Jump to content

My third SmartGauge...


MtB

Featured Posts

1 minute ago, Tony Brooks said:

Trouble is people with solar need to get up very early at this time of year to read the Smartguage before it gets light - they don't want to. Perhaps they should fit a switch in the panel to controller line so they don't have to but then they sacrifice some solar while they wait for the Smartgage to sort itself out.

I turn the solar on about 8am,but mainly because the fan in the controller wakes me up otherwise. 

Link to comment
Share on other sites

10 minutes ago, dmr said:

Its always been said, by the anti Smartgage faction, that they don't work with solar, but I thought somebody did some tests in real life scenarios and concluded that it was adequate??? Smartgage is an instrument for monitoring battery state during discharge so its not really fair to criticise its inaccuracy during charge.

Smartgage is not particularly good at telling when the batteries are fully charged, its purpose is to tell when they are part discharged or very discharged and I suggest it will do this even if solar is fitted.

Ive only got a little bit of solar by modern standards but my Smartgage gives sensible results.

 

Has your Smartgage given poor results with your solar installation?

 

..............Dave

Yes, the Smartguage is inaccurate when charging which is 50% of what you want it for!  When to start charging and when to stop!  My voltage gauge tells me how full the batteries are as long as I’m not pulling a great amount of amps out of them at he time!

Edited by Robbo
Link to comment
Share on other sites

7 minutes ago, Tony Brooks said:

Trouble is people with solar need to get up very early at this time of year to read the Smartguage before it gets light - they don't want to. Perhaps they should fit a switch in the panel to controller line so they don't have to but then they sacrifice some solar while they wait for the Smartgage to sort itself out.

No, they just need to buy grandson of smartguage, the Masterplex. 

 

It does everything they want, but they may need to sell the boat to afford one ...

 

Link to comment
Share on other sites

Just now, Robbo said:

Yes, the Smartguage is inaccurate when charging which is 50% of what you want it for!  When to start charging and when to stop!

No that really is not what the Smartgage is for, thats what the Ammeter is for. The CO meter measures CO, the water gauge tells you level in the water tank, the smartgage tells you how the batteries are discharging.

 

Actually the CO meter sort of tells you when the batteries are very fully charged but thats for another day. ?

 

...............Dave

Link to comment
Share on other sites

13 minutes ago, Tony Brooks said:

Trouble is people with solar need to get up very early at this time of year to read the Smartguage before it gets light - they don't want to. Perhaps they should fit a switch in the panel to controller line so they don't have to but then they sacrifice some solar while they wait for the Smartgage to sort itself out.

Surely you only need to get up early if you assess state of charge by looking at rested voltage, with a Smartgage you can sleep in as late as you like and still get a reasonable estimate when you do get up.

 

...............Dave

Link to comment
Share on other sites

Just now, dmr said:

No that really is not what the Smartgage is for, thats what the Ammeter is for. The CO meter measures CO, the water gauge tells you level in the water tank, the smartgage tells you how the batteries are discharging.

 

Actually the CO meter sort of tells you when the batteries are very fully charged but thats for another day. ?

 

...............Dave

The smartguage tells you what percentage the batteries are, apart from it’s off when charging!  So pretty useless then if you want to monitor when to stop charging, so you really need a amp meter.  So once you have that and a voltage meter, you have the gauges you need to correctly monitor the batteries and more accurately.

Link to comment
Share on other sites

8 minutes ago, Robbo said:

The smartguage tells you what percentage the batteries are, apart from it’s off when charging!  So pretty useless then if you want to monitor when to stop charging, so you really need a amp meter.  So once you have that and a voltage meter, you have the gauges you need to correctly monitor the batteries and more accurately.

But a voltmeter measures voltage, not state of charge, whilst a smartgage measures and displays state of charge. Are you suggesting that several times every day I turn everything off, including the solar panels, wait several minutes, measure the voltage at the batteries, then use a little look-up table to find the state of charge???? Been there and done that but now I find it much easier, quicker and safer just to look at the Smartgage. ? 

 

I do of course use my ammeter to indicate when to stop charging but its not very accurate if there is any sulphation.

 

.............Dave

Except for mtb is there anybody here who actually has a smartgage and doesn't find it useful/essential?????.

 

.................Dave

Link to comment
Share on other sites

5 minutes ago, dmr said:

But a voltmeter measures voltage, not state of charge, whilst a smartgage measures and displays state of charge. Are you suggesting that several times every day I turn everything off, including the solar panels, wait several minutes, measure the voltage at the batteries, then use a little look-up table to find the state of charge???? Been there and done that but now I find it much easier, quicker and safer just to look at the Smartgage. ? 

 

I do of course use my ammeter to indicate when to stop charging but its not very accurate if there is any sulphation.

 

.............Dave

Except for mtb is there anybody here who actually has a smartgage and doesn't find it useful/essential?????.

 

.................Dave

A smart gauge uses the voltage to gain SoC, nothing else!   To be accurate you need to know what the amps are doing.

 

MTB only knows his smart gauges are off because they all say different things - if it was good they would be roughly the same percentage.

Edited by Robbo
Link to comment
Share on other sites

11 minutes ago, Robbo said:

MTB only knows his smart gauges are off because they all say different things - if it was good they would be roughly the same percentage

Or perhaps he compared them to a calibrated meter. 

Link to comment
Share on other sites

18 minutes ago, Robbo said:

A smart gauge uses the voltage to gain SoC, nothing else!   To be accurate you need to know what the amps are doing.

 

MTB only knows his smart gauges are off because they all say different things - if it was good they would be roughly the same percentage.

 

Yes, if the smartgage had included a shunt and its clever voltage prediction method then it would have been the perfect instrument (Simarine claim to have done this?). Trouble is people already say smartgage its a bit costly so you can't win.  I note that many boaters use a low cost shunt based Chinese ammeter/amp hour counter for battery monitoring (are you aware of the "12 volt boating group" on Facebook?).  I note that some variants have a current accuracy of worse than 5% so are probably less accurate than the Smartgage without a shunt.

 

.............Dave

Link to comment
Share on other sites

6 minutes ago, dmr said:

 

Yes, if the smartgage had included a shunt and its clever voltage prediction method then it would have been the perfect instrument (Simarine claim to have done this?). Trouble is people already say smartgage its a bit costly so you can't win.  I note that many boaters use a low cost shunt based Chinese ammeter/amp hour counter for battery monitoring (are you aware of the "12 volt boating group" on Facebook?).  I note that some variants have a current accuracy of worse than 5% so are probably less accurate than the Smartgage without a shunt.

 

.............Dave

I currently have the simarime, at the moment I wouldn’t recommend as it’s meant to be a little more than a gauge with all the smart features, but they are not their yet.   Give it a year or two.

 

im aware of the 12v Facebook group, but didn’t care for it for one of the admins.  Facebook is poor for forum based content as well so that doesn’t help.

Edited by Robbo
Link to comment
Share on other sites

7 minutes ago, dmr said:

Yes, if the smartgage had included a shunt and its clever voltage prediction method then it would have been the perfect instrument...

It’s been available for several years now and it’s called Datacell but it’s 10 times the cost of a SmartGauge. To be fair, it does a lot more than simply give you the SoC. Obviously, boaters aren’t its target market. 

Link to comment
Share on other sites

7 minutes ago, Robbo said:

I currently have the simarime, at the moment I wouldn’t recommend as it’s meant to be a little more than a gauge with all the smart features, but they are not their yet.   Give it a year or two.

 

im aware of the 12v Facebook group, but didn’t care for it for one of the admins.  Facebook is poor for forum based content as well so that doesn’t help.

12v group is a bit of a one man show but he does know his stuff, a few other posters talk a bit of the usual rubbish. The Simarine is a bit costly but looks to have potential but like you I think its not really ready yet. I sort of get the impression it is not advancing at present????. Nice concept and graphics but too much peripheral stuff and not enough battery monitoring?  I will keep watching.

 

..............Dave

Link to comment
Share on other sites

12 minutes ago, dmr said:

Simarine claim to have done this?

I believe they claimed that they ‘would’ do this. And then they discovered how difficult it is to do and went quiet about it. SmartGauge remains the only accurate SoC gauge currently on the market anywhere in the world. As noted in the manual it can be up to 10% adrift when charging (but is usually closer than that), so measure tail current to determine ‘full’. 

1 minute ago, dmr said:

12v group is a bit of a one man show...

It is now since the guy with odd ideas who used to post here has left the group. I’m chipping in here and there. 

Link to comment
Share on other sites

1 minute ago, WotEver said:

I believe they claimed that they ‘would’ do this. And then they discovered how difficult it is to do and went quiet about it. SmartGauge remains the only accurate SoC gauge currently on the market anywhere in the world. As noted in the manual it can be up to 10% adrift when charging (but is usually closer than that), so measure tail current to determine ‘full’. 

It has its odd moments, sometimes it sticks (falls slowly) for a bit then sort of catches up with itself but I like the concept of an instrument that corrects itself. We often do a longish but low and steady current draw which might give it a hard time, though the fridge is always cycling. I am still impressed with it, it does its job well. I really don't care if my batteries are at 48% or 52%, as long as I have every confidence that they really are between 48 and 52!

 

.................Dave

Link to comment
Share on other sites

2 minutes ago, dmr said:

12v group is a bit of a one man show but he does know his stuff, a few other posters talk a bit of the usual rubbish. The Simarine is a bit costly but looks to have potential but like you I think its not really ready yet. I sort of get the impression it is not advancing at present????. Nice concept and graphics but too much peripheral stuff and not enough battery monitoring?  I will keep watching.

 

..............Dave

It does battery monitoring okay and the last update was a big step compared to the ones last year which were more frequent but probably more bug fixing.   Historical graphs, amp counter and ability to connect to a WiFi router are the features I’m waiting for.

Link to comment
Share on other sites

3 minutes ago, Robbo said:

It does battery monitoring okay and the last update was a big step compared to the ones last year which were more frequent but probably more bug fixing.   Historical graphs, amp counter and ability to connect to a WiFi router are the features I’m waiting for.

I thought it did amp-hour counting already? I am not so bothered about built in historical display because I hate working with small screens, I would want to download the historical data to a PC, but wifi or a cable would be fine.

 

............Dave

Link to comment
Share on other sites

Just now, dmr said:

I thought it did amp-hour counting already? I am not so bothered about built in historical display because I hate working with small screens, I would want to download the historical data to a PC, but wifi or a cable would be fine.

 

............Dave

It doesn’t show it currently, but it is coming.  Amp counters for other devices as well.  At the moment the gauges are very much current used now where it should also be how much is used historically.  Remember this is a device to monitor individual items like fridges and lights and daily usage is more relevant and useful rather than current usage.  Export of data would be nice as well, I’ll add that to the list as well.

Link to comment
Share on other sites

33 minutes ago, WotEver said:

It’s pretty, I’ll give it that :)

Cue bad words about pretty user interfaces being developed before core functionality. 

 

If I had a pound for every time I saw that I would have errr about a hundred quid! 

 

 

Link to comment
Share on other sites

4 hours ago, jonesthenuke said:

I These units do not measure anything but the battery voltage. They they cannot infer SoC from just the voltage at a point in time as you would also need to measure the battery temperature. The algorithm must therefore calculate SoC from the changes in voltage over time. Consequently,a fixed error will not matter, its much the same as operating the battery at a different temperature.

 

Now if the error is varying over time than that would indeed make a difference.

 

Perhaps one day the algorithm will be published and then we will know how it works?

 

 

If you understand multi linear regression modelling, it is not difficult to work out how the model works.

The SG DOES NOT measure only voltage. It measures voltage and time. The SG has therefore at any instant the voltage, the change in voltage with time and likely the change in change of voltage with time (the second differential). To work out the algorithims, you test lots of batteries in discharge and charge and measure voltage, time and estimate SoC. You then do some fancy stuff with modelling software to give you linear relationships between SoC and the variables (as above). The test data is called the training set. You then get some more data and test the linear models with the new data to make sure its working. I would expect the model uses actual voltage and its differentials. For example, if the unit sees the voltage at 12.5V, steady for an hour, it will assume it is not charging or discharging and therefore be able to predict an SoC. If it sees as varying rate of reduction in volts, it will assume the batteries are discharging and by comparison to the data (the linear relationship lines), it will be able to predict a decreasing SoC. It will therefore be very important to have an an 'accurate' voltage measurement otherwise its steady voltage wont mean anything and will give poor SoC readings. Of course the model will also be able to spot the voltage varying with charging if there is a load on the batteries...likely most of the time. It can work out what is happening.

This type of model however depends on the data used to build the training sets. To me it is no surprise the model is accurate on discharge as Gibbo (I understand) did thousands of tests to destruction. It will have been much easier to get discharge data. Likely he had much less charging data and as different charge devices perform differently when getting to 95% SoC  then the training set was likely not as robust. For example, an alternator will kick out 14.4V with an nice low tail current till the cows come home. Pretty easy to model. My battery charger, and everything else goes into float at the drop of a hat and that will be very difficult to get enough data. Remember, it is easy to model an individual system - I could probably do it for my own boat - but to have a GLOBAL model that is good for everyone and all the different systems out there, needs a huge amount of data and particularly for the charging side. Global models and multi linear regression are usually  not used in the same sentance. It is no surprise there will be a lack of accuracy up in the high 90%s during charge.

Will it work with Solar? Yes probably to a point. The SG doesnt care if its solar. If enough data with solar charging was in the training set then it will give reasonable results. It will be like all the other 'controllers' that go to float too soon.

My cynicism of the SG is that I dont believe global models work well with multi linear regression .....and that seems to be what is happening when charging over 90% SoC. It looks like however the SoC prediction under discharge is much easier to model (so it works) and I do like the fact that during the night when the voltage is constant, it can do an 'at rest' calculation or an estimate if say there is a contstant 1A draw (like my boat - inverter on, fridge compressor not running). I can do this by a glance at the voltmeter in the winter but never get up early enough in the summer. The SG does however need a fairly accurate voltage reading.

I do have a lot of experience of these types of models. We used them to run our Steam Crackers/ethylene plants and polyethylene plants. We used them to measure RON and MON of petrol in the refinery before the product was released for sale.....some far more difficult problems than batteries. Hats off to Gibbo for doing all the testing that allowed him to construct the models.

This isnt rocket science.

 

 

Edited by Dr Bob
Link to comment
Share on other sites

You are missing a crucial point here though.

Yes the smartgage is most likely using rate of change of voltage as well as voltage to deduce battery state of charge, I also suspect its curve fitting (oops doing regression analysis) to the the exponential voltage changes in response to current changes, But, the reason it struggles with charging is not because Gibbo did not have enough charging data, its because charging voltage is controlled by the alternator regular Not by any battery characteristics so there is simply no real battery related information available, the instruction book actually says this. So its just a best guess from basic generic information..."at 14.4 volts a 400Ah bank will take 3 hours to charge" sort of thing.

 

..............Dave

Link to comment
Share on other sites

5 minutes ago, dmr said:

You are missing a crucial point here though.

Yes the smartgage is most likely using rate of change of voltage as well as voltage to deduce battery state of charge, I also suspect its curve fitting (oops doing regression analysis) to the the exponential voltage changes in response to current changes, But, the reason it struggles with charging is not because Gibbo did not have enough charging data, its because charging voltage is controlled by the alternator regular Not by any battery characteristics so there is simply no real battery related information available, the instruction book actually says this. So its just a best guess from basic generic information..."at 14.4 volts a 400Ah bank will take 3 hours to charge" sort of thing.

 

..............Dave

Yes indeed. I believe the instruction book reads something like “SmartGauge will be measuring the charger, not the battery”

Link to comment
Share on other sites

1 hour ago, Dr Bob said:

If you understand multi linear regression modelling, it is not difficult to work out how the model works.

The SG DOES NOT measure only voltage. It measures voltage and time. The SG has therefore at any instant the voltage, the change in voltage with time and likely the change in change of voltage with time (the second differential). To work out the algorithims, you test lots of batteries in discharge and charge and measure voltage, time and estimate SoC. You then do some fancy stuff with modelling software to give you linear relationships between SoC and the variables (as above). The test data is called the training set. You then get some more data and test the linear models with the new data to make sure its working. I would expect the model uses actual voltage and its differentials. For example, if the unit sees the voltage at 12.5V, steady for an hour, it will assume it is not charging or discharging and therefore be able to predict an SoC. If it sees as varying rate of reduction in volts, it will assume the batteries are discharging and by comparison to the data (the linear relationship lines), it will be able to predict a decreasing SoC. It will therefore be very important to have an an 'accurate' voltage measurement otherwise its steady voltage wont mean anything and will give poor SoC readings. Of course the model will also be able to spot the voltage varying with charging if there is a load on the batteries...likely most of the time. It can work out what is happening.

This type of model however depends on the data used to build the training sets. To me it is no surprise the model is accurate on discharge as Gibbo (I understand) did thousands of tests to destruction. It will have been much easier to get discharge data. Likely he had much less charging data and as different charge devices perform differently when getting to 95% SoC  then the training set was likely not as robust. For example, an alternator will kick out 14.4V with an nice low tail current till the cows come home. Pretty easy to model. My battery charger, and everything else goes into float at the drop of a hat and that will be very difficult to get enough data. Remember, it is easy to model an individual system - I could probably do it for my own boat - but to have a GLOBAL model that is good for everyone and all the different systems out there, needs a huge amount of data and particularly for the charging side. Global models and multi linear regression are usually  not used in the same sentance. It is no surprise there will be a lack of accuracy up in the high 90%s during charge.

Will it work with Solar? Yes probably to a point. The SG doesnt care if its solar. If enough data with solar charging was in the training set then it will give reasonable results. It will be like all the other 'controllers' that go to float too soon.

My cynicism of the SG is that I dont believe global models work well with multi linear regression .....and that seems to be what is happening when charging over 90% SoC. It looks like however the SoC prediction under discharge is much easier to model (so it works) and I do like the fact that during the night when the voltage is constant, it can do an 'at rest' calculation or an estimate if say there is a contstant 1A draw (like my boat - inverter on, fridge compressor not running). I can do this by a glance at the voltmeter in the winter but never get up early enough in the summer. The SG does however need a fairly accurate voltage reading.

I do have a lot of experience of these types of models. We used them to run our Steam Crackers/ethylene plants and polyethylene plants. We used them to measure RON and MON of petrol in the refinery before the product was released for sale.....some far more difficult problems than batteries. Hats off to Gibbo for doing all the testing that allowed him to construct the models.

This isnt rocket science.

 

 

It isn’t quite as simple as just a model, there is also a “learning” feature whereby the various coefficients (or variables) in the equation can be tweaked to improve prediction. So for example with the SG measuring a certain dV/dt and probably d2V/dt2 it will assume a certain relative discharge rate and so ramp down the SoC as a consequence. But when the voltage flatlines it can do an empirical measure of SoC and, if there is some difference between this instantaneously measured SoC and the previously calculated one, it can tweak the coefficients so that next time the rates of voltage change will be interpreted better.

 

The way the SG measures voltage is by timing the charge (or is it discharge?) of a capacitor - this gives a very high resolution measure, but not a particularly accurate one. The high resolution is of course required to calculate small rates of change.

Link to comment
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
  • Recently Browsing   0 members

    • No registered users viewing this page.
×
×
  • Create New...

Important Information

We have placed cookies on your device to help make this website better. You can adjust your cookie settings, otherwise we'll assume you're okay to continue.