Pressure Sensor Circuit - advice/critiques appreciated!

[rockrockmcrock]

Hi guys,

It was suggested I post here after being booted off another forum because this has a classic car related application (hope you guys don't mind?)

I'm a (modern) electronics newbie - but am slowly learning the basics for this project. Loads of peeps have already been very helpful on helping me get started, but I still seem to have loads to learn, so any advice is hugely appreciated.

The problem:
I want to retain some old bi-metallic gauges in one of my classic cars (got to keep it original on the dash right?). These older gauges are "bi-metallic electronic". However modern gauges work in a different way (i.e. potentiometer based). This means that modern 'sensors' for gauges are mostly 'reversed' in the way they work (more on this later). Of course without the right signal from the sensor the gauge will not work. Is there a simple circuit to make modern sensors work with old bi-metallic gauges?

How bi-metallic gauges and sensors work:
All gauges are 10v. The gauges work by a 'sensor' which acts as a variable resistor and limits the current on the gauge circuit in line with pressure. The current heats a coil wrapped around a bi-metallic strip in the gauge with responds by moving the needle on the gauge. Low pressure = high resistance (240 ohms) => no heating of coil & gauge needle reads low, high pressure = low resistance (33 ohms) => coil is at it's hottest & gauge needle reads high.

How modern sensors work:
Most sensors are 10v. Modern sensors work the opposite way round but over roughly the same range. In other words, Low pressure = low resistance (33 ohms) => gauge needle reads low, high pressure = high resistance (240 ohms) => gauge needle reads high. If you put this modern sensor on the old bi-metallic gauge setup the reading's would be reversed (i.e. the needle would read high when at low pressure and low when at high pressure.

What the project circuit would need to do:
Work from a 12v car battery, produce a 10v regulated sources and then reverse the modern sensors output - i.e. when the sensor gives high resistance, allow more current to flow on gauge circuit and visa versa.

The Current Project Circuit
So far (with a lot of help from others - thanks H!) I have come up with this circuit - but will it work?:

Here's my understanding of the above circuit and how it works (I might be sooo wrong though!):

• Voltage regulator produces 10v & up to 1A (easily enough) at 2 & 6 for 10v rated gauge and RSensor
• Voltage divider (2,3,4) effectively inverts RSensor output, by producing a voltage at 3 which is proportional to RSensor resistance (i.e. increasing RSensor resistance, increases voltage at 3). RTrimmer can be used to vary the sensitivity for the op amp (ie "high-low" voltage change delta - is that correct?)
• No current is carried across the Op Amp (3,5), this is the need for the second circuit 6,7,8,9 - the voltage at 3 is simply used to proportionally vary the current across 6,7,8,9.
• The Op Amp triggers the NPN transistor to control the current and the feedback loop 8 to the Op Amp provides the reference input to the Op Amp (comparison of 8,3).
• To ensure the right current in 6,7,8,9, RSense is used, such that current across RSense (8,9) must equal current across Gauge (6,7). Does this mean I need to know the resistance of the Gauge (Load) though to be accurate?

OK, assuming I'm not completely wrong about the above - any advice the following hugely appreciated:

1. NPN Transistor needs to carry the current and be triggered by a low current - hence a high gain, 400mA+ specification is ok?
2. The Op Amp - there seems to be loads of variables to consider here....totally confused at present, but slogging on! All help appreciated!
3. Any improvements to the circuit

Thanks again guys and to all in the community (internet and real-world) who have helped on this already. This seems to be a very widespread issue in the classic car world, but no one has found an easy solution that I can see - if we can get this solved then the classic car community would be in your debt!

RRMR

 

[MikeMl]

Welcome. There is Newbie - Needs help with electronic variable resistors - Custom fuel gauge already a thread running on this topic....

 

[rockrockmcrock]

Welcome. There is Newbie - Needs help with electronic variable resistors - Custom fuel gauge already a thread running on this topic....

Hi Mike - thanks, took me a little time to find it as that link doesn't work for me - hopefully the one below does for others.

https://www.electro-tech-online.com...-variable-resistors-custom-fuel-gauge.143938/

 

[MikeMl]

...
I want to retain some old bi-metallic gauges in one of my classic cars (got to keep it original on the dash right?)....
How bi-metallic gauges and sensors work:
All gauges are 10v. The gauges work by a 'sensor' which acts as a variable resistor and limits the current on the gauge circuit in line with pressure. The current heats a coil wrapped around a bi-metallic strip in the gauge with responds by moving the needle on the gauge. Low pressure = high resistance (240 ohms) => no heating of coil & gauge needle reads low, high pressure = low resistance (33 ohms) => coil is at it's hottest & gauge needle reads high.

To proceed, you need to hook up this simple circuit and make the specified measurements:

 

[rockrockmcrock]

Cheers Mike - OK - I can certainly do that, which I guess will give me the gauge's response curve. However, to date I've approached the problem from the other end and looked at the response curve of a properly matched pressure sensor/transducer for the gauge. Ohms law should hopefully give me the current flowing around the circuit.

The 'static' resistance of the gauge itself is 25 Ohms when the coil is cold and (surprisingly) when hot also!

The response of the original pressure sensor (borrowed over Xmas from a mate) for this gauge is:
0 psi = 352 - 364 ohms
28.5 psi = 207 - 234 ohms
57 psi = 101 - 142 ohms
85.5 psi = 42 - 83 ohms
114 psi = 21 - 56 ohms

Which matches well with some historical data available on the internet for an EN056 sensor (my only reference data before Xmas):
0 psi / 0 bar = 350 - 367
28.5 psi = 205 - 233
57 psi / 4 bar = 105 - 144
85.5 psi = 46 - 86
114 psi / 8 bar = 22 - 56

However, others (on the internet and real world) have reported that American marine standard 240-33 Ohm pressure sensors work well with these gauges and approximate the response of the original sensor above. The problem is getting hold of these sensors in the UK!

So how can a 240-33 ohm sensor work fine when the original spec was for a 350-50 ohm sensor? And does it matter?

Well these types of gauges/sensors are inherently inaccurate due to mechanical hysteresis (amongst other things!). However, an accurate measure of oil pressure is not a required - it is not critical if it is out by some degrees - what matters is that you can tell a sudden loss in pressure, have a rough idea of minimum pressure, and that you can follow a trend over time. Add to that the fact that my gauge only shows "L,N,H" on the dial (for Low, Normal and High) and basically, for this application, an accurate response curve (within obvious limits) is not so critical, as I'll be calibrating it using a mechanical pressure gauge anyway

 

[alec_t]

Why the big variation in resistance for a fixed pressure? For example, at 57 psi you have 101-142 Ohms .

 

[rockrockmcrock]

Why the big variation in resistance for a fixed pressure? For example, at 57 psi you have 101-142 Ohms .

Mechanical hysteresis because of the way the sensors are constructed - as I said accuracy is not that important

Also, the variance at the mid-point increased slightly with increasing cycles in succession - I did 20 cycles from memory, stopping when I was pretty sure the historical figures weren't going to be breached (which probably explains why the figures agree so well )

Here's a general hysteresis cycle to give you an idea of the way the variance....er...varies :
**broken link removed**
Note that maximum variance is at the mid-point of the cycle (i.e. 57psi point in my figures)

 

[MikeMl]

Follow my thought experiment:

The newer cars do indeed use a resistively-heated bi-metal to mechanically move the pointer on the gauge. These gauges are not your father's D'Arsonval meter movement used in early cars (and airplanes). The latest cars use stepping-motor movements.

The bi-metal gauge (I'm calling it a bmg) responds to the temperature of the bimetal. To a first order, the bending of the bimetal is proportional to its temperature. The temperature rise of the resistive heater above ambient is proportional to the power dissipated in the heater. Assuming the resistance wire that heats the bi-metal is constant (a big assumption if the heater wire is NiChrome), the power in the bmg's heater is proportional to the square of the current through it.

To eliminate ambient temperature effects, bmgs usually have three terminals: One tied to a regulated 10V, a second which goes to the sender, and a third to ground. I suspect that there are two heaters in the bmg; the pre-heater is wired between 10V and Gnd, and is used to "bias" the temperature above ambient, and the sender-heater is modulated by the sender resistance to create the needle deflection. The pre-heater might actually utilize a negative tempco in the wire to make the pre-heat start-point temperature somewhat independent of ambient temperature?

This little sim shows what I am talking about:

I am varying the sender resistance from 30Ω to 240Ω in steps of 3Ω (X-axis).

The upper plot pane is power dissipated in the bmg heater vs sender resistance at different heater resistances from 200Ω (Green) to 600Ω (Violet) in steps of 100Ω.

The lower plot pane is the current through the heater or sender vs sender resistance. (Same color scheme).

Note that the linearity of the power plot gets better at higher heater resistance, but is there enough heating of the bi-metal at ~100mW to make the needle move?

Now you see why it is so important to actually characterize the existing gauge before we start talking about circuits...

 

[ccurtis]

Voltage divider (2,3,4) effectively inverts RSensor output, by producing a voltage at 3 which is proportional to RSensor resistance (i.e. increasing RSensor resistance, increases voltage at 3). RTrimmer can be used to vary the sensitivity for the op amp (ie "high-low" voltage change delta - is that correct?)

Actually, it won't be proportional. Voltage at 3 will be (10 x Rsensor)/(Rsensor+Rtrimmer). The larger Rtrimmer is compared to Rsensor, the closer to proportional the voltage at 3 will be to the sensor resistance. Something to be aware of, even though precision is not a big consideration here. Current through the meter (gauge) will be (voltage at 3)/(RSense).

 

[MikeMl]

... Current through the meter (gauge) will be (voltage at 3)/(RSense).

but that says nothing about needle deflection (which I suspect is proportional to gauge power), and that we know nothing about the spacing of the markings on the meter face...

 

[ccurtis]

but that says nothing about needle deflection (which I suspect is proportional to gauge power), and that we know nothing about the spacing of the markings on the meter face...

True. If just going on the assumption that his existing gauge is known to work with a specific resistive sensor (old and no longer obtainable) in series with 10V and the only difference is that the new sensor has the resistance flipped, then how the gauge works should not be a consideration to my thinking.

 

[rockrockmcrock]

Actually, it won't be proportional. Voltage at 3 will be (10 x Rsensor)/(Rsensor+Rtrimmer). The larger Rtrimmer is compared to Rsensor, the closer to proportional the voltage at 3 will be to the sensor resistance. Something to be aware of, even though precision is not a big consideration here. Current through the meter (gauge) will be (voltage at 3)/(RSense).

Thanks - yes I get what you mean I think, as RSensor varies for a set RTrimmer resistance in use the proportionality (and hence accuracy) of the output at 3 will vary.

but that says nothing about needle deflection (which I suspect is proportional to gauge power), and that we know nothing about the spacing of the markings on the meter face...

I wouldn't worry too much about face markings.....

Follow my thought experiment.....To eliminate ambient temperature effects, bmgs usually have three terminals.... I suspect that there are two heaters in the bmg; the pre-heater.....

No this has a single coil, without a preheater - I can absolutely confirm this having just restored this gauge and had it in bits for a while (see photo above). This is a pretty basic gauge even by classic standards! You're bang on that many VW/Willys gauges used a preheater to remove ambient temperature effects, but this one is no way near that 'advanced' - we're talking a coil/bimetallic strip assay attached to a pivot pin and tensioner - seriously this is not an accurate gauge at all

I do see what you mean with your thought experiment though.

True. If just going on the assumption that his existing gauge is known to work with a specific resistive sensor (old and no longer obtainable) in series with 10V and the only difference is that the new sensor has the resistance flipped, then how the gauge works should not be a consideration to my thinking.

That is kind of where I was coming from but don't have the electronics knowledge to know if I'm right or not to make this assumption...

 

[MikeMl]

True. If just going on the assumption that his existing gauge is known to work with a specific resistive sensor (old and no longer obtainable) in series with 10V and the only difference is that the new sensor has the resistance flipped, then how the gauge works should not be a consideration to my thinking.

There is no reason to think that the resistance versus whatever (pressure, temp, etc) of the new sender is the inverse function of the old one...

 

[rockrockmcrock]

There is no reason to think that the resistance versus whatever (pressure, temp, etc) of the new sender is the inverse function of the old one...

Actually that's a very good point - but look at my gauge - does it practically matter in this case? I don't know, so I'm looking for advice here, but my gut tells me that it is probably not critical for my particular gauge.

It's worth noting that all these senders are made to either Euro or US standards and are supposed to be reasonably interchangeable within the standard resistance / pressure bandings (eg. US marine standard 240-33ohms 100psi).

 

[KeepItSimpleStupid]

We just had a discussion not too long ago about another meter which compensated for the difference in battery voltage using two coils. Yours has a regulator.
Yet other systems I know about had a pulsing regulator for the gas gauge and fluid in the meter. There are others who might be using GM stepper type motors and a computer (Arduino) to replace gauges.

I'd also call your attention to:
https://www.linear.com/product/LTC2983#demoboards

With the reminder that you want to keep things simple.

A sensor type-changer is an idea I've been toying with, but not for this application.

==

PS: How did you do the gauge restoration? That's gorgeous.

 

[rockrockmcrock]

Yet other systems I know about had a pulsing regulator for the gas gauge and fluid in the meter.

Actually, mine also in its original form - most classic cars use the system. Basically, in order to get a 'regulated' voltage supply for gauges, they used a bimetallic oscillating strip to open and close contacts to provide an effective 10v supply from the 12v source. This oscillating regulator would play havoc with my 'modern' circuit so I've used a modern equivalent.

There are others who might be using GM stepper type motors and a computer (Arduino) to replace gauges.
I'd also call your attention to:
https://www.linear.com/product/LTC2983#demoboards
With the reminder that you want to keep things simple.

Always tempted by tech - in fact, if the gauge had had proper pressure markings on it I might have stripped out the guts and replaced them with a cheap RC micro servo (which I've done before). Despite having no idea about analogue electronics, circuits etc., I do know a fair bit about hacking digital systems & programming MCs (ASM preferred over higher level languages like C for this approach whatever anyone says!). Weird, that I can explain how to decode and programme a chip to interface with I2C bus comms, but had no idea what a voltage divider does - modern world for you I guess, but digital reference circuits and test boards have a lot to answer for!

PS: How did you do the gauge restoration? That's gorgeous.

Patience and a steady hand mostly - took 2 hours, but the results are always worth the time. Luckily the guts of this gauge were sound - except for a false widow spider I had to evict! Then its all just soapy water, cotton buds and very fine paint brushes for the gauge face. Actually it's not that hard - anyone can do it if they have the right attitude to life (i.e you get out what you put into it ).

 

[MikeMl]

I'm working on the interface between the old gauge and new sensor, but I really need to know how many mA it takes to make the old gauge read 10% of scale (just right of L), 50% of scale (N), and 90% of scale (just left of H).

 

[rockrockmcrock]

Wow! Thanks mike - its just that I'm currently ill with flu (hence the reason I've been on here 24/7 rather than at work or out) - unless I can find a way to wing it with what I already have, a trip to the shops to pick up some components is a bit beyond me at present

 

[rockrockmcrock]

OK I've finally managed to get hold of a pot of the approximately right value (1K) - the measurements are below, bear in mind they are very approximate given the gauge markings (or lack of them!)

High (100%) 4.67v 183.6ma
Mid (50%) 5.94v 137.7ma
Low (0%) 8.68v 70.2ma

R2 measured out at 22.3Ohms on my meter.

Interestingly it would seem that my gauge is of a different spec to others like it from Jaeger - doing some sums (R=V/I) gives me a resistance range of 123.6 - 25.5 ohms - very unexpected as my mate has a gauge exactly like this one that comes out completely differently!

RRMR