Cheap LiFePO4 BMS?

[Dr Bob]

1 hour ago, nicknorman said:

I think the difference between our approaches is that I intend to charge to a specified SoC rather than to a specified voltage, but for that to work the system has to have a reasonably accurate measurement of SoC.

Yes, I understand!

I dont think I would ever trust the BMV SoC reading. I've fiddled with the paramaters to try and get it so it syncronises well, but have given up. Control via voltage is working for me at the moment!

[peterboat]

2 hours ago, Dr Bob said:

I assume you mean the SoC has reduced?

Over the past month (hooked up to shore power in a marina) our Li's have been isolated at around 40% SoC. Most of the time I have had both the BEP auto switch live, the BMV live and the cell monitor live. The battery voltage has dropped around 0.10V over 21 days so that is 0.01V every few days. A week ago I turned the BEP switch off and disconnected the cell monitor and the system has stayed at 12.95V for 7 days. The current drain from the BEP switch and the cell monitor dont show up on the BMV. The cell monitor is wired directly onto the cells and not via the shunt but I am sure the BEP is wired via the shunt. Once I isolate the bank, the SoC or Ahrs used remain constant.

On my system, a drop of 0.1V is equiv to around 50Ahrs of capacity and I find that after a couple of months out and about with not going over 80% SoC, when I charge to full to syncronise the BMV I usually have to put in 50Ahrs ish past when the Ahrs out gets back to zero. I guess this is the 0.01V loss from the system every few days to 'power the system'. It could however be a BMV issue. I've given up worrying about it as 50Ahrs is half an hour of engine running!

BMVs are a thing of mystery.

I am losing power on my drive batteries the midnite controller is connected but solar is all ro the domestic batteries, so whatever the controller uses plus 30 BMSs running equals lost amps 

[nicknorman]

3 hours ago, Dr Bob said:

Yes, I understand!

I dont think I would ever trust the BMV SoC reading. I've fiddled with the paramaters to try and get it so it syncronises well, but have given up. Control via voltage is working for me at the moment!

To be clear my intention is to terminate charge based on SoC AND ensure that no cell goes over 3.6v. So I don’t think it will be “dangerous” but it might mean that when I want to charge to 80%, 50% or whatever, it isn’t that accurate. If that turns out to be the case I will have to implement the KalMoomin filter! Or something else. At least I will be able to refine the code “on the hoof” once it is installed and I see how the batteries behave when charged at 1/4C or whatever.

Edited December 7, 2020 by nicknorman

[tree monkey]

I pop into the thread to try to keep up but to be honest I got lost at page 2, actually I lie page 1, anyway my question.

 

As I understand most of the thread is geek level interest in how to manage and understand the battery to a level most of us won't ever really care about, the overall picture seems to be once set up they are a sort of wunderkind of the battery world, so what are the basic problems or difficulties of managing a bank once set up properly? 

 

 

[MoominPapa]

52 minutes ago, tree monkey said:

I pop into the thread to try to keep up but to be honest I got lost at page 2, actually I lie page 1, anyway my question.

 

As I understand most of the thread is geek level interest in how to manage and understand the battery to a level most of us won't ever really care about, the overall picture seems to be once set up they are a sort of wunderkind of the battery world, so what are the basic problems or difficulties of managing a bank once set up properly? 

 

 

Once set up, there are no problems. You really can treat them like a bucket of electricity. You can fill the bucket full, but no further, once the bucket is empty the lights go out. In our case we use between a quarter and a third of a bucketful per day, and fill it at about 10% per hour running the genny or the engine on tickover, and about 15% per hour when the engine is also moving the boat. You can go for weeks or months without ever completely filling the bucket and it won't complain, and our bucket is the same size it was when we installed it two years ago.

 

MP.

 

[peterboat]

36 minutes ago, MoominPapa said:

Once set up, there are no problems. You really can treat them like a bucket of electricity. You can fill the bucket full, but no further, once the bucket is empty the lights go out. In our case we use between a quarter and a third of a bucketful per day, and fill it at about 10% per hour running the genny or the engine on tickover, and about 15% per hour when the engine is also moving the boat. You can go for weeks or months without ever completely filling the bucket and it won't complain, and our bucket is the same size it was when we installed it two years ago.

 

MP.

 

Thats how I explain it to other boaters

[Craig Shelley]

8 hours ago, nicknorman said:

I have got the idle power consumption down to 3 mA which is not quite as good as I'd hoped, but acceptable I think bearing in mind the CANBUS interface is live. It's about 15 mA with the display on, but that times out after a while.

 

Which brings me to a question for BMV712S users. Having disconnected everything from the load side of the shunt (this is all still in my house, not connected to the boat), with 200Ah of batteries at around 40% SoC, the SoC indication on the BMV has crept up slightly, roughly by 0.1% per day. I have tried the "zero current calibration" - item 09 on the setup. Which doesn't seem to help. I select setup item 09, I press "select" again, and it comes back with "ERROR" so I am not convinced it is doing anything. I also notice that if I set the "current threshold" - item 07 - to zero, the current display shows fluctuations between 0 and 0.02A. Maybe not a lot but 20mA is 1/2Ah per day which is also not a lot if one is routinely synchronising the BMV by charging to 100% as one would do with LA  batteries. But of course that is not the plan. 0.25Ah per day for 6 months is 87Ah which definitely is significant. 6 months is the battery manufacturer's recommended interval for a full charge/discharge cycle for the batteries.

That is an impressively low idle power consumption. Is that also counting Ah in/out of the battery?

 

I always wondered how the BMV achieved zero-drift. It's very difficult to achieve the required dynamic range i.e. being able to measure 300-400A at the top end whilst accurately measuring sub 10mA at the bottom end. The zero calibration can be thrown out by a few microvolts at the shunt resistor - easily created by a thermocouple junction given the different metals involved.

 

This makes my RPI solution look rather laughable at 80mA!

[nicknorman]

6 minutes ago, Craig Shelley said:

That is an impressively low idle power consumption. Is that also counting Ah in/out of the battery?

 

I always wondered how the BMV achieved zero-drift. It's very difficult to achieve the required dynamic range i.e. being able to measure 300-400A at the top end whilst accurately measuring sub 10mA at the bottom end. The zero calibration can be thrown out by a few microvolts at the shunt resistor - easily created by a thermocouple junction given the different metals involved.

 

This makes my RPI solution look rather laughable at 80mA!

No my scheme doesn’t count Ah directly, I already have a device doing that - Mastervolt Mastershunt. And maybe the BMV712. My device gets I, Ah,  SoC etc over CANBUS from the mastershunt, and possibly from the BMV over VE.direct (haven’t written the code for that yet, but the hardware is in place).

 

The strategy is that one doesn’t need to sample something fast, if it can’t change fast. The AD7280 battery monitoring chip takes a few mS to power up, a few mS to read the cell voltages and send to the micro over SPI, then it powers down again. It uses about 10mA when it is actually on, but for 99% of the time it isn’t on. I have set the sample rate to once a second whilst the display is active, but then once every 8 seconds after the display times out.

 

Although I’m running the PIC micro at 8MHz, it has various power saving modes and also allows any unused peripherals to be powered down. I’m using idle mode in which the CPU clock stops but any powered up peripherals keep working. It’s woken up either by a timer interrupt (1 sec or 8 secs) or by an interrupt such as the CANBUS interface receiving some wanted data, or the little push button to bring the display on. If it ever activates the emergency low cell voltage disconnect relay it goes fully to sleep (power consumption is about 1mA) until the push button is pressed. Obviously in the event of a low voltage disconnect it’s important not to drain the battery further after everything else has been disconnected.

 

The BMV has a fudge parameter, “ignore any current below x mA”. The default is 0.1A which tells you something about their expectations of zero accuracy! But then it is fairly cheap for what it does. Our boat has Empirbus DC distribution and control so even when the virtual master switch is off, it still takes 10mA or so and we also have a GSM remote control receiver for putting the heating on etc, that takes about 20mA. So I will need to set the threshold down at around 20mA for it to take these into account - which might or might not work out!

[Craig Shelley]

3 hours ago, nicknorman said:

To be clear my intention is to terminate charge based on SoC AND ensure that no cell goes over 3.6v. So I don’t think it will be “dangerous” but it might mean that when I want to charge to 80%, 50% or whatever, it isn’t that accurate. If that turns out to be the case I will have to implement the KalMoomin filter! Or something else. At least I will be able to refine the code “on the hoof” once it is installed and I see how the batteries behave when charged at 1/4C or whatever.

I've tried to implement this, but due to various effects our long-term Ah counting has turned out to be not good enough. I'm hoping to make improvements over the next few weeks, but there's no guarantee that will overcome all of the long-term drift error which is inherent with this method. Charging to 100% periodically to reset the meter is one obvious solution, but you might not want to do that as often as is needed.

 

The problem we found when using estimated SoC to stop the charging was that the numbers ended up looking too trustworthy and we lost our mental reference of SoC. We only realised that the SoC gauge had drifted out when the vacuum cleaner tripped the low voltage protection switch. This is despite the fact that we still had a trust-worthy volt meter. For some reason it was more satisfying to trust the Ah based SoC gauge. We then decided to charge to a known point just to see how far down they really were. The engine ended up running for over 6 hours! It was reminiscent the feeling of "suddenly realising you're lost" after figuring out that you've been misreading a map... for most of the day.

In the end we opted for a simple approach of charging until a predetermined voltage, and then resetting the counter. This method is not perfect, but it is easy to live with the "known imperfection". The result is a variation in cut-off point due to temperature, charge current, and other factors but at least this is a variation on a fixed datum as opposed to something which can gradually drift with time.

 

I was hoping to implement a SoC percentage gauge based on the battery voltage, which could then be used to null out the long-term drift in the Ah measurement. It turns out that this is non-trivial since the various factors are involved, primarily the history of charge/discharge currents. There are countless papers on this topic in the engineering/science journals.

 

It's easy to overlook the fact that we've got by using just a simple volt meter for the first couple of years. My father used to take a glance at the meter first thing in the morning before the solar started to kick in. Using this approach he somehow got an approximate feel for the SoC. Also without realising it he also used to wait for an appropriate moment in the fridge compressor cycle. I imagine this is possibly the simplest approach to implement in software as instead of having a complex model to compensate for the effect of current, it's easier just to eliminate the effect by taking the voltage measurement at a carefully chosen moment.

[nicknorman]

Was this using your own measurement of current / integrating to get Ah and SoC, or some proprietary device like a BMV?

I think I will have to suck it and see, we will have a relatively large battery bank (600Ah) and of course for a given accuracy of current measurement the percentage error in SoC (rate of accumulation of the error) is less the larger the bank. The point about the “allure” of the SoC display is well made!

 

Id have to look back in this thread but Moominpappa has implemented a successful SoC calculation by a hybrid of Ah counting and looking at voltage with an estimation of internal resistance. So that would be a fallback position. The AD7280 only has jitter of about +-1mV in the cell readings, which is a good starting point (I’ve set it to do x8 over sampling, which helps). One annoying thing is that whilst both the Mastershunt and BMV send out a lot of data, and the Mastershunt has a lot of parameters that you can adjust over CANBUS, SoC is not one of them. So I would be stuck with my own device having one estimation of SoC whilst the other displays showed something else.

Edited December 8, 2020 by nicknorman

[Dr Bob]

13 hours ago, Craig Shelley said:

I was hoping to implement a SoC percentage gauge based on the battery voltage, which could then be used to null out the long-term drift in the Ah measurement. It turns out that this is non-trivial since the various factors are involved, primarily the history of charge/discharge currents. There are countless papers on this topic in the engineering/science journals.

I dont think it is difficult.

I put the graph below together 2 years ago based on readings at rest. Ignore the stuff in the red bit as they were values with no current flowing but just after the bank was isolated after charging (the voltages decay to the at rest value).

The graph is still valid 2 years later. I know the effect on voltage of a current draw of 5A (typical daytime draw with no fridge) or 9A (with fridge compressor on) so a glance at the voltmeter tells me the SoC % within 10%. You dont need it any better than that.

[Dr Bob]

17 hours ago, MoominPapa said:

Once set up, there are no problems. You really can treat them like a bucket of electricity. You can fill the bucket full, but no further, once the bucket is empty the lights go out. In our case we use between a quarter and a third of a bucketful per day, and fill it at about 10% per hour running the genny or the engine on tickover, and about 15% per hour when the engine is also moving the boat. You can go for weeks or months without ever completely filling the bucket and it won't complain, and our bucket is the same size it was when we installed it two years ago.

That is a very good analogy. The key is not to overflow the bucket but that is done by setting it up properly. Bad things happen when the bucket overflows.

 

That is the simplistic view. There is another level of detail that I would introduce to help those who dont have a clue. It is not really one bucket, its four all connected together. They fill and empty together but if they are not balanced then there could be a problem of overfilling one. You can avoid this by only ever filling to 4/5ths full. That's what I do. You do that by setting it up right. Once set up right then all you need is to glance at the voltmeter once per day to see how many days left of charge you have.

My 4 buckets are the same size as they were 2 years ago.

If you want to fill the buckets to full then you can do that by just keeping an eye on them as they get to almost full. I do that every 3 months but you dont really need to. Almost muppet proof!

[MoominPapa]

1 hour ago, Dr Bob said:

 They fill and empty together but if they are not balanced then there could be a problem of overfilling one. You can avoid this by only ever filling to 4/5ths full.

OK, it's time to go from simple summaries, straight back into the nitty-gritty. I think the sentence above may be dangerous oversimplification.

 

More specifically, it confuses two things. The first, which is certainly true, is that significant imbalance is not a problem if you never enter the nearly-full and nearly-empty regions. If you never go above 90% and below 10%, then (oversimplifying slightly) 10% imbalance is not a problem, so you don't need a balancer to correct it.

 

The second thing here, which is sort-of implied, but not often said, is that if you avoid the top and bottom of the discharge curve the amount of imbalance will not change, so you don't need a balancer for that either. That's not necessarily true, and it has implications. If your batteries are never balanced, and their balance continues to drift, then eventually the amount of imbalance will be more than the margins you are leaving, and a cell will over or under voltage. 

 

Cells get out of balance for two broad reasons: difference in self discharge rate, and difference in coulombic efficiency. I don't see how avoiding 90%+ and 10%- can effect the first of these, and whilst it might effect the second, I've not seen any evidence of that.

 

TL;DR limiting the charge and discharge margins makes you safe from imbalance, it doesn't stop imbalance from happening, and getting worse, so that eventually your margins are too small. If you don't have balance problems because you keep margins, that doesn't mean that you cell balance is not getting worse over time, and you may eventually exceed those margins.

 

MP.

 

 

 

[Dr Bob]

2 hours ago, MoominPapa said:

The second thing here, which is sort-of implied, but not often said, is that if you avoid the top and bottom of the discharge curve the amount of imbalance will not change, so you don't need a balancer for that either. That's not necessarily true, and it has implications. If your batteries are never balanced, and their balance continues to drift, then eventually the amount of imbalance will be more than the margins you are leaving, and a cell will over or under voltage. 

You are of course right that the simple analagy isnt good enough for the Li newbie and that is why I had to go down to the next level of complexity for TM and introduce balancing.....because if you dont - then it can go badly wrong. Setting up the system at the start is the essential bit to stop overcharging each of the 4 buckets by i) having a total voltage limit on the bank (so isolating at 13.8V or wotever) and ii) having a cell voltage limit (so isolating if any goes over 3.8V). TM was asking what needs to be done once set up properly - so I think my advice still stands. If a system went way out of balance to the point that one cell got into the knee while the others were still at 80% SoC then the cell monitor would isolate and the user would be back on here in quick time to find out why.

It is interesting however that looking at the cruiser forum which has a thread running for 8 years on Li's, I have never seen ANY posts on peeps with anything near as out of balance as that. If it did though, the minimum DIY BMS should catch the problem. From what I read on the cruiser forum, the major cause of inbalance is due to people charging up well into the high voltage knee ( ie the more you are in that region, the more the cells diverge - not a clue why) - so keep out of there and they retain balance better.

Have your batteries drifted? Mine have now been in for 2 1/2 years and my balance is just as bad as it ever was after my first cycle. My cell one reaches around 3.60V as the other 3 get to 3.50V whereas cell one also gets into the bottom knee first. The performance has been pretty much the same for the last 18 months. I tried to balance cell one to get it so it was better than 100mV but depleting cell one at the top end just sends it into the bottom knee earlier. Cell one obviously has slightly less capacity but the 'poorish' balance is not changing. I only go above 80% about 4 times a year.

Am I worried? No. My system isolates at 13.8V so I am well away from the high voltage knee and if the balance did drift, then the cell monitor would isolate at 3.8V.

The key to the above is getting the setup done properly. With a good set up then you shouldnt need to worry about balance and the system will tell you if the balance goes wrong. The bit in red from your post is a key point. To set up properly - you would always balance when installing!

My advice of the previous post still applies. Keep below 80% and you wont have a balance problem if you have a decent set up - the cell monitor will catch any balance issue.

 

[nicknorman]

I’ll just mention that 13.8v (3.45v/cell) is quite a high voltage in terms of SoC. Charts on the internet suggest that if the batteries stabilise at 13.8v that is over 95%. I think in your case you stop charging when the voltage as measured somewhere, hits 13.8v with still a lot of current going in? So it will depend on where you measure the voltage and what the charge current is, as to what the SoC is when you hit 13.8v.

 

When I was charging my 200Ah banks at 5A (which is very slow, obviously), 13.8v measured at the cells was definitely nearly fully charged and so it wouldn’t take much imbalance for a rogue cell to start climbing up the knee. In other words, unbalanced battery health is not assured by having a stabilised charge voltage of 13.8v. It is giving a few % margin, but not 20%.

[MoominPapa]

10 minutes ago, Dr Bob said:

You are of course right that the simple analagy isnt good enough for the Li newbie and that is why I had to go down to the next level of complexity for TM and introduce balancing.....because if you dont - then it can go badly wrong. Setting up the system at the start is the essential bit to stop overcharging each of the 4 buckets by i) having a total voltage limit on the bank (so isolating at 13.8V or wotever) and ii) having a cell voltage limit (so isolating if any goes over 3.8V). TM was asking what needs to be done once set up properly - so I think my advice still stands. If a system went way out of balance to the point that one cell got into the knee while the others were still at 80% SoC then the cell monitor would isolate and the user would be back on here in quick time to find out why.

It is interesting however that looking at the cruiser forum which has a thread running for 8 years on Li's, I have never seen ANY posts on peeps with anything near as out of balance as that. If it did though, the minimum DIY BMS should catch the problem. From what I read on the cruiser forum, the major cause of inbalance is due to people charging up well into the high voltage knee ( ie the more you are in that region, the more the cells diverge - not a clue why) - so keep out of there and they retain balance better.

So you're saying that imbalance is caused in the knee? Presumably because the coulombic efficiency drops off there? There's a potential runaway effect here. If the cells are out of balance, one goes into the knee earlier. But then you'd expect it's efficiency to fall (some energy diverted to heat?) so that would tend to balance things out.  I just don't think it's sensible to rely on avoiding the knees to keep things balanced, as opposed to making the system resistant to some imbalance. You still need, at minimum, cell voltage monitoring and cut off to tell you when the balance has gone too far, and if you want a true "bucket of electrons" system, you need automatic balancing so that the innocent are not woken in the night by alarms they don't understand and can't fix.

10 minutes ago, Dr Bob said:

Have your batteries drifted? Mine have now been in for 2 1/2 years and my balance is just as bad as it ever was after my first cycle. My cell one reaches around 3.60V as the other 3 get to 3.50V whereas cell one also gets into the bottom knee first.

 

Mine are pretty much perfectly top balanced now. I got them most of the way there by hand, tweaking things each time I got to charge termination. I've since automated that, so the system does a post-charge balance based on the relative voltages at termination by itself. That got them perfect and keeps them there.  What's not balanced is the capacity of the parallel cell groups. I have one  which always always drops below 3v first, two which are higher but not far behind  and one with sufficient excess capacity that it's still at 3.2v . Short of dismantling the whole thing, capacity testing each cell in isolation and grouping them for the most equal capacities, there's not much I can do about this. My usable capacity to 400Ah, from nominally 480Ah second hand cells. so I'm not inclined to bother. What I have if fine for my use.

 

MP.

 

MP.

 

[Dr Bob]

29 minutes ago, MoominPapa said:

So you're saying that imbalance is caused in the knee?

No. I am saying the info in the cruiser forum suggests the majority of inbalance issues takes place in the knee.

I dont know. I have been successful in not seeing inbalance getting worse - perhaps due to me keeping below 80% if you believe the cruisier forum -which has 8 years of Li experience.

29 minutes ago, MoominPapa said:

I just don't think it's sensible to rely on avoiding the knees to keep things balanced, as opposed to making the system resistant to some imbalance. You still need, at minimum, cell voltage monitoring and cut off to tell you when the balance has gone too far,

That's exactly what I said. Set it up properly. TM specifically asked about how to run the system ONCE IT WAS SET UP PROPERLY  hence my response.

29 minutes ago, MoominPapa said:

and if you want a true "bucket of electrons" system, you need automatic balancing so that the innocent are not woken in the night by alarms they don't understand and can't fix.

Now this is a lot more difficult. Automatic balancing is not easy. My system wouldnt work on a typical autobalancer as in the 20-80% range my cell one is lowest whereas once in the knee is the highest. Mine could only be balanced once it got into the knee. Under normal operation, the balancer would push capacity into cell one and this would increase the inbalance in the top knee. If it was a clever balancer, then it would remember that cell one was high at the top and extract capacity in later after charging. Problem then is cell one gets in the bottom knee first. Maybe if it was an even cleverer balancer, it would work out that cell one was a lower capacity and do nothing.....like I do!?

I think for what we are doing - using 2nd hand LiFePO4 cells - that there is no clear answer. A autobalancer on mine just wouldnt work. For you it does. I think that the adopter needs to know about balancing - hence my 4 bucket approach for TM - but not necessarily implement auto balancing in the set up......but understand why the alarm might go off in the middle of the night.

Li's are not fit and forget for the non battery person and diy Li's will never be. The user must have some degree of Li knowledge if buying 2nd hand Li's and be able to assess the risks.

 

Edit - I dont think the 'innocent' should install 2nd hand Li's with DIY BMSs.

Edited December 8, 2020 by Dr Bob

[Dr Bob]

33 minutes ago, nicknorman said:

I’ll just mention that 13.8v (3.45v/cell) is quite a high voltage in terms of SoC. Charts on the internet suggest that if the batteries stabilise at 13.8v that is over 95%. I think in your case you stop charging when the voltage as measured somewhere, hits 13.8v with still a lot of current going in? So it will depend on where you measure the voltage and what the charge current is, as to what the SoC is when you hit 13.8v.

 

When I was charging my 200Ah banks at 5A (which is very slow, obviously), 13.8v measured at the cells was definitely nearly fully charged and so it wouldn’t take much imbalance for a rogue cell to start climbing up the knee. In other words, unbalanced battery health is not assured by having a stabilised charge voltage of 13.8v. It is giving a few % margin, but not 20%.

My isolation is set at 13.8V. That is around 90% SoC on my system rather than 95% .

Normally I never get that high as the B2Bs turn themselves off at a lower voltage (13.6V ish) and usually showing -50Ahrs on the BMV (+ another 20-30Ahrs due to out of sync) so typically I am using 20-80%. In the 2.5 years use I have never triggered the self isolation in normal use. NEVER.?

[peterboat]

15 minutes ago, Dr Bob said:

My isolation is set at 13.8V. That is around 90% SoC on my system rather than 95% .

Normally I never get that high as the B2Bs turn themselves off at a lower voltage (13.6V ish) and usually showing -50Ahrs on the BMV (+ another 20-30Ahrs due to out of sync) so typically I am using 20-80%. In the 2.5 years use I have never triggered the self isolation in normal use. NEVER.?

I ran my system right down the other day literally down to 9 volts, I then charged it up slowly over 3 days popped it on the puter and all is well, batteries all in balance. I have a 24 volt system but to save confusion I am using 12 volt battery terms, like you I consider 13.8 volt 80% as I think my batteries from memory can go up to 14.6 volts.

What I can say is that whilst the batteries at low levels of voltage can power up the heating system on the whispergen they cant start it just not enough amps to spin it over. As I have said before all the power in these batteries is above 12.8 volts

[nicknorman]

I’ve been away for a few days. First thing I noticed is that the BMV SoC reading hadn’t changed. So my strategy to do the zero current calibration with with the backlight off, has worked.

 

Second thing I noticed is that 3 of the cells were showing 3.291, one cell was showing 3.281, ie suddenly 10mV lower. Which may not sound a lot, but is significant in terms of Li batteries. Battery terminals checked for tightness and generally wiggled - no effect. Then I unplugged and replugged the BMS. Suddenly the 10mV was restored.
 

So clearly a bad connection was dropping 10mV, despite the connectors still being in my house and never having seen a damp boat. I had use a fairly ordinary 5 way PCB connector with tin plating on the contacts. Tin plating is good for high currents, bad for very low currents such as voltage sensing. Fortunately in my redesigned board which is now winging its way back from China, I decided to use connectors with gold plating which are ideal for very low currents.

 

The point of all that is that it seems very easy to have a sub-optimal connection between cells, or between cells and voltage sensors, and that can throw off measured voltages significantly.

[MoominPapa]

13 hours ago, nicknorman said:

I’ve been away for a few days. First thing I noticed is that the BMV SoC reading hadn’t changed. So my strategy to do the zero current calibration with with the backlight off, has worked.

 

Second thing I noticed is that 3 of the cells were showing 3.291, one cell was showing 3.281, ie suddenly 10mV lower. Which may not sound a lot, but is significant in terms of Li batteries. Battery terminals checked for tightness and generally wiggled - no effect. Then I unplugged and replugged the BMS. Suddenly the 10mV was restored.
 

So clearly a bad connection was dropping 10mV, despite the connectors still being in my house and never having seen a damp boat. I had use a fairly ordinary 5 way PCB connector with tin plating on the contacts. Tin plating is good for high currents, bad for very low currents such as voltage sensing. Fortunately in my redesigned board which is now winging its way back from China, I decided to use connectors with gold plating which are ideal for very low currents.

 

The point of all that is that it seems very easy to have a sub-optimal connection between cells, or between cells and voltage sensors, and that can throw off measured voltages significantly.

I use the same connections from the cells to the BMS for voltage sensing and the 200mA balance currents. This means that the measured voltages are wrong when balancing is taking place, which is a bit of a pain, but maybe there's a hidden advantage in keeping the contacts clean. Do you fuse the wires from the cell terminals to the BMS? Are you going to use gold plates fuses and holders in that role, following this logic?

 

MP.

 

[Dr Bob]

1 hour ago, MoominPapa said:

The point of all that is that it seems very easy to have a sub-optimal connection between cells, or between cells and voltage sensors, and that can throw off measured voltages significantly.

 

1 hour ago, MoominPapa said:

I use the same connections from the cells to the BMS for voltage sensing and the 200mA balance currents. This means that the measured voltages are wrong when balancing is taking place, which is a bit of a pain, but maybe there's a hidden advantage in keeping the contacts clean. Do you fuse the wires from the cell terminals to the BMS? Are you going to use gold plates fuses and holders in that role, following this logic?

 

MP.

 

Is this really an issue? I guess to Nick it is - as 'loosing 0.01V' may have a big effect on his system. I'm with MP here though as I cant afford gold plated fuse holders on the four fused cell voltage sense systems. In practice I have my BMV as the main voltage sensor to isolate the bank which seems to work quite well. The individual cell monitoring only kicks in if a cell gets over 3.8V so I guess 0.05V out wouldnt hurt that much - and I hope I never get to that as the charge sources will cut out long before the BMV isolation which is long before the cell monitor isolation.

[peterboat]

9 minutes ago, Dr Bob said:

 

Is this really an issue? I guess to Nick it is - as 'loosing 0.01V' may have a big effect on his system. I'm with MP here though as I cant afford gold plated fuse holders on the four fused cell voltage sense systems. In practice I have my BMV as the main voltage sensor to isolate the bank which seems to work quite well. The individual cell monitoring only kicks in if a cell gets over 3.8V so I guess 0.05V out wouldnt hurt that much - and I hope I never get to that as the charge sources will cut out long before the BMV isolation which is long before the cell monitor isolation.

Bob we have all lived with our systems for a number of years, we have discovered that cell balancing is for the most a natural event.  Recently I took my batteries down to 9 volts and recharged them slowly over a number of days, for the first time in a long while I have connected my puter to them,  they were all well balanced and showing no ill effects of being taken down to 9 volts, BMSs are flashing green so they will do many more years of reliable service 

[nicknorman]

2 hours ago, MoominPapa said:

I use the same connections from the cells to the BMS for voltage sensing and the 200mA balance currents. This means that the measured voltages are wrong when balancing is taking place, which is a bit of a pain, but maybe there's a hidden advantage in keeping the contacts clean. Do you fuse the wires from the cell terminals to the BMS? Are you going to use gold plates fuses and holders in that role, following this logic?

 

MP.

 

Yes the wires from the cell terminals to the BMS are fused, except for the 0v one. I pass them through a small PCB (cut off from my original BMS board design) which has 4 surface mount fuses. So they are soldered in place. Obviously it will be a bit of a pain to replace a blown fuse but that should not happen unless I’m very careless, in which case serves me right!

 

Yes Im sure passing 200mA through the contacts will help a lot. I plan to have the MOSFETS and resistors remote from the BMS, with separate wires going to the cells. This was based on a presumption that for 600Ah of batteries I would need a fairly hefty balancing system capable of several amps. However the perceived wisdom on here seems to be that balancing is not much required so maybe I’ll scale that back a bit and follow the MP strategy of “little and often”.

41 minutes ago, Dr Bob said:

 

Is this really an issue? I guess to Nick it is - as 'loosing 0.01V' may have a big effect on his system. I'm with MP here though as I cant afford gold plated fuse holders on the four fused cell voltage sense systems. In practice I have my BMV as the main voltage sensor to isolate the bank which seems to work quite well. The individual cell monitoring only kicks in if a cell gets over 3.8V so I guess 0.05V out wouldnt hurt that much - and I hope I never get to that as the charge sources will cut out long before the BMV isolation which is long before the cell monitor isolation.

It is an issue if I am to have an accurate assessment of SoC based on a hybridised parameter using Ah counting (the BMV or Mastershunt) and rested voltage vs SoC data. Gold plating may sound posh but the amount of gold is miniscule and in terms of the additional cost, it is a few 10s of pence for the gold plated connector vs the tin plated one. Choice of contact plating is about the best plating for the application, it is barely at all about cost.

[MoominPapa]

45 minutes ago, Dr Bob said:

 

Is this really an issue? I guess to Nick it is - as 'loosing 0.01V' may have a big effect on his system.

If you pull 200mA down the wires the effect is more like 0.05V, which has a significant effect on the algorithm which decides which cells to balance. It's easy to work round: the computer runs the balancing resistors for four minutes, then switches them off for a minute, checks the cell voltages and decides which cells to balance for the next cycle.

 

MP.