 Okay. Well, first off, I want to thank a few of you here who found your way up here. That was the first challenge this morning. My name is Brent Roman, and I'm here to tell you about the Environmental Sample Processor, which is a microbiological laboratory that we at the Monterey Bay Aquarium Research Institute deploy in the oceans of the world to study microbes in the water. And what you're viewing here is an image of a few of the team actually hoisting one of these units off the end of a ship for deployment. We'll talk briefly about what the ESP, the Environmental Sample Processor, what it is and what it does, how it works and why we built it. And then we'll go into a more detailed discussion about how we got to the point where we could deploy this kind of unit for six months to a year on the batteries that you see below. And that turned out to be an evolving challenge that took many years to accomplish. So let's get started. First off, some background. A number of you, in fact, I overheard a couple of people talking about the Monterey Bay Aquarium. So the Monterey Bay Aquarium was founded by David Packard's two daughters. She had three daughters and two of them became marine biologists. And so David Packard challenged his daughters to come up with some project that would really make a difference in their chosen field. And that project became the Monterey Bay Aquarium. And the fundamental difference between, or what's really kind of unusual about the Monterey Bay Aquarium is that it showcases local species found right there in the Monterey Bay itself as opposed to bringing in a lot of exotic species. So it encourages people to appreciate the ocean they're near as opposed to viewing it as a zoo. But as you might also be aware, David Packard also was running, keyword Packard at the time, and so he was an engineer's engineer and he was really disappointed and frustrated with the quality and the sophistication or lack thereof of the instruments that were being used to study the ocean. And so he resolved to found the Monterey Bay Aquarium Research Institute, which is separate from the aquarium, whereas the aquarium is about art outreach and education about the ocean. The Research Institute is more about getting engineers and scientists putting them together so they can work together closely to improve the instrumentation and the tools that scientists have to study the ocean. The Monterey Bay, for those who aren't familiar with the geography, it's about 100 kilometers south of San Francisco. And Monterey itself is at the south end of the bay. Both the aquarium and the Research Institute were located there in Monterey, but very soon the Research Institute was moved to Moss Landing. Monterey is a town of about 30,000 people. Moss Landing has 700 people in it. There's nothing there. There's a harbor, a few fishing boats, that's it. What we have, however, is one of the largest marine canyons in the world. And this amazing underwater geography makes it possible for a scientist taking a boat from Moss Landing to be in 2,000 or even 3,000 meter water and get back in time for dinner. So it's an excellent opportunity to study the deep ocean, and that's where we are located, where we're at. The group I work with at the Monterey Bay Aquarium Research Institute, which I'll call Umbari for short, the group I'm with at Umbari studies plankton in the ocean, the microbial ocean. Plankton supply over half of all the world's oxygen. They provide the base for the ocean's food web, and they regulate the CO2 in our air, just as trees do on land. But they also have some rather perplexing negatives. One of them is that they regularly will bloom, and often when they bloom, they will secrete neurotoxins. And these neurotoxins get first consumed by shellfish, then as other animals consume the shellfish, they get concentrated up the food chain, until finally they can result in these large fish die-offs, and not only they affect fish, but they also can affect birds and mammals. Now traditionally, the problem here is traditionally, when these kind of intense blooms occur, they discolor the water, making it kind of a reddish-brown, that's why we typically call them red tides. But not all red tides are dangerous. Sometimes you have these massive algae blooms, and there's no poison. So to determine this, so we don't sound a bunch of false alarms, traditionally what you would do is you'd send a water sample to a lab on shore, and in a few days you get the all clear, or you get the indication that yes, you have problems with poisons in the water. The trouble of course is by then the damage may be done. So the environmental sample processor was initially designed to detect those sorts of harmful algal blooms in hours as opposed to days, by not only sampling the water, but actually performing some analysis right there in situ. So the fundamental problem, aside from all the processing and all of the neat robotics that you can see here, the fundamental problem is to put the sample processor where it needs to be in the ocean and to keep it there. And for that we have various moorings. This is one of the first ones Mbari created for use in the Monterey Bay, which is a fairly sheltered environment, not very strong waves. So in this case we were able to put the ESP right about 10 meters below the surface, which is where the algae is that we want to sample. And all we have is a simple, there's an underwater cable about 15 meters, goes up to a float, and we're close enough to shore that we can just use a regular cell modem to communicate in this particular environment. Development on this began in 1996, so this has been a long time coming. And the key innovation that happened about the turn of the century was the creation of these pucks. These are filter pucks. Each of these represents a pair. There's a top and a bottom, and each of them snaps together. They all have different internal dimensions, different internal and different functions, some filter raw water, some preserve animals so that they can be later studied in a lab, some actually facilitate processing for genetic identification of species, which we'll talk about a little bit later. But they all have the same, once they snap together, they all have the same exact external dimensions, and that's what facilitates the robotic handling. So we can build a robot that essentially is a jukebox for handling these pucks and shifting them around in position inside the sample processor. We'll see a very quick movie about how it works in practice. The first thing it has to do is to get a puck and clamp it in position, and then once it's clamped into position, we start drawing raw water from the outside through the filter in the puck, and thereby we collect cells on that filter. The next part is to lice those cells, to heat them, and to chemically treat them so we break open the cells. Now we have this extract bug juice that we can then spread over another puck that is treated with a special filter that has spots on it with different DNA markers. Those DNA markers are chosen usually so that they can be unique for specific species of interest, specific algae that we're looking for. And in the end, you get this monochrome image, which we then radio back to shore. The fiduciary marks on that image allow us to correlate individual spots or groups of spots with specific algae that we're interested in, or other species. A lot of people have gotten involved in funding this, and we can go on here. So let's look a little bit at the robotics itself. The nuts and bolts would make this all possible. The ESP consists of about ten servo motors, small servo motors, and eight rotary valves. We've got about 20 solenoid valves. Here you can see a solid model of the arm, the robotic arm that is moving pucks from the storage carousel where they're stored in large stacks, about 132, and we move them from these clamps where they're processed, where we can actually draw water or other reagents through the filters, and then putting them back in the carousel once they've been processed. This is just a cutaway of that clamp, one of the clamps. So the ESP is not just for detecting algae, although that was what it was originally designed for. By simply changing the chemical markers that we put on those pucks, we can detect not only algal species, but we can detect deep water bacteria, and also we can detect human pathogens. So we put the ESP in some deep water deployments as well, and we're not going to talk a lot about those, but this one in particular, this is actually a one meter wide titanium sphere, and this is rated down to 4,000 meters. So we've dropped this down to the bottom of the ocean looking for deep water bacteria. Another application for the ESP that's become popular in the past few years has been water quality, things like, you might have heard of some of the problems we've been having in the Great Lake regions with contaminated water due to algae and also due to human pathogens in the water. So we put the ESP on a pier and just draw the water up from the surface. Now in these kind of applications, power isn't so much of a problem. Usually even in the case when we go on the bottom of the ocean, we can bring a lot of batteries or we'll have a cable in place. But our most common, still our most common deployment mode is these shallow water deployments for you looking at algae in the Monterey Bay or other coastal areas. In this case the only power we have to work with is our battery and in this case it's really about waiting for the algae to come to the ESP. So we have to be able to stay on station and simply wait until the environmental conditions indicate that we should take a sample. Keep in mind that the ESP only has a certain number of pucks, it has a certain number of reagents, so it's not just battery power that we have to manage all the other consumables as well. We wanted to keep this out for as long as six months because that was the duration of a typical season. And when it's deployed you can see in these images what we're seeing here is some of the team are actually dropping that core ESP you saw on the previous slide, they're dropping it into the sealed container. This is hoisting that over the back of the ship and finally we have a diver that will remove some pins and get everything checked out. Once we drive away from it, we don't interact with it for the next few months. That's the goal. So you might think we use some really exotic battery technology. This is it. Alkaline batteries, 360 D cell batteries. We've looked at other technologies, but we keep coming back to this one because they're really energy dense. Most people don't realize that alkaline batteries have as much energy per weight as your lithium ions. They just can't be recharged. But on the other hand, they're extremely inexpensive. They are recyclable, so this is what we end up using. We have two of these containers. Each of them weigh almost 40 kilograms and each one has 180 of these D cells in it. So each one is 3 kilowatts, 3 kilowatt hours. The total energy budget for all the months are going to be deployed as 6 kilowatt hours. So our first concern when we're doing this is trying to minimize the amount of energy that we use to actually process pucks when we were working, when we were actually doing the sampling and doing the analysis. And to that end, one of the first things we ran into was we've got these 10 servo motors. And we're trying to do... We're a small group. We're trying to do this. We're only going to make... We have a few dozen of these that we're going to make, so we're not doing big production runs. We're trying to minimize the amount of new engineering that we have to do. But unfortunately, we discovered that at least in 2002 when we first did this design that all of the microcontrollers that were designed for DC servos, they were using buses like CAN and RS485, and they had quiescent power in the watt range. And so if we had 10 of those things, we would have a 10 watt quiescent load, which would blow our budget very quickly. So we developed our own microcontroller, that servo controllers, and we got that down to 70 mW per every two motors. Now we had this sort of under control, less than half a watt for control of the motors. And that told a pretty good story because to do one of these... When you did one of these HAB identifications where we're trying to identify harmful algal bloom species, that takes four pucks, about three hours, to do the whole thing, and we only consume 25 watt hours in all that. And remember, we only have 33 sets of these pucks because we only can store 132 pucks anyway. So when we look and do the arithmetic, we're using about 800 watt hours to do all the processing we need to do. That's when the ESP was put in the water, if it just started processing pucks and didn't do anything else. The trouble is, of course, we have clearly enough battery capacity for that, but the problem is the weighting gain, and that's what kills us. We don't want to just suddenly... We could consume all of our pucks and reagents in a week or less, but we want to be out for as much as six months. And to do that, we try to choose when we take our samples, when we burn our reagents, when the environment, the chlorophyll, the temperature, the salinity, we want to choose the time so that when these environmental sensors indicate that we're likely to have algae present in the water right now, that's when we want to take a sample. So our main processor stays awake and continuously monitors these environmental sensors to determine when's the best time to take the next sample. Scientists also want to be able to interact with the ESP at any time if they see something interesting on a satellite photo, for instance, that indicates a bloom is present, they might want to fire off a sample just on the basis of that. So in the end, we end up with a system that draws about three watts while we were idle. And keep in mind, this is designed in 2002. In 2002, we really didn't want to design our own processor board, so we went out and found what at the time was about the lowest power option we had, and at the time, only a few companies were putting arm processors on a PC104 bus. This board is our mother board. It's a custom board. This board is an off-the-shelf product, and it was a lot better than the Intel products we typically had at that time, but it still was much too high power for our purposes. We were burning basically 75 watt-hours a day, and when you do the arithmetic, we only could stay out for less than 70 days, and we burned through the five kilowatt-hours that we had available. So this is far short of our goal, and what are we going to do about that? Well, it turned out not to be as bad as you might think. This 70 days was enough to do a lot of good science, and in practice, if you can remember what the ESP is, 10 servo motors, a bunch of robotics, do you think it worked the first time? Reliably, for months at a time? No. We were lucky if we got three weeks initially, if we got two or three weeks, we were really happy before something jammed or something leaked. But after a few years, it got to be pretty reliable. We got to the point where, typically, our main concern was we can't stay out longer because we don't have enough batteries to do the job. So finally, and after 12 years, we finally turned around and bit the bullet and said, I was looking for a processor on the PC104 bus because we didn't want to change everything. I was looking for a processor that I could just drop into place on that PC104 bus and would get us lower power consumption. In the end, we kind of had to make our own. We didn't make it from scratch. We took a dim module that's produced by embedded artists, and we took that very low power arm module and we mounted it on our own PC104 carrier board. That became a drop-in replacement for the previous board. And it had dramatic, obviously, 12 years is a long time in processor evolution, especially if you get to design the carrier board yourself. So we went from almost two and a half watts to a quarter watt for Linux idling, and that's full up Linux. It's 64 megabytes of memory. We're not really cutting any corners that way. Now we have a system that even with the radio and the rest of the system up and running, we're looking at one watt for our idle power while we're monitoring all the environmental sensors that we need to. And low, we're there, 200 days. And that was a great accomplishment. Everybody's quite happy about that, but we have to take a step back and look at how we did it and the limitations of it. One thing I'm going to show you in a moment is that we aren't the only people using the ESP. In our environment, the ESP is in a fairly protected bay. We don't have to cope with large, heavy seas. We can simply have a straight cable up to the float. And one of the ways we keep the power consumption down is we run RS-232 over that cable, which you think, RS-232, my God, who does that anymore? But for oceanographic research, it's often a good choice. It's very low power. And you can run it a long distance if you're willing to slow it down. But what we found is, while we could run it at 15 meters or so that we needed to, we couldn't, for instance, go 20 meters on that cable. And another thing to mention is that the spec on RS-232 is ridiculously, if you ever work with it, it's ridiculously conservative. If you look at the spec, it'll tell you that we really should only be able to run this at about 192 kB, but in fact, we run it at 115 kB. It's just ridiculously conservative. But we are at the practical edge. Well, our collaborators on the east coast of the United States, a large oceanographic institution called Woods Hole Oceanographic Institution, they were early adopters of ESP technology. They wanted to deploy it, and they did deploy it regularly off the coast of Maine, and that's a much rougher environment. They couldn't put the ESP at 10 meters. If they did, it would be pummeled by the waves. It wouldn't last a month. It would just be destroyed in the seas. So they have to put it on a larger mooring about 20, 25 meters below the surface. Now they have a problem. The algae aren't there. The algae are still way up above. So they developed a system with a stretch hose, a rubber hose that connected this larger float, and when they need it to do a sample, they actually have to pump the water from the top to the ESP and run it through the processing. They also want their environmental samples sampling. They also want the environmental sampling up here on the surface as well. To do this, they have to go through this stretch hose, and that has a cable in it. That cable has spiraled up the stretch hose. You can't have a straight cable, because if you had a straight cable, when the hose stretched, it would just break it. So that cable is 65 plus meters long. It's terrible in terms of electrical properties. It won't pass anything like Ethernet. It's too long for RS232. They ended up using DSL, which you guys might remember a decade ago. That killed their power. Basically, they now are using all their power for the DSL and Ethernet translation there. So even when we retrofitted these units with a new low-power CPU board, it wasn't relevant, because we went from 8 watts to 6 watts, and they went from 60-day duration to 80-day duration. It really didn't get them much. And just to reiterate, this is kind of the take-home message if I have a take-home message from this talk. It's that when we start looking at processors that are in this sub-watt range, down in the half watts and more, we can get that today. But at that point, we have to rethink how we connect those processors, because even something as ubiquitous as Ethernet becomes a huge power hog relative to the actual processors. Ethernet, a 1-gig link or a 100-base T-link, that's a watt. That's four times more than our processors using. Now you can slow it down. You can go back to the old 10-base T that nobody uses. It helps a little bit. And high-speed serial links are a real problem. So we also worked with another institution, Scripps over in San Diego. Scripps wanted to... They had the same problem. They wanted to put the ESP off the west coast. They needed to keep it lower. And so they opted for the stretch hose as well. But in this case, we put the... I'm sorry, in this case, we used RS-422 to go up and down the stretch hose. That's a variant of RS-232 that allows you to go further. It draws a little bit more current, but it allows you to get the job done. And that looked like... One of the main reasons for this particular mission was to get a six-month time series of the evolution of algae off the coast. So that was one of the primary goals for the mission. What happened was, we're hanging this ESP literally in a parking lot. This is hanging in a parking lot. This is a 10-meter stretch hose. And it's hanging like it will hang from the float when it's deployed. And as we're testing this thing, we're finding out if you ever looked at bungee jumping, you know how sensitive the system is to mass. So we find out that, guess what? We can't have two battery packs. You don't get that. If we do that, we're going to break this thing. The only thing we could do... I only had about months before the deployment. The only thing we could do is take off a battery. And then I'm just, what do I do now? Because I just had my budget. I was able to make it. But now we went from having five kilowatts for that idle power. Now we only had two kilowatts for that idle power. And we're going to be depleted in just about 85 days. So the first thing I thought was, well, let's do something quick. Maybe we can just suspend the CPU to RAM. And again, this is one of these issues where when you're doing a cell phone, you can suspend a RAM and you'll save a lot of power. You'll save maybe 20 to 1. But when you take a quarter watt processor and you suspend a RAM, you don't save that much. Because the total proportion of the amount of energy you're using to actually keep the RAM image refreshed is a fairly big part of the total. So in this case, doing that would only give us about 15 days more duration. We'll get to 100 days. We're still short. Well, let's think about suspending the disk, at least briefly. And the usual problem here, we only have an SD card. It's going to be slow. We're going to have problems of flash wear as we do this hundreds of times. But the real stopper was the 2.6 kernel that we had at the time just didn't implement it. Didn't implement it on the ARM kernel. The ARM side has no code for that. And I wasn't about to try to make hibernation work on the ARM. So that was stopped right there. So at this point, it's time to kind of punch. Went back to the scientists and I said, well, look, you want six months. You just took away half my energy. Let's make a deal. So I got them to agree that for this particular deployment, we could have the whole system be time-based. Just have the system come up at specific times and run the processing. For this particular deployment, that was deemed acceptable. What they insisted was that the system be able to be woken at any time from a radio command. So if they see something really interesting, or they're going to take a boat out there and ground truth it, going to interact with the system with the boat, they can wake it up. But the radio, if you recall, draws about a half a watt. So in fact, that's become the power hog in the whole system. So how am I going to... So I need to power management. I need to power manage the radio. I need this, and I switch everything off. Just the radio is going to be a 140-day duration. We're close, but not there. So I looked at what we'd have to do to address that. Turns out, as in... You'll see in oceanographic systems, we're about 10 years behind the times. I think you're seeing that here. These radios are the guts of 2G modems. They're the guts of 2G flip phones. If you remember your 2G flip phones, they could sit on the counter. They worked for a week. They had really good standby. Turns out these modems had that standby mode buried in their AT command sets. We found that, and then we put the system... We put them into that mode. And to get this odd situation where it has no data connection, but we can use these modems as pagers. We just send... We call it like it was a voice call, and the modem outputs literally ring, the old-fashioned AT command, the old-fashioned AT ring. That made it possible for our microcontroller here. With a little clever electronics, we have the microcontroller able to monitor the output of the modem after this microcontroller is very limited in what it can do. But as the host processor is shutting down, it puts the radio into the standby mode. The microcontroller then is just monitoring it for a ring. It doesn't know why it's happening. It just knows, I saw ring. And when it sees ring, it powers everything back up. Now, that gets you an impressive... impressive sleeping power consumption. Now we're down to 200 mW. We can stay out for more than a year. But, of course, we gave up a lot. We're not monitoring our sensors. So that was a real trade. This is just going on how the system powers back up. The only really tricky part here is we actually have some muxing electronics so that we can mux the serial port from the microcontroller to the host processor. But other than that, it's pretty straightforward. Final mooring I want to talk about is a really nice one that I thought that the University of Washington put together. And they wanted to monitor... They wanted to deploy ESPs, and they have been deploying ESPs for the last couple of years, off of Seattle, off the coast of Seattle. And again, same problem. They're not in a protected area, so they have to deal with wave action. They have to get the ESP down below 20 m. But rather than use a stretch hose, they got wise and they realized that the stretch hose is quite expensive. And instead, they used a really long CAT5-rated cable. And that turns out to be a good choice for a lot of reasons. One is, it's CAT5 finally. So we can run Ethernet over this cable if we really want to. Why not? Well, Ethernet isn't... We saw how we could run Ethernet slowly for a half a watt, and we could make that work. But what you find out in practice with Ethernet is you need a computer at both ends because we can't just plug a serial device into an Ethernet cable. We have to have another computer out there to do the breakout. And we need a cell modem, which now, instead of being a modem, is going to be a cellular router. Those devices typically use two, three watts. And if you shop real hard, you get it down to two watts. So really, that would cost us another three or so watts to the total system. We're back into this mode where we don't have enough power to get out as long as we want. So I got this idea. I've had it kicking around for a while. Why not use USB? If we have this cable... I actually tried this with the old... with the old stretch hose and it failed miserably because the cable simply couldn't support the high bandwidth. But if you've got a CAT5 cable, why not just run USB up it? Well, obviously, aside from the... the obvious issue of distance, we're going to use a little bit more power. What we get there is we could put Wi-Fi on the float. We could put weather instruments on the float. It allows us to have expandability up on the float. And we don't have to have multiple computers. We have everything driven by the ESP down below. But the big problem here is we need to span 40 meters. We're not going to do that with USB. But we have CAT5. And fortunately, there are a lot of devices. There are a number of CAT5 USB extenders on the market. They're not that expensive. Some of them are quite unreliable. But we found this one from ICRON that reliably goes through 50 meters of cable. And this system worked pretty well. It doesn't support high-speed USB. But as you find out, when you start analyzing these systems, we don't care about high-speed. All we care about is power. And 12 megabits a second is plenty fast for us. This worked well. We had to make a change to the actual... We had to dig into this device and change it a bit because we had to share our cable with other units, other systems on the mooring. So we only had access to six of the eight conductors. These guys wanted access to all... They wanted all eight conductors. Fortunately, I found that two of the conductors were used just to tell when a remote device... a remote USB device was unplugged and the USB needed to be re-enumerated. So you could detect, oh, I've unplugged it. I need to re-enumerate when I plug it back in. Well, these systems never get unplugged. The whole thing's sealed. You power it up once, it stays powered up. So logically, I thought, we can just not use those particular connections. Get it down to six wires, and we have enough to make this work. It did work for the first 10 days. And then a lightning bolt. That's literally what happened. We don't get... If you know weather on the West Coast, we don't get lightning storms. They are really strange. We got a doozy. And it reset everything on the float, including this guy, which immediately wanted to re-enumerate. And couldn't because it couldn't signal back to the low end. So make a long story short. We waited. We actually could tell that the system was still working. We knew that we were going to go out and recover the whole thing, but the rest of the morning came back. That is the other parts of the morning came back. And they had a sensor that could see that the pump that we were operating to sample was running at the right time. But the ESP was actually running, although we couldn't communicate with it. So he decided to sit tight and wait. And sure enough, after 10 days of nervous waiting, this thing synced up again. And we had communication. Everything was all right. When we went out the next time this year, in fact, the system is deployed again right now, we learned our lesson. We have a mechanism for resetting both ends of this connection. So if we get another bolt from the blue, we might survive it better. I was going to talk... I think most of you realize there are problems with power management. The theory and the practice are a little bit... There's a big gulf between theory and practice. We found that the most practical method was simply to cut off the USB power. If we needed to power manage a device, there are all the right way to do things. They don't work. They just aren't supported. The way that works is to simply shut off the USB power, cut off its plus five, and then the USB stack handles it beautifully. When you shut off the power, it sees it as a disconnect. You turn on the power, it sees it as a connect. It doesn't know that this is happening. It's not being coordinated anywhere in the kernel, but it works. There's no need... We found, luckily, there's no need to splice the high-speed data lines because those are impedance control. That would be tough. But we didn't need to do that at all. Finally, let's talk a little bit about energy harvesting. I mean, I think... I'm sitting here worrying about two milliwatts here, five milliwatts here. I take off LEDs off of circuit boards because, well, that's three milliwatts that I don't need to spend, and there's nobody in that dark case to look at it. So why not just solve the problem with a little solar panel? And in fact, we could do that. It wouldn't be that hard. We've never done this yet, but we're looking into it. A quarter-square meter of solar panel would give us 50 watt-hours a day, and that would allow us to run forever. We could sit there and monitor. We wouldn't give up monitoring the environment. We could sit there and operate forever. But our little floats that we have, those would probably have to be a little larger because even that small solar panel on those tiny floats would tend to tip them over in the wind, and if they tip over, then the antennas don't work so well. So to recap, we got there. We got to this point by, first of all, identifying a very low-tech, but very high-energy battery type, and then doing some custom electronics. So first we were concerned about the active power consumption, but probably we spent too much worrying about that. We should have been more worried about the passive power consumption, that is the power consumption between samples. We finally addressed that years later. But in all of this, the real message that I got out of it that surprised me most was how important it is to look at the actual communications power consumption. In modern designs, I think that's safe to say that's going to be, more and more, it's going to be the focus of energy management. And to get to the rest of it, I mean basically, I don't think I need to go through this again. We talked about it. And if you really want to stay out there indefinitely, you're going to have to do something to get the energy out of the environment. We have to do some energy harvesting. So hopefully I have a few minutes for questions. Yeah, I did all right. Questions, comments, raw fruit. Yes. Do we want to do the microphone thing? Are we small enough group? We can just talk. Tides? Okay. Yes. We could use wind. Wind is excellent, and we could use currents. All those mechanisms involve, well, they're mechanisms. They convert physical energy to electrical energy. You've got moving parts exposed to the ocean. And they're not likely to support, they're not likely to support, thank you. They're not likely to support, the thought was that they wouldn't likely last a year without any maintenance. So we thought, we really think that the solar has a better chance. In fact, we have it in Bari, a group that is trying to do exactly that. And so we know how challenging that is. There may be some instances where it works, but in general, we found that it's a mechanical system and it can break. But they don't live in the ocean. They don't actually live in the ocean, which corrodes everything, and they're not exposed to weather. So I would be willing to consider it, but that was our initial reaction. We can talk more about it. Yes. Oh, that's a good question. Because we could have, but when we tried the system based on RS-422, the problem with that is it needs a lot of conductors. It uses two conductors for every one conductor that RS-232 uses. It's a differential signaling, and we didn't have enough conductors. We only had six lines available to us. So with the handshakes that we needed for the modem and everything, we just ran out of conductors. And also the University of Washington would like the USB idea or a network because they could then add additional devices on the float. If we got it to work with RS-422, it's pretty closed. You can't just throw another device on there. Yes. Do I have any? One vandalism. Vandalism. Oh, you know, luckily we have, we had one that was struck by something that suddenly pulled it down 20 meters. And we don't know still what did it. It could have been a whale. But no, we haven't had any confirmed vandalism. The people in Washington, University of Washington, they had their mooring stripped clean and they still don't know what did it. So there's some suspicion there, but we don't have proof. Yes, that's an issue. That's part of the reason we want to keep the float small. If they're that small, the seals just can't get on it. If they're goes away, it's no fun for them. Okay. Yes, in the back. No, it was not easy to solve it. It wasn't the focus of the talk. The motors, we developed our own low-power motor controller and that whole networking, we networked them together with I squared C on a multi-master bus. That took months. That took four months of my time. So that was many years ago. I'm glad it's behind me. Yes. A broader reach. The question is, do we look at a broader reach Ethernet? I'm not aware of that. I guess you can talk to me about that. I would love to find out if any of you folks know about copper line transceivers that will cover 100 meters or 50 meters and don't draw more than a watt or half a watt, I'd be really interested because the strange situation we have right now is I can take a wireless signal and broadcast it 100 meters with less power than I need for Ethernet. It's not intuitive, but that's the fact. So any other questions? Yeah. No, that has just as many. The 485 is a party line system that with the 485 you could conceivably have multiple drops and that would be a way to add expandability to the float. It's not nearly as standardized as USB, but it could be done that way. But each drop needs to have the minus and the plus for the data, so it has just as many wires. It can be a duplexer or not, but in all cases, for every data signal you've got two wires, a plus and a minus, which USB has by the way, too. They managed to put everything on just one pair. What would be? You know, I admit I've heard of it, but I haven't looked at that really well and if you like one wire and you think I should look at it, we can talk. I don't know. The USB is faster than we need. If we had reliably 100 kilobits a second up to the float, that would be enough. Okay. Anybody else? Excellent question. There are about 25 of these in existence and at this moment we have four that are deployed. There's two off the coast, off the west coast. I think there are three off the west coast and there's one in the Great Lakes. A couple of years ago there was a Toledo Ohio had to shut off their water because they had so much algae in the lakes and this happens every year. Now there's actually government money to investigate that and so there's a government group that's using it to monitor the lakes. There's one out there. The scientists want to do it every 20 seconds. We have this push and pull. We're sitting there penciling out look. You'll need a Honda generator out there to do that. We usually can agree on every five minutes, 10 minutes depending on how... It's strange. The algae are in little waves and so it's not that the algae are growing really fast but the water is moving and so we're in one spot and the currents moving by us so we kind of need to know when to grab, when to take that sample. I was very skeptical as you are about needing to sample every five minutes and then I saw the plots. I saw the raw data and it's there. You sample it 10 minutes and you can see that you're getting Nyquist effects. You're just missing things. They made me a believer. I think... How are we doing for time? Well, it's up to you guys. We'll have a few more minutes. Are we good? Yep. So you're talking about sampling the environmental sample? Yes, we do statistics on the samples. Actually the environmental sampler, the electronic sampler, the thing that's looking at chlorophyll and looking at salinity, that is a little microcontroller in its own right and it's doing a lot of statistics and we take that information and do some more statistics. I haven't mentioned it in this talk but the whole system is programmed in Ruby of all things. That allows the scientists to actually write these scripts and so they can write scripts where they specify what we call trigger conditions and they can write out in pretty much English when the chlorophyll is greater than this threshold and the salinity is less than this threshold, proceed and do a sample. So we do all our basic statistics. Nothing too technical but the basic stuff. In the back, right here. For everything that we've done with open source in terms of the kernel, we've done a little work on the kernels but these are old kernels. I just put that out there. I'm a total believer. For the stuff that we've modified, the Ruby interpreter, all that information is online. You can look up Mbari Ruby patches and you'll find all that work on a Google page and with all the code. And there's a Git repository for that but for the actual Ruby scripting, what the scientists write to actually deploy and run a mission, that's considered proprietary because it's really their work. I think you work the hardware. Oh, the hardware. The issue there is if you are interested in the hardware Mbari is a research group. They're a nonprofit. We really can't sell it but we are always looking for partners. In fact, the ESP in its entirety is made by a group in the West Coast, a separate company. So if you're interested in this hardware and want to market it or work with it, all I can say is talk to me and likely we would be willing to license it to you. But the hardware is not open source in terms of it is proprietary, it is ours but we don't intend to market it and we'd love to find somebody else who would like to do that. Okay, I might be a good place to stop. Thanks for your attention.