Will this work (optocoupler over shunt)?

[stenespen]

See attachment.

Trying to detect broken lamps in a simple circuit.

Will this work?

[Benta]

No.

Lots of things wrong:
CNY17 LED voltage drop
CNY17 CTR
CNY17 max.output current
...

[stenespen]

Which optocoupler should I choose?

[HB9EVI]

no optocoupler will do that; they use infrared led, which actually have quite a low forward voltage, but not low enough for a shunt drop voltage, additionally it will need a considerable amount of current to drive the led, so no way.

the common way is to amplify a shunt voltage to a level easier to operate with.

the maximal current of the phototransistor also has to be taken into consideration; the magic word also here: amplifying

[jmelson]

Yes, I have used this to detect overcurrent in a servo amplifier.

Jon

[HB9EVI]

well, over 2V drop can drive the IR, but isn't the whole circuit silly:

- burning almost 4,5W over shunt, what is pretty much 1/5 of the loads power rating
- loose over 2V to drive the bulb which are missing to drive the load

who would do that in real world?

[Zero999]

The trouble is you'l lose >1.2V tuning on that opto-coupler.

Does it need to be isolated? You already have a relay so there seems little point in having an opto-coupler.

Here's a crude discrete circuit, similar to one I've used before.

The threshold current is about an Amp and the voltage across the sense resistor is clamped by D1, thus limiting the voltage loss, when the current greatly exceeds the threshold. If the current will never significantly exceed the threshold or the additional loss is acceptable, then it's fine to omit D1.

EDIT:
I've just realised the original poster's relay coil has a resistance of 100R, not 500R, in which case R2 should be reduced to 1k. It will also increase the sensitivity of the circuit, since the voltage drop on D2 will be a bit higher.

[spec]

Hi stenespen,

Afraid that your circuit has a few issues but, attached below, is a schematic showing a circuit that will do your job.

The two diodes are to catch the back EMF of the relay when it switches. I have just put two diodes in because I don't know what the favored snubber arrangements are on EEV- it's always a subject of great debate.

If you want to know anything about the circuit please ask me and don't be discouraged by any negative statements you may get from certain parties. That is not to say that I don't welcome genuine critics who point out real problems with any designs.

[T3sl4co1l]

Right, to get a smaller offset voltage you need some kind of gain; the best we can do for single devices is the antiquated germanium bipolar transistor, with a Vbe around 0.3V.  To do any better (with silicon or otherwise), we need more transistors/diodes, to adjust the voltage drop smaller.  Typically a differential amplifier is used, to remove the temperature and bias dependence that a Vbe drop has.

Physics presents us with other opportunities, which can be more generally useful.  If we apply the Hall effect, we can generate a sense voltage proportional to a magnetic field, in turn proportional to a current flow.  The current flow can be isolated, AC or DC, and with a little circuitry we can see what's going on.

Hall effect sensors are cheap and abundant; though "cheap" is relative (a few bucks).  You're likely better off with a resistor, reference and comparator, if they are suitable.

For AC only, we can use Faraday's law to induce a sense voltage, adding an arbitrarily small voltage drop to the sensed line.  In other words, a current transformer.  (Obviously, this doesn't work so well at DC, so doesn't apply to the present case.)

So, we must consider suitability.  Is power always available?  (Kind of a silly question maybe, but important when we're looking at whether the lamp circuit itself is live, versus just the lamp.)  Is it always in the same direction?  Do we have to worry about it being plugged in backwards or anything?  (If the sense were on the lamp itself, we might need protection to deal with the lamp being plugged in wrong.  Depending on type of course.)  Given answers to these and other questions, we can solve for a perfect solution -- if the requirements are very modest, then the above examples already given will be close already.   Likely any better-suited solution will still follow the same plan, just with different sorts of comparators, or output signaling, or signal filtering, or etc.  Simple stuff.

Tim

[StillTrying]

An infra-red photo transistor pointing at the bulb.

[Mark Hennessy]

This can be done with a few turns of wire around a reed switch. I like this idea because it minimises losses, is nice and simple, and reed switches are cheap.

About 25 years ago I found a module made by Audi that monitors brake lights using this technique - I probably still have it somewhere. The only added complexity was a couple of transistors to latch the faulty state on IIRC.

Depending on your application, you might need a transistor to invert the logic, but if you don't need that, you might not need any other parts. Here's a reed switch that is good for 500mA, so should drive the relay coil (or possibly even the load you want to switch with the relay?) directly:

https://uk.farnell.com/hamlin/mdsr-4-22-38/switch-reed-spst-no-0-5a-200v/dp/2103636

[spec]

While it is obviously good idea to reduce the senor voltage drop, the OP seemed quite happy with the voltage drop of 2V in his original circuit. Don't forget that the battery voltage varies from car to car and according to battery state of charge and temperature. There is another point too, if the bulb is incandescent the filament resistance has a strongly positive temperature coefficient and the absolute  cold resistance of the filament will have a wide spread from one bulb to another and one manufaturer to another. So it is very important to put all these theories in perspective, especially as the human perception of light is logarithmic. 

The -2.2mV deg C temp coefficient of a Si forward biased junction is not significant in this application where the difference being measured is 0A and 2A. The effect of this coefficient is often overstated: it only amounts to a change of voltage of -0.37% Deg C

The only thing the OP asked is will his circuit work, and afraid to say that the answer is no. But, if a low forward drop were required, it would be a simple design task to achieve that. One cheap opamp or comparator would do the job for next to nothing.

The current activated reed relay approach is clever, but it does introduce an electro-mechanical part, and a hall effect sensor would be another approach as already mentioned. The infrared detector is the Rolls Royce approach, because, rather than by proxy, it measures the actual parameter of interest, the light output.

[ArthurDent]

Here's a circuit that has been around for some time. I haven't tried it but it should work. The advantage of current sensing in the supply going to the bulb is that it can be done remotely whereas a light sensing circuit needs the sensor next to the bulb which may not be possible as in auto taillights.

If you can put a photo diode or photo resistor near the bulb the circuit becomes very simple.

http://www.discovercircuits.com/DJ-Circuits/bad-bulb.htm

[spec]

The advantage of current sensing in the supply going to the bulb is that it can be done remotely whereas a light sensing circuit needs the sensor next to the bulb which may not be possible as in auto taillights.

Good point.

[spec]

UPDATE #1  2018_12_06 (decoupling and filtering added)

Attached below is the schematic for a bulb current monitor with a voltage overhead of 50mV. It uses a colossally expensive precision dual comparator costing 33 pence.

The 25mR resistor could just be a four terminal resistor made from a piece of wire.

[ArthurDent]

spec - "Attached below is the schematic for a bulb current monitor with a voltage overhead of 50mV."

Even less voltage drop than the previous circuits but is the optocoupler and the second 12VDC supply necessary? Seems like the two halves could be combined with the comparator driving Q4

[spec]

spec - "Attached below is the schematic for a bulb current monitor with a voltage overhead of 50mV."

Even less voltage drop than the previous circuits but is the optocoupler and the second 12VDC supply necessary? Seems like the two halves could be combined with the comparator driving Q4

Of course, but that is not what the OP has asked for. We can only assume that the OP has a reason for having two isolated supplies. The resolution would be to ask him.

Another aspect of simplifying the design is to ask what the relay is for. Maybe that could be eliminated too.

[spec]

spec - "Attached below is the schematic for a bulb current monitor with a voltage overhead of 50mV."

Even less voltage drop than the previous circuits but is the optocoupler and the second 12VDC supply necessary? Seems like the two halves could be combined with the comparator driving Q4

AD, is this what you had in mind >>

[ArthurDent]

spec - "AD, is this what you had in mind >>"

That is what I had in mind and I couldn't see why there would need to be 2 supplies. The only question I have with the circuit is the input connections on the unused comparator. I think connecting the unused inputs directly to ground is recommended as in section 8.1 of this application note.

http://www.ti.com/lit/ds/symlink/lm393-n.pdf

[spec]

spec - "AD, is this what you had in mind >>"

That is what I had in mind and I couldn't see why there would need to be 2 supplies. The only question I have with the circuit is the input connections on the unused comparator. I think connecting the unused inputs directly to ground is recommended as in section 8.1 of this application note.

http://www.ti.com/lit/ds/symlink/lm393-n.pdf

Well well well- I have been doing it wrong for all these years! Thanks for the info, but I am not too happy with tying both unused inputs to 0V because, it seems to me, that then the comparator would be in an undefined state and likely to oscillate (the inputs are still active at 0V input).

UPDATE #1 2018_12_06 I just sent a report to TI about the recommendation to tie both inputs of unused comparators to 0V- be interesting to see there response, if it's from and applications engineer that is. 

[Zero999]

spec - "AD, is this what you had in mind >>"

That is what I had in mind and I couldn't see why there would need to be 2 supplies. The only question I have with the circuit is the input connections on the unused comparator. I think connecting the unused inputs directly to ground is recommended as in section 8.1 of this application note.

http://www.ti.com/lit/ds/symlink/lm393-n.pdf

Well well well- I have been doing it wrong for all these years! Thanks for the info, but I am not too happy with tying both unused inputs to 0V because, it seems to me, that then the comparator would be in an undefined state and likely to oscillate (the inputs are still active at 0V input).

Some hysteresis would also be a good idea.

[spec]

I think you just made an oscillator 

There is no real need for hysteresis in this application, because the inputs are never close to one another.

[Zero999]

I think you just made an oscillator 

How? Please take another look.

There is no real need for hysteresis in this application, because the inputs are never close to one another.

The inputs do get quite close, about 12mV between them, when the bulb is working and C5 makes it worse, as it increases the length of time when the comparator will be in its linear region.

I do agree, hysteresis not essential, as the relay has some hysteresis built-in, there's no hysteresis on any of the discrete designs posted here, but the relay will last longer with it and it's only two extra resistors, so why not?

[ArthurDent]

Hero999 - "Some hysteresis would also be a good idea."

Although it probably isn't necessary, if I were adding hysteresis I would put a 1M resistor between pin1 (output of LM393) and pin3 (+ input) instead of the way the schematic shows it between pin2 (- input) and the collector of Q2, which is the noisiest point in the circuit. I also might go with .1uF for C5 and C6 with C6 moved to be across R15..

[Zero999]

Hero999 - "Some hysteresis would also be a good idea."

Although it probably isn't necessary, if I were adding hysteresis I would put a 1M resistor between pin1 (output of LM393) and pin3 (+ input) instead of the way the schematic shows it between pin2 (- input) and the collector of Q2, which is the noisiest point in the circuit. I also might go with .1uF for C5 and C6 with C6 moved to be across R15..

Yes, that was my first instinct, but then I noticed Q2's base clamps the comparatot's output voltage to around 700mV, which would make it less effective, so I opted for the collector and pin 2 instead.

How is Q2's collector noisy? It just goes to the relay coil? There's no active noise source.

[ArthurDent]

I would, as I said, follow the datasheet/application sheet information and, if included, would put the hysteresis resistor between pin1 (output of LM393) and pin3 (+ input) and ground the unused comparator inputs. Next I would replace the BC337 with a MOSFET and eliminate your concern of the base-emitter clamping although a resistor in series with the base would work as well.

If the collector of Q2 was connected to a resistive load I wouldn't be concerned but the relay more or less requires the added 2 diodes to eat up spikes (sorry for the technical jargon  ). Granted the circuit will rarely be triggered judging by the life of most lamps but I think using an inductor for the feedback point isn't wise.

We are talking in somewhat general terms here and if an actual circuit were to be built and tested I'm sure we'd find a number of other mods that would make sense.   

[Zero999]

I would, as I said, follow the datasheet/application sheet information and, if included, would put the hysteresis resistor between pin1 (output of LM393) and pin3 (+ input) and ground the unused comparator inputs. Next I would replace the BC337 with a MOSFET and eliminate your concern of the base-emitter clamping although a resistor in series with the base would work as well.

If the collector of Q2 was connected to a resistive load I wouldn't be concerned but the relay more or less requires the added 2 diodes to eat up spikes (sorry for the technical jargon  ). Granted the circuit will rarely be triggered judging by the life of most lamps but I think using an inductor for the feedback point isn't wise.

We are talking in somewhat general terms here and if an actual circuit were to be built and tested I'm sure we'd find a number of other mods that would make sense.   

Ideally I'd try to find a relay with a lower current draw <50mA, replace the LM393 with the LM311 and drive the relay directly, but a MOSFET is a good suggestion.

How is taking the feedback from the inductive relay coil, rather than a resistor a problem? When Q2 turns on, the collector voltage instantly falls near 0V and when it turns off, the collector voltage rises to a diode drop above the power supply voltage, until the flux in the inductor is depleted. Also only one diode is required. D7 is superfluous and will never conduct.

[ArthurDent]

Here is my final variation of the detector that may address some concerns and, like I said, if any of these various circuits were actually built and tested there could be minor or major changes that might come to mind. I think stenespen now has enough information so they can pick and choose what they think might be best for them and try that circuit. 

[Zero999]

Here is my final variation of the detector that may address some concerns and, like I said, if any of these various circuits were actually built and tested there could be minor or major changes that might come to mind. I think stenespen now has enough information so they can pick and choose what they think might be best for them and try that circuit.

That will not overcome the issue of the transistor clipping the comparator's output.

If you want to use the comparator's non-inverting input for positive feedback and don't want to use a MOSFET, then use a PNP BJT, but the logic will be reversed.

You seem not to believe me when I said the relay coil's inductance doesn't generate any noise. Where did you get that idea from? It's compete nonsense!

Here's a simulation showing an inductor being pulsed on and off with a transistor. Note that the waveform on the collector is a fairly decent squarewave. The overshoot beyond the supply voltage, when it turns off, is no problem.

Still don't believe me? Try building the circuit.

[spec]

Below is the schematic for a precision version, with hysteresis for H99, and an  NMOSFET for AD.

The trouble is that it uses non jelly-bean components, which may not suit the OP.

[spec]

Here is my final variation of the detector that may address some concerns and, like I said, if any of these various circuits were actually built and tested there could be minor or major changes that might come to mind. I think stenespen now has enough information so they can pick and choose what they think might be best for them and try that circuit.

Living dangerously there AD.

[StillTrying]

Isn't anyone going to do an Arduino version.

[Zero999]

Below is the schematic for a precision version, with hysteresis for H99, and an  NMOSFET for AD.

The trouble is that it uses non jelly-bean components, which may not suit the OP.

Here's a version which uses jell bean components. The turn on and turn off currents can be independently set if necessary. V2 can be the TL431 or similar voltage reference. It uses a PNP output but it can easily be modified to work with an NPN BJT, N-MOSFET or P-MOSFET.



It's a variant of a circuit I posted earlier this year, which is linked below.
https://www.eevblog.com/forum/beginners/the-simplest-thermostat-with-programmable-thresholds/msg1432025/#msg1432025

EDIT: I haven't included any filtering, which would be a good idea.

[spec]

Here is my final variation of the detector that may address some concerns and, like I said, if any of these various circuits were actually built and tested there could be minor or major changes that might come to mind. I think stenespen now has enough information so they can pick and choose what they think might be best for them and try that circuit.

That will not overcome the issue of the transistor clipping the comparator's output.

If you want to use the comparator's non-inverting input for positive feedback and don't want to use a MOSFET, then use a PNP BJT, but the logic will be reversed.

You seem not to believe me when I said the relay coil's inductance doesn't generate any noise. Where did you get that idea from? It's compete nonsense!

Q2 e/b junction is vulnerable here, as is the open collector transistor in the comparator. Also, you would be losing more drive to the relay than with a saturated collector arrangement, like the original.

Attached is a schematic showing  a simple and safe way to maximize drive to the relay and provide hysteresis (functionally, the TS391 comparator is just a singe comparator version of the original dual twin comparator LM393):

[T3sl4co1l]

Vulnerable to what?

I wouldn't recommend running the relay coil wires to the outside world where ESD and maybe surge may be a problem.

Tim

[T3sl4co1l]

Attached is a schematic showing  a simple and safe way to maximize drive to the relay and provide hysteresis (functionally, the TS391 comparator is just a singe comparator version of the original dual twin comparator LM393):

What's R22-C4 for?
Q2 doesn't turn off, needs B-E resistor.
What's D6 for?
D7 is oversized, can be a 1N914 or less really.  I never got the fascination with using huge rectifiers on tiny relays... kids probably read it first in Forrest Mims' or something and kept doing it the rest of their lives?

Tim

[spec]

I thought somehow that you would pop out of the woodwork

Vulnerable to what?

If you need to ask- you would not understand the reply

I wouldn't recommend running the relay coil wires to the outside world where ESD and maybe surge may be a problem.

Hell- I would never have thought of that!

[spec]

Back again! Why don't you post a circuit for a bulb monitor yourself instead of nit picking my circuit- Oh yes, I remember now, you did once post a circuit after a lot of pushing and we saw the results of all your theories. It is a great shame that you do not apply your high standards to your own work. Tell me what are you?  Are you a design engineer who has actually designed products for the market, or are you a perhaps a consultant of some kind who gives advice and has no responsibility for the end product. I don't like your aggressive tone either- there are ways of asking questions politely.

What's R22-C4 for?

I don't believe that anyone who knows the first thing about electronic design would  ask that question!   

Q2 doesn't turn off, needs B-E resistor.

You are nit picking. It is the architecture that is important. But as a matter of interest, show me your calculations of how much current Q2 would conduct... 1uA, 1mA, ... 100A?

What's D6 for?
D7 is oversized, can be a 1N914 or less really.

So what. It is not up to you to define what components I can specify. Besides I have had this argument before from theoreticians who like to display there superior knowledge. Those of us that design products that go into the market know that it is ridiculous to design right on the edge. I have seen quite a few 1N4148/1N916 fail when used for relay snubbing and inductance catching. The 1N4148 is designed for fast low level signals. Using your criteria practically every component in circuits are overkill (a much abused word anyway).

The other thing is that the 1N400x physical construction is more robust and less likely to fracture when fitted to the pins of relays, motors, etc. By contrast the IN4148 is made of brittle and flimsy glass and with thin weedy self- leads the 1N4148 is quite simply not suited, or intended for the job of snubbing/catching. You need to consider the practical as well as the theoretical aspects of design.

And finally, what is the downside of fitting a 1N400x diode... none. So what you are saying is just a load of hot air.   

I never got the fascination with using huge rectifiers on tiny relays... kids probably read it first in Forrest Mims' or something and kept doing it the rest of their lives?

What a patronizing fellow you are- I recon you still read Forrest Mims, whatever that is.

I don't know what your problem is, but on a general note, can you please stop following me around EEV, which you have been doing since I recently started posting again. You, and a couple of others, are making a nuisance of themselves and wasting my time- like now.

What I can't understand is that there are hundreds of circuits posted on EEC, some with obvious problems, where you could display your wide knowledge,  but you say not a word. 

[T3sl4co1l]

Back again! Why don't you post a circuit for a bulb monitor instead of nit picking my circuit- Oh yes, I remember now, you did once and we saw the results of all your theories.

I still don't get your thing about circuits...

You are nit picking. It is the architecture that is important. But as a matter of interest show me your calculations of how much current Q2 would conduct... 1uA, 1mA, ... 100A?

If architecture is important, the OP opened with it; indeed, I corroborated that from an analytical perspective.  Just a matter of implementation then.  Which means... yes, arguing about component choice as well as value.

It's about 1mA by the way, and no, it's very unlikely to cause a problem; but it's a poor design pattern that's very easily corrected, so it's always near the top of the list.

(One could also save even more with a "digital"/"prebiased" transistor, but the OP probably isn't making a million of these so normal discretes will be fine of course.)

So what. It is not up to you to define what components I can specify.

Did I?

I'm hardly writing a dictionary over here... (cough, though I've probably posted a few of them's worth here over the years..)

Besides I have had this argument before from theoreticians who like to display there superior knowledge.

Is that what you think I am, an armchair troll?

(Do you also make a habit of ignoring anyone who is right and outspoken, because clearly they must be an egoist?...)

Those of us that design products that go into the market know that it is ridiculous to design right on the edge. I have seen quite a few 1N4148/1N916 fail when used for relay snubbing and inductance catching. The 1N4148 is designed for fast low level signals. Using your criteria practically every component in circuits are overkill.

Hm, did I define a criteria?

I like the implication that I haven't made or sold anything, though.

I don't know what your problem is, but on a general note, can you please stop following me around EEV, which you have been doing since I recently started posting again. You, and a couple of others, are making a nuisance of themselves and wasting my time- like now.

Coincidence?  I don't care to read or reply to every single thread, just those that look of interest of course.

Indeed, when I first noticed some of your posts, I thought it very interesting -- new design blood is an uncommon sight here.  It's probably more than coincidence, if we're drawn to the same sorts of threads more than chance.

I'm sure this is my fault -- I enjoy a good rousing critique myself, and easily forget how sensitive others can be about their work.  I also tend not to emote very well in written form.  If it helps, try to read my posts with a disinterested voice.  I generally want to be helpful, or try to be; eh, not that that necessarily means anything (I can think of more than a few bad people who claimed the same thing.. the context here is hopefully very different, at least).

What I can't understand is that there are hundreds of circuits posted on EEC, some with obvious problems, where you could display your wide knowledge,  but you say not a word. 

I probably tend to avoid threads with problems so obvious that the subject gives it away.  Hmm, clearest example that comes to mind: an ESL* struggling with a technical subject AND a language barrier.  I just can't be of much help when it's apparent that I won't be understood. Fortunately, the demographics here are quite diverse, and someone with better knowledge of the language barrier will be of greater help than I can.

*English Second Language.

Or, there's only so much talking I can do to a brick wall.  That 30V 3A bench supply megathread is a classic example...

Cheers,
Tim

[Zero999]

Here is my final variation of the detector that may address some concerns and, like I said, if any of these various circuits were actually built and tested there could be minor or major changes that might come to mind. I think stenespen now has enough information so they can pick and choose what they think might be best for them and try that circuit.

That will not overcome the issue of the transistor clipping the comparator's output.

If you want to use the comparator's non-inverting input for positive feedback and don't want to use a MOSFET, then use a PNP BJT, but the logic will be reversed.

You seem not to believe me when I said the relay coil's inductance doesn't generate any noise. Where did you get that idea from? It's compete nonsense!

Q2 e/b junction is vulnerable here, as is the open collector transistor in the comparator. Also, you would be losing more drive to the relay than with a saturated collector arrangement, like the original.

I thought somehow that you would pop out of the woodwork

Vulnerable to what?

If you need to ask- you would not understand the reply

Someone else might. I was going to ask the same question: what are the base-emitter junction of Q2 and the internal transistor inside the comparator vulnerable to? Please explain what you meant.

Back again! Why don't you post a circuit for a bulb monitor yourself instead of nit picking my circuit- Oh yes, I remember now, you did once post a circuit after a lot of pushing and we saw the results of all your theories. It is a great shame that you do not apply your high standards to your own work. Tell me what are you?  Are you a design engineer who has actually designed products for the market, or are you a perhaps a consultant of some kind who gives advice and has no responsibility for the end product. I don't like your aggressive tone either- there are ways of asking questions politely.

What's R22-C4 for?

I don't believe that anyone who knows the first thing about electronic design would  ask that question!   

Q2 doesn't turn off, needs B-E resistor.

You are nit picking. It is the architecture that is important. But as a matter of interest, show me your calculations of how much current Q2 would conduct... 1uA, 1mA, ... 100A?

What's D6 for?
D7 is oversized, can be a 1N914 or less really.

So what. It is not up to you to define what components I can specify. Besides I have had this argument before from theoreticians who like to display there superior knowledge. Those of us that design products that go into the market know that it is ridiculous to design right on the edge. I have seen quite a few 1N4148/1N916 fail when used for relay snubbing and inductance catching. The 1N4148 is designed for fast low level signals. Using your criteria practically every component in circuits are overkill.
The other thing is that the 1N400x physical construction is more robust and less likely to fracture when fitted to the pins of relays, motorrs, etc. By contrast the IN418 is made of brittle and flimsy glass and with thin weedy self- leads is quite simply not suited, or intended for the job.   

I never got the fascination with using huge rectifiers on tiny relays... kids probably read it first in Forrest Mims' or something and kept doing it the rest of their lives?

What a patronizing fellow you are- I recon you still read Forrest Mims, whatever that is.

I don't know what your problem is, but on a general note, can you please stop following me around EEV, which you have been doing since I recently started posting again. You, and a couple of others, are making a nuisance of themselves and wasting my time- like now.

What I can't understand is that there are hundreds of circuits posted on EEC, some with obvious problems, where you could display your wide knowledge,  but you say not a word. 

• Don't take it personally. Can't handle people nitpicking your circuits? Don't post them!
• Why not just answer the question, rather than making a sarcastic comment? It wouldn't have taken any longer. I believe R22 & C22 are for filtering the supply to the comparator. They're not needed in this, as it's not a linear amplifier which would be susceptible to a noisy supply, hence the question.
• He's right. Q2 won't turn off. The collector current will be dependant on the Hfe. R27 dominates and the Hfe varies widely so there's no point in including the other ones in the calculation. I_B = 12/560k = 21µA. Assuming Hfe = 400, I_C = 8.6mA, which isn't much, but it will cause Q2 to heat more and the Hfe to rise. It shouldn't cause thermal runaway because the Hfe will stop increasing at safe current, but it's still bad design.
• The freewheeling diode only needs to be rated to carry the coil current. There's nothing wrong with overrating but don't be surprised when someone questions it.

You've also not said what D6 is supposed to do? It will never be forward biased!


I repeat, when posting a circuit, expect questions and criticism: both valid and invalid. I've learned when I saw someone's circuit which didn't look right, asked questions and criticised it, only to discover I was wrong, but it helped me so I don't regret it.

[MK14]

The idea of this circuit, is to be SIMPLE and made out of jelly bean parts. It is partly based on the OP's original circuit.
It does have approximately 1.5 to 2 Volts drop across the diodes, but on the other hand, simplifies things for beginners.
Edit: Updated resistor values. Feel free to adjust resistor values as necessary.

[spec]

Back again! Why don't you post a circuit for a bulb monitor instead of nit picking my circuit- Oh yes, I remember now, you did once and we saw the results of all your theories.

I still don't get your thing about circuits...

You are nit picking. It is the architecture that is important. But as a matter of interest show me your calculations of how much current Q2 would conduct... 1uA, 1mA, ... 100A?

If architecture is important, the OP opened with it; indeed, I corroborated that from an analytical perspective.  Just a matter of implementation then.  Which means... yes, arguing about component choice as well as value.

It's about 1mA by the way, and no, it's very unlikely to cause a problem; but it's a poor design pattern that's very easily corrected, so it's always near the top of the list.

(One could also save even more with a "digital"/"prebiased" transistor, but the OP probably isn't making a million of these so normal discretes will be fine of course.)

So what. It is not up to you to define what components I can specify.

Did I?

I'm hardly writing a dictionary over here... (cough, though I've probably posted a few of them's worth here over the years..)

Besides I have had this argument before from theoreticians who like to display there superior knowledge.

Is that what you think I am, an armchair troll?

(Do you also make a habit of ignoring anyone who is right and outspoken, because clearly they must be an egoist?...)

Those of us that design products that go into the market know that it is ridiculous to design right on the edge. I have seen quite a few 1N4148/1N916 fail when used for relay snubbing and inductance catching. The 1N4148 is designed for fast low level signals. Using your criteria practically every component in circuits are overkill.

Hm, did I define a criteria?

I like the implication that I haven't made or sold anything, though.

I don't know what your problem is, but on a general note, can you please stop following me around EEV, which you have been doing since I recently started posting again. You, and a couple of others, are making a nuisance of themselves and wasting my time- like now.

Coincidence?  I don't care to read or reply to every single thread, just those that look of interest of course.

Indeed, when I first noticed some of your posts, I thought it very interesting -- new design blood is an uncommon sight here.  It's probably more than coincidence, if we're drawn to the same sorts of threads more than chance.

I'm sure this is my fault -- I enjoy a good rousing critique myself, and easily forget how sensitive others can be about their work.  I also tend not to emote very well in written form.  If it helps, try to read my posts with a disinterested voice.  I generally want to be helpful, or try to be; eh, not that that necessarily means anything (I can think of more than a few bad people who claimed the same thing.. the context here is hopefully very different, at least).

What I can't understand is that there are hundreds of circuits posted on EEC, some with obvious problems, where you could display your wide knowledge,  but you say not a word. 

I probably tend to avoid threads with problems so obvious that the subject gives it away.  Hmm, clearest example that comes to mind: an ESL* struggling with a technical subject AND a language barrier.  I just can't be of much help when it's apparent that I won't be understood. Fortunately, the demographics here are quite diverse, and someone with better knowledge of the language barrier will be of greater help than I can.

*English Second Language.

Or, there's only so much talking I can do to a brick wall.  That 30V 3A bench supply megathread is a classic example...

Cheers,
Tim

If I have misunderstood you, please accept my apologies. And I had no idea that English is your second language. Maybe that is partly the problem. But do try to see it from my point of view too. You and another regularly contradict/give the impression of contradicting what I say. This has two bad effects. The OP doubts what I say and it takes me ages to reply to all the adverse comments. Quite often I have had to just leave a thread because of this. And the annoying thing is that people will make an adverse remark and when you prove that are wrong, they do not apologize or even discuss it.

I hate to say this but in your post, there was only one thing that was of any significance, that is about the base resistor. You are absolutely right, but do you think that I would not have known that. And to be specific the current that could flow through the transistor (worst case) will have no effect on the operation of the circuit.

You could have just said: 'Just in case, a resistor across the e/b transistor would be good design practice'.

I am not getting at you now, but while all these obvious points are being aired, the real problems are being missed. This has happened on a number of occasions with my circuits. By the way, errors are all part of design.

So sorry Tim, I have obviously misread your intentions.

Having said that, I genuinely appreciate it when someone points out an error of significance or gives new information.

[spec]

Here is my final variation of the detector that may address some concerns and, like I said, if any of these various circuits were actually built and tested there could be minor or major changes that might come to mind. I think stenespen now has enough information so they can pick and choose what they think might be best for them and try that circuit.

That will not overcome the issue of the transistor clipping the comparator's output.

If you want to use the comparator's non-inverting input for positive feedback and don't want to use a MOSFET, then use a PNP BJT, but the logic will be reversed.

You seem not to believe me when I said the relay coil's inductance doesn't generate any noise. Where did you get that idea from? It's compete nonsense!

Q2 e/b junction is vulnerable here, as is the open collector transistor in the comparator. Also, you would be losing more drive to the relay than with a saturated collector arrangement, like the original.

I thought somehow that you would pop out of the woodwork

Vulnerable to what?

If you need to ask- you would not understand the reply

Someone else might. I was going to ask the same question: what are the base-emitter junction of Q2 and the internal transistor inside the comparator vulnerable to? Please explain what you meant.

Back again! Why don't you post a circuit for a bulb monitor yourself instead of nit picking my circuit- Oh yes, I remember now, you did once post a circuit after a lot of pushing and we saw the results of all your theories. It is a great shame that you do not apply your high standards to your own work. Tell me what are you?  Are you a design engineer who has actually designed products for the market, or are you a perhaps a consultant of some kind who gives advice and has no responsibility for the end product. I don't like your aggressive tone either- there are ways of asking questions politely.

What's R22-C4 for?

I don't believe that anyone who knows the first thing about electronic design would  ask that question!   

Q2 doesn't turn off, needs B-E resistor.

You are nit picking. It is the architecture that is important. But as a matter of interest, show me your calculations of how much current Q2 would conduct... 1uA, 1mA, ... 100A?

What's D6 for?
D7 is oversized, can be a 1N914 or less really.

So what. It is not up to you to define what components I can specify. Besides I have had this argument before from theoreticians who like to display there superior knowledge. Those of us that design products that go into the market know that it is ridiculous to design right on the edge. I have seen quite a few 1N4148/1N916 fail when used for relay snubbing and inductance catching. The 1N4148 is designed for fast low level signals. Using your criteria practically every component in circuits are overkill.
The other thing is that the 1N400x physical construction is more robust and less likely to fracture when fitted to the pins of relays, motorrs, etc. By contrast the IN418 is made of brittle and flimsy glass and with thin weedy self- leads is quite simply not suited, or intended for the job.   

I never got the fascination with using huge rectifiers on tiny relays... kids probably read it first in Forrest Mims' or something and kept doing it the rest of their lives?

What a patronizing fellow you are- I recon you still read Forrest Mims, whatever that is.

I don't know what your problem is, but on a general note, can you please stop following me around EEV, which you have been doing since I recently started posting again. You, and a couple of others, are making a nuisance of themselves and wasting my time- like now.

What I can't understand is that there are hundreds of circuits posted on EEC, some with obvious problems, where you could display your wide knowledge,  but you say not a word. 

• Don't take it personally. Can't handle people nitpicking your circuits? Don't post them!
• Why not just answer the question, rather than making a sarcastic comment? It wouldn't have taken any longer. I believe R22 & C22 are for filtering the supply to the comparator. They're not needed in this, as it's not a linear amplifier which would be susceptible to a noisy supply, hence the question.
• He's right. Q2 won't turn off. The collector current will be dependant on the Hfe. R27 dominates and the Hfe varies widely so there's no point in including it in the calculation. I_B = 12/560k = 21µA. Assuming Hfe = 400, I_C = 8.6mA, which isn't much, but it will cause Q2 to heat more and the Hfe to rise. It shouldn't cause thermal runaway because the Hfe will stop increasing at safe current, but it's still bad design.
• The freewheeling diode only needs to be rated to carry the coil current. There's nothing wrong with overrating but don't be surprised when someone questions it.

You've also not said what D6 is supposed to do? It will never be forward biased!


I repeat, when posting a circuit, expect questions and criticism: both valid and invalid. I've learned when I saw someone's circuit which didn't look right, asked questions and criticised it, only to discover I was wrong, but it helped me so I don't regret it.

I wondered how long it would be before you popped out of the woodwork too. So here we go again. Your rhetorical question shows that you have not read or understood a word of my post. And who are you to tell me what and when to post. You presume too much.

As for answering the questions directly. I gave answers. There was nothing sarcastic at all. I did mean that if I needed to explain some of those questions that the answer would not be understood. Surely you can see that as the questions were so elementary.

But any way I have replied to the third party's post, now that he has explained his position.

[spec]

• Don't take it personally. Can't handle people nitpicking your circuits? Don't post them!
  
  
• Why not just answer the question, rather than making a sarcastic comment? It wouldn't have taken any longer. I believe R22 & C22 are for filtering the supply to the comparator. They're not needed in this, as it's not a linear amplifier which would be susceptible to a noisy supply, hence the question.
• He's right. Q2 won't turn off. The collector current will be dependant on the Hfe. R27 dominates and the Hfe varies widely so there's no point in including it in the calculation. I_B = 12/560k = 21µA. Assuming Hfe = 400, I_C = 8.6mA, which isn't much, but it will cause Q2 to heat more and the Hfe to rise. It shouldn't cause thermal runaway because the Hfe will stop increasing at safe current, but it's still bad design.
• The freewheeling diode only needs to be rated to carry the coil current. There's nothing wrong with overrating but don't be surprised when someone questions it.

You've also not said what D6 is supposed to do? It will never be forward biased!

Take a holiday - you just never let go do you. 

[spec]

About the base emitter vulnerability. That is a surprising question but if I gave an answer, like I did about the diode, you would not accept it and there would be endless posts going round and round, just like with the diode.

Another point is that, although you are all over EEC and assume the position of resident expert on all matters, you are just a member like everyone else and it is not your place to tell me that I should expect this or that or that I am obliged to explain anything.This is an example of your circular arguments. Although if someone does post a genuine question politely I am only to pleased to help out- as you must have seen on many occasions.

[Zero999]

About the base emitter vulnerability. That is a surprising question but if I gave an answer, like I did about the diode, you would not accept it and there would be endless posts going round and round, just like with the diode.

I would not accept it, because you didn't provide any explanation as to why the base-emitter junction is vulnerable. I admit i missed the part about the higher on losses, which is true, an emitter follower will have higher on losses, than a common emitter configuration. I suppose one could say the power dissipation will be higher, thus making the transistor more vulnerable to overheating, but in this case the power dissipation is only 110mW, which is well within the the rating of the BC327.

Here is my final variation of the detector that may address some concerns and, like I said, if any of these various circuits were actually built and tested there could be minor or major changes that might come to mind. I think stenespen now has enough information so they can pick and choose what they think might be best for them and try that circuit.

That will not overcome the issue of the transistor clipping the comparator's output.

If you want to use the comparator's non-inverting input for positive feedback and don't want to use a MOSFET, then use a PNP BJT, but the logic will be reversed.

You seem not to believe me when I said the relay coil's inductance doesn't generate any noise. Where did you get that idea from? It's compete nonsense!

Q2 e/b junction is vulnerable here, as is the open collector transistor in the comparator. Also, you would be losing more drive to the relay than with a saturated collector arrangement, like the original.

Someone else asked the same question, but you didn't say why.

Vulnerable to what?

If you need to ask- you would not understand the reply

Another point is that, although you are all over EEC and assume to position of resident expert on all matters, you are just a member like everyone else and it is not your place to tell me that I should expect this or that or that I am obliged to explain anything.This is an example of your circular arguments. Although if someone posts a genuine question politely I am only to pleased to help out- as you must have seen on many occasions.

Although I post regularly, I'm no expert on all matters, for example I know very little about programming, so I leave that to others.

You're right that no one is obliged to explain anything, but people will not accept your answers to their questions, without a clear and logical explanation.

[spec]

Your reply above indicates that you havn't taken in a word that I have said.

Please keep out of my hair and I will do the same to you.
________________________________________________________________________________________________________

[Zero999]

The idea of this circuit, is to be SIMPLE and made out of jelly bean parts. It is partly based on the OP's original circuit.
It does have approximately 1.5 to 2 Volts drop across the diodes, but on the other hand, simplifies things for beginners.
Edit: Updated resistor values. Feel free to adjust resistor values as necessary.

Yes that'll work, but the logic is reversed. R3 and D3 could go in the emitter, to make the LED off when the lamp is blown and on, when it's working.

I agree, the circuits have become outlandishly complex and overkill, but that's feature creep and people having fun, posting different solutions.

[MK14]

Yes that'll work, but the logic is reversed. R3 and D3 could go in the emitter, to make the LED off when the lamp is blown and on, when it's working.

I agree, the circuits have become outlandishly complex and overkill, but that's feature creep and people having fun, posting different solutions.

I guess the OP, hasn't necessarily been 100% clear on, if they want it on or off, when faulty.

I based the logic, on a diagram, similar to the following one, shown below.
I.e. It ONLY lights up when the bulb is blown/faulty and needs replacing.

[T3sl4co1l]

If I have misunderstood you, please accept my apologies. And I had no idea that English is your second language.

Not me, I mean examples of threads by original posters (OPs) that are.

Maybe that is partly the problem. But do try to see it from my point of view too. You and another regularly contradict/give the impression of contradicting what I say.

It is easy to misread criticism as contradiction, but please do not jump into a defensive position when you see such a statement!

All that I have said about myself, probably applies in part to Hero999 as well.  He's been here longer than I have, and I don't recall any particular disagreements.  If I may be so bold: you seem to be the common denominator here, so please, take it easy!

I hate to say this but in your post, there was only one thing that was of any significance, that is about the base resistor. You are absolutely right, but do you think that I would not have known that. And to be specific the current that could flow through the transistor (worst case) will have no effect on the operation of the circuit.

You could have just said: 'Just in case, a resistor across the e/b transistor would be good design practice'.

Could.

I'm very thorough in my work, and I try to present that here in the hopes that others will pick up on it.  The design process can go infinitely deep, but you only need to go finitely far to get an arbitrarily good result.  That is, it is a process where successive approximation yields excellent results (as opposed to some processes that diverge -- sometimes this happens in SPICE for example, with subsequent hair-pulling..).

So, in that process, rather than leave it at the zeroth or first approximation, I do a quick second-order check to verify that the first degree is, in fact, the correct stopping point, and that no big surprises are likely to show up.

And so, in that process, I make observations, look for weird things.

If I see something that doesn't have an obvious purpose, I ask.  I mean, asking questions are how we find out about things, right?  Can I never ask you a legitimate question ever?  That's really unfair, man.

To paraphrase Freud, sometimes a question is just a question.

Sometimes a question reveals insight.  Is there actually no reason for R22-C4?  Might as well remove it, then.  Fewer parts in a discrete design, easier to assemble, less to go wrong.  That's valuable.  Or replace it with something better suited, like if inductive spikes and reverse protection are a concern, then a TVS diode would be even better.  (And if this is an automotive application, both are likely!).

And so our circuits can improve, and we learn more breadth of knowledge in the process.

(Professionally, a typical response to a question like this is: "it was copied from a previous design".  That's a possible warning sign -- many designs build up technical debt just as much as software or facilities do.  But it also at least strongly implies that it's passed functional and regulatory checks before, and is worth consideration.  Another typical response is: "yeah, we're not really sure there, we just pulled from the appnote", and useful improvements can be made -- and often prove necessary!)

Tim

[spec]

Your long reply above indicates that you haven't taken in a word that I have said.

Please keep out of my hair and I will do the same to you.

[Simon]

spec - "Attached below is the schematic for a bulb current monitor with a voltage overhead of 50mV."

Even less voltage drop than the previous circuits but is the optocoupler and the second 12VDC supply necessary? Seems like the two halves could be combined with the comparator driving Q4

Of course, but that is not what the OP has asked for. We can only assume that the OP has a reason for having two isolated supplies. The resolution would be to ask him.

Another aspect of simplifying the design is to ask what the relay is for. Maybe that could be eliminated too.

The OP has used a simulation program te try a concept out and has likely been a bit "lazy" with circuit diagram conventions. They should perhaps be specific as to whether or not an isolated supply is required as if this is for use an a car then it will be several quid for the likely unneeded DC/DC converter.

[Zero999]

spec - "Attached below is the schematic for a bulb current monitor with a voltage overhead of 50mV."

Even less voltage drop than the previous circuits but is the optocoupler and the second 12VDC supply necessary? Seems like the two halves could be combined with the comparator driving Q4

Of course, but that is not what the OP has asked for. We can only assume that the OP has a reason for having two isolated supplies. The resolution would be to ask him.

Another aspect of simplifying the design is to ask what the relay is for. Maybe that could be eliminated too.

The OP has used a simulation program te try a concept out and has likely been a bit "lazy" with circuit diagram conventions. They should perhaps be specific as to whether or not an isolated supply is required as if this is for use an a car then it will be several quid for the likely unneeded DC/DC converter.

Well they've used a relay, as well as an opto-coupler but only one is necessary for isolation. Both a relay and opto-coupler could be used together to provide the sum of withstand voltages of both parts, but this is only 12V and it's generally cheaper to use only one isolation component, with the appropriately rated isolation voltage.

[MK14]

The source of the original circuit design, seems to come from here (speculation, I could be wrong, but it seems amazingly similar):

https://www.electroschematics.com/11918/auto-12v-bulb-failure-warning/

The Opto-isolator, seems to have been put in (as regards the link, above), for future options, rather than because it is needed now.

I presume, the OP, based their circuit on the above linked circuit, and have added a relay and changed it, for reasons unknown. I can only guess at what they are trying to do.

For anyone who is curious, schematic copied here (from link, above):
Then below it, is what the final unit would look like: