My new Boost/EGT Gauge

[300SD81]

I've posted bits and pieces of this elsewhere, but here it is all together...

Presenting the homemade digital boost/exhaust/EGT gauge!



(Note Mercedes Star etched into circuit board)

Total cost was $10 with generous use of free samples available from almost every semiconductor company. The large chip on there is a PIC microcontroller programmed with Microchip's USB serial port library and about 300 lines of my own C code. It uses a MAX6675 0-1024C thermocouple to digital converter for EGT and 2 MPX4250GP 0-36psi pressure sensors for boost/exhaust.

The display,

(oil/transmission sensors are in the works, temperature sensor chip outputs full scale when no thermocouple is conneted)

written in C#, simply reads the serial port (raw binary ADC input from microcontroller), converts the data to PSI/degF and outputs it on the gauge.

I haven't installed it in the car yet, but will be installing the boost gauge function only first. Don't have the money for new manifold/turbo gaskets at the moment, or a tap to thread the manifold. For now its going to be read on a laptop in the passenger seat, but this is going to be mounted under the glovebox as soon as I can find a old flatscreen that can be taken apart and mounted in the drivers sun visor.




I also have parts on order for an accelerometer, which I should be able to program to give me 0-60 times fairly easily (velocity is the integral of acceleration, should be easy to calculate). I can't get the gauge design for it to look good, which is why its not shown.

[Matt L]

Originally Posted by 300SD81

I also have parts on order for an accelerometer, which I should be able to program to give me 0-60 times fairly easily (velocity is the integral of acceleration, should be easy to calculate). I can't get the gauge design for it to look good, which is why its not shown.

While it is true that velocity is the integral of acceleration (from 0 to t, specifically), have you considered how you would compute the error bound?

[Actros617]

O i wish i had VB, but cant afford it

[pawoSD]

Would you be willing to share the code you wrote for that? I have VB and could use it....

[TheDon]

I have a 15" laptop screen if you want it.. just cover shipping

[pawoSD]

Its going to have to be a pretty small and narrow LCD to fit in the drivers sun visor.....not to mention the wiring fun that would bring with it.

[300SD81]

Originally Posted by Matt L

While it is true that velocity is the integral of acceleration (from 0 to t, specifically), have you considered how you would compute the error bound?

Haven't really thought about it, but the accelerometer has a 2.5ms sample rate and the microcontroller is at least twice that. It should be accurate enough for a rough 0-60, at least more accurate than the stopwatch method.

Originally Posted by pawoSD

Would you be willing to share the code you wrote for that? I have VB and could use it....

The code is in C#, but can be translated to VB. Its really simple code anyway, get yourself a copy of Dundas Gauge and have a look at the .NET SerialPort class. My code is basically:
Data = SerialPort.ReadByte();
Pressure = Data * ScaleFactor;
Gauge.Value = Data;

Its a bit more complex because the scale is actually a 6th order polynomial, and I have to determine which gauge the data goes to with a bitmask (data is 12 bit, the first 4 bits of the 16 bit integer are ID bits) The Dundas Gauge library takes care of all the graphics, I just feed values into it.

You'll need a good bit more than VB to build it though, at least MPLAB and some knowledge of C and a PIC programmer to program the microchip. All my circuits are surface mount, and the thermocouple chip is surface mount only. (I solder parts down to size 402 (0.04" × 0.02") with a $5 radioshack iron, so it is very possible to do it at home.)


Anyone interested in a kit? You'll have to buy/sample the chips yourself though. I'll offer pre-programmed PICs.

Originally Posted by TheDon

I have a 15" laptop screen if you want it.. just cover shipping

Way too big, but thanks anyway. I'm looking for a 10.4" max widescreen or 8" non-widescreen. Laptop LCDs are impossible to drive anyway, the cheapest VGA-laptopLCD converter is like $200, since laptops generate the LCD type signal directly on their motherboards, theres no VGA input to them.

Originally Posted by pawoSD

Its going to have to be a pretty small and narrow LCD to fit in the drivers sun visor.....not to mention the wiring fun that would bring with it.

I'm not planning to use the casing. I'm going to strip it down and use the bare LCD itself, so thickness is not an issue (think about half the thickness of laptop screen), and wiring shouldn't be either, if they can fit it in a laptop hinge, I can through the metal thing on the visor.

[pawoSD]

What metal thing on the visors? The power to the visor comes through the securing clip and not the moving shaft.....so theres no where to run wires....

[Matt L]

Originally Posted by 300SD81

Haven't really thought about it, but the accelerometer has a 2.5ms sample rate and the microcontroller is at least twice that. It should be accurate enough for a rough 0-60, at least more accurate than the stopwatch method.

You're guessing. Say it takes ten seconds to get to 60. How many samples is that? Each sample has descritization error, and a computation like an antiderivative compounds the error as you go along. I'm sure that you can visualize this, but there are college courses that cover computing error bounds.

What if you have a ten-second error bound?

[300SD81]

Originally Posted by pawoSD

What metal thing on the visors? The power to the visor comes through the securing clip and not the moving shaft.....so theres no where to run wires....

Ehh... that does present an issue.. and theres nowhere else in the car I can mount a LCD and have it look decent..


Looks like I can still use the visor, drill a little hole in the plastic thing that screws onto the car and run the lcd wires along there. Shouldn't stand out, especially if I paint the cable. I want the screen in a place where it can be hidden, but still easily visible while driving (using it for gauges, GPS, etc)...don't want the car being the first target for some gangster wanting a new LCD display...

Originally Posted by Matt L

You're guessing. Say it takes ten seconds to get to 60. How many samples is that? Each sample has descritization error, and a computation like an antiderivative compounds the error as you go along. I'm sure that you can visualize this, but there are college courses that cover computing error bounds.

What if you have a ten-second error bound?

You just went over my head.. I just graduated highschool 2 months ago and took AP calculus my senior year (and got a 5 on the test)... Never even heard of an error bound until this thread.

[Matt L]

To prove that a particular discrete computation models the continuous reality, you can usually not show that it is completely accurate. Rather, you have two functions, F(t) and its numerical approximate G(t), and you show that |F(t) - G(t)| < E(t) where E is called the error bound. If E(t) is sufficiently small for your application, G is a good candidate for numerically computing F.

[bgkast]

Originally Posted by Matt L

To prove that a particular discrete computation models the continuous reality, you can usually not show that it is completely accurate. Rather, you have two functions, F(t) and its numerical approximate G(t), and you show that |F(t) - G(t)| < E(t) where E is called the error bound. If E(t) is sufficiently small for your application, G is a good candidate for numerically computing F.

How does the G-tech do it?

[Matt L]

Originally Posted by bgkast

How does the G-tech do it?

You would have to ask them.

I am not trying to imply that it is impossible, just that you need to be careful with your computations. Things are not always as they seem.

I would imagine that they have proved that their formulas to be correct within a certain tolerance (error bound) and publish this tolerance in the specifications. Perhaps it is empirical data used to validate their system, in which case it may well be very wrong if it is used on a very fast (or very slow) car.

[lowriderdog37]

Once you get yours up and running, I would be interested in a kit. So many gauges I want, but just not enough room in the car. How much would something like this cost?

[Deni]

How will you correct the acceleration squat? As the car leans towards the rear the accelerometer will think it's accelerating faster than it actually is.