You must be careful putting an inductor in the input circuit of a regulator. Properly done it will do a good job of eating short fast transients. However, there is a risk of producing a high-Q resonance that can generate an over-voltage what would not otherwise exist. The greatest peril is generally when the input voltage is applied abruptly (e.g. switched from a low-impedance source - like a car battery) and the capacitance at the input to the regulator is small and has good high-frequency characteristics. Even the inductance of long input wires can be a killer. The solution is to "spoil the Q" (dampen) the resonance. This is usually readily accomplished by using electrolytic in parallel with the ceramic. The large capacitance along with the ESR of the large cap is usually quite effective in damping the resonance. If an electrolytic is not allowable, an RC in parallel with the main ceramic cap might do if the inductance is not too large. The same sort of problem can arise anywhere LC filters are used in power circuits.

The other thing that must be considered when using a discrete inductor is that the energy stored there-in has to go somewhere if the load on the output of the regulator drops abruptly. Again, a bulk cap can prevent over voltage.

===
The 7800 series originated as the µA7800 series from Fairchild. National introduced their LM340 series after that.

 

For a typical automotive power filter the calculations were primarily about cost and size. The inductance and inductive choke require that you have a reasonable knowledge of how much series resistance and voltage drop your application can accept. That determines the wire size, and once you select a core, it sets a limit on the number of turns, which defines the inductance. For a less monetary method, consider that the impedance of the inductor and the capacitor form a voltage divider to noise voltages. Given that the noise rise time equates to an effective frequency, which you may be able to know, usually the limit is as much inductance as you have room for and as big a capacitor as you can afford.But always keep in mind the voltage drop of the inductor's resistance on the load current that you are drawing.
Sorry about not having an easy and simple to use formula.

 

Your schematic looked pretty damn good to me, and very right! Except for the TVS diode showing a forward biased diode that will pop the fuse instantly. My apologies for my incomplete posts. You guess right about the L and C values I suggested. The values I have quoted are just what I would start with. To make a properly designed filter the noise to be filtered needs to be enumerated and the maximum noise that can be tolerated needs to be known as well. Then the values can be defined properly. If you use the values I suggested you should have at least a good chance of no major problems with electrical noise and the 100uH inductor is likely to remain inductive over the frequency band of interest. A higher value will have a lower self resonant frequency and therefore will become capacitive at a lower frequency as well.
The fuse will need to be a slow blow or time lag type because of the inrush current into the caps. For example:
https://www.digikey.com.au/product-detail/en/bel-fuse-inc/0659C2000-12/507-2005-ND/5843989
The caps around the 7808 should be an electro and a ceramic in parallel on the input and output. I'd suggest 100nF and 100uF.
For the TVS, have a look at
https://www.digikey.com.au/product-detail/en/littelfuse-inc/SMBJ15A/SMBJ15ALFCT-ND/285981
or if you want something with more current capacity
https://www.digikey.com.au/product-detail/en/littelfuse-inc/5.0SMDJ15A/5.0SMDJ15ACT-ND/1835377
My apologies again for the lack of detail in my posts. Luckily you are also an engineer and clever enough to correctly fill in the blanks I leave. If my CAD (Altium) had not died I would be using schematics more and the details would be there.

I've redone the schematic and because of the lack of a TVS I used a regular Zener diode. I guess for the schematic and simulation will do.
I assume the only way to actually know the noise and the frequency I have to eliminate is by putting an oscilloscope in there. Two problems here: I don't have an oscilloscope (but will definitely get one soon), there is no "in there" (the engine I'm rebuilding is in pieces now down to the last bolt).
The last resistor is the sensor which for pressure is 8-184Ohm and for temperature is 4kOhm to 15Ohm.

With 100uH and 2200uF I get a cut-off frequency of ~340Hz and an impedance of ~0.2ohm. I literally have no clue what this means. I assume frequencies above 340Hz will not make it to the regulator, but that's about all I can think of. I don't have a clear (or half clear) understanding of the real-world effects. I'm perfectly happy to take your word for it and when I'll have the engine working and an oscilloscope maybe I'll understand a bit more.

I hope I got the caps right, 100nF (0.1uF) ceramic and 100uF electrolyte. Does it matter which comes first if they are in parallel?

The fuse is another big question mark. I'll have small currents (or hope to have) so I probably first need to see the max current the sensor will draw from the regulator.

 

I've redone the schematic and because of the lack of a TVS I used a regular Zener diode. I guess for the schematic and simulation will do.
I assume the only way to actually know the noise and the frequency I have to eliminate is by putting an oscilloscope in there. Two problems here: I don't have an oscilloscope (but will definitely get one soon), there is no "in there" (the engine I'm rebuilding is in pieces now down to the last bolt).
The last resistor is the sensor which for pressure is 8-184Ohm and for temperature is 4kOhm to 15Ohm.

With 100uH and 2200uF I get a cut-off frequency of ~340Hz and an impedance of ~0.2ohm. I literally have no clue what this means. I assume frequencies above 340Hz will not make it to the regulator, but that's about all I can think of. I don't have a clear (or half clear) understanding of the real-world effects. I'm perfectly happy to take your word for it and when I'll have the engine working and an oscilloscope maybe I'll understand a bit more.

I hope I got the caps right, 100nF (0.1uF) ceramic and 100uF electrolyte. Does it matter which comes first if they are in parallel?

The fuse is another big question mark. I'll have small currents (or hope to have) so I probably first need to see the max current the sensor will draw from the regulator.

You are doing very well with your understanding and with my posts which are a bit short on details.

Your calculation of resonant frequency is correct. At the resonant frequency the noise voltage is reduced by 1/sqrt(2) which is a factor of 0.7 or so. For every decade increase in frequency the attenuation will be increased by 40dB which is a voltage attenuation factor of 0.01. so at 3.4KHz any noise voltage is reduced to 1% of its original value (incident on the input to the filter) and at 34kHz it will be 1% of 1% or a factor of 0.0001, and so on. Unless the perturbation originates in the alternator, it is very unlikely that a voltage spike with a bandwidth below a few kHz can be propagated on the automotive power supply as other loads and the battery should be able to absorb the energy of the spike.

The 0.2 Ohm you quote is the filter impedance and an indicator of the Q of the filter. Unless you get an inductor that is exceptional (and very expensive) the real Q of this filter will be nowhere near that number and besides all the other reasons for the filter not suddenly producing huge voltages at its resonant frequency, the fact that it is a series resonant filter means that the current at resonance will be maximum which will cause the inductor to saturate and hence no second order resonance. So don't worry about resonance and high voltages generated in LC circuits as some have predicted.

I think you also have some advice on winding your own inductor. I do not recommend taking that course of action without a lot of further research. There are just so many details missing from those instructions I don't know where to start to correct any of it (and I have several years experience designing magnetic components). What I do suggest, is an inductor that can handle the inrush current without damage (does not need to maintain inductance, it just needs not be damaged by it) and handle the DC load current without losing too much inductance (they all drop some with current flowing). The resistance of the inductor will reduce the Q of the filter (good) and help limit the inrush current (also good). A toroidal type with a powdered iron core for higher load currents or a smaller, shrouded or unshrouded ferrite type otherwise, whatever is cheapest and available.

The filter I suggested should be a good and safe starting point until you can define what you are dealing with properly.

The order of the caps is only important in terms of physical placement. It is good practice to put the ceramic caps as close to the regulator as practicable and the electrolytics not too far away either. The result should be the ceramic caps have the shortest and least inductive connections to the regulator pins that you can manage but don't bust a gut doing it. It is important but that importance has been a bit overplayed on here in my opinion. The electro cap connections to the regulator pins are less critical than the ceramic cap connections. If you are developing a PCB for your application then connecting the caps to the pins with a ground plane for the 0V node and a 0.04" or wider trace for the input and output sides of the regulator should see you right.

The fuse, in this case needs to be a high enough value that the inrush current into the 2200uF cap does not pop the fuse and small enough (if possible) to pop before the TVS device pops in an over voltage situation. My suggestion is to start with a 1AT, which is a 1A fuse with a Time lag characteristic. If it does not handle the inrush and normal load currents ok increase the value until it does. Prototype testing is probably your best bet and the easiest way to define what fuse rating you need. If the TVS does not survive popping the fuse there are two possible solutions to that problem. You could include inrush current limiting which could be done fairly simply given you have an arduino in the box or isolate the TVS from the input cap with a diode and change the TVS to a thyristor type instead of a zener type. They come in the same packages but the thyristor type can handle 100's or 1000's of amps (peak) to the zener types 1's to 10's of amps. If you do use the thyristor type, I suggest a higher voltage rating than a zener type because if it fires it will pop the fuse, and without question will it do that.

I hope I have given you enough details there. Overvoltage protection and fusing is actually a big topic in itself and covers circuit details and parameters that are usually considered mundane if considered at all.

 

In a spark ignited engine there are some predictable frequencies as well as their harmonics. Those are the spark rate frequency , the alternator RPMx number of poles frequency, and the brush noise frequency of all the small motors. There is also the clock frequency of each of the many processor crystals, and those harmonics. In addition there is the resonant frequency of each wire that connects to something that sparks. The short summary is that there is a lot of electrical noise on the wiring in a car. So really the filter should extend down to a very low frequency, which implies a larger inductor. If you can get one from an older scrapped car radio that will be a great start. A henry or two and a low resistance, and sometimes a magnetized core that yields a higher effective inductance than the number of turns would suggest. Thus a much lower resonant frequency is better.

 

In a spark ignited engine there are some predictable frequencies as well as their harmonics. Those are the spark rate frequency , the alternator RPMx number of poles frequency, and the brush noise frequency of all the small motors. There is also the clock frequency of each of the many processor crystals, and those harmonics. In addition there is the resonant frequency of each wire that connects to something that sparks. The short summary is that there is a lot of electrical noise on the wiring in a car. So really the filter should extend down to a very low frequency, which implies a larger inductor. If you can get one from an older scrapped car radio that will be a great start. A henry or two and a low resistance, and sometimes a magnetized core that yields a higher effective inductance than the number of turns would suggest. Thus a much lower resonant frequency is better.

I am sorry but I am going to have to pull you up here MrBill. While some of what you say is pertinent and more or less correct a good deal of it is in fact wrong. I can no longer ignore it as I cannot ignore misinformation, myth and legend presented as technical facts.

It is true that there is electrical noise generated at a frequency given by the rpm of the motor. It is not true that the spectrum of that noise is only the frequency given by the rmp of the motor. the noise is largely higher frequency noise generated in bursts at a rate given by the rmp of the motor. So the fundamental frequency may be low, but the energy in the fundamental is also very low. The high frequency energy is the problem area.

A 1 or 2 Henry inductor to carry even a small DC current without saturation and to be effective at higher frequencies is a monster of a choke. the magnetic bias you refer to is used to push the operating point along the BH curve in the opposite direction to that which the DC current will do and thereby extend the DC current the choke can handle without driving the core into magnetic saturation.

Your statement about the number of turns in this and other posts indicates to me that you know very little about magnetics at all and it would be a better thing if you were to avoid presenting yourself as an expert in this way and on this topic. I have several years experience designing magnetic components so believe me I do know what I am talking about here and I do not take any pleasure from this post but I can not ignore your misguidance to this OP and followers of this thread. If I could have said this to you in a PM I would have but the damage is now public so a PM is not an option.

I am sorry and I apologise for doing this. I derive no pleasure from it. You are an enthusiastic poster, I can see that and I applaud it, but go easy with the presentation of facts please.

Now you've made me sound like a pompous git!

 

Unfortunately (for me) the sensor I have to work with is an automotive one. There is no other/better option available.

When the system is fired up, the calculated current is 400mA and drop to 85mA when it reaches the working point (1bar, sensor becomes a 47ohm resistor).

If I'm getting this right, in normal working conditions, the voltage regulator will have to drop 13.4V (alternator on) to 8V at 400mA.
This give me an incredible 2.16W to heat things up.

That would be 14.4V-8=6.8V
6.8V x 0.4A = 2.72W per hour losses.

If the sensor is like a resistor, then you need a different set up. You are saying that the sensor is powered by 8V, but can go up to 24V? Is the 6V to 24V the power supply or the output from the sensor? Is this a voltage controlled or current controlled sensor?

 

I guess someone already asked this, but it took a lot of time to read so many posts.

 

I am sorry but I am going to have to pull you up here MrBill. While some of what you say is pertinent and more or less correct a good deal of it is in fact wrong. I can no longer ignore it as I cannot ignore misinformation, myth and legend presented as technical facts.

It is true that there is electrical noise generated at a frequency given by the rpm of the motor. It is not true that the spectrum of that noise is only the frequency given by the rpm of the motor. the noise is largely higher frequency noise generated in bursts at a rate given by the rpm of the motor. So the fundamental frequency may be low, but the energy in the fundamental is also very low. The high frequency energy is the problem area.

A 1 or 2 Henry inductor to carry even a small DC current without saturation and to be effective at higher frequencies is a monster of a choke. the magnetic bias you refer to is used to push the operating point along the BH curve in the opposite direction to that which the DC current will do and thereby extend the DC current the choke can handle without driving the core into magnetic saturation.

Your statement about the number of turns in this and other posts indicates to me that you know very little about magnetics at all and it would be a better thing if you were to avoid presenting yourself as an expert in this way and on this topic. I have several years experience designing magnetic components so believe me I do know what I am talking about here and I do not take any pleasure from this post but I can not ignore your misguidance to this OP and followers of this thread. If I could have said this to you in a PM I would have but the damage is now public so a PM is not an option.

I am sorry and I apologise for doing this. I derive no pleasure from it. You are an enthusiastic poster, I can see that and I applaud it, but go easy with the presentation of facts please.

Now you've made me sound like a pompous git!

My intention was to point out that these are some of the noise sources, and I thought that I made it clear that the harmonics were there as well. Brush noise certainly includes spark noise and it's spectrum, which I may not have adequately explained. My conclusion, I thought, was that the automotive electrical system is a very noisy environment. The inductor-choke filters that I suggested are mostly "large" by current standards, if you examine a mid-sixties car radio you will see that the hugeness is not really so big, except compared to current surface mount devices. And the current drawn by the system being discussed in this post is not that very great, and so the wire size for the choke would not be huge, except relatively.
It may be that a whole one henry filter choke is more inductance than required for the application, but it is a fair point to start an analysis as to what compromises may be needed. But a choke of a few microhenries is probably not going to block much noise in the lower frequency spectrum.