Why not just use a cheap current clamp?

[Faringdon]

Hi,
Why ever would anyone use the opamp current clamp instead of the TLV431 based one? (LTspice's and PNGs as attached)
The TLV431 cct is so much cheaper and just as good.

(i mean, the TLV431 version looks so much cheaper and better that i wonder why anyone would use the opamp version? There is  the bias current in
the TLV431 which ends up in the clamped current too, but its just a few mA)

[Zero999]

The op-amp circuit has better current regulation because it doesn't have a resistor connected from the input to output via the TLV431. It's also possible to add a potential divider to the op-amp circuit to reduce the reference voltage and thus the voltage drop across the sense resistor, which would reduce the drop-out voltage.

[Faringdon]

Thanks,
So anyway, we need an opamp with common mode input range which includes VCC+.
Also, it must be able to take a 0-20V supply voltage.
Since it must perform the attached rough current clamp function.
However, the cheapest opamp that can do this is LT1498, which is $4.

Our customer cannot afford this, and they wish to use a cheap jelly bean opamp instead, the TL084CDT,
which is $0.3
Their circuit , which does the attached, works fine.
They wonder why they should pay $4 when their $0.3 circuit works just fine.
Their volumes are a few thousane a year or so.

LT1498 ($4)
https://www.analog.com/media/en/technical-documentation/data-sheets/14989fg.pdf

TL084CDT   ($0.3)
https://www.futureelectronics.com/p/semiconductors--analog--amplifiers--general-purpose/tl084cdt-stmicroelectronics-4593950?msclkid=0426691494681f3ba6230ca37a437b06&utm_source=bing&utm_medium=cpc&utm_campaign=A-MPN%20EMEA%20STMICROELECTRONICS&utm_term=TL084CDT&utm_content=Adg%20-%20TL084CDT

[Zero999]

It depends on the requirements.

The LT1468 has much better characteristics, than the TL084. The greatly lower offset voltage is the most relevant one here.

Why not use the LM317 and a resistor?

[Faringdon]

The LT1468 has much better characteristics, than the TL084. The greatly lower offset voltage is the most relevant one here.

Thanks, you meant the "LT1498" i think, typo.
But surely the most important one is the input common mode range, which the TL084CDT certainly fails on.
Would you not concur?

[Zero999]

The LT1468 has much better characteristics, than the TL084. The greatly lower offset voltage is the most relevant one here.

Thanks, you meant the "LT1498" i think, typo.
But surely the most important one is the input common mode range, which the TL084CDT certainly fails on.
Would you not concur?

A lower offset voltage will reduce the error.

The TL084 is not guaranteed to have a common mode range up to the positive supply, but the datasheet says it typically does. This means most devices should work with their inputs close to the positive supply rail. Your circuit has the inputs 1.2V less than the supply so it's highly likely to work, especially if it's not operated at extreme temperatures.

What are you trying to do? You haven't said why the much simpler LM317 circuit isn't suitable. There are other ways to build this circuit which don't require an op-amp with a common mode range all the way up to the positive supply.

[Faringdon]

Thanks, its just an overload current clamp.
Must roughly clamp to approx 80mA.
Will check LM317 datasheet...have done one now with the time old TLV431 based current clamp with TLV431 into series_pass NPN base.

The TL084 is not guaranteed to have a common mode range up to the positive supply, but the datasheet says it typically does. This means most devices should work with their inputs close to the positive supply rail. Your circuit has the inputs 1.2V less than the supply so it's highly likely to work, especially if it's not operated at extreme temperatures.

Thanks, the TL084CDT input common mode range  can be only up to Vcc-4V at the lowest.

So 1.2V below Vcc+ is well below this.
It could suffer phase reversal.

Would you agree, phase reversal is a possibility for any opamp with one or both inputs outside of the common mode input range?...unless the datasheet specifically states that phase reversal will not happen in such a case.

LM317_____________________________________________________________________..________
https://www.ti.com/lit/ds/symlink/lm317.pdf?ts=1706603309583&ref_url=https%253A%252F%252Fwww.ti.com%252Fproduct%252FLM317%253Futm_source%253Dgoogle%2526utm_medium%253Dcpc%2526utm_campaign%253Dapp-lp-null-44700045336317554_prodfolderdynamic-cpc-pf-google-eu_int%2526utm_content%253Dprodfolddynamic%2526ds_k%253DDYNAMIC%2BSEARCH%2BADS%2526DCM%253Dyes%2526gad_source%253D1%2526gclid%253DCj0KCQiA2eKtBhDcARIsAEGTG41BtwKUhaQDcFpdh_NXNZbpJ81z2qsdcZPy9crwUC8HZdPkt2-_aCIaAqY0EALw_wcB%2526gclsrc%253Daw.ds

Unf'y, LM317 doesnt come in nice small SOT23..or if it did, its 3v headroom would mean it dissipating 240mW at 80mA clamp.

So we prefer SOT23 TLV431 into NPN base...we will re-use an NPN thats used much elsewhere in the BOM. Clamp current oscillates when rail is hard shorted...at 7MHz.(its a sine oscillation)..this can be reduced to 80mA +/-6mA with a 680pF from Collector to base of the NPN, which is PBSS4540Z. Any more good tips on reducing the oscillation further nicely appreciated. A hard short is a fault so the 7MHz isnt the end of the world....the kit would nt be expected to work if someone hard shorts the rail anyway...but would appreciate any other tips on reducing the 7MHz if you may?

[Zero999]

If it's just a rough current limiter, then just use a single PNP with an emitter resistor and a couple of diodes to limit the base voltage.

Here's an example driving an LED.


If you need the precision then use the TL084H, which is has a common mode range which extends so just above the positive rail
https://www.ti.com/lit/ds/symlink/tl082h.pdf

[Faringdon]

Thanks, that double diode/PNP cct is a nice one for us.

I  thought id leave the  TL084H out of the talk.......its literally the only <$3.5 opamp in the world that can do Rail Vcc to 20V and common mode input range up to the rail (Vcc+)...as such there were concerns about second source situation. Do like it though.

[Zero999]

You could go for a Howland. This has the sense on the positive side and reference (V2) on the negative. The voltage across R5 is equal to V2*R1/R2. The ratio of R1:R2 must be closely matched to R3:R4, in order for the circuit to work really well.

[Zero999]

Thanks, that double diode/PNP cct is a nice one for us.

I  thought id leave the  TL084H out of the talk.......its literally the only <$3.5 opamp in the world that can do Rail Vcc to 20V and common mode input range up to the rail (Vcc+)...as such there were concerns about second source situation. Do like it though.

There are loads of cheap op-amps rated to <15V with the common mode extending to the positive rail, so why not use one of them, powered by the reference (V2 in the simulation can be a posh reference, or just the cheapo LM79L05)?

[MarkT]

So anyway, we need an opamp with common mode input range which includes VCC+.

Not the case.  you can use a differential amp topology to bring the opamp inputs down to a fraction of the sense resistor voltage, but watch out for matching resistor values accurately.

Or simply use the right component, a jelly-bean comparator for the first circuit (which is a crowbar, not a clamp).
For an actual clamp almost any opamp can do this with the differential circuit, or if you are willing to have a diode-drop's worth of dropout use the classic transistor current limit circuit used in every old-school bench supply.

[Faringdon]

Or simply use the right component, a jelly-bean comparator for the first circuit (which is a crowbar, not a clamp).

Thanks, you mean in the  top post?..(the TLV431 circuit) ..the first circuit in the top post clamps the current to approx 1.24V/15R.
Is that what you meant?...A crowbar as i understand it is a circuit which shuts off current flow completely when a threshold is breached.
(which is a behaviour that we dont wish to have)

The two opamp ccts kindly provided by Zero999 (in  posts #9 and #10),  appear to tbe the best way for us if using an opamp, since they garantee not going outside the common mode input range of the opamp, and therefore they garantee
not going into phase reversal.
Though the TLV431 cct (of top_Post  and post #6) looks the cheapest. But when the output is shorted , it does suffer the 7MHz oscillation in the clamped current as discussed in #6 above.
This 7MHz oscillation is reduced well by adding 680pF from Base to collector of the NPN, but any better ways much appreciated...

Having said that, any current over 70mA is an overload so it means the customer is at fault, and has overloaded the circuit, so as long as it clamps to approx 80mA , then thats fine, and the 7MHz oscillation doesnt matter.....(because the 7MHz osc only happens if the bus is hard shorted , and if the customer does that, then the thing doesnt work anyway)

Anyway, we have all well & kindly  discussed the evil that is the common mode input
range of an opamp. Its so easy to accidentally make a circuit on the bench that works even though it violates the common mode input range of the opamp........and so it then goes into production, and then  some unknown time later, a batch of opamps is used which goes into phase reversal
due to the common mode input range violation.....and the company goes bust!

[Zero999]

Why not have it cut-off? That's how over-current protection normally works.

The circuits posted so far will not drive a short circuit for a long period without blowing up. A power transistor would need to be used, rather the TO-92 packaged devices shown on the schematics.

Here's an example of a current limiter which cuts off the current. If it's going to protect against fire, then it needs to have a real fuse as well, or use a fusible resistor for R5.


There are also various tricks to combine the E-fuse circuit, with a time delay linear current limiting, so it limits the short circuit current to several times the rating for a certain period of time, before cutting the power. I'll post it if, there's any interest.

[Faringdon]

A power transistor would need to be used

Thanks yes, we are indeed using a SOT223 BJT, gap padded to the heatsink.
Thanks Zero999, really great circuit!
(looks  like it latches off, ill have another check later)

[PCB.Wiz]

Hi,
Why ever would anyone use the opamp current clamp instead of the TLV431 based one? (LTspice's and PNGs as attached)
The TLV431 cct is so much cheaper and just as good.

I think you mean Current regulator.
You can also buy single current regulator diodes if you seek minimal BOM costs

(i mean, the TLV431 version looks so much cheaper and better that i wonder why anyone would use the opamp version? There is  the bias current in
the TLV431 which ends up in the clamped current too, but its just a few mA)

The TLV design has many components, and it has a high minimum overhead voltage.

These days, there are dedicated LED current drive solutions, so you should check those too, to see if your BOM and board area can improve.

[Zero999]

A power transistor would need to be used

Thanks yes, we are indeed using a SOT223 BJT, gap padded to the heatsink.
Thanks Zero999, really great circuit!
(looks  like it latches off, ill have another check later)

Yes, it latches off, due to the action of D1. It can be reset by cycling the power or briefly connecting U1's output to 0V.

Here's a circuit which implements the ideas I mentioned in my previous post.

D3 & D4 limits the peak current to around 200mA, by clamping the voltage on Q1's base, to 1.4V, below +V, thus limiting the voltage across R5 to 0.6V.

The tripping current is still 80mA, but there are time delays, as explained below.

C1 delays the off time. It's about 500ms in this case. It depends on the values of R1 to R4 and how much current is being drawn. If it's lightly loaded, the delay will be longer, than loaded up to the maximum.

C2 limits the off time, about 1.5s, thus converting U1 into a monostable, rather than a bistable latch. Again the current affects this delay. D2 allows C2 do discharge, when U1's output goes low.

C3 ensures the circuit oscillates when the output is continuously subjected to an overload, by introducing positive AC feedback. It might not be required in real life. I just found, it acted as a linear regulator when simulated, with the load resistance continuously set to 50R.

Note the off time is much shorter, when the overload is prolonged and there isn't chance for the C1 to discharge.

The circuit is simulated by first applying a 500R load, drawing around 40mA. After 4s, the load resistance is reduced to 50R for a further second, which would try to draw 400mA, but it's limited to 200mA, until the delay times out and it switches off. At 5s the load resistance is increased to 500R again, but no current flows, until it auto-resets. At 10s the load is reduced to 50s again and is left there to demonstrate the hiccup mode and short 200mA pulses pass at a low duty cycle.

[Zero999]

I've just had another idea: use a shunt regulator with the input referenced to the positive side such as the LM385 or LM4041-ADJ.