The breadboard prototype and final build.

The switch controls 2 levels of brightness for daytime and nighttime use. The finished unit will probably have a jack for auxiliary O2 signal monitoring but that won't go in until the case panels are complete. Edit: I ditched that idea for now.

Some design features, plus some thoughts that might be helpful for those with enough general electronics knowledge to do this but maybe not familiar with some of the specific parts needed:

I ordered parts from www.mouser.com and www.digikey.com. For simplicity I include Mouser part numbers here although most or all parts should also be available at Digikey. Both are good sources for electronic parts. I only slightly favor Mouser, for various reasons.

I used half of a Busboard Prototype Systems POW3U project board. Mouser part# 854-POW3U. This layout allowed for 3 banks of LEDs and eliminated the need to hand wire 27 of the 30 LED cathodes to the chip pins. I prefer strip board for this custom board work rather than perf board to minimize wiring and related node interconnects. The board is rather expensive at just under $11 but it would be possible to split the board in half and build two, or use the other half for something else. Alternatively something like the Radio Shack 276-150 strip board would be good for one LED bank but 3 (~$7) would be needed for this project, requiring much inter-board wiring and complexity. So I bit the bullet and bought the best tool I could find for this job :-)

3 position SPDT switch (ON-OFF-ON) controls LED brightness, dim, off and bright. It switches different values of the R1 resistor. Here I used 20 ohms and 432 Ohms. The two R1 resistors are plug replaceable on a DIP8 socket due to some uncertainty over optimum brightness levels, especially night time. I don't want it to be too distracting :-)

The case is a Hammond RM2015S. It is one of the few off the shelf project boxes that are wider than deep and with suitable end panels of around 1" height. The search for a well fitting streamlined case was tougher than I first thought. I was very happy with this case, especially the under $5 cost. Mouser part# 546-RM2015S (even cheaper from Digikey). The board is screwed down only in the rear, with the front "floating" but held in place by the LED panel cutout. In practice that seems to work well.

My layout is such that the board is oriented upside down for the standard left-->right bar display. That puts the case "bottom" facing up, against the bottom of the dashboard (as I currently have the box positioned in the car).

The right angle 20 pin DIP sockets were more or less mandatory for a LED bar display perpendicular to the board but are tough to source. I used Mill-Max 299-93-320-10-001000, Mouser part# 575-332010 at $7.35 a pop. This significantly added to the cost but the cleanliness of the layout and the ease of construction were too compelling. Using some sort of DIP socket allows for plug and play replacement of the LEDs to change colors or maybe to a better LED bar some day. Somewhere out there may exist LED bars that come in different colors of better matched brightness for a given current level (details below).

My selected board was designed to accept a 2x30+ pin slide in-slide out header connector (see the board's datasheet). It might be possible to source something similar, with a right angle configuration and for as little as a couple of dollars. But I never found one with the required 0.300" spacing needed for the DIP20 LED bars.

The heat sink is Avid/Thermalloy #577202B00000G, Mouser part# 532-577202B00. Just something I had laying around, leftover from old projects and a good size for this board, case and (barely adequate for the) thermal needs. In a roomier, thicker case I would have used a larger sink.

The controller chip is an LM3914 in a DIP18 package. This appears to be a cannoical LED controller chip, used by every example of this meter I could find.

The 10 bar LED banks are Kingbright #DC10GWA, Mouser# 604-DC10GWA (green), available in Red, Green and Yellow. Or Lite-On TA-1000HG, a seemingly identical and plug-replaceable bar. There are others but these two were the cheapest at under $2 each. I don't like the variation of brightness between colors of the actual LEDS I sourced, which was a combination of Kingbright and Lite-ON due to stock status issues. I currently use 2 reds and a yellow (Lite-ON) in the finished unit.

There is a method described in the LM3914 datasheet to individually tune LED brightness via the resistor on Pin7, but when using a resistor in series with the LED supply, as suggested in the datasheet, the pin 7 resistor no longer has an effect. The result is that I was unable to simultaneously provide some sort of brightness control AND tune the brighter colored LED. If you find a way, let me know :-).

I would have much preferred a fully variable brightness control (just like the vehicle's dash dimmer) but the required ~500 ohm pot in a suitable size (and cost!) eluded me. If you find a suitable pot please let me know!

There are 3 trimmer resistors (two optional), each effectively setting the reference voltage level on LED #10 in each respective bank. This allows for an "expanded scale" on the outer two banks. For example, the standard linear arrangement would have bank 1 ending at 0.33V and bank 2 ending at 0.66V. I might want to tune bank 1 down to 0.2V, providing 20mA resolution on the lean end. I'm going to run this fully linear for now and see if that is not a solution in search of a problem but I saw the expanded scale setup in at least one design I ran across and wanted to be able to implement that without future board modifications.

Resistor R11 is plug replaceable. This resistor is in series with the last trimmer. Switching between linear and expanded scale mode dramatically changes the resistance needed by trimmer R4, which sets the voltage reference on the final (30th) LED and therefore the overall maximum votlage range of the device. My spreadsheet model of the voltage dividers suggested this would be good insurance that one 100K R4 trimmer would work for all configurations. An alternative would be to overkill the R4 trimmer value (maybe 200K Ohm or more) but using too large an R4 value could cause problems with the precision of the high end scale.

Trimmers R2 and R3 (if used) parallel the internal 10K voltage dividers on LED banks 1 and 3, allowing for the optional expanded scale display. A pair of 2 pin SIP strips make this optional, shorting the SIP pins with shorted SIP headers puts these trimmers into the circuit. If my R2/R3 values prove insufficient then a series resistor can be soldered across the SIP header "short plug". I have some reason to believe the 25K I spec'd for R3 is not quite sufficient for some possibly useful expanded scale configurations. The 20K specified for R2 should be fine for any seemingly useful configuration I modeled.

Be aware that my expanded scale voltage divider is the simplest possible, yet has the disadvantage of all 3 trimmers interacting together. The datasheet provides alternative methods of non-interacting expanded scale adjustments, but quite a bit more complex and not fully in context of our needs here for a 30 LED setup. There may be a better way to do this.

I used a 5V fixed regulator (LM7805) in a TO-220 case with an appropriate sized heat sink. My 5V rail drives the entire circuit but there are viable alternatives.

My reference voltage divider is tied to the regulated 5V, providing consistent accurate calibration on the work bench or in the car. In principle I could have used the internal 1.25V reference provided by LM3914 pins 7 and 8, and dividing down from a lower voltage level might have made the trimmer selections easier but I found that constant voltage source to act strangely with the cascaded controller arrangement. It was intuitively simpler to use the LM7805 output, which I fully understood. At the point I made that decision I also wanted to "reserve" the pin 7 resistor selection to optimize LED brightness but that did not work out well either. The datasheet suggests that trying to do it all (divider ref and LED brightness control, cascaded chips) with pins 7 and 8 requires an additional opamp. I never tried that because I didn't want the additional complexity and board space.

The LED supply voltage could, in principle, be set to battery voltage (~14V) or the 5V output of the regulator. In daytime mode the LEDS cumulatively draw 200 ma or more (up 600 to 750 ma max per their spec of 20-25ma each!), depending on the brightness you configure. All that power needs to be dissipated somewhere. I decided to let the regulator take the brunt of this, simply because the heat sink best allows for it. Low (bright) values of R1 also drop some heat across that resistor, requiring something in the range of 1/2 to a full watt dissipation for a bright LED output. The 1/2 watt Vishay/Dale resistors I used are conservatively rated and generally considered good for up to a watt and in practice that 20R value seems to work.

Omitting the R1 resistor results in a very bright display but the regulator will get very well above hot to the touch and I personally would avoid that. I don't like to push these TO-220 devices much above 50C or so.

The board includes a slow blo resettable 500 ma PTC fuse (Mouser part# 650-RXEF050). My intent here was not to protect the vehicle's electrical system but to shut the thing down in the event the LED current draw for some reason exceeded my calculations, but less than a suitable general fuse value of an amp or two. Given this is a DIY project that is inherently subject to design and build flaws I would SERIOUSLY consider fusing the 12V power connection at the source of your tap (not on the case end). I used a fast blow 2 or 3 amp inline ATO fuse at the RX7's B23 connector where I got all 3 required taps. That 12V tap is fused via one of the mission critical engine/ECU fusible links. The idea is to make very sure that in the event of a short your inline meter fuse blows before that link; otherwise your vehicle could just just die while on the road.

The schematic and build includes an "optional" 2.2 uF Tant capacitor across the Vled supply. According to the datasheet this is required in the event that "the leads to the Vled supply are longer than 6 inches". I interpret the "Vled supply" to be the path all the way back to the battery/alternator. In researching this project on the net I think most people misinterpret it to mean the length of the wires between the controller chips and the LEDs, which in this case are less than an inch or at worst a couple inches. I don't know who is right here but the part is only a buck and change so I would suggest using it.

It is possible to add in a bar/dot switch option but for a 30 LED display that would have to be a 3PDT switch, with 7 wires running from it to the board (3 switch pins wired to a common 5V wire to the board). I didn't want that badly enough to go through that effort and didn't think the switch through until after 2 rounds of parts orders. But it could be done :-). I think dot mode *might* make better sense of the strange voltage curve of the narrow band sensors, and possibly be less distracting on a day to day basis after the novelty wears off.

A final note: it is considered good practice to tie the unused end of a pot to the wiper and maybe I should have done that just for proper form.

Although the datasheet suggests the internal divider is "10K" ohms, I found in practice it is slightly above 11K, which had a significant impact in the resistors I selected for the final external divider circuitry. That was one reason for my concerns over adequate range of the trimmer pots and that plug in series resistor on the high side.