PIP-4048MS and PIP-5048MS inverters

[coulomb]

But it's a big update (some 224 kiB, less some unused space), so that's expected over a slow serial link.

I suspect that they send the hex file undecoded, so that's almost 500,000 characters to be transmitted in an update. To transfer in under 8 minutes (480 seconds), they'd need to send around one character per millisecond, so that implies 9600 bps.

[coulomb]

So the new firmware enables higher AC/DC charged rate and a higher Mppt solar reg charge rate? Or a higher combined rate?

Was it not 30A PV and 30A AC charging 60a combined on the old firmware?

Yes, I think it was 30 A mains (utility) charging, with 60 A total charge current. That would imply 30 A maximum SCC (solar) charge current.

[ Edit: I got the above wrong; see Weber's explanation here. ]

The new firmware gives up to 60 A utility charging and 120 A combined charge current. I'm guessing that the SCC takes care of its own maximum current. So basically the new firmware doubles all the charging limits. Not bad for a software upgrade, but of course the hardware has to be capable of it. Perhaps they didn't have the new firmware ready when the new hardware started shipping, so since the release of the 52.30 firmware, some models can get a performance increase. Possibly they had issues with stability or the like, and didn't want to release higher current firmware until they were satisfied it wouldn't cause too many problems.

Edit: Weber pointed out to me in email that it's more likely that the higher current capable PIPs shipped with high current capable firmware, and some problem unrelated to higher current necessitated the release of the 52.30 firmware. The warning was needed for people not having the higher current hardware. So the higher current might have happened around firmware 52.28, for example. It would be nice to know exactly which firmware version was the first one with the higher current capability.

[offgridQLD]

That's interesting one. As my unit was advertised as 3000w PV charger but from memory the charge limit was 30A on the PV settings (perhaps I am remembering that wrong) though I do remember thinking 60A was the max combined.

The fan speed was my main reason for upgrading.

I'm about to test the pip out running a small inverter style split system AC. I know it runs a 2.5kw old (non inverter) split system fine as i tested one a few months back. I had thoughts the variable speed drives might trip up the hi frequency style PIP. Somthing like it (some times) did with the imiev's variable charge ramp.

Will report back.

Kurt

[solamahn]

I have one site with 3 4048's in single phase parallel running 5 a/c no problem.

[offgridQLD]

OK, are they modern inverter style spit system AC units?

Kurt

[solamahn]

No. Just split system but not inverter. Also just completed a job yesterday with 2 4048in single phase parallel running 3 of those big coke coolers with glass doors and 2 500l deep freezers and the 4048's were able to start up all of them together.

[offgridQLD]

OK it's the inverter style spit systems that (potentially) might have issues. It's not the size of the load it's how the load varies up and down that (could) be a issue.

Though have to see.

Kurt

[weber]

I just had a go at updating the firmware on my PIP4048 to 72.40

What was the original version?

[paulvk]

I now have a temperature probe in the fins of the 230v inverter it gets to 48c with a light load of 300watts I think they may have gone too far with the fan slow down.

[weber]

I now have a temperature probe in the fins of the 230v inverter it gets to 48c with a light load of 300watts I think they may have gone too far with the fan slow down.

What firmware version first introduces the fan slowdown?

[paulvk]

Firmware 52.30 compared to the older one that had the fans running flat out under a light load Note I have a number of inverters have not updated any firmware just have one with the older version before 52.30

[offgridQLD]

"What was the original version?"

I had 52.25 and updated it straight to 72.40.

I had it running some loads this afternoon around 700-1000w for 1hr or so. I did notice the fans where very slow just a very gentle air flow that you could just feel and it was noticeably warm. No temp measurements sorry.

Are the fan speeds load based (I think that was the conclusion befor) or temp triggered as well.

Kurt

[paulvk]

No have not changed firmware two different inverters but they are identical other than version of firmware.

[Tjadenw]

Hi

I have the 2.4 KW model running a VFD controlled pool pump and a Inverter Aircon (12000btu).

No issues at all.

[weber]

Coulomb is a hacking legend!

That's hacking in the good (and original) sense of the word.
https://en.wikipedia.org/wiki/Hacker_culture

Despite having spent several days in intensive care after a serious kidney infection recently, and several weeks in hospital after that, Coulomb has been working for months on understanding why the low-cost Taiwanese PIP-4048MS inverter/charger/MPPT does some of the strange things it does, in the hope of being able to fix it and adapt it to work sensibly with LiFePO4 batteries, since the manufacturer has not done so despite us pointing out some of these problem to them at the beginning of this year. The manufacturer has nothing to fear from Coulomb's efforts, except maybe more sales.

Coulomb has now reached the point where he can practically make a PIP sit up and beg!

Yesterday he patched the firmware to fix a longstanding bug that affected all kinds of batteries, not just Lithium. If the charge current available from the PV array at the beginning of the day did not ramp up to greater than one fifth of the maximum total charge current setting (parameter 2) within 50 seconds, it would prematurely go to float mode and never achieve absorb voltage that day. This made it almost impossible to achieve automatic top balancing of LiFePO4 cells.

And once in float mode, it did not return to boost mode unless the voltage fell more than 4 volts below float voltage for 10 seconds. This is fine for lead-acid batteries. But we float a 16-cell LiFePO4 battery at 53.8 V, and 4 volts below that is very low SoC indeed. So he changed it so it only has to go 1 V below float voltage before it will go back to boost mode (and hence absorb voltage, which we set to 55.2 V). [Edit: We have changed that to 1.2 V below float in later versions.]

Today, he made it so the low voltage cutoff can be set as high as 52.0 V. The previous maximum setting of 48.0 V was way to low to protect a 16-cell LiFePO4 battery, hence all those extra contactors I had to add to the Black Monolith, controlled by the BMS. Not needed now! [Edit: Although we still must have a main battery contactor controlled by the BMS, as last ditch protection in case any single cell goes out of its safe operating area, including temperature.]

And then for an encore, and so no one could possibly mistake this for the manufacturer's firmware, he did this.



That's the Main CPU version number display. What a freakin' legend!

[offgridQLD]

Yes it sure can knock you about. I wouldn't wish it upon anyone. I remember going from a bundle of energy one day to being walked to the car like a 100 year old man and driven to hospital in what felt like a very short space of time, very debilitating.

I hope this effort is a sign he is feeling a lot better now.

What can I say all this hacking and just making electronics do what you like. Sure is a skill I can only wish I had.

Well done Coulomb.

PS, What do you want for xmas for a copy of it?

Kurt

[coulomb]

Thanks for the kind words, Weber. But a lot of thanks is due to you, too. It was a team effort, over many months (low part time, thankfully).

This started when I observed that the firmware upgrades must contain a copy of the actual firmware. There are two sets of firmware; one is for the main DSP (Digital Signal Processor) that drives the inverter proper, and one for the SCC (Solar Charge Controller), which is a stand alone MPPT (Maximum Power Point Tracking) battery charger. These have very different processors; the DSP is a TMS320F2809 fixed point 16/32 bit processor, while the SCC uses a Freescale 8-bit HCS08 microcontroller. I happen to be familiar with the TMS DSP chip from other work; the HCS08 is somewhat similar to a computer I once used that was housed in a shoe. That latter one is a long story, not for this forum.

It so happens that there are good reverse engineering tools available to disassemble binary images such as these, and it also happens that I have some experience in this area. That doesn't make it easy, just possible.

One thing that helped is that the manufacturers kindly, or inadvertently, left some debugging information in some of the firmware upgrade files. This gave me some valuable clues as to what some of the thousands of memory locations were doing.

Weber and I became interested in examining the firmware, because it seemed to be doing some things that really weren't what is needed for LiFePO4 batteries. These batteries have an extremely flat voltage versus state of charge curve, which makes it virtually impossible to even estimate their state of charge from their voltage. This is in marked contrast to lead acid batteries, which vary sufficiently in their voltage from full to empty that it is possible to use only voltage to control them. Only a few things needed to be changed, and so we asked if these changes could be made so that everyone could benefit. It is documented in this thread that unfortunately the manufacturer thought that these changes couldn't be done.

The main problem we wanted solved is one that became evident with Black Monolith number one. With this bug, the only workaround was a high float voltage, which meant the battery would either end up bypassing all day, or virtually never enter bypass. (Bypassing is when the cells are at a relatively high state of charge, and some of the charge current is made to bypass the highest voltage cells so as to top balance them). This problem would also beset any future monoliths (solar electric energy systems with battery storage). We either had to fix this problem, or find an alternative inverter that didn't have this problem. The problem is, these alternatives are either way more (6-8 times more) expensive, or turned out to be vapour-ware.

Many theories were formed about how the two processors co-operated when charging the battery. It turns out that the algorithms in the DSP and the SCC are very different. For example, the SCC has a three stage charge algorithm, with the absorb stage lasting 10 times as long as the bulk stage lasted. There is a graph in the PIP-4048 manual stating this. What we found was, this is completely overridden, at least in the cases we examined. When the SCC is connected to the DSP inside a PIP-4048 (as opposed to running stand alone in some of the MPPT charger products), it is forced to operate in step with the DSP. It turns out that the DSP, while effectively a three stage charger, internally combines the bulk and absorb stages. We ended up adding snooping circuitry to the serial connection between the processors, snooping both directions at once, to establish this. So it became evident that the problem is that the DSP charger exits bulk/absorb (going to float) way too early most of the time. As Weber has mentioned, this is a real bug that probably affects many PIP-4048 inverter-chargers.

In essence, the DSP terminates its bulk/absorb mode when the charge current drops to a certain value. That value is a bit complex; it depends for example on the number of PIP units connected in parallel (using special paralleling hardware so they can talk to each other and co-operate properly), and also on the maximum charge current setting. The current has to fall below the limit without exception until a certain counter reaches the value 1500, which turns out to be around 50 seconds. The problem is (and Weber figured this part out), that it might still be ramping up towards maximum charge current when this criterion is fulfilled! The other problem is that the current limit is the maximum of two values, one of which is the maximum charge current setting divided by 5. If this setting happens to be 120 A, then the limit is 24 A. It could easily be the case that near sunrise, the PV current doesn't exceed 24 A for 50 seconds at a time, so that means that for that whole day, the charger will stay in float mode. This is the exact problem that we were seeing with black monolith number one.

One solution might have been to increase the time that the current is seen to be below the limit. But still many days would not have sufficient PV charging current available to stay in bulk/absorb stage till the PV current became high enough.

So Weber suggested that the best solution would be to add an additional criterion to the decision to exit bulk/absorb mode. The battery voltage should be at higher than a particular value for the duration of the 50 seconds. This value should be just below the CV voltage setting; we used half a volt below. So now if the solar day starts weak and the current starts low, the charger will stay in bulk/absorb mode until the battery voltage is near the CV set point, AND the current falls below the threshold. We also decided to change the threshold from the higher of five amps (per paralleled unit) and the maximum charge current divided by 5, to the maximum of one amp and the maximum charging current divided by 30. These values are more suited to LiFePO4 batteries. Most of the changes are by changing immediate constants in the operands of DSP instructions, but the changed algorithm was implemented by calling out to a new patch function. It turns out that there is a fair bit of flash space available for this sort of purpose.

So now we have a patched version of the 72.40 firmware that works well with LiFePO4 batteries. Note that if you have a different lithium chemistry battery, then their terminal voltage will probably vary a lot more than with LiFePO4, so you likely don't need or want this patch. (The exception is the actual bug; hopefully the manufacturer will soon fix this problem). So this is a LiFePO4 patch, not a lithium patch. Hence the display in the post above. The QVFW command (Query Version of FirmWare) cannot return anything other than a pair of numbers, so it reports 72.43. I hope that there aren't many firmwares in the field with this version number to be confused with this patch. That was the other reason for the display above; you can't mistake this for the original firmware!

Hopefully in the next few days I'll be able to post the firmware patch with instructions. Actually, the instructions will be the same as the official instructions. Although we performed several SCC firmware patches in developing this patch, there is no need to update the SCC firmware to get the benefits of the LiFePO4 patch.

In summary, here are the patch changes:
1) Version number changes to identify itself.
2) Float to bulk/absorb now happens at float voltage setting - 1.0 V, not float voltage setting minus 4.0 V.
3) The exit bulk/absorb to float transition uses different charge current thresholds now. It was
MAX(5 A, maximum total charge current setting divided by 5)
and is now
MAX(1 A, maximum total charge current setting divided by 30) .
When multiple PIP units are paralleled, the charge current settings are added.
4) The time constant associated with the exit from bulk/absorb to float has been increased from 50 seconds to 120 seconds.
5) The additional exit bulk/absorb to float criterion is added, i.e. in addition to the charge current falling below the threshold stated in 3) above, the battery voltage has to equal or exceed (CV voltage setting minus 0.5 V).
6) The Battery Under Voltage Cutoff setting voltage range has been changed from 40-48 V to 44 to 52 V (i.e. 4 V has been added to the lower and upper limits). When this limit is exceeded (on the low side) for some 20 seconds, the inverter turns itself off to protect the battery from excessive discharge. It seems to take about 15 minutes of exceeding this limit plus 1.0 V (was plus 4.0 V) before the inverter comes on again.

We hope that this patch will be of use to many users, and if it leads to sales of more PIP inverter chargers, then that's great too.


[ Edit1: clarified that the bypassing all day was not a result of the bug, but a consequence of the only workaround we could think of. ]
[ Edit2: clarified that the DSP charger is effectively three stage, but internally it combines the bulk and absorb stages. ]
[ Edit3: Added link for black monolith #1 ]

[rhills]

Hi Coulomb,

... a computer I once used that was housed in a shoe. That latter one is a long story, not for this forum...

Hey, this is the Lounge! Do you really think you're going to get away with that little tease?

Sorry to hear about your illness but pleased that you're back on deck again.

[T1 Terry]

Great news Weber & Coulomb, and great news you are on the mend Coulomb.
Will this upgrade work on the other PIP inverters such as the 24v units? It certainly saves a lot of aux control, our fix was to cut the solar supply every hr so it sensed a restart when hopefully the solar conditions were better. We eventually gave up and went to a different solar control method as our arrays are generally within 10% of the target voltage so very little advantage from an MPPT control.

T1 Terry

[offgridQLD]

" the SCC has a three stage charge algorithm, with the absorb stage lasting 10 times as long as the bulk stage lasted."

I know that's how the instruction manual list it. Though I always just assumed that was crazy and perhaps a typo with the *.

If it took 1hr to reach absorb voltage then it would spend 10hrs at absorb voltage. What if it took 3hrs to reach absorb voltage 30hrs in absorb is crazy!

Edit" looking at the instructions again it also mentions max 8hrs. Though to reach that you would only need 48min of bulk and that still sounds crazy to me.

Because of this I thought that perhaps it was actually.

Time spent at Absorb = time to reach absorb voltage +10 min. Effectively giving you a minimum absorb time at 10 min.

Though I could be wrong. As I have never actually used the solar charger long enough to watch the bulk/absorb stage and time it.

[coulomb]

Will this upgrade work on the other PIP inverters such as the 24v units?

You certainly can't use this patched 48 V firmware directly on a 24 V inverter. The firmware is customised for a particular voltage. But if you could get hold of a 24 V firmware image, it's quite possible that the logic is similar enough that the same sorts of changes could be made to a 24 V unit.

Actually, I did grab a 24 V firmware at one point, and had a quick look at it. The 24 V inverters use a HCS08 processor (similar to the one in the SCC), and it happens that the symbols aren't available with that image, so it's much harder going. Perhaps something for the far future.

We eventually gave up and went to a different solar control method ...

Yes. It's a shame, as the hardware seems to be pretty good.

[coulomb]

[ shoe computer ] Hey, this is the Lounge! Do you really think you're going to get away with that little tease?

Well, in a former life, I was interested in the mathematics of gambling. I got started on a newsgroup called rec.gambling (people that know what I'm talking about are showing their age ). There was also a book about beating the roulette wheel, "The Eudaemonic Pie" (edit: it might haven been the later and/or British version, called "The Newtonian Casino"):
      

Here is a short page on it, even mentioning their shoe computer: http://physics.ucsc.edu/people/eudaemons/eudaemons.html

It was a great read. Somehow, a friend of mine got hold of some notes from someone else who attended a seminar by one of the Eudaemons, I think Doyne Farmer. They used a simplification to make the maths easier to solve. We figured out (with the help from some maths geeks from the internet) how to solve the equations without making the simplifying assumption. But it required some moderate calculations, so we built a HC11 processor into the heel of a shoe, and had a microswitch near a toe that could be quietly operated. It buzzed some timing information as the response. This would tell the operator what bet to make, and this bet had to be made at the last possible second or three. I suspect that Weber still has the shoe somewhere; when he heard about it he wouldn't let me throw it out.

Alas, we got cold feet (so to speak), and never used it on a real wheel. However, we progressed very far, even having a full-sized roulette wheel turned on a giant lathe, with optical sensors to gather data. In the end, we found a manual method that was perhaps half as good as the computer method, but much safer. Even that didn't get much use, as we tired of the discipline required. The technique worked only on wheels that happened to be tilted. Casinos are supposed to balance them every few weeks, but they get lazy. The thrill was in in the chase, so to speak. It was a fun time.

Any further discussion on this needs to go to a separate thread, please.

[ Edit: Rejevenated second image. }

[coulomb]

" the SCC has a three stage charge algorithm, with the absorb stage lasting 10 times as long as the bulk stage lasted."

I know that's how the instruction manual list it. Though I always just assumed that was crazy and perhaps a typo with the *.

Oops. Perhaps I didn't make myself clear. That's the behaviour of the SCC on its own, not being controlled by the DSP. We changed a branch in the SCC firmware to decouple the two chargers, and saw this behaviour. But under normal circumstances, the DSP overrides this behaviour completely, and you get the combined bulk/absorb behaviour, much like when it's mains charging or combination mains and SCC charging.

If it took 1hr to reach absorb voltage then it would spend 10hrs at absorb voltage. What if it took 3hrs to reach absorb voltage 30hrs in absorb is crazy!

Yes, I always thought that 10x was way too long.

Edit" looking at the instructions again it also mentions max 8hrs.

We never tested it, but the code seemed to imply a maximum of 3 hours. But that's still pretty long to have hydrogen bubbling from lead acid cells. Though I'm no expert with lead acid.

[coulomb]

This patched firmware is now obsolete. Please see the Firmware section of the index post to find the latest patched firmware for your machine.



Here is the LiFePO4 patch. NOTE: this patched firmware is intended for 48 V PIP-4048MS inverter chargers manufactured 2014 or later. It should work on 2013 models (we tested the patch on a 2013 model) as long as you leave the mains (utility) maximum charge current setting at 30 A or lower. NOTE: if you try this patched firmware on a 2013 model and don't keep the maximum mains charge setting at or under 30 A, you will likely blow it up! Anything very recent (latter half of 2015) may be too new, and you will probably at least lose USB functionality. If you are using a lead acid battery, you should probably be using the other patch here.

To install: you will need a Windows computer and a USB to serial adapter, or an older computer with a real serial port. Use the serial to RJ45 cable that came with your inverter-charger. Some serial adapters (or possibly the combination of serial adapter and Windows version) seem to be more suitable than others; see earlier posts on this.

Power up your PIP inverter charger. You can use a battery or the mains. Use the inverter switch to turn it on, if necessary. Make sure that your computer has power that won't go off during the reflash process.

Find a suitable folder on your computer (perhaps create a new one under your downloads folder) and unzip the attached zip file. The zip file has no folder inside it; it will extract files to the current folder, so best to start with an empty folder. Make sure you are not running any software that could be using the serial port, in particular, exit the WatchPower application. Don't just close the application; it seems you have to use the right mouse button on the WatchPower system tray icon, usually at the bottom right of Windows:



The file name the reflash tool looks for is fixed, it's always "dsp.hex" in the same folder as the reflash tool. So you don't have to tell it what file to read, but you have to make sure that you have the right dsp.hex file. It's best to not trust any file named dsp.hex; make a copy of the relevant file with a meaningful name to dsp.hex just before you run the reflash tool.

Finally, double click on the ReflashTool_Xseries.exe icon. It should bring up the small reflash application. If necessary, change the serial port selection to the one that connects to your PIP. It won't give you the full name of the serial port (e.g. "Prolific USB-to-serial..."), just e.g. "COM4", so if necessary use another application (e.g. TeraTerm) to be sure you choose the right serial port.

Click on the Update button, and answer Yes to "Are you sure". Nothing will appear to happen for 10-20 seconds, so don't panic. Eventually, new text should start appearing in the box under the progress bar, saying that erasing has been successful and so on. It will take some 8 minutes to complete the flash programming. At the end, just click OK and exit the reflash application.

If you get an error to do with the serial port (e.g. invalid serial port, or the serial port doesn't appear in the drop down window), it's probably because you have something running that is keeping the serial port in use. Double check that you have exited WatchPower and any terminal emulation programs like Tera Term. Weber has a netbook with Windows 7 that somehow doesn't see the USB to serial port adapter. So you may have to try another computer.

The standard instructions say you should revert all the settings to standard, then change them to suit your needs. As long as you had been running a recent firmware (52.30 or later), this should not be necessary. But if you were running an earlier firmware, or if the settings seem strange after the update, use WatchPower to reset to default settings, then change them to suit your needs.

The instructions say to turn off the inverter after the firmware update. This doesn't seem to be necessary.

That's it. To check that you have the patched version, scroll through the settings with the up or down button. In place of the U1 setting should be "Li FE PO4".

However, if something goes wrong, the original 72.40 firmware file is now included in the zip file. To replace the patched firmware with this official firmware, first you need to do a little work with Windows. Open a file explorer at the folder where the contents of the zip file was extracted to. Find dsp.hex and rename it to something else, e.g. "dsp.hex.patched" (without the quotes). Then rename "dsp_original_72.40.hex" to "dsp.hex" (this is necessary, because the reflash tool won't let you choose a hex file to load). It may be necessary to restart the inverter by cycling all power (disconnect the battery, mains, and PV power). (This may be why some people have had trouble recovering from a failed firmware update.) Now merely repeat the reflash process; this time it will reflash with the original firmware.

For convenience, I repost here the summary of the changes from the official firmware version 72.40:
1) An additional criterion is added for exiting from bulk/absorb to float. In addition to the charge current staying below the threshold stated in 3) below, for 50 seconds, the battery voltage has to equal or exceed (CV voltage setting minus 0.5 V). This is what fixes the bug where it would go straight to float mode without ever having gone near the absorb voltage.
2) Float to bulk/absorb now happens at float voltage setting - 1.0 V, not float voltage setting minus 4.0 V.
3) The exit bulk/absorb to float transition uses different charge current thresholds now. It was
MAX(5 A, maximum total charge current setting divided by 5)
and is now
MAX(1 A, maximum total charge current setting divided by 30) .
When multiple PIP units are paralleled, the charge current settings are added.
4) All "I" and "R" characters now render as "i" and "r" for arguably better readability.
5) The Battery Under Voltage Cutoff setting voltage range has been changed from 40-48 V to 44 to 52 V (i.e. 4 V has been added to the lower and upper limits). When this limit is crossed (on the low side) for some 20 seconds, the inverter turns itself off to protect the battery from excessive discharge. It seems to take about 15 minutes of exceeding this limit plus 1.0 V (was plus 4.0 V) before the inverter comes on again.
6) BatLowVolt (voltage at which a battery warning indicates) is now at EEBatteryVoltUnder+0.5 V (was +2.0 V).
7) Version number changes to identify itself. Shows as 72.44 on WatchPower or using the QVFW command. Shows "Li FE PO4" on the LCD display (in the "Main CPU version checking" screen).

[ Edit: lopezjm2001 found that WatchPower won't allow you to change the Battery Under Voltage Cutoff setting to more than 48.0 V, so you'll have to use either the LCD or the PSDV command with a terminal program to set this. Anyone handy with a Java decompiler should be able to patch the WatchPower program to allow this. ]

At Weber's suggestion, I'll later attempt to put together a version of this patch suitable for lead acid batteries. In other words, it will only have changes 1) and 4) mentioned above. [ Edit: it's here. ]

dsp_LiFePO4_patch3b.zip

[lopezjm2001]

Thanks for the patch Weber and Coulomb. Has the patch been tested for a period of time with no problems observed?