diy solar

diy solar

New Daly "Smart" BMS w/ Communication. (80-250A)

The voltage seen across the shunt is a 120 hz ripple on top of a dc bias.

So, here’s the question:
The average of a ac current is zero. The rms value is greater than zero. In the case of the signal I described, is the current just the dc bias, or greater than that. In other words, the rms value.
Can’t edit previous post. Here’s an example.

The current shunt is showing a dc bias equivalent to 50A. On top of the dc bias is a sine wave at 100A p-p (100A peak to peak). The BMS will be showing currents in the range from 0A to 100A. The average current is 50A. The rms current, I believe is 50 + 0.7 * 50 = 85A.
 
The DC bias is caused by the current flowing through the shunt. This voltage is what is being measured to determine the current. The 'meter' knows the resistance so it can just use ohms law to get current I=V/R.

The ripple is noise caused by the inverter. (120hz is 2x60hz so I would bet you are in North America). Unfortunately, this ripple can cause problems with the reading the meter is getting. Even a one millivolt ripple can create a large swing in the measured current.

Example:
300A-75mV shunt has .075/300 = .25mohm resistance.​
+1 mV: .001/.00025= 4 Amps increase in measurement​
- 1 mV: -.001/.00025= 4 Ampls decrease in measurement​
The total swing between the high and low peak is 8 amps.​

It seems like a lot of inverters (all?) create this noise but not all set-ups see a problem. I have seen the Chargery BMS have this problem with a Victron Multiplus compact, but the ALI meter on the same set-up did not have the problem.

In another thread there was speculation that the sample rate compared to the noise frequency can create an amplification of the problem...???.
 
After thinking it over, battery current is the average value of the waveform and not the rms value.

That means a RC filter should be able to give a correct steady state average.
 
BTW: Because of this problem, I use a large bank of power resisters to generate a DC load (without the inverter) when I am calibrating the Chargery.
 
After thinking it over, battery current is the average value of the waveform and not the rms value.

That means a RC filter should be able to give a correct steady state average.

That seems accurate to me.... as long as the sample rate is not a harmonic of the ripple. In that case your average could be constantly low or high. However, it seems very unlikely that the sample rate of the meter would just happen to be a harmonic of the noise from the inverter.
 
BTW: Because of this problem, I use a large bank of power resisters to generate a DC load (without the inverter) when I am calibrating the Chargery.
OK, you’re the filter guy. Design a filter that attaches to the shunt to eliminate the ripple.

Edit
Frequency is 120 hz and input to adc is high impedance.
 
OK, you’re the filter guy. Design a filter that attaches to the shunt to eliminate the ripple.

Edit
Frequency is 120 hz and input to adc is high impedance.
Well.... I did control circuits on industrial water filters, so it is not quite the same.....

I have thought a bit about what it would take to build a filter but not tried to put pencil to paper.

The best place to fix the problem is in the inverter.... but that is not an option. The next best place is to rebuild the meter to ignore the ripple, but that is not feasible either. That leaves two approaches

1) Filter out the ripple (Next best thing to fixing the inverter). The problem with this approach is that we are dealing with very large currents so the components for the filter will have to be large. The two 'simple' filters is a capacitor in parallel and/or an inductor in series. With the current we are dealing with, an inductor would have to be massive, so that is out. A very large electrolytic capacitor might help but I have not tried to figure out what 'very large' means. (Note: all inverters already have big banks of capacitors....)

2) Filter the ripple out of the signal between the shunt and the meter (Next best thing to fixing the meter). The current and voltages on this signal are very small, so building a filter might be a bit easier to do. However, if the load changes quickly, the filter would tend to 'hide' the changes. Consequently, you would want to be careful not to over-do it with the filter. (You would want a low pass filter that only targets the ripple frequency and above.) The other issue is that if the filter introduces any voltage drop between the shunt and the meter, the meter would constantly read low. (With something like Chargery where you can calibrate the system you might be able to compensate for this)

I'll ponder on this some, but it is going to require dusting off my college training in filter design from a *long* time ago. Consequently, I am not sure I will do it.
 
The voltage seen across the shunt is a 120 hz ripple on top of a dc bias.

So, here’s the question:
The average of a ac current is zero. The rms value is greater than zero. In the case of the signal I described, is the current just the dc bias, or greater than that. In other words, the rms value.
Surely there are ripple filters at the DC input of the inverter , unless the designer could not have bothered . Also , this ripple superimposed on the DC should be very small in amplitude. I have never measured this with a scope and would be very interested to see a photo.

Sorry guys , I posted this without reading all the posts after my previous comment.
I am still wondering about the ripple freq of 120Hz. Will it not be a 60 Hz sine wave with the bottom half inverter to create a 120Hz ripple , so ripple is not a sine wave ? Cannot remember that far back to college days ;) Else , how does the 120Hz signal originate ? Is it just noise that is a by-product of the inverter operation and is it also present on the output waveform ?
 
Last edited:
Finally the seller answered about SOC calculation. After changing design capacity in BMS we have to go to engineer menu in the program and press "restart BMS". Now it is calculating well.
It is so hard to write a small user manual from factory.
 
Last edited:
Well.... I did control circuits on industrial water filters, so it is not quite the same.....

I have thought a bit about what it would take to build a filter but not tried to put pencil to paper.

LI'll ponder on this some, but it is going to require dusting off my college training in filter design from a *long* time ago. Consequently, I am not sure I will do it.

Sorry, mistaken identity. Water filters.... my impression was way off base.

You got the right ideas for a low pass filter. It still needs to be reasonably responsive to changing current. And the BMS probably requires new current calibration. As you mentioned, the inverter already has a huge capacitor bank on the input. Little can be done there.

Depending on the number of measurements the BMS takes per 120 hz cycle, current measurements could be accurate or way off base.
 
Surely there are ripple filters at the DC input of the inverter , unless the designer could not have bothered . Also , this ripple superimposed on the DC should be very small in amplitude. I have never measured this with a scope and would be very interested to see a photo.

The current ripple when operating my microwave (160 A @ 12V) is more than trivial. It is substantial.

I’ve taken a scope plot at the shunt. Don’t recall 100%, but believe it’s 120 hz sinusodial, not rectified 60 hz. I’m traveling and don’t have access to that data.
 
The current ripple when operating my microwave (160 A @ 12V) is more than trivial. It is substantial.

I’ve taken a scope plot at the shunt. Don’t recall 100%, but believe it’s 120 hz sinusodial, not rectified 60 hz. I’m traveling and don’t have access to that data.
Much appreciated -- looking forward to the scope plot. I am sans scope presently , so cannot do the measurement.
 
You can use a basic RC filter on the shunt but the perfs wouldn't be great as it is a first order filter.

Assuming the ADC has a 1 M input impedance and you want an additional error of 0.1 % maximum the filter series resistance should be 1 k or less.

With a 1 k resistor and let's say a 50 Hz cut-off frequency you would need a 3.3 µF capacitor. It would halves the ripple at 60 Hz and divide it by 4 at 120 Hz and the step response would be decent at less than 10 ms to reach 90 % of the input value.

If you want more attenuation you can use 5 of them in series with a R of 180 Ohms and a C of 18 µF for example, but at that point I'd recommend to use LC filters as C goes high because R needs to be low.

To have more attenuation you can also lower the cut-off frequency of the filter but by doing that you'll increase the step response delay.

If the BMS has short-circuit protection you can't use a low pass filter as it'll introduce far too much delay (SC protection should act in µs and we're already at 10 ms with a single RC filter...)
 
If the BMS has short-circuit protection you can't use a low pass filter as it'll introduce far too much delay (SC protection should act in µs and we're already at 10 ms with a single RC filter...)
Excellent point!!!
 
Last edited:
but at that point I'd recommend to use LC filters as C goes high because R needs to be low.

Is it practicle to build an LC filter for this? If we are trying to filter something as low as 120hz, wouldn't the inductor value have to be crazy large?
 
Thanks, I’m thinking LC filter. L might get too big?
Is it practicle to build an LC filter for this? If we are trying to filter something as low as 120hz, wouldn't the inductor value have to be crazy large?

It all depends on the value of the cap. Also a good thing is that we don't have any current so the inductor can have a very low saturation current (smaller, less expensive).

Another good thing is that a LC is a 2nd order filter so it has a slope twice better than the RC filter.

So for a 50 Hz cut-off frequency and a 100 µF cap we would need a 100 mH inductor (don't go past 100 mH as prices tend to skyrocket beyond that). I checked and you can have 100 mH inductors for less than a dollar on Mouser for example, so no problem here. And same for the cap ;)

This inductor has a 235 Ohms DC resistance which means you can't put more than 4 filters of you want to stay under the 1 kOhms we fixed earlier. But that would be a 8th order filter so pretty good already (equivalent to 8 RC filters...).
 
  • Like
Reactions: Cal
Is this the circuit you are thinking?

1594667735300.png

I wonder what the 235 ohm resistance will do to the accuracy? If the input imedence of the meter/bms is 1Mohm it is only .2% of the input impedence so it should not be to bad. Either way, I would want to calibrate the BMS with the filter installed.
 
  • Like
Reactions: Cal
Back
Top