Hi,
I need to build an in-circuit board tester.
Basically I need to measure:
1) resistors (2 Ohm to 1 MOhm) 2) capacitors (10 nF to1000 uF) 3) inductors (10 uH to 10 mH) 4) diodes polarity 5) transistors base-emitter polarityI don't need to do "Open Pin Scan".
I'm thinking to use the following chip from Analog: AD5520
The bed-of-nails and the contacting system is ready.
Any help or hint will be appreciated
Enrico
Production, development, or for batch qauality full audit.
Circuit powered or unpowered?
Circuit idle or in use?
For how many points (impetus/response) at same time total number of test points total number of tests per point
How often does the unit have to be calibrated and to what standards?
You do have a full test spec to work to for all the tests? Including lower and upper limits?
How many and how many different ranges?
You do know what the circuit loading variations are going to be at all points.
You do know how to calculate on the fly the new upper/lower limits if measuring at a point where one resistors variance could affect another reading.
No in circuit capacitance to allow for measurement settling time.
What response time.
That is a wide range to test, and requires some form of loading circuit, to get approximate idea as the rest of the circuit capacitance will also be measured.
Load may affect circuit.
Over what frequency range.
What response time.
Same as per capacitors
Why?
Run diode test on some types of very small signal diode especially reverse voltage can blow them. This means you will have to have programmeable drive and range selection.
Again why, and this is only a diode test.
This is always useful to check your probes are connected properly and right orientation to the circuit. Garbage In Garbage Out.
That is meant for parameteric testing of semiconductors with 64 pin LQFP per active impetus and response pin of test. I.E. to check under test conditions a batch of devices function within all limits, over the full temperature range of the device.
Even semiconductors undergo functional test only for production as full parametric testing takes too long, when you have a 10 inch wafer with 50,000+ parts on it. Non military/avionic usually at room temperature only.
Unless you are designing for space applications where 100% test is required, this seems overkill.
I assume with a huge bank of relays to configure which sets of pins are going to be active at any point in time, for each test. Otherwise you will need a 64 pin LQFP for EACH 'nail', which makes test setup and changeover lengthy and have to be done in particular sequence to avoid circuit damage. Controlling that will be a nightmare.
How many 'nails' do you have as this seems overkill fo a circuit test.
No doubt you will have to run open-pin scans to check that each test is configured properly and a relay has not stuck/failed.
Also as you will have to have banks of relays for matrix switching (to ensure connection path impedance does not effect measurements) this means lots of cabling and connections, which means more points of failure.
It takes ages to setup and test/calibrate a tester like that.
I have designed and commissioned ASIC testing systems for wafer probing to full packaged parts, connecting to units being tested on full military/avionics temperature range. These are for production and still only do functional testing at limit points.
What you are describing seems overkill.
The control, vectors and pass/fail limits for this are going to be massive, and with measuring capacitance and inductance, will extend testing time to make this unprofitable for any size circuit. The larger the board the more tests are required.
--
Paul Carpenter | paul@pcserviceselectronics.co.uk
PC Services
Timing Diagram Font
GNU H8 - compiler & Renesas H8/H8S/H8 Tiny
For those web sites you hate
Hi Paul,
thanks for your answer.
PREFACE
--------------- Our boards usually have: one microcontroller, passive components, one DC/DC converter, some analog, leds, push buttons, and relays.
Our boards' power supply is +24 Volt DC.
We are a small company. We do the actual design, we route the board, and we write the software for the microcontroller.
We are just approaching the world of Tests and Measurements.
So far, I've been doing just Functional Tests. But I would like to do both: in-circuit test and functional test
We also make the contacting system and the bed-of-nail (we use a piece of PCB as a substrate)
What I would like to make is a lab-made in-circuit tester to test our products.
What I know is:
1) In circuit tests are carried out with the board NOT POWERED 2) Accoding to Teradyne's standards, signals used during in-circut tests are in the +/-200 mV range, in order not to power up the boardI've been looking around for almost a year for a cheap In-circuit Test but couldn't find anything I can afford. Here are some prices:
Spea 3030 =3D 30.000 to 35.000 Euros Keithley 2600 Source Meter + Analog Mux + Relay matrix =3D 20.000 Euros German Company (*) =3D 25.000 Euros (*) I can't remember the name
I actually need to build an instruments like the "smart tweezers":
This instruments cost 315.00 $
ANSWERS
----------------
Production. My volumes are 1000 to 5000 boards at a time.
Unpowered
Unpowered
30 to 50 test points.
Once a year.
I just need to know the value of Resistors, Capacitors, Inductors and the diodes' polarity. I don't want to do full parametric or margin tests.
The total number of resistors are 10 to 20. Different ranges are 5 to
10
I don't understand the question. Maybe because I don't have experiences in tests.
I don't understand the question. Maybe because I don't have experiences in tests.
I don't understand the question. Maybe because I don't have experiences in tests.
I need to know if the diode is correctly oriented.
Open Pin Tools costs 5000 to 10000 Euro My production volumes are not big (1000 to 5000) therefore I can't afford one.
The bank of relays is not that huge: 1 Eurocard =3D 128 relays
30 to 50 nailsThe system I got in mind is:
1 board (Euro card size) with 128 small signal RELAYS 1 board (Euro card size) with MICROCONTROLLER + ADC + AD5520The MICROCONTROLLER does the measurements and sends to the PC the results.
The challenge is to optically isolate the sytem from the PC.
OPEN PROBLEMS
---------------------------- I don't know how to measure capacitors and inductors I don't know how to guard a node of a circuit?
The measurements methods are a secret, I couldn't find a single book about in-circuit tests.
QUESTION
---------------- Is the AD5520 suitable to build a system like that?
Enrico
So you are likely to need=20
a) different bed of nails for each product variation b) probably different relay matrix board for each product variation =20
Like a lot of companies. =20
What are you trying to acheive beyond your existing tests? =20
You have not said why you need this level of test and what you are=20 trying to achieve beyond functional test. =20
That spec is mainly for semiconductor testing, in your cicruit, you may or may not be able to use higher voltages.
Which also means you need to keep cable lengths SHORT, and path=20 resistance (cables, relays, nail connection and measurement circuit) to minimum to avoid affecting the measurements.
Especially for low resistance ( I've been looking around for almost a year for a cheap In-circuit Test
Search Circuit Cellar for a LC meter using tweezers for surface mount components. There was an article in December 2008 issue - 'Component=20 control' that did just this. =20
How much time do you expect the tests to take per board?
You would have to a high failure rate due to component assembly problems to justify this. Maybe you are fixing the wrong problem.
Work out what you think for one board the FULL list of TESTS and expected results, then try doing the tests using=20 a multimeter and get a GUIDE time for each test.
This will compare with the relay turn off delay and next relay turn on time, impetus setup and settling time, and measurement=20 sampling time. =20
How are you going to calibrate the unit? You will need to calibrate bearing in mind the cabling, relays etc.. =20
Resistors, capacitors and inductors have tolerances, and combined in a circuit they will have a combined tolerance, so you will have an upper and lower limit for EACH measurement.
There will always be some form of margins for the test.
That is a complex range switching for max 200mV, along with switching out= =20 for other types of tests. =20
Maybe basic circuit theory, there WILL be other impedances in=20 parallel with what you are testing, so you need to know the variations of all parts loading the point of test. =20
If one measurement is affected by the tolerances of another component being near one of its limits, could affect other measurements. =20
Basic circuit theory, what is maximum capacitance in parallel with an impedance test will create an RC delay with your tester so will take a variable amount of time (due to size of C) to settle to the level you expect to measure. =20
You must have a high failure rate on board assembly to justify this.=20 =20
Sorry you are wrong you need to sample various points to make sure your board is orientated correctly and making correct contact, otherwise you=20 will get a lot of false positives (failures that are actually test=20 equipment malfunction).
It needs to be built into your testing procedure to interpret results correctly, NOT a specific piece of equipment.
You are forgetting all the signal path integrity and crosstalk problems that can occur for small signal routing. =20
Needs cables to bed of nails
Bad of nails is also a potential point of failure
Needs cable to relay board.
Needs either on board software for each product to test or means of selecting tests, or a host (PC) programme to control the tests. =20
So you have to develop a host programme for each product test procedure along with means of identifying the board set capabilities, product to=20 test. Let alone know the board is correctly orientated =20
That is the least of your worries, and I reckon for an unpowered circuit hardly necessary. Ground loop problems maybe. =20
Plenty of resources for correct Google search.
What do you mean by 'guard a node', I have a suspicion you mean something different to everybody else.
I suspect you expect to isolate each component to measure it as if it was a discrete component. This will not be possible if parts are assembled on the board, without a secondary assembly process of adding lots of links to put the components back into=20 circuit, which you will then have to test that all over again to check the links are all in. =20
Probably because you are looking for the wrong thing, i.e. one book to answer all your questions.
There are plenty of resources including books, papers and other things describing measurement methods.
Overkill for what you are doing.
Work out how each type of measurement is made, then work out what you=20 need.
I personally think you need to contract someone else to design this to meet your company requirements.
--=20 Paul Carpenter | snipped-for-privacy@pcserviceselectronics.co.uk PC Services Timing Diagram Font GNU H8 - compiler & Renesas H8/H8S/H8 Tiny For those web sites you hate
thanks for your answer.
PREFACE
--------------- Our boards usually have: one microcontroller, passive components, one DC/DC converter, some analog, leds, push buttons, and relays.
Our boards' power supply is +24 Volt DC.
We are a small company. We do the actual design, we route the board, and we write the software for the microcontroller.
We are just approaching the world of Tests and Measurements.
So far, I've been doing just Functional Tests. But I would like to do both: in-circuit test and functional test
We also make the contacting system and the bed-of-nail (we use a piece of PCB as a substrate)
What I would like to make is a lab-made in-circuit tester to test our products.
What I know is:
1) In circuit tests are carried out with the board NOT POWERED 2) Accoding to Teradyne's standards, signals used during in-circut tests are in the +/-200 mV range, in order not to power up the boardI've been looking around for almost a year for a cheap In-circuit Test but couldn't find anything I can afford. Here are some prices:
Spea 3030 = 30.000 to 35.000 Euros Keithley 2600 Source Meter + Analog Mux + Relay matrix = 20.000 Euros German Company (*) = 25.000 Euros (*) I can't remember the name
I actually need to build an instruments like the "smart tweezers":
This instruments cost 315.00 $
ANSWERS
----------------
Production. My volumes are 1000 to 5000 boards at a time.
Unpowered
Unpowered
30 to 50 test points.
Once a year.
I just need to know the value of Resistors, Capacitors, Inductors and the diodes' polarity. I don't want to do full parametric or margin tests.
The total number of resistors are 10 to 20. Different ranges are 5 to
10
I don't understand the question. Maybe because I don't have experiences in tests.
I don't understand the question. Maybe because I don't have experiences in tests.
I don't understand the question. Maybe because I don't have experiences in tests.
I need to know if the diode is correctly oriented.
Open Pin Tools costs 5000 to 10000 Euro My production volumes are not big (1000 to 5000) therefore I can't afford one.
The bank of relays is not that huge: 1 Eurocard = 128 relays
30 to 50 nailsThe system I got in mind is:
1 board (Euro card size) with 128 small signal RELAYS 1 board (Euro card size) with MICROCONTROLLER + ADC + AD5520The MICROCONTROLLER does the measurements and sends to the PC the results.
The challenge is to optically isolate the sytem from the PC.
OPEN PROBLEMS
---------------------------- I don't know how to measure capacitors and inductors I don't know how to guard a node of a circuit?
The measurements methods are a secret, I couldn't find a single book about in-circuit tests.
QUESTION
---------------- Is the AD5520 suitable to build a system like that?
Enrico
Hello Enrico,
I think you are looking at this the wrong way - if you need in circuit componet test go out an buy a tester - it will cost you more to build than to buy once you factor in the cost of development of the hardware and software. If you buy in you get a system with all the problems sorted and enough help/training to get it going (or your money back !!).
But do you NEED in circuit component test ?
If your board assembler uses automatic optical component checking then you can be pretty sure that every joint is soldered and every part the same on every board. If you have some very critical areas in your design then make them testable by adding testpoints, on board self test, whatever.
Far better to put some work into your design than into re-inventing a test system which you could buy.
Michael Kellett
Enrico,
Do yourself a great favor, buy a in-circuit tester. I will have to admit that I am a partially biased voice on this issue, as I onced managed the development of In-circuit test equipment (Zentel/Teradyne). I have also had the experience of working for a company that after seeing a Zentel product (Z1800), had decided in would be very easy to build a clone of the product for their own testing needs. Long story short, they failed! (several years later, and thousands poorer, they could not even test resistors). I was very glad that I was never asked to make in work for them.
A quick glance of the used/referbished market shows that a low-end Z1800 could be purchased for ~$18K. (I'm sure that a deeper look could find a better deal, try ebay) In the long run this would be much cheaper, and far more reliable than a home built unit.
With a more capable unit, you could do things like test multiple boards at once, test with power on, maybe even gang program devices......
Good luck Gene
--------------------------------------- Posted through
I'm wondering why you had such a bad experience.
Years ago I was hired to solve a specific test problem at a ballot counting company. They were unable to align and test the optical readheads for their main product. In one month I had a prototype working that would give a graphical display of the mechanical alignment of each of the 4 lens assemblies, to .001" and select the proper resistors for gain. In three months, they had 3 production testers. The readhead test department went from 7 people to one. They only needed one tester, but wanted to have a spare and a spare for the spare (:.
Given my experience, it's hard to believe that after several years, your associates couldn't test resistors.
admit
the
also
the
Jim,
I think that the biggest mistake that people make in this situtation is that they under estimate the full scope of the job. What looks like a very simple requirement (In-circuit connection and testing of a hand full components) gets increasely more difficult as you get further down the stream. What about guarding of connected component? What about charge build up in capacitors? Further more, it was quite clear that the original poster was not too clear about how to even test a capacitor.
I too know of people who were able to successfully build their own test equipment. I know that sometimes it is the only way to go. But more times that not, the time and cost of rolling your own tester will cost you more in the long run.
In the case of the co-workers that I refered to, they completely under estimated the task at hand. Not only that, they were far to stubborn to realize that they were in over their heads and re-evaluate the situtation.
Gene
--------------------------------------- Posted through
I'll second that notion, but for different reasons. Bluntly put, testing equipment shall never be built, nor designed, by the same people who make the device under test, period.
You need different people designing the tester to avoid etting false OK results slipping through because your home-grown tester made the same mistakes it's supposed to be checking for. You need a different plant to build the testing equipment to ensure that they don't both have the same, slight mechanical ill-alignment.
Oh, and designing that test equipment requires all kinds of knowledge that by your own admission you currently don't have. By the time you've acquired that knowledge (and are 101% bet-the-farm-on-it _sure_ you've got it right), it's almost certain you'll have spent more than a turn-key solution would have cost.
Testing equipment, like all measurement devices, has to be of higher quality than the stuff it's supposed to test, all across the board. That's the only way you can ever tell apart actual effects from flawed measurements. The usual way people check that condition is by telling you to get your testing equipment from companies who are known to know what they're doing on that front, and will vouch for calibration, long-term stability and such. The companies in that market get to call up the kind of prices that you consider so forbiddingly high because their machines are worth just about every single cent.
And if, as I suspect, it's your customers' QA people that are driving this requirement for an ICT step, either they or your liability insurer will quite likely _insist_ that you use third-party testing gear anyway, for all the above reasons. They simply won't believe that the results of a tester you made for yourself are significantly more reliable than your production process already is.
Ultimately, if your production volumes are so small that you can't seem to justify the costs of an ICT machine, although your quality requirements suggest you should have one, I would suggest you thoroughly re-investigate whether you should be populating your own PCBs in the first place. You might be better off farming that job out to contractors --- they have ICT.
Hi guys, thanks for your valuable answers.
I would like to summarize what was said so far:
I read carefully how the "SmartTweezers" measures the values of components and realized that's exactly what I need.
Look at page 3 of:
If they sell the "SmartTweezers" as an OEM product (that is, no LCD, no cover, just the PCB), one can build up a portable IN-CIRCUIT TESTER
Just to make an example, this is the Bill of Material of my last design (I skipped connectors and IC's):
Fuse - 2 A - fast = 2 NTC - 33 Ohm - in-rush current limiter = 1 Capacitor 1 nF - 1 kV - ceramic = 2 Capacitor 100 nF - 1 kV - poly = 1 Capacitor 100 nF - 50 V - ceramic = 8 Capacitor 100 uF - 35 V - Elec. = 1 Capacitor 22uF - 400 V - Elec. = 1 Common mode choke - 100 mH = 1 Transformer for flyback converter - 4 windings - 2 mH = 1 Diode Schottky = 2 Diode 1N4007 = 1 Diode BAS 16 = 1 Transistor - BCX41 = 2 Resistor - 220 Ohm - 2 W = 1 Resistor - 0603 - from 220 Ohm to 10 kohm = 20 Relay - DPDT - 2 A = 1 Led - 0603 - red = 6
I bet 100 $ that the SmartTweezers are able to measure'm all :-)
ciao, Enrico
Which depends on circuit configuration on how accurately you can measure them and how long the measurements take.
You still do not show basic circuit theory for when two impedances are in parallel (even if you are not aware of that impedance). You do not show any understanding about time of measurements for parallel capacitances (intentional or remainder of circuit).
But it takes some knowledge to work out how to us them and how interpret the readings into pass/fail or correct/incorrect.
How are you going to determine a diode polarity if the diode is a commutating or flyback diode across the relay coil? Plenty of other configurations of diodes and transistors where you will meet problems.
How are you going to determine the value of one resistor with another resistor in parallel?
How long are you going to wait for a resistance measurement?
How will measure a circuit that has a bleed resistor across it? Or for that matter the bleed resistor, not uncommon for very large capacitors or other charge storing circuits.
To what accuracy (percentage) are you expecting to measure the components?
What accuracy do you expect from the measurement system?
The IC's and connectors may exhibit parallel loads to your testing (let alone shorts).
It will be able to take measurements, whether they will be correct to match the component in there, or you know how to interpret the results is another matter.
I suggest you make one and try measuring ALL the components on several models of working products and see what readings you get. Simpler still just use a multimeter to measure resistance to do the resistance, diode and b-e tests, then see what answers you get.
You show little knowledge of basic circuit theory to believe you actually understand the simple points everyone has put to you in the past, and are just looking for the cheapest solution and believing it will work without understanding.
You are blinded by expecting a technology silver bullet to just happen.
--
Paul Carpenter | paul@pcserviceselectronics.co.uk
PC Services
Timing Diagram Font
GNU H8 - compiler & Renesas H8/H8S/H8 Tiny
For those web sites you hate
.....=20
Without reading it properly.
It works on components NOT ASSEMBLED TO THE BOARD. =20
....=20 =20
The IMPORTANT part you did not read and understand, which shows signs of not understanding basic circuit theory -
"You don?t have to remove and replace a part because you don?t know if it?s the right part. This goes a long way in reducing stress, especially when a circuit is not working correctly. Of course, you need to know how the component is connected in the circuit to determine if its measured value is being affected by the surrounding components."
The sentence starting "Of course.." is the bit everyone has told you about and you are not listening to. This is on Page 6 of the PDF right=20 hand column.
When you show signs of understanding that, then we might believe you stand a chance.
--=20 Paul Carpenter | snipped-for-privacy@pcserviceselectronics.co.uk PC Services Timing Diagram Font GNU H8 - compiler & Renesas H8/H8S/H8 Tiny For those web sites you hate
Hi,
One measure per second would be good.
I spoke to an ATE expert and he told me:
When the impedance is complex, that is R//C or R//L or any other combination, there are many approaches possible. The real part of the impedance, that is R, could be measured applying a constant voltage and wait for some time. Another method to get R and C is applying a sine voltage ad a certain frequency and measure the sine current and then get the phase difference and the attenuation.
The ATE expert told me: there's no way to distinguish 2 capacitors in parallel. The ATE system will always measure the sum of the two.
First of all, during In-circuit test, the board is powered off, therefore no voltage comes from the mains. In Flyback power supplies, the diode is not in parallel to the coil. In parallel to the coil I have an RCD clamping network therefore the diode's nodes can be accessed by 2 nails.
The ATE expert told me: there's no way to distinguish 2 resistors in parallel. The ATE system will always measure the parallel of the two.
In certain cases the surrounding circuitry can be "GUARDED", that means that using a third or a fourth probe you can isolate the Device Under Test from the surrounding circuitry. Take a look at the following 2 documents:
1) SCHEIBER, S. F. (2001) Building a Successful Board-Test Strategy pages 62-63 2) US Patent 4774455 - Method for automatic guard selection in automatic test equipmentone measure per second
The ATE expert told me: large capacitors are measured by injecting a constant current, for some time and sampling the voltage. The voltage is allowed to rises up to 200 mV for not powering up the surrounding circuitry. By using the equation: ic = C (dV/dt) one can get the value of the capacitor.
One can even test if the capacitor is correctly oriented (I still don't know how)
The SmartTweezers measures are accurate, take a look at the specs. I expect to get the same level of accuracy.
ATE systems never power up the surrounding circuitry. Some Teradyne's documents state the measuring voltage must be in the range of +/- 200 mV
I'm trying to save 20.000 USD :-)
The specs say that it may work even on parts assembled on the board. It may fail because the "SmartTweezers" are not a portable ATE system and because they are software programmed to accomplish certain tasks. In my case, the hardware and the software is under my control.
This is my Linkedin profile:
ciao, Enrico
You had to speak to an expert to work that one out?
You missed the word IF in the sentence, just because one board is OK, others may not be as there may be other impedances in parallel to a diode.
You had to speak to an expert to work that one out?
....
So you have not specifed your requirements first?
So, I have seen all sorts of loading even at those loads, as some pins of devices were actually a low impedance (< 500R in some cases
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.