brick still applies as a pack, you just have a system voltage pack.

As long as the voltages are the same on the outside ends of the packs as they are connected in parallel, the BMS will work to keep the internal cells balanced with the others.

You don't really need to try to balance one string with another. They are sorta independent of each other, even if they are connected in parallel on the ends. So you really only need to worry about the BMS keeping its own cells balanced and within spec of each other

Click to expand...

Hi Korishan,

Exactly this worries me... Internally a pack will maintain a correct balance, but the output can differ in millivolts, which will create currents between packs... Only in the case of paired power mosfets, we can assume the same internal resistance leading to exactly the same output voltage.

Chances of being so lucky with Chinese manufactured BMS controllers is same as the luck you need to land 3 first stage rockets safely back to earth after successful launch of FH from 39A pad... (Wish I was there on Playalinda during launch).

With Bacnet/IP I can create a parameter to control the output voltage of each pack, but these BMS controllers are there to keep internal balance, not external balance between multiple BMS controllers... They are unaware of each other..

The design of the datacenter will be for max 14,000 servers, which means theoretically max 1,330,000 Watts of power usage. This means loads of these packs will be there in an endless scheme of adding battery packs... Same applies to means of power generation...

I am worried about creating a crowd where each individual shouts something different to others in the crowd, resulting in a mass battle among each other...

You see my point?

Fred

P.S. 1 datacenter in Lelystad, NL and the other in Titusville, FL... (case you wonder why I know Playalinda)

many here use 14s instead of 13s because the voltages line up better with 48v lead acid equipment. plus you have a 16 to 18 volt discharge range
Later floyd

just wondering... why are you looking to power part of your datacenter with a homebrewed project? I know you mentioned your datacenter is using 1.3 MW so you're definitely not using this powerwall to power the whole datacenter. Besides, I don't know how i'd feel with datacenters claiming an SLA uptime with our soldered packs in it.

But it seems you're trying to power some POE application by directly connecting to the battery. The BMS only balances the voltage differences between the cells. To make 48V output through direct conenction to your midspans, you'll need some voltage regulators to buck/boost voltage to maintain a constant 48v regulated voltage.

AaronHancock said:

A pack of cells in parallel don't need balancing - if a cell has a slightly higher voltage than another cell in parallel, it will provide slightly more current, and the parallel pack will stay at the same voltage regardless of load (assuming there's no failed cells). A pack of cells in parallel can be treated as one bigger cell, and a bank of batteries in parallel can be treated as one big battery.

A battery of cells (in series) needs balancing. All cells must allow the same number of electrons to flow through them, so a cell with a lower capacity will go to a lower voltage faster than the other cells in the battery. Let's say you have 2 cells in series, and one has 2000 mAh capacity while the other has 1000 mAh capacity. When the 2S battery has expended close to 1000 mAh, the smaller capacity cell approaches its cutoff voltage (let's say 3.0V), but the other cell has only expended half its capacity, so it's at its nominal voltage (let's say 3.7V). The battery now has a voltage of 3.7V + 3.0V (6.7V) which is above the cutoff voltage for a 2S pack (6.0 V), so the protection circuitry would keep the load connected, and the smaller battery would over-discharge. A BMS would check each cell and disconnect the load if the cell dropped below its cutoff voltage. It would also artificially drain the higher-capacity cell to keep them at the same voltage during discharge.

You need a BMS for each set of series-connected packs. If you have 13 cells in series per pack, and then connect 6 packs in parallel, you would need 6 independent BMSs to keep each pack balanced. The upside is, if one 13S pack is disabled, the other 5 can continue to operate and provide power.

If you instead make packs of 6 parallel cells, and then put 13 of those packs in series, you only need a single BMS. The downside is, if any of the parallel packs dies, the whole battery is disabled.

If you want to add independent 13S6P packs in parallel, no BMS is needed for the bank, but the pack itself will need an internal BMS.

On another note, 802.3bt is the spec you're looking for (4PPoe) and when using type 4, you can only guarantee 71W at the endpoint, and the endpoint will need LLDP to negotiate power above the ghost power provided by default on the line.

Click to expand...

Hi Aaron,

The answer I gave to Korishan helps to understand my worries...

If I add servers up to 14,000 and bind that to increasing #'s of battery packs/bricks and alternative power generation sources, I need to create a must-be-controlled crowd.

Otherwise you create a cacophony of 'individuals' each shouting something different...

If there is 1 mV difference between each pack, we are talking about 0.001 percent difference. On a theoretical total of 1,330,000 Watts,
it is a lot of Watts between them...

In my old datacenter I was dealing with syncing 3 phase parallel AC UPS systems, balancing load on individual phases, harmonics, effectivity of server power supplies and useless heat created by keeping the crowd under control... Losses where great and a PUE of 1 does not exist... Anyone who tells you any different deserves a one way trip to Mars...

So I can assure you that I rather deal with DC problems...

But that doesn't mean we can neglect any potential dangers which can blow up in our faces and create a datacenter performing a Limbo dance...

So, I guess I need in- and out controllable BMS units and MPPT charge controllers to maintain a crowd which sings the same song in sync...

I didn't came across these type of controllers on the internet yet...

Linear produces a chip which does max 95 Watts PoE++

Fred

FredArendse said:

AaronHancock said:

A pack of cells in parallel don't need balancing - if a cell has a slightly higher voltage than another cell in parallel, it will provide slightly more current, and the parallel pack will stay at the same voltage regardless of load (assuming there's no failed cells). A pack of cells in parallel can be treated as one bigger cell, and a bank of batteries in parallel can be treated as one big battery.

A battery of cells (in series) needs balancing. All cells must allow the same number of electrons to flow through them, so a cell with a lower capacity will go to a lower voltage faster than the other cells in the battery. Let's say you have 2 cells in series, and one has 2000 mAh capacity while the other has 1000 mAh capacity. When the 2S battery has expended close to 1000 mAh, the smaller capacity cell approaches its cutoff voltage (let's say 3.0V), but the other cell has only expended half its capacity, so it's at its nominal voltage (let's say 3.7V). The battery now has a voltage of 3.7V + 3.0V (6.7V) which is above the cutoff voltage for a 2S pack (6.0 V), so the protection circuitry would keep the load connected, and the smaller battery would over-discharge. A BMS would check each cell and disconnect the load if the cell dropped below its cutoff voltage. It would also artificially drain the higher-capacity cell to keep them at the same voltage during discharge.

You need a BMS for each set of series-connected packs. If you have 13 cells in series per pack, and then connect 6 packs in parallel, you would need 6 independent BMSs to keep each pack balanced. The upside is, if one 13S pack is disabled, the other 5 can continue to operate and provide power.

If you instead make packs of 6 parallel cells, and then put 13 of those packs in series, you only need a single BMS. The downside is, if any of the parallel packs dies, the whole battery is disabled.

If you want to add independent 13S6P packs in parallel, no BMS is needed for the bank, but the pack itself will need an internal BMS.

On another note, 802.3bt is the spec you're looking for (4PPoe) and when using type 4, you can only guarantee 71W at the endpoint, and the endpoint will need LLDP to negotiate power above the ghost power provided by default on the line.

Click to expand...

Hi Aaron,

The answer I gave to Korishan helps to understand my worries...

If I add servers up to 14,000 and bind that to increasing #'s of battery packs/bricks and alternative power generation sources, I need to create a must-be-controlled crowd.

Otherwise you create a cacophony of 'individuals' each shouting something different...

If there is 1 mV difference between each pack, we are talking about 0.001 percent difference. On a theoretical total of 1,330,000 Watts,
it is a lot of Watts between them...

In my old datacenter I was dealing with syncing 3 phase parallel AC UPS systems, balancing load on individual phases, harmonics, effectivity of server power supplies and useless heat created by keeping the crowd under control... Losses where great and a PUE of 1 does not exist... Anyone who tells you any different deserves a one way trip to Mars...

So I can assure you that I rather deal with DC problems...

But that doesn't mean we can neglect any potential dangers which can blow up in our faces and create a datacenter performing a Limbo dance...

So, I guess I need in- and out controllable BMS units and MPPT charge controllers to maintain a crowd which sings the same song in sync...

I didn't came across these type of controllers on the internet yet...

Linear produces a chip which does max 95 Watts PoE++

Fred

Click to expand...

Hi Fred,

If you have 2 packs that are 1mV out, your power will still be fine. As the load on a single battery pack increases, the internal resistance drops the output voltage slightly until the voltage drop across the load is equal on all batteries. In other words, the greater the load, the greater the voltage drop within the battery itself.The one with the higher voltage will provide a little more power than the others as a natural reaction, and they will keep their voltages in sync.

802.3bt Type 4 can output 100W at the switch port, but with 12.5 ohm drop allowedon the cable, it may be as low as 75W at the end device. This depends on wire gauge and distance, of course.

MPPT controllers are really only useful for solar arrays where the power generation efficiency is dependent on load, temperature, and other factors. Depending on your power source, you may consider creating multiple breaker-protected circuits, with a group of parallel packs on each circuit, and each circuit then powering a subsetof the PoE midspan switches. If you had a charge controller on each circuit, and a passive BMS in each 13S6P battery pack/brick, that should keep everything safe-ish.

not2bme said:

just wondering... why are you looking to power part of your datacenter with a homebrewed project? I know you mentioned your datacenter is using 1.3 MW so you're definitely not using this powerwall to power the whole datacenter. Besides, I don't know how i'd feel with datacenters claiming an SLA uptime with our soldered packs in it.

But it seems you're trying to power some POE application by directly connecting to the battery. The BMS only balances the voltage differences between the cells. To make 48V output through direct conenction to your midspans, you'll need some voltage regulators to buck/boost voltage to maintain a constant 48v regulated voltage.

Click to expand...

Hi not2bme,

You don't want to deal with AC problems. I can assure you... Running on DC eliminates lots of problems and creates less losses which translates in $$$

A mix between new and used cells would be fine as long as you control the whole crowd... The lifespan of a cell on the street is shorter than the lifespan of a cell in an absolute constant temperature in a datacenter. In the coming years we will be overwhelmed with millions of cells which can be revived for years to come with some of our TLC.. That is the general idea of second life storage, right?

Maintaining customers running requires more than single servers. It requires farms of services divided over many devices. Same story with power supply.

I used to have a rack in a datacenter called Ancotel in Frankfurt, Germany. Their uptime was 99.999999 percent, until a techy accidentally dropped his screwdriver during scheduled maintenance in an UPS system, responsible for AC feed A in each customer rack... After the fireworks, AC feed A was gone, resulting in a power out on loads of servers... For hours...

In AMS-IX location 1 (NIKHEF & SARA) complete GRID power was out after someone hit a 380 kV transformer at a power station with a forklift truck. All systems went on local provider backup batteries before generators could kick in. That revealed that 300 telco providers neglected changing their batteries in time... So, all fiber went dark...

So, the soldering work of you and me is less risky than you mentioned as long as you maintain a NNNNNNNNNNNN situation...

Fred

---

AaronHancock said:

FredArendse said:

AaronHancock said:

A pack of cells in parallel don't need balancing - if a cell has a slightly higher voltage than another cell in parallel, it will provide slightly more current, and the parallel pack will stay at the same voltage regardless of load (assuming there's no failed cells). A pack of cells in parallel can be treated as one bigger cell, and a bank of batteries in parallel can be treated as one big battery.

A battery of cells (in series) needs balancing. All cells must allow the same number of electrons to flow through them, so a cell with a lower capacity will go to a lower voltage faster than the other cells in the battery. Let's say you have 2 cells in series, and one has 2000 mAh capacity while the other has 1000 mAh capacity. When the 2S battery has expended close to 1000 mAh, the smaller capacity cell approaches its cutoff voltage (let's say 3.0V), but the other cell has only expended half its capacity, so it's at its nominal voltage (let's say 3.7V). The battery now has a voltage of 3.7V + 3.0V (6.7V) which is above the cutoff voltage for a 2S pack (6.0 V), so the protection circuitry would keep the load connected, and the smaller battery would over-discharge. A BMS would check each cell and disconnect the load if the cell dropped below its cutoff voltage. It would also artificially drain the higher-capacity cell to keep them at the same voltage during discharge.

You need a BMS for each set of series-connected packs. If you have 13 cells in series per pack, and then connect 6 packs in parallel, you would need 6 independent BMSs to keep each pack balanced. The upside is, if one 13S pack is disabled, the other 5 can continue to operate and provide power.

If you instead make packs of 6 parallel cells, and then put 13 of those packs in series, you only need a single BMS. The downside is, if any of the parallel packs dies, the whole battery is disabled.

If you want to add independent 13S6P packs in parallel, no BMS is needed for the bank, but the pack itself will need an internal BMS.

On another note, 802.3bt is the spec you're looking for (4PPoe) and when using type 4, you can only guarantee 71W at the endpoint, and the endpoint will need LLDP to negotiate power above the ghost power provided by default on the line.

Click to expand...

Hi Aaron,

The answer I gave to Korishan helps to understand my worries...

If I add servers up to 14,000 and bind that to increasing #'s of battery packs/bricks and alternative power generation sources, I need to create a must-be-controlled crowd.

Otherwise you create a cacophony of 'individuals' each shouting something different...

If there is 1 mV difference between each pack, we are talking about 0.001 percent difference. On a theoretical total of 1,330,000 Watts,
it is a lot of Watts between them...

In my old datacenter I was dealing with syncing 3 phase parallel AC UPS systems, balancing load on individual phases, harmonics, effectivity of server power supplies and useless heat created by keeping the crowd under control... Losses where great and a PUE of 1 does not exist... Anyone who tells you any different deserves a one way trip to Mars...

So I can assure you that I rather deal with DC problems...

But that doesn't mean we can neglect any potential dangers which can blow up in our faces and create a datacenter performing a Limbo dance...

So, I guess I need in- and out controllable BMS units and MPPT charge controllers to maintain a crowd which sings the same song in sync...

I didn't came across these type of controllers on the internet yet...

Linear produces a chip which does max 95 Watts PoE++

Fred

Click to expand...

Hi Fred,

If you have 2 packs that are 1mV out, your power will still be fine. As the load on a single battery pack increases, the internal resistance drops the output voltage slightly until the voltage drop across the load is equal on all batteries. In other words, the greater the load, the greater the voltage drop within the battery itself.The one with the higher voltage will provide a little more power than the others as a natural reaction, and they will keep their voltages in sync.

802.3bt Type 4 can output 100W at the switch port, but with 12.5 ohm drop allowedon the cable, it may be as low as 75W at the end device. This depends on wire gauge and distance, of course.

MPPT controllers are really only useful for solar arrays where the power generation efficiency is dependent on load, temperature, and other factors. Depending on your power source, you may consider creating multiple breaker-protected circuits, with a group of parallel packs on each circuit, and each circuit then powering a subsetof the PoE midspan switches. If you had a charge controller on each circuit, and a passive BMS in each 13S6P battery pack/brick, that should keep everything safe-ish.

Click to expand...

Hi Aaron,

Cable length between powered device and midspan will be 10 meters or less... Losses are small and TCO will be low with just a single copper cable responsible for power and Ethernet...

Fred

not2bme said:

just wondering... why are you looking to power part of your datacenter with a homebrewed project? I know you mentioned your datacenter is using 1.3 MW so you're definitely not using this powerwall to power the whole datacenter. Besides, I don't know how i'd feel with datacenters claiming an SLA uptime with our soldered packs in it.

But it seems you're trying to power some POE application by directly connecting to the battery. The BMS only balances the voltage differences between the cells. To make 48V output through direct conenction to your midspans, you'll need some voltage regulators to buck/boost voltage to maintain a constant 48v regulated voltage.

Click to expand...

Hi not2bme,

I have another of these unbelievable outages for you... I am doing this datacenter stuff since 1994 and was behind the keyboard since 1976... So, I have some creepy stories for you during a campfire...

There where I had my 1000 m2 AC powered datacenter in Rotterdam in a huge telecom building, one of my neighbors was Versatel (later Tele2). They have 2000 m2 datacenter on 2 floors. When there was a GRID outage in the area, usually their standby gensets would start after 3 minutes. To overcome these 3 minutes, they had a huge no-break system with a really big battery farm.

After 3 minutes I saw people running to the ground floor, because the generators refused to start...

power was out more than a minute before that, because someone forgot to calculate the actual backup time on older existing batteries...

after the generators were manually started, problems were not solved since fiber equipment didn't start up after years of running without reboot...

either core OS on flash was corrupt, and/or the burst/sparks these gensets caused blew up some AC-DC 48 Volt rectifiers...

2 full days it took 20 engineers to recover from that short 10 minutes GRID outage and next day two 40 foot trucks were delivering new batteries...

All I did was shutdown non-essential services to save power...

Fred

Any two or more cells/batteries that are connected in parallel will not have different voltages. It's not like AC power. So you don't have to worry about the output voltages.

If you put a pack into service that is slightly out of voltage sync with the others, when you connect it up, the voltages will become the same. It'd be like connecting a water container in parallel with another water container. They will balance each out to the same height on the water line. It's just how it works.

In this case, if it's higher, then it will discharge into the other packs. If it's lower, it will be charged by the others till it's the same voltage. As far as the BMS is concerned at this time, it's being charged or discharged for those few moments and it will report it as such.

I think you're making a bit of a mountain out of a mole hill on the voltage issue of packs. It's not that critical as it is with AC. It won't slowly oscillate between voltages.
But, at least you're asking to gain knowledge!

Fred Arendse said:

2 full days it took 20 engineers to recover from that short 10 minutes GRID outage and next day two 40 foot trucks were delivering new batteries...

Click to expand...

Wow! That sux!

not2bme said:

just wondering... why are you looking to power part of your datacenter with a homebrewed project? I know you mentioned your datacenter is using 1.3 MW so you're definitely not using this powerwall to power the whole datacenter. Besides, I don't know how i'd feel with datacenters claiming an SLA uptime with our soldered packs in it.

But it seems you're trying to power some POE application by directly connecting to the battery. The BMS only balances the voltage differences between the cells. To make 48V output through direct conenction to your midspans, you'll need some voltage regulators to buck/boost voltage to maintain a constant 48v regulated voltage.

Click to expand...

Hi not2bme,

Last of the Halloween stories...

A new designed datacenter was equipped with an in-building free cooling system based on a horizontally mounted wheel of a diameter of 6 meters packed with a huge amount of copper elements. Half of that wheel was in the hot datacenter area and the other half was in a air cooled compartment.

That wheel revolved 6 times per minute thus picking up heat in the datacenter and cooling down the copper pack in the air cooled compartment.

It was sold to them being effective up to 24 degrees Celsius. Above that traditional Freon AC would kick in. In The Netherlands that would be days per year...

They got the city to co-sign a millions bank loan, since one of their customers would be a new huge hospital...

The first 6 months after opening they had a PUE of 1.1 . Great result on power effectiveness... But then spring/summer came...


Actually, the one who told them that the free cooling wheel would be effective until 24 C was wrong... At 22C the traditional cooling started up...

Since theydesigned the datacenter for 650 kW, the cooling took a bigger part of the available power...

the ONLY way to solve that until final resolution was to start one of the generators to isolated deliver power to the AC Cooling...

But the generator exhaust was designed right next the cold air intake for the 'magic' free cooling wheel... With certain wind direction the hot generator exhaust blew right into thecold air free-cooling intake...

I have more trust in soldering revived batteries, more trust the sun will shine and more trust in wind being there...

Fred

Korishan said:

Any two or more cells/batteries that are connected in parallel will not have different voltages. It's not like AC power. So you don't have to worry about the output voltages.

If you put a pack into service that is slightly out of voltage sync with the others, when you connect it up, the voltages will become the same. It'd be like connecting a water container in parallel with another water container. They will balance each out to the same height on the water line. It's just how it works.

In this case, if it's higher, then it will discharge into the other packs. If it's lower, it will be charged by the others till it's the same voltage. As far as the BMS is concerned at this time, it's being charged or discharged for those few moments and it will report it as such.

I think you're making a bit of a mountain out of a mole hill on the voltage issue of packs. It's not that critical as it is with AC. It won't slowly oscillate between voltages.
But, at least you're asking to gain knowledge!

FredArendse said:

2 full days it took 20 engineers to recover from that short 10 minutes GRID outage and next day two 40 foot trucks were delivering new batteries...

Click to expand...

Wow! That sux!

Click to expand...

Hi Korishan,

Better ask than sit on blisters, right?

I knew that I wanted a DC datacenter and the community of used 18650 cells enthusiasts filled in the blanks how to do it...

You read the post about low PUE? That can and will happen if people don't know anything about HVAC. Beside design temperature problems, nobody kept count with condensation on that wheel pushing up the relative humidity in the datacenter...

the only way to get rid of that was start up forced cooling/condensation... In that chain of events people throw away lots of money in TCO...

The 'wheel of fortune of somebody else' was turned which resulted in a price tag of 2 million for a not working solution...

Fred

P.S. There isn't one day in my life I discovered or learned something new...

Turned FredArendse said:

Korishan said:

Any two or more cells/batteries that are connected in parallel will not have different voltages. It's not like AC power. So you don't have to worry about the output voltages.

If you put a pack into service that is slightly out of voltage sync with the others, when you connect it up, the voltages will become the same. It'd be like connecting a water container in parallel with another water container. They will balance each out to the same height on the water line. It's just how it works.

In this case, if it's higher, then it will discharge into the other packs. If it's lower, it will be charged by the others till it's the same voltage. As far as the BMS is concerned at this time, it's being charged or discharged for those few moments and it will report it as such.

I think you're making a bit of a mountain out of a mole hill on the voltage issue of packs. It's not that critical as it is with AC. It won't slowly oscillate between voltages.
But, at least you're asking to gain knowledge!

FredArendse said:

2 full days it took 20 engineers to recover from that short 10 minutes GRID outage and next day two 40 foot trucks were delivering new batteries...

Click to expand...

Wow! That sux!

Click to expand...

Hi Korishan,

Better ask than sit on blisters, right?

I knew that I wanted a DC datacenter and the community of used 18650 cells enthusiasts filled in the blanks how to do it...

You read the post about low PUE? That can and will happen if people don't know anything about HVAC. Beside design temperature problems, nobody kept count with condensation on that wheel pushing up the relative humidity in the datacenter...

the only way to get rid of that was start up forced cooling/condensation... In that chain of events people throw away lots of money in TCO...

The 'wheel of fortune of somebody else' was turned which resulted in a price tag of 2 million for a not working solution...

Fred

P.S. There isn't one day in my life I discovered or learned something new...

Click to expand...

Worked in one building around 34 floors, generators on top of the roof due to the exhaust and being in the middle of the city. Power goes out, UPS covers until the generators start, which run for a few minutes before stopping and shortly after all the UPS's powered down. Turned out the fuel tank could not be placed on the roof (obvious in certain regards) and the one fuel pump (single point of failure) on the ground floor had failed so the units on the roof just ran out of fuel and stopped. Short test they had done in the past just used a small bit of the fuel up top....

I like the idea of creating bricks with self-contained BMS that can report via BACNet, as that way it ties into your building management system and can be easily managed from a single console. Any one brick showing fault can easily be hot plugged our and a new one plugged it.

Since all you are going at that point is adding bricks to a parallel bus making it infinitely scalable, increasing amp limits and capacity with each additional brick.

On top of that, with something like an Anderson disconnect found on your typical APC unit, you could literally hot plug these things in and out in place during use one at a time to test individual brick voltages.

Your goals seem obtainable. I would also consider a 7s configuration or higher as you still need to get some form of synchronous buck/boost DC-DC to stabilize output voltages unless you have some other supply that is running on 24v that I didnt pick up in my reading.

Something similar was actually my original plan for a Powerwall project, because I can always add modules/bricks as I grow, but for that capability you pick up complexity and cost on the BMS management side for that flexibility. Most here build very large parallel packs and build a 7s configuration with those. The downside being you need to harvest for much longer times before you can increase capacity, also making maintenance more challenging.

Crimp Daddy said:

I like the idea of creating bricks with self-contained BMS that can report via BACNet, as that way it ties into your building management system and can be easily managed from a single console. Any one brick showing fault can easily be hot plugged our and a new one plugged it.

Since all you are going at that point is adding bricks to a parallel bus making it infinitely scalable, increasing amp limits and capacity with each additional brick.

On top of that, with something like an Anderson disconnect found on your typical APC unit, you could literally hot plug these things in and out in place during use one at a time to test individual brick voltages.

Your goals seem obtainable. I would also consider a 7s configuration or higher as you still need to get some form of synchronous buck/boost DC-DC to stabilize output voltages unless you have some other supply that is running on 24v that I didnt pick up in my reading.

Something similar was actually my original plan for a Powerwall project, because I can always add modules/bricks as I grow, but for that capability you pick up complexity and cost on the BMS management side for that flexibility. Most here build very large parallel packs and build a 7s configuration with those. The downside being you need to harvest for much longer times before you can increase capacity, also making maintenance more challenging.

Click to expand...

I'm working on a pack level bms system that would also communicate with a central mcu. I plan on having parallel packs with each pack running at system voltage. This would allow for true hotswap without any down sides, or very little.

One way I thought of connecting them was to use the same kind of connecting like that is used on power tools. There's a pair of blades with a curved end and accepts a straight blade connection. Have 2 - 4 of these on each end and have a din rail to hold the packs you could quite literally hang and plug at the same time. And maybe even make the din rail be the primary bus bar as it is quite large (or have the main bus bar running directly above/below with the din bar having the contacts; but the din bar would be live)
The other style was to have 4 pins on one end of the pack, 2Pos and 2Neg, top and bottom. Then when you slide the pack into place on the shelf, they clip into place.
Obviously that's for power connections and I'd need another set of contacts for the communications portion.