Multichannel buffer IC with current limiting and overcurrent detection ?

[Unixon]

Can anyone suggest a digital or analog buffer IC with multiple channels that can

1) limit current independently for every output;
2) detect and report [with an additional digital signal] overcurrent condition for every output;

My current option for (1) is just a regular op amp in voltage follower configuration and a window comparator or XOR gate (input^output) for (2).
Impedance of a non-oveloaded buffer output must be low, so simply putting a current limiting resistor at every output is not an option.

If a suggested IC does only (1), this must have at least 4 channels per package.
If a suggested IC does both (1) and (2) then even 2 channels per package would be fine, but 4-6-8-... is better.
Anyway, I'd like to know even about single channel devices that do both current limiting and fault detection and reporting.

[S. Petrukhin]

What current and voltage should be switched? There are a lot of PMIC chips, for example, for 5V is STMPS2151STR.
https://lcsc.com/product-detail/PMIC-Power-Distribution-Switches_STMicroelectronics_STMPS2151STR_STMPS2151STR_C135998.html

[Unixon]

What current and voltage should be switched?

Not switched, but transmitted to a DUT which may have its inputs shorted to some other signals including power rails.
I want to protect multiple digital/analog outputs without limiting their driving capability unless DUT consumes too much current.
Source signals are either 5V/3.3V CMOS logic levels up to 20-30mA or 5V/3.3V unipolar analog up to 60-100mA;

[David Hess]

There are ATE pin driver parts like the MAX9963:

https://www.maximintegrated.com/en/products/analog/amplifiers/MAX9963.html

Linear Technology has the LT1970 and LT1970A:

https://www.analog.com/en/products/lt1970.html
https://www.analog.com/en/products/lt1970a.html

Another way to do it is to add a current driven diode bridge in series with the logic output driver.  This had the disadvantage of drawing the maximum output current at all times but is very fast and can protect against high output voltages.

[Unixon]

The best specialized piece of hardware I found on my own that fits the description for digital signals is ST's L6374, but it is too slow.
I need 10-100MHz range covered and L6374 can handle only 350kHz signals.

MAX9963 is nice part, it's fast and does a lot of things, still about $7 per channel and huge package, however, sufficiently fast op amps may cost about the same. Good, went to memos.

LT1970 is unacceptably expensive for this purpose. LT makes a lot of fancy parts and it's a tragedy they rarely meet price target.

[Unixon]

OK, maybe my way if thinking is different to what is should be and I'm approaching the problem from a wrong direction.
What is the typical practice to protect test bench signal outputs from a faulty DUT which can possibly harm "the brains" of the test bench?

[David Hess]

I need 10-100MHz range covered and L6374 can handle only 350kHz signals.

Some ATE (automatic test equipment) pin drivers are that fast but common linear amplifiers are not.

OK, maybe my way if thinking is different to what is should be and I'm approaching the problem from a wrong direction.
What is the typical practice to protect test bench signal outputs from a faulty DUT which can possibly harm "the brains" of the test bench?

Most test equipment does not bother protecting outputs beyond short circuit current limiting, and often not even that.  Still, it can be done.  Protected outputs that are also fast just makes things more difficult because common current limited linear amplifiers have limited bandwidth.

My earlier comment covers one solution, and probably what I would do for operation beyond 20MHz:

Another way to do it is to add a current driven diode bridge in series with the logic output driver.  This had the disadvantage of drawing the maximum output current at all times but is very fast and can protect against high output voltages.

The idea is that a diode bridge is added between the output connector and a digital driver capable of sinking and sourcing the full output current at the exact voltage levels required.  Then the diode bridge is fed with source and sink currents which set the maximum output current.  Now if the output is shorted, the output current is limited by the source and sink currents driving the diode bridge.  If the output is shorted to a higher or lower voltage, then the current is still limited, one diode becomes reversed which is fine, and one current source or sink is pulled to the output voltage which it must be designed to handle.  (1) The logic drivers sees none of this.  Switching time is only limited by the reverse recovery time of the diodes, which can be very fast.

Usually this circuit is used for protecting inputs at 100s of MHz but it can also work for outputs.  Below are two examples where Tektronix used this idea in the 485 and 7A29 oscilloscope vertical amplifiers.

(1) The current source and sink must support the voltage driven back into the pin without breakdown however could shut down to limit the output current during overload relaxing the requirements for high power dissipation.

[jonpaul]

Hello: Your requirments is not specific, total number of channels needed, V levels, packages acceptable, PHY std (eg RS-485?) , loads, 3.3 or 5V etc.

 Suggest that ON semi, TI,  Maxim ST, and Intersil RS485 drivers may be useful.

 
AM26LV31 Low-Voltage High-Speed Quad Differential Line Driver
DS8921/DS8921A/DS8921AT Differential Line Driver Receiver Pair 5 V
TI AM26C31  5V Quadruple  Differential  Line Driver   RS422 5V     
TI CDCV304 200 M SE 3.3 V driver
Lin Tech  LTC1688CS
Intersil ISL3259EIBZ 

Many can have outputs paralleled.

You could Use a  current limit circuit  in series with the Vcc pin.

Most line drivers can withstand a short as they have builtin current limit and Zo.

Jon

[Unixon]

Your requirments is not specific, total number of channels needed, V levels, packages acceptable, PHY std (eg RS-485?) , loads, 3.3 or 5V etc.

Well, I agree that my request is somewhat vague, but this is because I don't have all test cases on hands at the moment.
I've mentioned above a couple of specs that would definitely play out: 5V/3.3V CMOS logic levels up to 20-30mA or 5V/3.3V unipolar analog up to 60-100mA;
Since there are very few parts that do the job solely on their on, I don't want to narrow selection prematurely, especially on packages.

Number of channels depends on DUT, I would prefer to have it in blocks of 1,2,4,6,8-channel buffers, later maybe 12,16-channel instead of 2x6/2x8 for something bigger.
I can easily make 1,2,4-channel driver with a regular fast quad op amp in single small package, but fault reporting part, if implemented, takes up a lot of board space.
Also, an integrated solution should not be significantly more expensive than discrete solution built from 1,2,4-channel op amps and comparators.
Maybe ATE drivers or single-ended line drivers is the way to go, I have to see what parts are available and read their datasheets.
 

AM26LV31 Low-Voltage High-Speed Quad Differential Line Driver
DS8921/DS8921A/DS8921AT Differential Line Driver Receiver Pair 5 V
TI AM26C31  5V Quadruple  Differential  Line Driver   RS422 5V     
TI CDCV304 200 M SE 3.3 V driver
Lin Tech  LTC1688CS
Intersil ISL3259EIBZ 

Thank you, I need some time to review these.

You could Use a  current limit circuit  in series with the Vcc pin.

I've though about that. The downside is that one faulty line can bring whole buffer IC down and then it is impossible to tell which line that was.

Most line drivers can withstand a short as they have builtin current limit and Zo.

Yes, thanks, I know, I was about going to research available line drivers in general.

[Unixon]

The idea is that a diode bridge is added ...

Yeah, that's clever. I guess no one bothered to turn this into an IC ?

[jonpaul]

Hello: You can Sense either Vcc or ground return current.

 multi channel driver IC has COMMON Vcc and GRD, so you can sense ONLY the entire package current.

If you can describe the entire system and what the end goal is you will get better responses.

Jon

[Unixon]

Hello: You can Sense either Vcc or ground return current.
 multi channel driver IC has COMMON Vcc and GRD, so you can sense ONLY the entire package current.

I guess I have to sense both, since which rail is overloaded depends on voltage difference between signal source (MCU/DAC/...) and buffer output connected to DUT and this can go both ways.

If you can describe the entire system and what the end goal is you will get better responses.

Among other tasks I'm developing multiple test jigs for assembled electronic modules from small DIY breakouts to more complex boards, but in all cases these are just PCBs with soldered parts and connectors. Depending on a type of board being tested, procedure may require driving digital or analog interface with test patterns and measuring some output or intermediate signals and comparing them with correct values or sequences/patterns. The problem is that if a faulty DUT somehow breaks test jig circuit this could be troublesome to timely fix or replace it. Therefore I'm trying to prevent this from happening in the first place by designing test equipment functional blocks with necessary protection circuits to make the whole system more reliable.

[jonpaul]

Hello again: Just put the drivers/buffers on sockets, use DIP parts and easy replacement if blown.

All RS-485 drivers have built in transient and overload protections.

Jon

[Unixon]

Hello again: Just put the drivers/buffers on sockets, use DIP parts and easy replacement if blown.

No, that's not the point. In case if something blows up, a test jig / bench, whatever you call it, will be out for servicing anyway.
Whatever needs to be replaced, servicing is done in a different place by different people from where testing occurs and this disrupts normal working process.
DIP sockets are not exactly required since the test system is already modular and could be repaired by replacing small individual boards.
So, I'm trying to reduce the very probability of a serious malfunction at least to give time to service a broken unit and return it back to backup/hot reserve.
Production batches aren't very large, so we're planning to have at minimum two jigs of a kind, one for regular use and one for backup.

All RS-485 drivers have built in transient and overload protections.

I wonder is it OK to use only one line of each pair if I don't need a differential line...
Wouldn't there be some current disbalance in the differential line driver that affects its operation?

[jonpaul]

All RS-232, RS-485 drivers are differential.

The output stages are oppsite polarity and independent.

We use just one side all the time.

Suggest to examine some TX spec sheets.

Here is a typical and excellent high speed transciever originally from Intersil.

https://www.renesas.com/eu/en/document/dst/isl3259e-datasheet

Bon Chance,

Jon

[David Hess]

The idea is that a diode bridge is added ...

Yeah, that's clever. I guess no one bothered to turn this into an IC ?

Some protected line drivers might do it this way.  The big advantage of using a diode bridge is that it can be extremely fast.  Multiple diode bridges could be driven from the same current source and sink with multiple outputs which saves some of the circuitry.

A slower alternative I have used to do this in the past is to use the current source and sink without the diode bridge.  The input signal switches which one is active, and the source and sink have a deliberately limited output compliance to set or limit the output voltages.  Essentially this is a logic output but using the output transistors configured as current mirrors.  The disadvantage is that it is slower because limiting the output voltage in this way causes the output transistors to operate in saturation.  In the past I have used 2N4401/2N4403 transistors this way to make fault protected high performance RS-232 drivers.

[S. Petrukhin]

You could Use a  current limit circuit  in series with the Vcc pin.

I've though about that. The downside is that one faulty line can bring whole buffer IC down and then it is impossible to tell which line that was.

You can install PMIC on the each VCC line for powered input sensor or each power line output chip - they are very cheap and tiny, the overload signal output is usually an open drain and this output can be connected in parallel to several chips.

[David Hess]

It just occurred to me that there may be another simpler option.  Use a driver which can provide the necessary output voltages and currents, this might be a bunch of logic outputs in parallel, and then put two back-to-back depletion mode MOSFETs in series with each output around a suitable low value resistor which sets the maximum output current.  This resistance will be in series with each output but each output will be fault protected against excessive current by the depletion mode MOSFETs up to their voltage and power limits.

https://electronics.stackexchange.com/questions/473310/using-a-depletion-mode-mosfet-as-a-current-limiting-device

[max_torque]

You could Use a  current limit circuit  in series with the Vcc pin.

I've though about that. The downside is that one faulty line can bring whole buffer IC down and then it is impossible to tell which line that was.

Er, the one with the fault is the last one you switched!  Monitor the voltage or current of the VCC/GND and if it's outside of spec, simply cycle your outputs till the fault occurs and that's your faulted line! As the normal state is "not faulted" this tiny additional to determine which line is faulted should that occur is not detrimental surely?

[tszaboo]

There are ATE pin driver parts like the MAX9963:

https://www.maximintegrated.com/en/products/analog/amplifiers/MAX9963.html

Linear Technology has the LT1970 and LT1970A:

https://www.analog.com/en/products/lt1970.html
https://www.analog.com/en/products/lt1970a.html

Another way to do it is to add a current driven diode bridge in series with the logic output driver.  This had the disadvantage of drawing the maximum output current at all times but is very fast and can protect against high output voltages.

I was searching for similar custom ATE ICs, came across these, and they seem to be super overkill even for most professional environments. I ended up using the LT1970A as my DUT power supply and MAX11300 as the I/O for the DUT. I dont see why there isn't a "middle of the road" ATE ASIC. These ones makes sense for the semiconductor business, where you want to save every microsecond of testing, but for a PCB functional test jig ?

[David Hess]

I was searching for similar custom ATE ICs, came across these, and they seem to be super overkill even for most professional environments. I ended up using the LT1970A as my DUT power supply and MAX11300 as the I/O for the DUT. I dont see why there isn't a "middle of the road" ATE ASIC. These ones makes sense for the semiconductor business, where you want to save every microsecond of testing, but for a PCB functional test jig ?

I doubt the market is large enough to support the economy of scale for something less expensive.  There are various types of fault protected drivers but the ATE market is tiny and the requirements are diverse across applications.

Anybody who really needs this sort of function can either adapt a more common ASIC or design a discrete implementation from scratch.

For instance a current limiter like I described above could be placed in series with the output of an LT1210 current feedback power amplifier before the feedback network to make a fast fault protected output.  How much simpler do you want?  I like the switched limited compliance current source and sink method but I have used it many times before.  The diode bridge method is used for the fastest applications.

[Unixon]

Usually this circuit is used for protecting inputs at 100s of MHz but it can also work for outputs.

I took a second look at this schematic and it appears that its standby current is also equivalent to current limit since all diodes from current source to current sink are forward biased.
Total static power consumption may become a problem for a device with many low current outputs or fewer high current outputs.

I've tested few fast op amps in voltage follower configuration and I'm quite disappointed with their stability without additional compensatory components with have to be tailored for certain working frequency bands.
So, for digital drivers I went for total cheap shit (see attachment)
Surprisingly this circuit works fast enough to save the buffer and doesn't introduce too much voltage drop at low currents.
I'm not entirely happy with 30 Ohms as one of these guys will still drop 0.6V at max driving current of +/-20mA before disable kicks in.

[David Hess]

Usually this circuit is used for protecting inputs at 100s of MHz but it can also work for outputs.

I took a second look at this schematic and it appears that its standby current is also equivalent to current limit since all diodes from current source to current sink are forward biased.

Total static power consumption may become a problem for a device with many low current outputs or fewer high current outputs.

That is right, it is not a low power method.  The big advantage it has is speed.

[tszaboo]

I was searching for similar custom ATE ICs, came across these, and they seem to be super overkill even for most professional environments. I ended up using the LT1970A as my DUT power supply and MAX11300 as the I/O for the DUT. I dont see why there isn't a "middle of the road" ATE ASIC. These ones makes sense for the semiconductor business, where you want to save every microsecond of testing, but for a PCB functional test jig ?

I doubt the market is large enough to support the economy of scale for something less expensive.  There are various types of fault protected drivers but the ATE market is tiny and the requirements are diverse across applications.

Anybody who really needs this sort of function can either adapt a more common ASIC or design a discrete implementation from scratch.

For instance a current limiter like I described above could be placed in series with the output of an LT1210 current feedback power amplifier before the feedback network to make a fast fault protected output.  How much simpler do you want?  I like the switched limited compliance current source and sink method but I have used it many times before.  The diode bridge method is used for the fastest applications.

But it could be used as a PLC input-output, 0-10V industrial input or output. Or for some interface. Or a bridge exciter. Or... many more things that need output.