Vanagon EuroVan
Previous messageNext messagePrevious in topicNext in topicPrevious by same authorNext by same authorPrevious page (July 2014, week 2)Back to main VANAGON pageJoin or leave VANAGON (or change settings)ReplyPost a new messageSearchProportional fontNon-proportional font
Date:         Wed, 9 Jul 2014 08:17:46 -0400
Reply-To:     David Beierl <dbeierl@ATTGLOBAL.NET>
Sender:       Vanagon Mailing List <>
From:         David Beierl <dbeierl@ATTGLOBAL.NET>
Subject:      First cleanup Re: Westy water tank level indicator problem
Comments: cc: Richard A Jones <>
Content-Type: text/plain; charset="us-ascii"; format=flowed

>1) You can test the input with a potentiometer and voltmeter. Input >lead is biased to B+ with one megohm, electrodes ground out >successive resistors in a voltage divider inside the little black >module by the tank. Result compared against similar network at the >left-side LM324 chip.

First unplug the green wire at the tank. You can plug and unplug freely with panel power on. None of the tank LEDs should be lit. Then ground the green wire. The green LED only should light. If either of these tests fail, skip immediately to the second section, called LM324s die a lot. Return to this section for info later. Otherwise continue from here:

To test the thresholds connect a digital voltmeter from the green wire to ground and either:

a) operate the sender float on newer ones (be *very* cautious; if you bend the rod the tiniest bit in the wrong direction it will break one or both reed switches inside because no slack was left when soldering them). The meter will pull the voltages down somewhat, see NOTE 1 below and method/formula at c).

b) ground all the sender terminals on the older tanks to the bottom terminal and then remove them beginning from the top, same comments about meter as above. The tank should be empty. If you want to keep water in the tank you can disconnect the electrode wires and short them together in the right order, but mark them carefully before you disconnect them.

c) disconnect the green lead and connect to ground via one end and middle of a 10-20 megohm pot, preferably linear taper, and adjust the pot. With this method you can find the actual thresholds. A ten megohm pot and a ten megohm meter will give you about 10.5V maximum which is plenty. If you want to know what the actual applied resistance is at each threshold, unhook the green lead, switch the meter to ohms, write down the value and calculate Reffective = Rmeter x Rpot / (Rmeter + Rpot). Same method if you want to know how the meter is altering sender resistances in previous methods.

d) disconnect green lead and connect positive side of a variable power supply to it, negative side to ground. CAUTION: do not apply power with the panel turned off (if the panel power switch is pulling out of the solder, fix that first to avoid possibly losing panel power while your supply is driving the line). CAUTION: Do not exceed the measured voltage present on the green wire with panel on and nothing connected, and do not allow the voltage to go negative more than a few millivolts. Some variable supplies may go slightly negative when set to minimum, so check and beware. The chip can handle a very slight negative input; that's how the flame detector LED operates, the fridge thermocouple is the only voltage in the van that's officially negative to chassis ground. But the emphasis is on very slight.

e) connect various combinations of dry cells likewise. See CAUTION above.

f) connect individual resistors likewise ranging down from ten megohms. This is the hard way for sure, but if you have a bunch of resistors it works.

NOTE 1: with either method where you're actually supplying voltage, adding about 5,000 ohms in series will prevent any excitement resulting from shorting a wire to ground. The green wire itself can be grounded safely.

NOTE 2: If you're not actively supplying voltage you will get most useful results with a ten megohm (or higher) voltmeter. Most digital meters have ten megohm sensitivity on voltage ranges, but some cheap ones may be one megohm. A ten megohm meter will drag the green wire down to about 11.5V all by itself, and a one megohm meter will drag it to about 6.3V, severely limiting your testing.

Assuming the LED panel is being fed by 12.6V, the following voltages at at the end of the green wire should have the following results with a few per cent variation for differing resistor errors on the LED board (if all resistors were x% off in the same direction the thresholds would not change):

>12.6-9.45 --> No lights (no bolt jumper/not reachable with float >sender. Signals </= one liter in the tank) >9.45-7.35 --> Red (bottom sender bolt jumpered to next up/float >sender at bottom) >7.35-5.25 --> Yellow (add jumper to next bolt/float sender at middle) >5.25-0.00 --> Green (add jumper to top bolt/float sender at top)

OOPS: Exact values of above numbers are suspect, I'll have to recheck the spreadsheet.

Note that varying battery voltage will change the voltage thresholds proportionately; but will not cause them to shift relative to the operation of a particular sender. The circuit employs three of the four amplifiers contained in the left-hand chip on the panel(the fourth one is a spare). Each amplifier drives one LED, and independently drives that LED any time the output of the voltage divider formed by the various discrete resistances of the sender in series with one million ohms becomes even microscopically less than the output of a particular tap on a reference voltage divider (made up of five resistors on the panel).*** Since both the sender-side and reference-side networks are powered by the same battery voltage, the ratio between them does not change.

***Ordinarily that would mean that the lights would show red, yellow+red, green+yellow+red; but the LEDs are wired in trick fashion so the red one loses its ground when the yellow comes on, and the yellow loses its ground when the green comes on. This is done simply by using the connecting LED- of the red and yellow LEDs to the adjacent LED+ instead of ground. The amplifiers can hold the line either high or low (to say another way they can either source or sink current) so they can easily act as the "ground" for an LED.

Incidentally a different arrangement is used for the battery meter, because there you *want* the result to change when battery voltage does. There a very similar ladder of resistors becomes the sensing part of the arrangement; and the (resistor+changing resistor) pair used for the tank sender becomes instead a (resistor+zener diode) pair which changes output voltage very little with any reasonable change in battery voltage. As with the water side, lighting one LED removes the ground from the previous one.

If you were to add the 500k resistor I refer to in my schematic, the thresholds would change thus: 12.6-9.72 --> No Lights 9.72-7.56 --> Red 7.56-5.4 --> Yellow 5.4-0.00 --> Green

OOPS: Exact values of above numbers are suspect, I'll have to recheck the spreadsheet.

I made this change on my '84 to prevent the panel indicating red when it should have been showing yellow. Adding that single resistor changed all the thresholds because it altered the overall value of the string; to manipulate one threshold without altering the others could be done in various ways. The simplest would be to leave the original resistor string alone and simply re-route the input to that amplifier to a separate pair of resistors having the correct ratio. If the designers had envisioned a need to adjust the thresholds they would have used a separate divider for each amplifier, probably with a tweak adjustment included. It all costs money though, and board real estate.

>2) The LM324 chips die a lot, at least on this board. > >Seriously.

If the battery lights work ok but the water tank lights don't all go out when you unhook the green wire and the green (only) doesn't light when you ground it (or if the battery lights start to misbehave), your first action is to replace the left-hand LM324N with another LM324; because it's highly likely that the chip If Radio Shack still carries them that's fine, otherwise get one or a few from wherever. Chip price used to be in the dollar range; I haven't looked for a few years. If you aren't up to soldering (and this is about as easy a board to work on as exists, see below for method) get whoever does the job for you to install a socket on both sides instead, so you can easily change or swap them yourself later.

UNFORTUNATELY I just took a look at my photo of the board and there may be no room to put sockets in. What could be done is to add individual socket terminals through the board after enlarging the holes slightly. They would sit practically flush on top and wouldn't raise the chips much at all. However I can't really tell from the angle, and a low-profile socket might do fine. The controlling depth is the plane of the top of the switch body and the shoulders of the LED ends as they are what seats up against the rear of the metal panel. If someone measures that clearance depth I can get a better handle on this. Back to business:

The three troubles that afflict this setup (no special order):

1) Learning the hard way that the float-type sender as originally built (and I hope and believe that Trevor has arranged for some slack in the assembly of the new ones) is ABSOLUTELY INTOLERANT of being bent the slightest amount in the direction away from the foil side of the circuit board inside, where the reed switches are soldered on. One or both of the glass reed switches will burst because their leads will separate and glass doesn't stretch much. Repairing is often practical and once you're inside you can put some slack in the soldered leads so it won't happen again. Best way I found to open them was with a plumbing-type tubing cutter used very carefully and rejoined with a rigid oversleeve.** One of the reed switches is a simple N/O type with two terminals and is easily sourced. The other has three terminals and is a bit harder to locate. This circuit carries extremely low current so if a lighter/smaller switch presents itself it will be fine so long as the magnet in the float can operate it and you're even more careful to put slack in the leads. Larger ones the same except they also have fit into the tube which may not accept anything larger. Note which way the three-terminal switch is oriented since the N/O and N/C terminals must be correct. Test the board with the LED panel or ohmmeter and a magnet before assembling. The resistance changes should be in consistent order up and down as you move the magnet. I can't remember exactly how it's wired right now so I'm going to wave my hands vaguely. Anyone with a good float sender can provide correct ohmmeter readings.

1a) Getting a drop or two of water inside said float sender, either from a leaky joint or more likely through the top where the wires come out. This causes troubles where the tank reads too high, but the reading is likely to shift as the panel stays on for awhile. This can be repaired by drilling a small hole next to the bottom, breaching the seal around the wires if need be, and flushing through several changes of 190 Everclear or Denatured Alcohol or high-strength isopropyl alcohol dried with air or vacuum between changes ; then over-cementing assembly seams as well as closing the hole you made and the top seal.** But don't be surprised if you break a switch in the process -- so really it's better to ever so gingerly open it up as above and resolder one of the leads of each reed switch so the board can flex a bit without popping them. Check for corrosion/electrolysis eating up the copper and clean and solder over it if need be. Then dry out the tube and circuit board and put everything back together. And triple-protect the entry point at the top as I think that's where the water usually gets in. I personally would advise removing the wand from the tank and storing it safely before any sort of cleaning activity inside or outside the tank for both flexing and leakage reasons.

1c) Having to scrub off the older-type tank electrodes with green Scotch-Brite, not the wussy blue stuff. Maybe adding a few pinches of salt to a tank if the water's really pure.

** I'm sorry, you'll have to experiment or do some research for a cement that will bond properly. Cheri or Trevor can no doubt supply that info now; they weren't involved with the senders when I was fixing them.

2) Power switch working loose from the solder. It's a very small switch supported only by its leads and it has a hard life from bumping, vibration, something. I used to routinely re-solder mine any time I had the panel out.

3) LM324 gone bad for no obvious reason.

That's it. I have never heard of any other difficulty with them ever.

Ok, I will check the threshold exact numbers and take another pass through probably tomorrow some time. If you (anyone) notice any problems in the mean time let me know. I'm a bit sideways on sleep this past week and my proofreading has suffered.

>3) Some info and schematics here >


-- David Beierl --

Back to: Top of message | Previous page | Main VANAGON page

Please note - During the past 17 years of operation, several gigabytes of Vanagon mail messages have been archived. Searching the entire collection will take up to five minutes to complete. Please be patient!

Return to the archives @

The vanagon mailing list archives are copyright (c) 1994-2011, and may not be reproduced without the express written permission of the list administrators. Posting messages to this mailing list grants a license to the mailing list administrators to reproduce the message in a compilation, either printed or electronic. All compilations will be not-for-profit, with any excess proceeds going to the Vanagon mailing list.

Any profits from list compilations go exclusively towards the management and operation of the Vanagon mailing list and vanagon mailing list web site.