Tank Monitor replacement - digital

[Zeppelinium]

OK, I admit I mentioned doing this a couple years ago. One good thing about snow and cold--it encourages/allows you to get back in the lab.

Very short tutorial: the older 70s, and earlier, tank monitors use four wires to report the water level to a circuit card with small light bulbs that illuminate to crudely indicate fluid level. The newer 70s (I don't know when it changed, but some time between 72 and 75) use two wires to send a varying voltage (5 steps, which is a bit more resolution than the 4-wire system) which drives small analog meters. Although these methods are very different from an electrical standpoint, a very similar circuit can be built to detect the two different kinds of inputs and turn on LEDs in the same way the old system turned on the light bulbs.

Note: the following description is preliminary and will most likely be modified once I can make the first prototype. This information is posted so that others can make inputs, both improvements and corrections. I'd rather have a correction than a smoke test.

Two-wire system: this system has four sensor probes and one driver probe. Using resistors in series with the probes, the four sensor probes can provide 5 output levels (the fifth level is zero--no contact with water). Here's where I might need a little help--it's too cold here to experimentally fill and empty my tanks, so had to improvise in the lab. When I used 9 volts to drive the probes, I got the following outputs. If you measure and existing set of probes that are connected to 12 volts, you should get something like the numbers in the right column.

For reference, this is the schematic of the original meter circuit:



A quad voltage comparator can be used to detect the voltage input level and turn on an LED ladder. It's a very inexpensive circuit, around $6 per tank, plus the experimenters punch board from some place like Radio Shack. Laying out a printed circuit board is helpful in seeing how it might look, even if it's made by hand. It looks like the total board area is 3"x3". The LED columns are 1" appart and are 0.9" long.



The four-wire (older) version is very similar, but a bit simpler. As each sensor contacts water, its LED illuminates.



The board would function only when a "press to test" switch is pushed. I've considerd putting in a switch with a momentary function and an "on" function. All the LEDs would illuminate at the water level and below (eg, all the LEDs below the current level would also illuminate). Any input on alternate functionality will be considered. BTW, there is a reason for using 9V as the (+) power--the HF meters work off of 9V and I intend to rip a couple of those suckers apart to provide an indication of battery voltage and current. More on that mod later.

Zep

[RichHog]

Is this something you would manufacture and sell? If so I'm in as my 72 TW doesn't have much functionality on the control panel and it seems that info on it's components is missing on the forums.

Thanks for your efforts, could you use different colored LEDs, say like green for full, yellow for low and red for empty?

Richard

[Zeppelinium]

Yes, different LEDs - green, yellow, orange, red.

No, not sell. Once I check it out completely, I'll put the plans here. If you haven't soldered DIP-type integrated circuits, I'm sure you can find a ham or a computer kid in your area that can do it. It will be simple.

Zep

[RichHog]

Hey I built a Heathkit Color TV back when you had to solder ALL the components on the boards. Will give it a try.

[Zeppelinium]

The circuit is changed so the LEDs come on rather than go off as the water level rises. Also a bit of simplification to the input resistors. The breadboard works per expectations. No other guarantees, express or implied! It appears that this could be built on a single layer PC board, so there should be a minimum of jumper wires when it's built on an punch board.



The bread board tested 4-wire circuit will be along shortly.

Zep

[Zeppelinium]

Here's the 4-wire circuit and layout. The electronics are somewhat simpler and there is a spare comparator, in case you have an application where one more wire/indicator is needed.



I'm guessing that this system was used before 1975 and the addition of the gray holding tank. So you ony need two of the modules, not three. But just in case, the layout provides for three.

Zep

[Zeppelinium]

If you've got a 4-wire system and you have a broken wire, or if you want to add a sensor (stainless steel screw-in post, really) to provide an additional water level reading, you can easily convert your 4-wire system to two wires. Just use a small piece of insulated board to mount the resistors and some kind of connector for the external wires and mount it near the tank. You can now use the 2-wire monitor module for each converted tank.



I'm going to convert my Caravel and Overlander to the 2-wire system. I doubt I'll add another sensor, however, just keep a 3-level indication capability.

Zep

[Zeppelinium]

I've been assuming that anyone who would try this is familiar with DIP prototyping tools. If not, here's what it takes to get started. Radio Shack has the layout board for about $20. You'll also need a small 12V power supply, like one of the types where the whole unit hangs on the 110V plug. It's also good to know which end of the LM339 has pin #1.



When you've got the bread board model figured out, you'll want to make something a little more portable, something that might actually be useful (temporarily) in the Airstream. That requires soldering the components onto a prototyping board. There are three kinds--bare, isolated-dots, and bussed. The bussed boards are easiest to work with, but the finished prototype will be a little larger, due to the component placement restrictions of the busses, than the isolated-dot type board.



The odd shapes are the result other projects the nibbled away at the orginal board(s).

[Zeppelinium]

This prototype did not follow that PCB layout pictured earlier. In addition, during the prototype testing there were some changes made to the schematic, so use the one in this post, not any of the earlier ones. The series resistors for the LEDs were adjusted to get the apparent brightness of the colors to match. The red LEDs are more efficient than the green and yellow, so they require a bit higher resistance in order to dim them down a little.

One of the major differences in this prototype is the use of a 10-pin socket for the voltage divider resistors. This will allow changes for the higher conductivity of the black and gray water (see tables in the next post).





You can see that the test sensor strip is pretty simple.

Here's the correct schematic. The schematic is the same for all type tanks--fresh, gray, and black. But the input voltages are different due to the differing conductivities, so the resistor values for R1-R5 are different (see next post). The values here were nominal for fresh water in the lab and will most likely change for the tank(s).

[Zeppelinium]

The input voltage from the tank(s) is influenced by the conductivity of the water in the tank. Clean water is not very conductive. Impurities, especially salt, increase its conductivity dramatically. The following tables show the differing inputs that can be expected from the three types of tanks (don't ask how). Only actual road test will confirm these values. In addition, the type of tank may have some unexpected influence, eg, tall versus flat tanks.



The values aren't grossly different, but the different conductivity does move the switching voltage enough that the fresh tank resistors definitely wouldn't work for the black tank. I suppose this is why the meters in the original monitor had an adjustable resistor in the circuit. I'm going to think about whether a variable resistor (one for each monitor) would be a more general solution for this design, too.

One interesting phenominom is that the black tank voltage decays very much more slowly when the sensors are taken out of the water than for the other two samples (ugh). The empty voltage before the sensors are wet is around 2.4 volts, but when they are removed after being wet, the voltage slowly decays down to about 2.8 volts and then very slowly to lower values. This is probably not a problem, unless you're filling and emptying your black tank really quickly...

happy soldering, Zep

[goransons]

its just crazy.. I mean, is there anything you don't do?

[Zeppelinium]

Thanks, Scott. I will admit I'm not a very good ballerina and I don't call my mom often enough. The truth is, I do a lot of stuff, but none of them as a true expert. So beware.

Zep

[silverleeper]

Originally Posted by Zeppelinium

Thanks, Scott. I will admit I'm not a very good ballerina Zep

Some say you have the legs of a ballerina.

[Zeppelinium]

Here's the final monitor design. It's for the 2-wire system. When you look at the printed circuit you can see 3 modules on the right side that are intended to be cut off and installed at the tank to convert a 4-wire system to 2 wires.

The only change contemplated at this time is to spread out the LEDs by an additional 0.1". This design accomodates 6.5x3.5 millimeter rectangular LEDs, which are hard to find. The alternative is smaller rectangular LEDs or slightly larger round LEDs.

The cost of a single board with three monitors is $32 in small quantities. If more than 10 members are intersted, it would lower the cost to about $22. The order could be soon, but definitely before the end of December. PM me if you are interested.











Enjoy, Zep

[2xS]

Zep,

Nice post here! I was just getting started on this when looked on here to see if anyone else has tackled this project.

I noticed you used comparators and an R2R network to emulate the original Airstream design, and is that Autotrax you're using? My real question is: in your experience, did you notice the output voltages of the sensors to be more static or analog? Is it essentially just switches in the tank or is there a variable resistance to play with?


I ask because I would like to have a PIC microprocessor controlled AD/C to manipulate that data into a percent notation instead of the standard LED display. As long as there are voltage steps in between the 1/2, 3/4, full marks, it should be fairly easy. But if it doesn't work, I need to rethink my panel display.

What do you think?

[Zeppelinium]

What is AutoTrax? The software to generate the circuit and the PCB is free from Express PCB.

The resistance detected by the sensors as they touch the water is more digital than analog. (the slope of the solid intervals between the steps is exaggerated in the graph.) The general shape of the resistance curve has large steps as each sensor is submerged, with very small change of resistance as the tank level changes until the next sensor is submerged (or uncovered), like so:



The center value of each of these steps varies, depending on the level of impurities in the water. That variability is much more significant than the total variation across one step. Ergo, the resistance ladder for the fresh tank is different than that for the gray and black (I did not test the reistance ladder for black water, just gray--that test is coming in the actual tank...).

I don't think you can devise a suitable algorithm to convert the stepwise output into a smooth slope. It would be interesting to see if you could devise a learning algorithm that could do an estimate based on the step size, but I think it would take a lot of samples.

I would have gone with a digital solution if it hadn't been for the fundamental variability in conductivity between fresh water samples.

Zep

[Xbob2]

Originally Posted by Zeppelinium

What is AutoTrax? The software to generate the circuit and the PCB is free from Express PCB.

The resistance detected by the sensors as they touch the water is more digital than analog. (the slope of the solid intervals between the steps is exaggerated in the graph.) The general shape of the resistance curve has large steps as each sensor is submerged, with very small change of resistance as the tank level changes until the next sensor is submerged (or uncovered), like so:

Attachment 97904

The center value of each of these steps varies, depending on the level of impurities in the water. That variability is much more significant than the total variation across one step. Ergo, the resistance ladder for the fresh tank is different than that for the gray and black (I did not test the reistance ladder for black water, just gray--that test is coming in the actual tank...).

I don't think you can devise a suitable algorithm to convert the stepwise output into a smooth slope. It would be interesting to see if you could devise a learning algorithm that could do an estimate based on the step size, but I think it would take a lot of samples.

I would have gone with a digital solution if it hadn't been for the fundamental variability in conductivity between fresh water samples.

Zep

Zep, What if you wanted to go the reverse? I want a "Command center" in my '62 ... how do I use modern tank sensors to feed "BIG HONKIN ROUND ANALOGUE GUAGES!"? (Kan ya' point me?)
Bob

[Zeppelinium]

Originally Posted by Xbob2

Zep, What if you wanted to go the reverse? I want a "Command center" in my '62 ... how do I use modern tank sensors to feed "BIG HONKIN ROUND ANALOGUE GUAGES!"?

Assuming that you have the system shown in the circuit diagram in post #1, you just need to replace the original meters with the kind you like. Good luck. Have you seen the prices for analog mechanical round-faced 1 milliamp current meters lately (like the past two decades)? If you think the price-esthetics ratio is acceptable, go for it.

However, if you're talking about some other kind of "modern" sensor system, such as capacitive or pressure, I'd have to see the circuit diagram. All of these systems eventually produce a voltage that is proportional to fluid level, which can be displayed by an analog meter.

Zep

[Zeppelinium]

The prototype works as intended. Yippee. Now to install and see if the water and other fluids exhibit the same impedence as the fluids I concocted (you don't want to know). A few tank fills and empties will determine the final resistor ladder values for each type of tank.

The LEDs are on one side, which still needs a bezel. There will be three switches--bright/dim, pump on, and a momentary-off-on power switch. You can see the color logic for each of the indicators. The gray and black tanks go from green at empty to red a full and the fresh water tank is just the opposite. Still need a couple of LEDs for the gray tank indicator (in the middle). The four silver standoffs in the corners will position the front panel/bezel so that there is room for the swiches--that's why the LEDs are proud of the board by about 9/16ths".



The components and connectors are on the other side. This side will be accessible with the bezel installed. Note that this really is a prototype. The arrows point to strip sockets that will make it easy to change the resistor ladder if adjustments are necessary. The 8-pin chip on the far side is an NE555, which can be configured as a flasher. I have it hooked up to show that the pump is on; 3 seconds off, 1/8th second on. It's a bright white flash, which ought to get one's attention even the daytime. I always seem to forget to turn the pump off when traveling. This flasher will greatly reduce the battery drain of a pump light. As it is, the LEDs only use 5ma when they're on, so that's an average current drain of about .25 ma.




Installation may be a few days away--the snow has stopped, but it's cold and breezy.

If others are interested in this item, the PC boards can be available in three days--$55 setup and about $25 per board (the software is a free download and ordering is automatic over the Net). I'll gladly send anyone the artwork so they can order a board(s), or take orders and do a single order to save on the setup fee. I think it would take about 10 orders to make this a worthwhile effort.

If we get a minimum number of people interested, I'm going to make the layout a bit narrower in height. I found that the current board is almost the same height as the control panel opening in the early 70s trailers. Making it about 1" smaller in height will make mounting it in the existing opening/bezel easier.

[RichHog]

Though I don't fully understand all the circuitry, I love seeing the process unfold. I'm probably ignorant but I haven't seen much on the forum regarding the control panel repair, replacement or troubleshooting in general. Mine (72 TW) only has the water pump switch work, which is the only critical must have component, tank levels, battery level, panel light all shot and the plastic is a little ugly. So, how do I replace the whole gizmo, not just electronics behind the face plate? Coming from the Audio/Video install arena, we often custom punch out metal plates and then engrave the faceplate. Is that the next step to make this work? Has anybody out there already done this? I for one would love to have a nice state-of the art display panel there with some sexy lights and LEDS.