diy solar

diy solar

Using Super Caps after a DC-DC to supply inrush current on inverter

I don't know but searching this I also found many special automotive audio capacitors that look fancy and have build in volt meters. Mostly 1F - 4F...and one claiming 18F but in the comments on amazon it was revealed to be 0.18F I think.

Yep and there's probably more scams involving these "fancy audio capacitors". This guy says most of the tube is filled with sand/etc. in order to look sturdy and it costs 50-70 dollars.

 
The other and probably much cheaper option is actually connecting that second Orion to see if that gives it enough stable voltage to solve the problem! I do think my 4mm2 wire is capable of 20A at 20 meters without heating up....at least for a short while.
Actually unless your wire run is buried inside the framework of a vehicle or something, the easiest thing by far which presents a 'permanent' solution is simply to increase your wire size. That doesn't even necessarily mean removing the existing wiring because you could parallel another matching pair onto it and still cut the resistance of the wire run in half.

And talk about 20a without heating up.. yes, but that is over a long time frame and you already know the problem you are addressing only exists on a short time frame. The wire is NOT adequate in that short time frame.

What i mean by that is.. voltage drops proportionally to resistance. So if something is 1% of circuit resistance it drops 1% of the volts, if it's 90% of circuit resistance it drops 90% of the volts. During normal steady-state operation your wiring between the boost converter and the inverter may be dropping an acceptably low proportion of the volts. But, when you try to power a device with a large inrush current, consider the resistance. If voltage doesn't change, the only way you GET high inrush currents, is by presenting a lower resistance/ESR or whatever term you like. As the apparent resistance of the inverter drops, so does its proportion of total resistance. So even though the 4mm2 wire has the SAME resistance, it's proportion of the total resistance in that circuit goes way up, and it briefly drops much more of the voltage. That is what is causing the inverter to go into low voltage shutdown.

So in a sense the wire run is probably the main problem. But, if the majority of your use falls in a range that the wire can comfortably support, i understand trying to address this inrush issue through other avenues (which are more interesting by far than just 'throw more money/copper at it'). Just wanted to make you aware that the 20m of 4mm2 was not blameless or irrelevant to the issue.
 
If you really want to measure it, you need an oscilloscope to measure the voltage drop across a shunt. From this you divide by the known resistance of the shunt and get the current. You can use the other channel to simultaneously see the voltage drop of the 24V supply during the inrush current. Then you could measure the total energy of the current surge and size the cap to supply that same energy. I have a picoscope. The cheapest one they sell would work fine. But any oscilloscope would work really.

You could also consider adding an NTC thermistor when you use the ps5 and the charger. You put these in series with the device you are powering. It basically has a few ohms of resistance until it heats up in a second or two, then the resistance goes to almost zero. So at first the inrush current is reduced because of the resistance, but then it operates normally after a second. The total energy of the inrush would be the same (or actually slightly more energy because the thermistor is consuming a small amount of energy too) but it would be a lower current spread over a longer time.
 
I have never implemented an NTC inrush limiter diy, although ive opened a bunch of my devices and found them on there from factory.

What i dont know, which might be relevant here, is if all ‘inrush’ type of loads, are ok with being fed through an ntc and ‘starting up’ at low voltage. For example in a car you can have a starter battery dropping to 8 volts while cranking, and even if the engine reaches sufficient speed it may not start because the control module is not working on only 8 volts. A motor cares less about the voltage ramping up as the ntc’s resistance (thus voltage drop) drops off, but is it ok to feed all types of electronics through an ntc? I have this feeling it is like a brownout in reverse and some types of components may not be happy with it?

I dont know, that’s why im asking. Could be a made up problem. ?
 
Last edited:
What i dont know, which might be relevant here, is if all ‘inrush’ type of loads, are ok with being fed through an ntc and ‘starting up’ at low voltage.

You bring up a good point. I've been looking into this a little since I'm planning to have a refrigerator be the main load on a small solar system for a farm stand so I basically need to size the inverter in the system just to power that. Most people say just size it large enough to handle the input surge. I tested a little wine fridge with my clamp meter and it draws only about 50W when running but the peak input surge was 6A@120Vac measured with my clamp meter! That's 720W or 14x it's running watts. I think a commercial fridge might be several times larger. So I'm looking into ways to reduce it. There doesn't seem to be much info online about it, but one thing I read is that the compressor might need that much current and torque to start. If that's the case there's not much that can be done to reduce it. I'm going to hook up my scope and try to better measure the transient input surge and try some techniques to reduce it and see if it affects the fridge.
 
So I did some "lab" testing in my shed on a little 120V R600a Wine Fridge I picked up at a garage sale for $20. I didn't really mind if I killed it but so far it's still fine. I hooked a 0.5ohm power resistor in series with the neutral wire and measured the voltage across that resistor with my picoscope as the compressor started up. Be careful if you try to do this as the ground connector of oscilloscope probes is usually earth grounded and high voltages are involved. So you have to know what you are doing (I'm an electrical engineer). I then repeated the test with 2.5ohms of resistance in the neutral line. Now we can see the actual startup inrush over the first few seconds in great detail!

blue half ohm - red two and a half.PNG
The top plot shows instantaneous current on the y-axis and time in seconds on the x-axis. In the bottom plot I've done a calculation in the software to display RMS current from the two runs. The blue line is with just the 0.5ohm resistor in series with the neutral and the red one is the 2.5ohm resistor. I'll make some comments here:

1. The inrush current is huge! 7.8A for about a quarter second and then 5.5A for another 1.5seconds, then it drops down to it's final value of about 0.65A. The nameplate of the fridge says "Rated Current: 1.2A". hmmm.
2. This fridge compressor uses a Positive Temperature Coefficient thermistor to startup. The PTC initially has about 5ohms of resistance in series with the start coil. As it heats up from the current, the resistance increases into the kiloohm range and that limits the current in the start coil. The run coil is always on. At least that's how I think it's supposed to work. I believe the first plateau at 7.8A is both start and run coils engaged but before the motor has started moving, the second plateau at 5.5A is the both coils but the compressor turning and the final 0.65A is after the PTC turns the start coil off.
3. Adding 2 more ohms of resistance definitely reduced the peak inrush current by 12%. It took the PTC longer to turn off the start coil though. The compressor didn't have any problem starting but I'm sure the torque was reduced, so there could be an issue with a compressor that was only marginally starting wouldn't work with extra resistance. Another way to think about the extra resistance is it's like a 250ft of 16AWG extension cords. Ouch.

I'm not sure if using an NTC to reduce the start current is worth it, but I might order a few and test it out. The real solution is either a DC compressor or one of those new fridges that have inverters built in.
 
EXCELLENT info! Yeah the proportion of that inrush to the ‘running’ amps is ludicrous.

Its too bad its not quite so easy to trace rpm to see whether the compressor has reached its full speed far earlier than the ptc is ‘turning off’ the start coil. Although i guess any kind of spinning electromagnetic device like a motor is going to put some trace of ‘back emf’ effect on the current waveform that rpm could be extrapolated from somehow. Or at least where it flatlines. Im not proficient in scope use so i might be talking out my ass.

Given the ‘scale’ of the inrush it doesnt seem practical to greatly diminish it, and an external NTC in series with both windings is ultimately going to extend how long the ptc takes to heat up, which might be a net negative to the goal of getting past the inrush before an inverter cries about it. Its probably not practical to intercept the start winding since it’s probably paralleled off the leads inside the sealed compressor housing (is it?). So all you can really hope to do is ‘get lucky’ that a 12, maybe 20% reduction in inrush is enough to get it below the ‘failure threshold’ of a particular inverter, which is a LOT more fiddly and less reliable of an approach than just buying a larger inverter.

Trying to find the sweet spot of diminishing inrush might also be complicated by having to figure out if the conpressor controls are ‘smart’ enough to not try activating the compressor before the refrigeeant system pressures have equalized. I imagine in normal compressor cycling the pressure would have time to fully drop, making the ‘starting torque’ requirement pretty static. But if it allowed it to ‘short cycle’ you might also run into a situation where it will start against the ‘normal’ headwinds of the static system pressure, but be ‘locked rotor’ if it tries to start while the high side still has some pressure built up from the last compressor run. Bleh

Ill be curious to see any future NTC results but it’s looking like a tough road on this one. Still awesome that you did the test.
 
Last edited:
i guess any kind of spinning electromagnetic device like a motor is going to put some trace of ‘back emf’ effect on the current waveform that rpm could be extrapolated from somehow.
I looked for any sign of this in the waveform and couldn't see it. I think this is a synchronous motor which runs at whole number ratios of the mains frequency so 60Hz*60s/2poles=1800RPM so that would explain why you cant see any motor noise on top of the 60Hz wave.
Given the ‘scale’ of the inrush it doesnt seem practical to greatly diminish it, and an external NTC in series with both windings is ultimately going to extend how long the ptc takes to heat up, which might be a net negative to the goal of getting past the inrush before an inverter cries about it. Its probably not practical to intercept the start winding since it’s probably paralleled off the leads inside the sealed compressor housing (is it?). So all you can really hope to do is ‘get lucky’ that a 12, maybe 20% reduction in inrush is enough to get it below the ‘failure threshold’ of a particular inverter, which is a LOT more fiddly and less reliable of an approach than just buying a larger inverter.
Yeah agreed, but there may be quite a bit of margin built into these compressor motors so it may make an unusable system work in a few cases.
You can actually get access to the different windings by taking the PTC off. so we could put the thermistor only on the start winding but I don't think the end result would be much different and it's probably more difficult to connect everything this way.
20221102_081437.jpg
"M"otor (run) and "S"tart pins. The unlabeled pin is common.
Capture.PNG
Unfortunately this compressor and the PTC are both labeled with part numbers but a thorough google search brought up no datasheets.

Trying to find the sweet spot of diminishing inrush might also be complicated by having to figure out if the conpressor controls are ‘smart’ enough to not try activating the compressor before the refrigeeant system pressures have equalized. I imagine in normal compressor cycling the pressure would have time to fully drop, making the ‘starting torque’ requirement pretty static. But if it allowed it to ‘short cycle’ you might also run into a situation where it will start against the ‘normal’ headwinds of the static system pressure, but be ‘locked rotor’ if it tries to start while the high side still has some pressure built up from the last compressor run. Bleh
I could get this fridge to do that already by unplugging it while the compressor was running and plugging it back in. It would lock the rotor and draw 500W until the thermal protection disconnected everything and then cooled off. Then it would try to start the compressor again a few minutes later. So an NTC in series would probably make this condition worse.
Ill be curious to see any future NTC results but it’s looking like a tough road on this one. Still awesome that you did the test.
I ordered 3 different NTC thermistors off Ebay. It'll take about a month for them to come so I'll test them out then. :)
 
Nice! Sounds like you’re all over this. Im not an EE but i mentioned all the things i could think of and none of it is news to you, which is pretty awesome because i feel like im getting ‘the whole picture’ from a small number of very informative posts with no lingering questions on my end. Thanks for exploring this, it’s been super interesting! I’ll be patient for the NTC testing, but this is already great info. ?
 
After much delay I finally tested the fridge with various NTC thermistors. The results are pretty underwhelming and I don't think using a thermistor to reduce the inrush current of fridge compressors is that great of an idea, but I wanted to report what I found nonetheless. Again I'll mention that to do this test you need to have a scope probe that doesn't have an earth grounded shield or even better use a specialty high voltage probe, so don't try this at home unless you know what you're doing.

Here is a snapshot of my test setup. I'm using a pair of 1ohm power resistors in parallel to make a half ohm shunt on the neutral wire. I measure the voltage across the shunt to and use this to calculate the current. Then I soldered different combinations of thermistors to add to 2, 4, 6 and 8ohms in series with the half ohm shunt. Then I cycled the fridge many times and tested the peak current for each run. I used SCK202R0 and SCK204R0 for the thermistors. These can be purchased on ebay for $3.

20230122_175158.jpg

1675130533959.png

Results: Adding a large 8ohm NTC thermistor reduced the peak current from 7.8A to under 6A. But that lower current surely reduces the starting torque of the compressor so it might lead to the compressor failing to start in some conditions. I waited about a minute or two between runs to let the refrigerant equalize and the only time the fridge rotor locked up was when I didn't wait long enough in between runs. Even with 8.5 ohms of extra resistance, it started reliably in my 45F shed/lab.

Overall I was hoping for a more dramatic reduction in current. If you have a situation where your inverter can almost power your fridge or it runs it most of the time but it sometimes overloads the inverter, then an NTC may be able to help remedy that. But in general an NTC won't dramatically reduce the startup current of the compressor..

I found that the total length of time that the current stayed high (above 1A) didn't depend too much on the resistance, it depended more on how many times I ran the test since the PTC that heats up and removes the start coil was staying warm from the previous test and turning off sooner. So without waiting an hour between runs or something, I don't have an easy way to measure that.

So there you have it. Hope that's interesting to some people.
 
There are plenty of threads on this, I have been using Maxwell supercaps in parallel with LiFePO4 to enable the use of a smaller battery without going over its surge capacity while powering inductive loads.

Works as intended, also removes AC ripple from inverter charger to battery.
 
I'm curious if you've done the theoretical math on the energy involved in the start up surge for those devices vs the amount of energy in the super caps?

In the past I've wondered about various strategies involving capacitors to deal with current surges and concluded that the total power stored in the caps was a little low to be a real solution for a lot of scenarios.

I'm also no EE, but we can look up the units...
  • farad = the SI unit of electrical capacitance, equal to the capacitance of a capacitor in which one coulomb of charge causes a potential difference of one volt.
  • coulomb = the SI unit of electric charge, equal to the quantity of electricity conveyed in one second by a current of one ampere.
So I think putting that in terms we're more used to working with, your 24.5v supply spread over 6 series 1.5 farad caps is about 4 volts when they're fully charged (in this scenario). When the caps drop from 4v to 3v (to make the math simple, let's assume the inverter alarms at 18v), you might get 1.5 coulombs (amp-seconds) @ 24.5 volts out of the whole mechanism. 1.5 amp-seconds * 24.5v = 36.75 watt-seconds. (Realistically, I suspect you'd get less because that 1.5 farads is probably the ideal case when starting at 5.5v, but let's just go with that for now.)

I think that means the next question would be, given one of your loads, what is the area under the curve for the start-up wattage in watt-seconds and how does it compare to the 36 watt-seconds in the super caps?
 
I'm curious if you've done the theoretical math on the energy involved in the start up surge for those devices vs the amount of energy in the super caps?

In the past I've wondered about various strategies involving capacitors to deal with current surges and concluded that the total power stored in the caps was a little low to be a real solution for a lot of scenarios.

I'm also no EE, but we can look up the units...
  • farad = the SI unit of electrical capacitance, equal to the capacitance of a capacitor in which one coulomb of charge causes a potential difference of one volt.
  • coulomb = the SI unit of electric charge, equal to the quantity of electricity conveyed in one second by a current of one ampere.
So I think putting that in terms we're more used to working with, your 24.5v supply spread over 6 series 1.5 farad caps is about 4 volts when they're fully charged (in this scenario). When the caps drop from 4v to 3v (to make the math simple, let's assume the inverter alarms at 18v), you might get 1.5 coulombs (amp-seconds) @ 24.5 volts out of the whole mechanism. 1.5 amp-seconds * 24.5v = 36.75 watt-seconds. (Realistically, I suspect you'd get less because that 1.5 farads is probably the ideal case when starting at 5.5v, but let's just go with that for now.)

I think that means the next question would be, given one of your loads, what is the area under the curve for the start-up wattage in watt-seconds and how does it compare to the 36 watt-seconds in the super caps?
The startup time for the compressor on the little wine fridge ranged from about half a second to 2.5 seconds and averaged about 5A at 120V so that's roughly in the range of 300Watt-seconds to 1500Watt-seconds. My interest was primarily in being able to use a smaller inverter by reducing the inrush current drawn by the fridge, but adding some extra capacitance near the inverter would also help especially if the battery and/or wiring were having trouble putting out the amps that the inverter was drawing.
 
Supercaps are about 50 times more expensive per Watt-hour than lithium batteries so maybe it would help, but you're probably better off just buying more battery. The cheapest supercap I could find is this one 2.7V 500F for $5.87 each. That works out to about $11.50 per Wh. That generously assumes all the energy in the supercap is usable and you discharge them down to 0V. In reality unless you have a DC/DC converter circuit to harvest all the energy in the supercap, you can only discharge about a quarter of their capacity like you mentioned. Plus you need some protection and balancing circuits. So in reality, supercaps are on the order of $40-50 per usable Wh. Here's a useful calculator for running some numbers.
 
While the idea to use a super cap as a buffer still interest me....I did eventually give up because of the involved cost to make it effective. For me it's just experiments so no loss here and I learned a lot.

The whole idea actually might have come into my head from a similar technique in digital model trains. Since contact with the rail can get lost on dirty tracks the digital decoder loses power and resets or at least stops the motor and/or sound. To "solve" this issue you either clean the tracks every 10 minutes (exaggeration off course ;) ) or you install a "buffer capacitor" as they call it. Usually a 2200uF rated 25V combined with a simple resistor and diode in parallel the cap positive will slowly charge it in a few minutes and provide about 2 seconds of power. At least that's what the specs say....I have not actually tried it myself.

here is the circuit diagram: (just did a quick google search, but this image is from the manufacturer documentation)
images


I do not have the massive 500F ones but I will absolutely do some more testing in a few months to see what I can do with 10,000uF caps at 35V. They hold a lot of power and I was thinking of replacing some battery operated devices with a ultra-low-power regulator of sorts to use them as "batteries" that charge quickly and hopefully last really long. Not sure what kind of devices though....but it will be fun and educational which is all that counts. ?
 
I'm curious if you've done the theoretical math on the energy involved in the start up surge for those devices vs the amount of energy in the super caps?
Symbioquine has nailed it.
Supercaps are very expensive for the stored energy they contain, especially if you require a fairly constant output voltage under load.
Lithium batteries are not exactly cheap either, but they will easily out perform supercaps as far as energy storage over time (per dollar) goes.

The solution might be a second battery at the far end of your long cable, and it may not need to be all that large to provide sufficient surge capacity.
 
That is my non-EE 'gut reaction' as well.. that you simply need some small lithium cells in parallel which are NOT protected by a bms, or not as tightly, to deal with the very short but high-C discharge to keep the inverter input above the low voltage cutoff.

I've never crunched any numbers to see if the value per farad is horrible, but if you are talking about a 12v system large 'stereo' capacitors (for stereo amplifiers, which are basically inverters with a different control scheme in my non-EE mind) are readily available with nice terminals and voltage displays, mounting brackets etc which might simplify the installation enough to justify their flashy overwrought exterior designs and price, if it saves you the trouble of paralleling a bunch of smaller ones and making a bunch of connections and having to figure out a mounting system, etc.

Interesting results on the fridge compressor. Just seems to confirm the earlier thoughts that this approach adds potential problems at a faster rate than it addresses the original intent. But now can be stated with more confidence!
 
That is my non-EE 'gut reaction' as well.. that you simply need some small lithium cells in parallel which are NOT protected by a bms, or not as tightly, to deal with the very short but high-C discharge to keep the inverter input above the low voltage cutoff.
Four small AA Lithiums with shortest possible wiring right at the inverter input should do something. At least its worth a try.
 
The supercaps work because when you put a supercap in parallel with a LiFePO4 and then apply a large load the voltage drops the same across the two, but the initial current is almost entirely provided by the supercap.

If you put two LiFePO4 in parallel, they will roughly equally share the load, so the benefit isn’t as great.

This is definitely one of those circumstances where the exact use case will determine what is required.

One thing I have noticed over the years is that LiFePO4 doesn’t like high surge currents. Over the next 20 years or so there will be more information on this, for now I’m convinced that there is value in keeping a residential system below 0.5C.
 
Back
Top