diy solar

diy solar

Chargery BMS16T Teardown and circuit analysis? Anyone?

I don’t think getting scope measurements will help much. I would buzz it out and generate a schematic. The heart of the BMS is the cell voltage measurement.
The scope is to measure noise on the inputs and the ADC input and a few places along the chain. I don't really care about the schematic as much as understanding what is going in and what is coming out.

I have a 6Ghz low-noise scope with great probes which allows me to see before and after. If the bulk of the apparent noise is in the analog domain, I may just bodge a filter in there. Obviously, I don't need 6Ghz, but the low-noise spec is handy for this along with a pair of LeCroy differential amplifiers.

Like I have mentioned before - if the analog section sucks, the ADC can't bail you out.
 
Like I have mentioned before - if the analog section sucks, the ADC can't bail you out.
My tests have shown voltage measurements are fairly good. What’s wrong with 5 mV accuracy?

Of course we need to power the BMS externally to eliminate the load errors seen on the two end cells.

Thermal drifts also look good.
 
5mV accuracy is no problem, instability and noise sensitivity are a problem. I see numbers bouncing all over the place which means the system is noisy. A system that is sensitive to noise can make poor decisions and poor decisions can damage batteries or itself.

I don't think it is severe, but I also don't know what is happening in the microcontroller that can cause trouble. In general, I design control systems with critical control happening in the analog domain. The microcontrollers are there to monitor and adjust various set points, but the important stuff will keep working even if the micro controller is in non responsive.

It seems that the software is not attempting any sort of DSP on the ADC data, so perhaps some mods on the analog side could help feed it better data.
 
5mV accuracy is no problem, instability and noise sensitivity are a problem. I see numbers bouncing all over the place which means the system is noisy. A system that is sensitive to noise can make poor decisions and poor decisions can damage batteries or itself.

I don't think it is severe, but I also don't know what is happening in the microcontroller that can cause trouble. In general, I design control systems with critical control happening in the analog domain. The microcontrollers are there to monitor and adjust various set points, but the important stuff will keep working even if the micro controller is in non responsive.

It seems that the software is not attempting any sort of DSP on the ADC data, so perhaps some mods on the analog side could help feed it better data.
I have seen an increase in stability of the displayed voltage and current data with the 4.02 ... now the 4.03 update.
I am also getting very accurate cell voltage readings since I powered directly from the battery thru external the external power jack. I took the battery negative from the downstream side of the shunt so the BMS and relay current draws will be factored into the current measurements.
 
I have seen an increase in stability of the displayed voltage and current data with the 4.02 ... now the 4.03 update.
I am also getting very accurate cell voltage readings since I powered directly from the battery thru external the external power jack. I took the battery negative from the downstream side of the shunt so the BMS and relay current draws will be factored into the current measurements.
That is great news!
I have a trip coming up and will leave my unit as-is until after I get back. FW updates make me nervous and I don't have a backup that would arrive in time if my unit is bricked for any reason.

Another thing I am planning to do is physically move my shunt and the BMS very close to minimize the noise. The wiring would be very short and I can make a shield based on the input opamp reference. This will reduce the antenna nature of the long shunt wiring - giving the system a fighting chance of getting more accurate data at fairly low currents. Since I have a Victron Smart Shunt in the system, the Chargery shunt and MOSFET SSR are just a backup electronic fuse - I only refer to the SOC on the Victron unit.

I am very curious if this system can be accurate enough to believe the SOC meter on the display. I think it needs better accuracy and stability in the shunt measurement as well as the battery voltage to do that well enough to eliminate the need for the Victron unit.
 
Jason is saying the measurements will be greatly improved in the new product .... maybe even .05 ma resolution. I have been very disappointed by the SOC tracking up to now, but have seen a lot of improvement with 4.02.
Are you following this thread?
 
This Deep Dive Analysis is pretty interesting but I gotta ask guy's where are you going with this ?
What's the intent ?
Make suggestions for improvements in the electronics ?
Or to reverse engineer to make your own version ? {tricky ground there}
 
This Deep Dive Analysis is pretty interesting but I gotta ask guy's where are you going with this ?
What's the intent ?
Make suggestions for improvements in the electronics ?
Or to reverse engineer to make your own version ? {tricky ground there}

The intent is to take a look at how a complex system works because it is interesting to some - including me of course.

This is not an effort to reverse engineer to make a competitive product. Even if it was, that is not tricky ground. From the perspective of an electrical engineer, there is no magic here. There are no special custom ASICS or novel circuits. There are a nearly endless number of ways to design one of these and I wanted to better understand this one since I own it and use it. From the perspective of reverse engineering for profit, anyone could buy this and copy the hardware in a few days. The firmware would be the much bigger project to directly copy. Probably faster and cheaper to write the code from the ground up IMHO. Either way, by the time some shady company did that and got it to the market - the new whiz-bang Chargery model would be on the market which is likely to be better in many respects. There is little incentive to copy, even though it would be easy to do so. Even if ALL the IC markings were laser obscured - I could still reverse this design. It would just slow me down a few more days. Chargery wins because they are already moving on a new and improved design, they have the customer base, the back office operations, the supply chain, assembly/test, distribution, etc, etc. A newcomer would have an older design, no customers, and have to prime the pump on all the other elements that make a manufacturing operation work.

1. I would like to see if I can modify it to work better for my needs.
2. I would like to provide a detailed design review to Chargery in hopes that they can take a step up in the future.
3. Provide some satisfaction of curiosity to technically inclined forum members.

I love the concept of the system, but also want better performance and considerably more robustness if at all possible.
 
As you have Noticed Jason is accepting of critical analysis & suggestions which are practical & applicable to improve on the products and to be abl to create & deliver something that people need and with the features they want. THIS IS A GIFT PEOPLE ! most would not consider such a level of interaction with "clients" nor be as responsive. These were aimed at E-Bikes and EV's, I came across Chargery early last year when I was considering Volt/Bolt Packs and other things, then I introduced Chargery to this forum when it was still a "freshie" and the rest is history.

Anyone see ANY other BMS Makers & Producers interacting with folks here on such a level ?
Has anyone asked general BMS makers to change their product or make something new for us (see DCC) and actually see it come to fruition quickly ?

Consider the fortune of having such access which accepts input suggestions & acts on them... It IS UNCOMMON. Mess it up and everyone loses.
 
As you have Noticed Jason is accepting of critical analysis & suggestions which are practical & applicable to improve on the products and to be abl to create & deliver something that people need and with the features they want. THIS IS A GIFT PEOPLE ! most would not consider such a level of interaction with "clients" nor be as responsive. These were aimed at E-Bikes and EV's, I came across Chargery early last year when I was considering Volt/Bolt Packs and other things, then I introduced Chargery to this forum when it was still a "freshie" and the rest is history.

Anyone see ANY other BMS Makers & Producers interacting with folks here on such a level ?
Has anyone asked general BMS makers to change their product or make something new for us (see DCC) and actually see it come to fruition quickly ?

Consider the fortune of having such access which accepts input suggestions & acts on them... It IS UNCOMMON. Mess it up and everyone loses.

I would argue that fortune is with Jason for participating in this forum and getting actionable data (for FREE) to bolster his business interests. Other companies have the same option to join the conversation but choose not to for whatever reasons they have.

The very foundation of my own business is constant interaction with customers to best understand what they need. If I ignored customers, I would be guaranteed to fail. While I am certainly happy to see the direct benefit to the forum members, it should also be established that this is a MUTUALLY beneficial arrangement where everyone wins.

All I really want is a solid performing BMS that is robust at a cost that I can afford for personal use. Sure, I could design my own but that would take many many hundreds of hours plus big money testing and validating which is totally and completely off the table of options for me.

We are fortunate for Jason's participation.
Jason is fortunate to have our participation.

Professionally, companies pay $250-500 per hour of qualified consulation to optimize products. Chargery is getting free money. We get a better product. WINNING.
 
Last edited:
Here is some data of Cell 1 and Cell 4 voltage I took over 24 hours. Temperature scale is to the right in deg.C. The first thing you'll not that there's no thermal drift. Voltages are rock solid (with respect to temperature). Cell 1 has a noise floor of 2 mV while Cell 4 has a noise floor of 3 mV. Once in a while you'll see a larger spike. That may come from switch noise within the processor. Time scale is in minutes. It was a warm night, temperature went down to only 10C.

Do you have an issue with those measurements?

Chargery24hrs.jpg
 
That is really, really noisy. If a $10 DMM had data logging, you would likely see a nearly flat line.

Short term raw data would be more interesting to me to evaluate noise. It would also be more interesting to me to see what happens when I connect a voltage reference to the inputs along with a arbitrary signal generator to inject noise.

The system sensitivity to noise is my biggest concern over absolute accuracy - which I don't really care about. Stability, stability, stability. That is the name of the game in power electronics.

If I inject a 1khz signal in one of the cell inputs at a rather low level - I would expect to see nearly nothing by the time it arrives at the ADC input. Based on what I see in the hardware, I doubt that will be the case. The experiment on the bench is to setup a signal generator that is inductively coupled to the the inputs to inject pulses and noise so that I can measure the difference between the input and output. It should be nearly out of the range of a scope and in the realm of my vector network analyzer.

One interesting observation is to time correlate the actual analog inputs with the data on the output. It is possible the analog side is fairly clean and the noise is from the mux switching or some other timing issue. If the mux timing is not optimized with the ADC sampling - it will be reading settling or switch noise. Considering how low the bandwidth is - the final out put should never see any deviation over a mV. I see 20mV of deviation.

Also, the noise appears to be triggering rapid changes in the balancing scheme. That would also be a source of noise is physically very close to the LM358 opamp with no ground guarding or via stitching to protect it.

A look with the scope will help me better understand what kind of noise exists in the analog domain which is the first step. It will be time consuming to setup, but I will try in the next day or so.
 
Last edited:
I don't think the plot is all that bad. You need to look at the application. Cell charge or discharge slopes are extremely flat in the 100 to 200 AH cells we're using. That means voltage changes very slowly. The noise can be filtered. Noise is not a factor.

You're not going to take your daughter to kindergarten in a 12 million dollar tank when a Chevy would suffice. That's an old Confucius proverb.;)

Here's a plot of the first 10 minutes. Data is taken every 5 seconds.

Vcell1.jpg

I'm more concerned that Chargery uses the voltage sense wires to conduct 1.2A during balancing. I've already shown the 150 mA used to power the BMS contributes to a measurement error in cells 1 and 4.
 
I don't think the plot is all that bad. You need to look at the application. Cell charge or discharge slopes are extremely flat in the 100 to 200 AH cells we're using. That means voltage changes very slowly. The noise can be filtered. Noise is not a factor.

You're not going to take your daughter to kindergarten in a 12 million dollar tank when a Chevy would suffice. That's an old Confucius proverb.;)

Here's a plot of the first 10 minutes. Data is taken every 5 seconds.

View attachment 29867

I'm more concerned that Chargery uses the voltage sense wires to conduct 1.2A during balancing. I've already shown the 150 mA used to power the BMS contributes to a measurement error in cells 1 and 4.

Noise is a huge factor when the system is making decisions based on the signal buried in it.

I don't see anything fatal here when only looking at cell voltage measurement stability, I just see a LOT of room for easy/cheap improvements. Nothing expensive at all. When I look at a design and the first thing I see raises questions - I will not trust anything else in the system until it proves itself. It is rare that a design is terrible in one area and amazing in all others.

In almost any design, all the elements add-up to what level it will perform. Without taking a single measurement or even powering the system up, I can already see that some things are not good. Personally, I would reject this if one of my engineers brought it to me, regardless of the application or intended use. It is too easy to improve without changing a single component. Add a few jelly bean parts and it could be a lot better than that. If that corner was cut, what other corners? What about things that are not visible like firmware? I am about to connect $4k worth of batteries to it and take it out into the wild for a few weeks - do I trust it? Nope, not yet. Not when the easy stuff seems to be passed over.

Perhaps the differnce here is that you seem to fixate on specific details while I am at 30,000ft looking at the system as a whole.
 
Noise is a huge factor when the system is making decisions based on the signal buried in it.

If the noise is filtered in SW before the decision making then it's fine (obviously filtering at the HW level would be better...).


In almost any design, all the elements add-up to what level it will perform. Without taking a single measurement or even powering the system up, I can already see that some things are not good. Personally, I would reject this if one of my engineers brought it to me, regardless of the application or intended use. It is too easy to improve without changing a single component. Add a few jelly bean parts and it could be a lot better than that. If that corner was cut, what other corners? What about things that are not visible like firmware? I am about to connect $4k worth of batteries to it and take it out into the wild for a few weeks - do I trust it? Nope, not yet. Not when the easy stuff seems to be passed over.

I can't agree more with that :love:
 
Perhaps the differnce here is that you seem to fixate on specific details while I am at 30,000ft looking at the system as a whole.
Absolutely, I'm looking at specific details.

When Chargery neglects to measure current in the +/- 1A region it impacts the SOC display. When I'm out camping and Chargery says the battery is at 95%, but in reality it's only at 20%, it has major consequences.

When Chargery uses the voltage sense lines to power the BMS, it can have major consequences on cell voltage measurements.

I see no adverse operation effects from the noise seen in the plots.

You really haven't show that you're looking at the system as a whole.
 
Absolutely, I'm looking at specific details.

When Chargery neglects to measure current in the +/- 1A region it impacts the SOC display. When I'm out camping and Chargery says the battery is at 95%, but in reality it's only at 20%, it has major consequences.

When Chargery uses the voltage sense lines to power the BMS, it can have major consequences on cell voltage measurements.

I see no adverse operation effects from the noise seen in the plots.

The first thing I looked at (just random really) was the analog input chain from the cell sense lines. I saw problems, not necessarily fatal problems - but they were enough to assume the system has other issues. You just pointed that out yourself. Things that I have not even looked at seem to have problems by your own observations and measurements. This is my point. This is the value of looking at this from a system level.

That is why I worry about the little stuff being passed over. Not because 20mV of noise will make the system fail on its own, but because that is added to a number of other deficiencies in the design that will ultimately prevent it from doing its stated job - PROTECT THE BATTERIES.
 
Back
Top