This is the fifth post in a series documenting the installation and first year of operation of Pololu’s 305 kW solar array. Here are the previous posts:

• Part 1: Background starting in late 2022 and how we committed to the $650,000 project by the beginning of January 2023, with a target completion date of May 31, 2023.
• Part 2: Installation from January 2023 through first day of operation on October 5, 2023.
• Part 3: System failures and production results during the first year of operation.
• Part 4: Analysis of electrical costs before and after our system was installed.

In this final post in the series, I will go over the costs and benefits of our solar system, some lessons learned, and my thoughts on whether this has all been worth it. Continued…

This is the fourth post in a series documenting the installation and first year of operation of Pololu’s 305 kW solar array. Here are the previous posts:

• Part 1: background starting in late 2022 and how we committed to the $650,000 project by the beginning of January 2023, with a target completion date of May 31, 2023.
• Part 2: installation from January 2023 through first day of operation on October 5, 2023.
• Part 3: System failures and production results during the first year of operation.

In this post, we will look at how our solar installation affected our electricity bills. Because operations were so dramatically altered by the pandemic starting in 2020, we have to go back to 2019 for the best baseline for electricity consumption and cost. To maximize the scientific rigor of our observations, we changed as many variables as possible since then, including: Continued…

We just released several small reverse voltage protection and ideal diode boards that can protect your projects from reverse voltage application. We have reverse-voltage protection built into many of our products, and we usually implement it using a P-channel MOSFET, like this:

Reverse-voltage protection using a P-channel MOSFET.

This approach is usually more efficient than just using a diode since the MOSFET has a lower voltage drop across it. However, P-channel MOSFETs have worse on-resistances than N-channel MOSFETs of similar prices and sizes. This has not mattered much for our lower-powered products, but that limitation is becoming more apparent as we are developing more products with maximum operating voltages over 40 V. The next common MOSFET voltage above 40 V is 60 V, and at that voltage and with currents above around 10 A, it starts becoming more size-efficient to use an N-channel MOSFET plus an extra chip to manage the additional complexities of controlling the N-channel MOSFET in this kind of application. This is how that circuit looks:

Schematic diagram of the Pololu Reverse Voltage Protector.

Since we are planning on using this approach on several new products, we decided to make standalone product versions as well. Here are how the first products look, using 3×3 mm MOSFETs:

This lets us get up to about 10-12 amps continuous current and an operating range of 4-60 V, which is perfect for most of our products. We tried to make the board as small as possible, and for the input and output connections we are using a new slot approach that lets the boards work with standard 0.1" headers or connectors, 3.5 mm connectors, and 5 mm connectors.

Examples of various connectors that can be used with the Pololu Reverse Voltage Protectors (from left to right: 5mm terminal blocks, 3.5mm terminal blocks, 0.1″ headers).

Texas Instruments offers two similar parts for the MOSFET controller. The LM74500 offers the same functionality as the simple P-channel MOSFET, allowing current to flow in both directions as long as the polarity is correct. This is useful for applications such as motor drivers where we want power to be able to flow back from the motor into the battery. There is also the LM74700 version, which makes the circuit function as an ideal diode, allowing current to flow in only one direction. We are offering our boards with both controller options and with two MOSFET options, for a total of four product versions:

Pololu Item #	Max current	On resistance	Reverse current blocking	Price
#5380	10 A	< 10 mΩ	no	$1.49
#5381	12 A	< 5 mΩ		$1.95
#5382	10 A	< 10 mΩ	yes (ideal diode)	$1.75
#5383	12 A	< 5 mΩ		$2.25

The datasheets for the LM74500-Q1 reverse voltage protection controller and LM74700-Q1 reverse voltage protection ideal diode controller provide additional information about adding a transient voltage suppressor (TVS) diode across the input as part of a more general input protection circuit. We have pads for an SMB-size TVS on the back side of the board for those interested in adding this kind of protection:

As with our other electronics products, we make these at our Las Vegas, Nevada headquarters, so we can build custom versions with that TVS populated with a part of your choice (typical minimum quantities to make that worthwhile are around 200 pieces).

Are these interesting products? Would you want to see higher-current versions with bigger MOSFETs? Let us know in the comments or on our X and Facebook posts.

Introductory special discount! Try some out for as low as $1.16 each using our introductory special coupon, RVPINTRO (limit 5 per version)!

Over the last two days, I attended the 2024 Western Regional Meeting of ECEDHA, the Electrical and Computer Engineering Department Heads Association. It was held this year at UNLV, which is only 3 miles (5 km) away from Pololu, in the engineering department’s new Advanced Engineering Building that was just opened earlier this year.

UNLV’s new Advanced Engineering Building, November 2024.

I was there representing Pololu as one of five local industry sponsors. The larger companies there were treating it more like a recruiting event, and while we have several UNLV alums working at Pololu along with half a dozen student interns from UNLV, I looked at the event more as an opportunity to meet some of our customers. I also got to see some of UNLV’s new facilities for engineering students and researchers. Continued…

This is the third post in a series detailing our experience over the past two years installing and operating a 305 kW array of 630 solar panels on our building in Las Vegas, Nevada. Here are the previous posts:

• Part 1: background starting in late 2022 and how we committed to the $650,000 project by the beginning of January 2023, with a target completion date of May 31, 2023.
• Part 2: installation from January 2023 through first day of operation on October 5, 2023.

I left off with our first look at the SolarEdge monitoring site on October 5, 2023. It’s nice to see nearly real-time generation results and status. The SolarEdge P1101 optimizers connect to pairs of solar panels, so that is the resolution we can see in the array. Here is a close-up as I write this at 10AM on October 31, 2024, with a section affected by the shadow from an air conditioner circled:

SolarEdge monitoring site solar panel array close-up at 10AM on October 31, 2024.

The 1.0.72 pair of panels and 1.0.19 pair of panels at around 130 Wh so far today have generated only about half as much as the nearby panels not affected by the shadows. Continued…

This is the second post in a series detailing our experience over the past two years installing and operating a 305 kW rooftop solar system on our building in Las Vegas, Nevada. In the first post, I covered some of the background starting in late 2022 and how we committed to the $650,000 project by the beginning of January 2023, with a target completion date of May 31, 2023. This post covers how the actual installation went. Continued…

October 2024 marked one year of operation of our 305 kW rooftop solar power generation system. In this series of posts, I will reflect on our installation and operation experience over the past two years to try to assess whether it was worth it. This first post covers some background leading up to the project and the overall system design. I will detail the installation process, the first full year of operation, and the production and financial results in subsequent posts. Continued…

We recently added six new low-power variants to our Motoron line of basic serial motor controllers: four Mxx550 1- and 2-channel versions, as well as 3-channel versions for Arduino (M3S550) and Raspberry Pi (M3H550).

1- and 2-channel micro motor drivers

The new M1T550, M1U550, M2T550, and M2U550 are single- and dual-channel serial motor controllers in a micro footprint. With a maximum motor supply voltage of 22 V, the Mxx550 versions are a great way to control small motors powered by power supplies up to 12 V and battery packs up to 12 cells in series for alkaline, NiCd, and NiMH, or up to 4 cells in series for LiPo. These are lower-voltage, pin-compatible versions of the Mxx256 models we released earlier this year, which have a maximum motor voltage of 48 V and can deliver slightly more current but are otherwise almost identical.

Here is the full array of tiny Motoron options, including I²C and UART serial interface versions:

Motoron motor controllers micro versions
	M1T550 M1U550	M2T550 M2U550	M1T256 M1U256	M2T256 M2U256
Control interface:	I²C or UART serial
Motor channels:	1 (single)	2 (dual)	1 (single)	2 (dual)
Minimum motor supply voltage:	1.8 V		4.5 V
Absolute max motor supply voltage:	22 V		48 V
Recommended max nominal battery voltage:	16 V		36 V
Max continuous current per channel:	1.8 A	1.6 A	2.2 A	1.8 A
Available versions with I²C:	headers soldered headers included	headers soldered headers included	headers soldered headers included	headers soldered headers included
Available verions with UART serial:	headers soldered headers included	headers soldered headers included	headers soldered headers included	headers soldered headers included
Price:	$12.49 – $14.49	$15.95 – $17.95	$16.95 – $18.95	$23.95 – $25.95

3-channel motor drivers for Arduino and Raspberry Pi

We also released larger (but still small!), 3-channel versions in Arduino (M3S550) and Raspberry Pi (M3H550) compatible form factors. These again have a maximum motor supply voltage of 22 V and correspond to the 48 V max M3S256 and M3H256 versions we released in 2022. Here is the full line of larger Motoron serial motor controllers, including the even higher-power, dual-channel Motorons in full-size Arduino Shield or Raspberry Pi Hat form factors:

Motoron motor controllers Arduino and Raspberry Pi form factor versions
	M3S550 M3H550	M3S256 M3H256	M2S24v14 M2H24v14	M2S24v16 M2H24v16	M2S18v18 M2H18v18	M2S18v20 M2H18v20
Control interface:	I²C
Motor channels:	3 (triple)		2 (dual)
Minimum motor supply voltage:	1.8 V	4.5 V	6.5 V
Absolute max motor supply voltage:	22 V	48 V	40 V		30 V
Recommended max nominal battery voltage:	16 V	36 V	28 V		18 V
Max continuous current per channel:	1.7 A	2 A	14 A	16 A	18 A	20 A
Available versions for Arduino:	M3S550 assembled kit board only	M3S256 assembled kit board only	M2S24v14 assembled kit board only	M2S24v16 assembled kit board only	M2S18v18 assembled kit board only	M2S18v20 assembled kit board only
Available versions for Raspberry Pi:	M3H550 assembled kit board only	M3H256 assembled kit board only	M2H24v14 assembled kit board only	M2H24v16 assembled kit board only	M2H18v18 assembled kit board only	M2H18v20 assembled kit board only
Price:	$20.95 – $30.95	$34.95 – $44.95	$59.95 – $69.95	$115.95 – $124.95	$59.95 – $69.96	$95.95 – $104.95

The great thing about the Motorons is that you can easily string together or stack multiple controllers, mixing and matching sizes to fit your application. For example, you could use one high-power dual motor version for drive motors on a mobile robot and then add a smaller 3-channel motor controller for additional actuators. This arrangement with three stacked Motorons on an Arduino Uno allows simple control of up to 9 motors:

The common protocol between versions also makes it easy to change motor sizes and to reuse your code between projects. Want to make a bigger version of your first prototype? Just use a higher-power Motoron! Want to make a tiny robot next time? Use a tiny Motoron! Want to… you get the idea.

While the 3-channel boards are designed to stack on Arduinos or Raspberry Pis, they are also easy to use on breadboards:

It may be easy to view the six new Mxx550 Motorons as just lower-voltage versions of the previously available Mxx256 Motorons, but I am especially excited about them because we are able to offer them at a very low price, extending the legacy of the Dual Serial Motor controllers that were among our first products over 20 years ago. We are launching the 2-channel M2T550 and M2U550 at just $15.95, a lower price than the original Dual Serial Motor controller from 2001 (without even adjusting for inflation!).

The chip shortages of the past several years have made it especially difficult to introduce new products and to keep their prices down, but things are finally seeming to get better on that front. You can see in the tables above that the higher-power 2-channel Motorons are much more expensive; those prices are still elevated because we are limited on some critical components we use there and in our other products. We should be able to manufacture plenty of the new Motorons without being constrained in a similar way.

I am super excited to introduce our newest robot, the 3pi+ 2040. This robot combines the 3pi+ chassis, which we initially released in late 2020, with the power of the Raspberry Pi RP2040 microcontroller. Here is a quick overview of its features:

This summer will mark 15 years since we released our original 3pi robot, which was designed to be fast enough to be competitive in line following and maze solving events. The high speed offers interesting programming challenges not present in typical robot kits of that era; here is a video from back then in which Ben demonstrates his 3pi learning a maze and then going extra fast on longer straightaways:

Although we developed our first injection-molded parts (wheels, ball caster, and motor mounting brackets) for that design, it was still largely a “PCB on wheels” kind of robot. The next-generation 3pi+, with a chassis mechanically independent of any circuit board, had been in development for several years when the coronavirus pandemic hit in early 2020. We kept working on it throughout that year, culminating with the November release of the 3pi+ 32U4.

Original Pololu 3pi robot.

3pi+ 32U4 Robot.

The 3pi+ delivered the most-requested feature missing from the 3pi, wheel encoders, along with many other improvements including a full IMU, bumpers, and programmability over USB (the 3pi required an external AVR programmer). With its support in the Arduino environment, the ATmega32U4 continues to offer a good entry point for working with microcontrollers, but the 8-bit architecture and 32 KB of program memory feel increasingly outdated and constraining, especially with the new sensors available on the 3pi+.

That brings us to the new 3pi+ 2040, powered by the Raspberry Pi RP2040 microcontroller (32-bit dual-core Arm Cortex-M0+) with 16 MB (128 Mbit) of flash memory. The robot ships preloaded with a MicroPython interpreter, so you can get started right away by plugging into its USB C port and editing the included example Python programs with your favorite text editor. No special programmers or programming software are required, and you can write MicroPython code from practically any desktop or mobile operating system as long as it has a text editor and the ability to copy files to a USB drive. For a basic Python IDE that lets you run code interactively, we are recommending the Mu editor. (See the User’s Guide for instructions on setting it up.)

MicroPython drive showing 3pi+ 2040 demo programs.

The blink.py demo program in a text editor.

There are many other programming environments and languages that you can use with the 3pi+. Since it shares the same RP2040 processor as the Raspberry Pi Pico, anything that works for the Pico should be usable on the 3pi+, including C, C++, and the Arduino environment. We already include some basic C examples in our example code repository, and we plan to write more examples and expand the software support for this robot. Do you have a favorite IDE that works with the Pico? Is there some language or system you’d like to run on the 3pi+?

The menu of pre-installed demo programs on the 3pi+ 2040 Robot.

Early adopter special: We are initially offering the 3pi+ 2040 Robot as a limited release intended for advanced customers who have had some experience with robotics or Raspberry Pi RP2040 programming (e.g. with a Raspberry Pi Pico). The initial release is available with 30:1 MP motors (the “Standard Edition”), either assembled for 38% off or in kit form for 50% off. Early adopter robots will generally need to be backordered as they are built to order; we expect to ship within a business day of ordering. The robot hardware is finalized so the only changes we expect for the full product release are in the initial firmware configuration and pre-installed example programs. Documentation will also continue to be developed as we release the robot to a wider customer base. Early adopters who publicly share their 3pi+ 2040 experiences will be eligible for an additional robot with an extra $25 discount.

Wow, it’s been almost a year since my last update about how Pololu has been impacted by the global supply chain disruptions and chip shortages. And unfortunately, not much has improved. In today’s post, I will cover a few representative component stock histories and then go over other areas of our business that have been impacted and what we are doing to get through this situation.

Some parts on order since 2020 still have not shipped

In the case of one important part I mentioned last year, we are still waiting for an order placed in late 2020 without having received anything since a partial shipment in March 2021! Here is what our internal stock chart looks like for that component:

When I wrote about this component in November of 2021, we still had 461 units in stock, and the manufacturer was giving me specific updates about where we were in line and how I could expect parts by Q1 2022 or maybe even by the end of 2021. Well, we are now getting close to the end of 2022, and they are not even giving me updates anymore on when I can expect these parts that I ordered in 2020! We have gone almost a year without being able to make or sell the products that use that chip.

Some parts arrived in 2021 and early 2022, but we are out again

That first example of still waiting for an order from 2020 is not typical. Unfortunately, we are seeing more and more of this pattern:

This is a component we ran out of over the summer of 2021, but we received some shipments in August of that year, and then more in early February of this year. But since then, nothing, and we are about to run out again despite our attempts to carefully ration the parts. It’s been over 14 months since I placed my oldest unfilled order for these parts, and the current expected ship date is February 2023.

Shifting demand clears out stock of alternative components

Another pattern we are seeing more of looks like this:

Here, we were in a pretty good stock situation at the beginning of the year on a component we didn’t use that many of. However, as we raised prices on other products or ran out of stock completely, our customers moved to some of our recommended alternatives and cleared us out of those, and hence the sudden dropoff of those parts in April of this year. The additional problem with components like these is that we did not have as many on order because our historical usage was not that high, so it might take an extra long time to get that back to decent stock levels.

New “supply outlook” feature

We commonly use the same components in several different products. One of the main ways we are dealing with the shortages is to substantially reduce our inventory of completed products so that we can be sure the components we do have are going toward products that are getting sold immediately.

One big downside of reduced ready-to-sell inventory is that it’s difficult for customers to tell what is really, really unavailable because we’ve been out of parts for a year and what is actually available as soon as we make some more. To give you some automated guidance, we introduced a “supply outlook” feature to our website. Here is how that looks at the moment:

The calculations of what we can make are quite complicated given that we have thousands of different components going into thousands of different products, and the products (and the associated inventory) can be in various stages of production. Components stop being available once they are soldered onto a board, but that board might still go through many more processing steps before being ready and available for sale. The in stock and “in final production stages” quantities should be spot on, but we variously round down the “enough components” estimate to keep it conservative. The numbers can be outdated quickly since we are selling and making products all the time, but we regenerate those numbers several times a day to be as up-to-date as possible.

The supply outlook feature does not factor in components we have on order, though this year has proven that would be almost useless anyway (I’m not sure if I prefer the suppliers who give me no estimate of a ship date or those who have been saying “next week” for months). On our to-do list is to get more manual/human notes so that we can have updates like, “we are estimated to receive components in March 2024”.

I wish that last line was exaggeration. Unfortunately, I am getting more and more order confirmations with lead times of well over a year and estimated ship dates in late 2024. For parts I ordered early this year, we are approaching three-year lead time estimates for components.

Supply chain issues impact other aspects of business

Although the chip shortages are the most nerve-wracking aspect of the current environment, other aspects of our business are also affected by the supply chain problems, and it’s getting more and more uncomfortable.

• Waiting more than 9 months for commercial air conditioners - One literal example is the air conditioners in our building. We have over fifty of them, and dozens of them are over twenty years old, meaning they are inefficient and reaching the end of their useful lives. We have had several on order since the beginning of the year, and at this point we are hoping that maybe they will arrive by the end of this year. Fortunately, we made it through the summer, but several units did die recently, and it’s not clear that we can even have them replaced by next summer.

Old ACs on Pololu building roof, waiting for replacement. Las Vegas Strip in the background.

• Waiting more than 6 months for window film - We started applying special solar-blocking films to our windows to help reduce the energy use by the ACs. That project started in late spring, and although part of it got done over the summer, most of it has been delayed by at least six months waiting for more of the film to get manufactured.

Pololu window tinting, July 2022.

• 6-12 month lead time on compressors and nitrogen generators - We ordered a nitrogen-generation system earlier this year, and lead times on that are in the ballpark of a year as well. There are several components to the system and we get billed for them as they arrive, so I don’t think the manufacturer is holding back on any of them while waiting for the others. One component is a fairly standard (though nice) air compressor that I am expecting to use for the rest of manufacturing as our existing ones are getting kind of old. It’s scary to think some of our production or equipment could be out of commission for a year waiting for machines or components that normally are available within a few weeks.

Outlook

We have been very fortunate at Pololu because we have a broad range of products and do our own design and production, so we have been able to adjust what we make based on what components are available. I don’t understand how more small manufacturers are not going out of business, though I am anecdotally starting to hear more about companies facing financial difficulties. Contract manufacturers in particular have it tough when they have to pay for the components they can get while waiting forever for the last few components and not getting paid until they can complete the final product.

My main hope is that just as we could not see how bad the shortages would be, we cannot see how close we are to the end. If it took two years to get a part that shipped today, it might be reasonable to estimate it will take two years to get a part we order now, or even to tack on an extra year for good measure, but eventually things will be better. I expect inventories everywhere are building up (ours are, just not of the last few critical parts!), and the coming global recession that seems to be forecasted from all sides (e.g. by the IPC) could accelerate chip manufacturers finally catching up to the extra demand from the last few years.

Since we are a small business, broader economic downturns can sometimes work in our favor. Our customer base is such a tiny portion of the world, and some of them could do well even if on average the global economy does not. If the slowdown leads to parts we need becoming available sooner, that might overall be better for us. Some of our best supplier relationships came out of the 2008 downturn, when companies started caring about our business after losing some of their bigger customers. We also got a good deal on renting part of the building we are in after it sat vacant for a couple of years, and that served us especially well as we gradually expanded to the whole building over the past ten years.

It’s unsettling that after two years of parts shortages, it does not seem to be getting any better. The situation might even be worse than it was a year ago, but we won’t really know until we are out of it and things are actually good again. I know it has been difficult for our customers, especially those who built our products into their own products or curricula and are counting on us to keep their operations moving. Please know that we are working very hard to keep our stock and production levels up with the minimal possible disruptions, and thank you very much for your continued business and support.