Are "black boxes" a lost cause?

[tcv]

I very much appreciate the forum indulging me on some stupid questions. In keeping with tradition, I'd like to ask another.

Please feel free to skip the bottom.

I work in I.T. so I often come across "broken" equipment that will get tossed. Such a thing happened this week. A Netgear GS724T v3 switch was getting tossed because it "wouldn't power up."

Since I have been reading a lot about electronics and have been doing little experiments here and there, I like to use stuff like this as an opportunity to learn.

I focused on the power supply since the symptom reported was the unit "wouldn't power up." I attempted to look for any obvious scorching or problems on the power supply. Keep in mind that I barely know what I'm doing, but I can probably see a broken or burned pin, bulging cap, or leaked electrolytic if I'm careful. I never found anything of the like but did learn that the switch does have a fuse. Not really knowing how to get the fuse out of its hard white tomb, I simply tested continuity on both sides of the fuse and heard the appropriate beep.

Right, so I kept reading around and found that the main pcb in the switch is powered exclusively by 5-volts and that the switching power supply inside is a widely used model. I knew how the power supply connected to the pcb, but I didn't know how to measure the pins on the power supply's female connector.

Looking at the pins, though, showed two labels "GD" and "VO." I surmised that GD was "Ground" and "VO" was volts. I powered up the supply and measured VDC and found 5vdc to my surprise. I then connected up the PCB and found that the fan spun. Connecting up the LED board, however, showed no activity. Powering off the switch and powering it back on showed the faintest flicker of all the LEDs. I don't know if that means power is making it there though.

This is the point where I've reached the limits of any "knowledge" I might have. I think I've determined that proper voltage is exiting the power supply and reaching the board... at least enough for the fan to spin on the other side of the pcb. I know already that the switch was not switching or routing traffic the day it was pulled out of the rack which lends more evidence to something on the PCB rather than simply unresponsive LEDs. I don't really know, however, whether that one pin that shows 5vdc is enough for everything. Maybe other pins should be showing 5v as well and I don't know what I don't know.

I also don't know anything about the PCB. I see that I can look up some components, but not the PCB itself. None of the numbers bring up Google results. I can guess from some things that it's a 4-layer board and most but not all items are surface-mounted.

What I'm asking is this: How useful is it to a person learning electronics to bother with trying to repair sophisticated systems? To me, most everything in this unit is a black box. Can I ever find out what everything does and identify the component(s) that are bad? Or is there a likely and large absence of information and knowledge without which I cannot repair it? Can I use opportunities like broken equipment to learn?

Thanks, everyone.

[Alana]

Had something like that once - switch turned out to have overheated board-to-board connector between psu and mainboard. Visible marks of overheating are de-colored boards, wire insulation, plastics... it all had it, as long with your symptoms.

[JackOfVA]

Everyone must start from somewhere, but Michelangelo's first work was not on the Sistine Chapel ceiling.

You might be better off looking at some simple equipment that is known to be working (and nothing that is so valuable that it would be upsetting to have it damaged by accident) and discovering how it works. Preferably something with schematics.  And, with enough test equipment so you can know what is going on and correlate your observations and measurements with the schematic and the box's inputs and outputs.

Once you have a sense of what a good box does, you are in a better position to identify the fault with a bad box.

Not saying a real wiz at repair can't fix anything with a jackknife and a roll of electrical tape, but it's a lot easier with the proper tools.

 

[alxnik]

It's highly dependent on the nature of the beast. You have to know how it works in order to troubleshoot it.

Embedded computers (like the netgear box) have many things in common and if you focus on just switches they have even more stuff in common. If you learn the general principles of operation of a specific kind of gear you then can work with some general troubleshooting.

Having said that, I find that if the supply of power is not the cause (not just the power supply itself) it usually is not worth the time. These are very complicated things in a very small form factor and you might have a fault that only by having the datasheets of all the chips and a lot of time you can solve. Sometimes you might find a debug port in these things that might give you some more info on the nature of the fault (if they boot up that is), but other than that...

[Galaxyrise]

Do you have examples of the same switch which work?  You might be able to make some more progress by having a known good unit to compare against. 

[kxenos]

You can do the following test.
The output electrolytic caps of the power supply may be damaged without any visual sign. If this is what happened then there is a case that you measure about 5V with no load but when you connect the board and fan your output voltage goes to something like 2V which is not enough. To test this, connect all the stuff, power up and measure the voltage again. It would be better if you had a scope to see the output voltage but a simple DMM will do. Disconnect the fan and let the power supply and the main board connected. Measure the voltage again because the fan may be dirty/damaged and draws a lot of current. Finally you can power the main board and fan from a bench PSU and see if it draws a reasonable amount of current or not. Set the current limit of your bench PSU to the current rating of the power supply board (probably about 2A). If it current limits, that means there is a shortcircuit in the main board. If so, give it 5A from the bench PSU so that the shortcircuited part (along with some traces) will blow up.
Hope that helps

[pickle9000]

You may want to invest in a cheap ESR meter. Caps are a big problem.

[dfmischler]

In theory, a persistent and knowledgeable person can troubleshoot and repair anything.  But there are a lot of factors that make this difficult.

 1. Multi-layer boards can make it very difficult to probe specific circuits.

 2. Component packages with interior "pins" (e.g. BGA) can be difficult to probe.

 3. Components that can't be identified can make things really hard to know what to look at.  Custom parts are more and more common in high volume goods.

 4. Firmware has a huge impact on what can/should happen in a product, and this affects your ability to understand how the product works.

Couple all of these together and you might need tons of time and resources to reverse engineer a product.  Way beyond what is economically justified to repair it.  That's why checking fairly obvious things like power supplies is about as far as most people go.

[Rerouter]

one very big tip for mains style power supplies for devices is that the AC function of your multimeter can be invaluable for getting a good idea of the ripple on a DC rail, as its only likely to be 60 or 120Hz (for 60Hz areas) measure the DC voltage, and minus the AC voltage, this tells you just how low the voltage sags to at its low point, if a device is meant to run off 5, and measures 5V DC, but has 2V of ripple then your looking at a capacitance issue,

always suspect power rails and failed solder joints, generally at the same time, if a part is constantly getting hot and cooling down such as a CPU its more likely to work harden and begin cracks in the solder joints, same for the regulators as when the CPU is crunching away its an extra demand and a bit of extra heat it has to deal with,

if the device is over 5 years old and the device pins are visible give the pins a quick spray with air, or even a firm blow on, very tiny tin whiskers can form over time providing low resistance shorts / lift-ups / pull-downs where they are not intended, but they are very fragile thus easy to remove, 

and finally to kick this myth in the butt as it seems to be spreading in this thread, you dont have to understand something as a whole to repair it, you only need to have basic enough knowledge to figure out what doesn't looks right and to try fixes until you get a result, further knowledge and skill speeds the process up through narrowing down the options but it doesn't mean the end result is any different. if you gave a man a board with 10,000 connections and told him it had a bad joint even if he has never held a multimeter eventually he will work out one of them looks and measures differently (or get bored)

[AKM]

I think it is very useful, even if it is time consuming and frustrating experience. If you continue to search for the problem, you will find it and you will learn a lot on the way.

Try looking at ICs, google the part numbers, download the datasheets and application notes and you will probably find what they and most of the surrounding components do.

If there are regulators, try measuring the input and output voltages. If there are ICs, try to find the VCC and GND pins and measure the voltage; compare it with its datasheet value.
Take macro pictures of the board and take a look on the monitor, you may notice something you havent before.

[Crazy Ape]

It's always worth checking the console (if you can find it) for errors on embedded systems. There could be a message telling you why it's failed such as Bad flash block, firmware checksum error, etc.

[tcv]

Hey everyone,

Thanks for all the responses. Once again, you all have indulged me and I am very appreciative. I'll try to respond to everyone.

First, I think the general idea I am getting here is that while, yes, one could learn electronics through the troubleshooting and repair of parts, complicated systems are complicated. I don't know many things and I even don't know what I don't know. Can I do it? Sure. Is it going to take a LONG time, yes. 

Second, talking about the Netgear in particular. I angled my post as a more general Can You Learn In This Way, but many of you made specific suggestions and I did some further tests.

This switch is dead after probably one or more surges. A Comcast modem also bit the dust during the same incident.

There are six pins on the female portion of the power connector. Three are ground. Three are +5VDC. Powering up the entire switch does not change that output as kxenos asked me to check. I don't have a PSU that I can use to increase the amps. I haven't gotten anything that sophisticated yet. FWIW: The output listed on the power adapter is 5v 8a.

I can measure 5vdc along much of the upper left of the board. There seemed to be a coloring to the mask in which one part is +5 and one is ground. The coloring reaches a series of ... capacitors I believe (they're all listed as "C##" on the board) around an Intel chip which, I believe is the flash memory. I don't know if this means that the flash memory is powered first or something in run from flash. (I honestly don't know much about the chip or about how chips are powered and code is run from one.)

I am pretty sure that this is the datasheet for the Intel chip: http://www-mtl.mit.edu/Courses/6.111/labkit/datasheets/28F128J3A.pdf

I see a lot of items on the board listed as TP which, if it's anything like pinball PCBs are test points. The problem, though, is that I don't know what I am supposed to be reading on these TPs. Most of them around the exposed ICs (meaning the ones that don't have heat sinks) show 0.0v. Some of them, especially right below the capacitors and on the right side of that Intel chip show 0.25vdc. I don't know what that means, though.

Rerouter, I'm hoping you can give me more step-by-step instructions to look for that ripple you're referring to. The DC voltage is 5, wouldn't the AC be 110v?

I am going to look for a console, Crazy Ape. Good idea.

There are many things on the board that I don't know how to identify. For instance, there are a few vertical-standing donut-shaped items in which copper wire goes in and around them.

I wish I had a working sample, Galaxywise. That would be awesome. ;-)

Thanks again, everyone!

[AG6QR]

There are many things on the board that I don't know how to identify. For instance, there are a few vertical-standing donut-shaped items in which copper wire goes in and around them.

Those sound like toroidal inductors.  Google the term to verify.

Some repair rules of thumb:

Always check the DC power supply.  You can often replace it as a unit if needed.  Switches, relays, potentiometers and other components with moving parts commonly fail before semiconductors.  Electrolytic capacitors are among the most common failure points that have no moving parts -- an ESR meter can find and diagnose these even if you don't know what the electrolytic capacitor's function is within the circuit.  Replace a failed capacitor with one rated at least as high a voltage as the original, and close to the same capacitance (often it won't hurt to increase capacitance, but that depends on the function the capacitor is performing).  If you have to replace one failed electrolytic, and you see others of the same value nearby, it's probably worth replacing them all.  Transistors and other semiconductors located at inputs and outputs are other common failure points, especially if there may have been problems related to power surges or nearby lightning strikes.

More detailed troubleshooting can often be done if you understand the circuit and schematic, and have the right tools.  But with something like that ethernet switch, the tools and techniques required to diagnose and fix problems in the internals are not trivial.  Sometimes you check the basic stuff, reach the limit of you knowledge and tools, and give up.

[garak]

I don't know jack about the particular box you're working on, but I might be able to give you some hints.

1) PSU. As everyone says, this is always the first place you start. You've already done some voltage checks which seem to be giving you reasonable readouts, but that's not quite the end of things. If there are bad electrolytic caps on the output stage then these can cause large peaks and troughs which will upset sensitive electronics at the other end. The voltage the DMM shows might be 5V, but it's just a rolling average. The only real way to check this is with something like an oscilloscope or possibly an analogue meter. Jitter on the readings of a DMM can also indicate this, but it's generally pretty sketchy.

1a) Many embedded computer PSUs also provide a signal called Power Good. This is a signal that indicates to the circuit being powered that the PSU has stabilised and it is safe to take the CPU out of reset. If there are large spikes on the supply line like I mentioned then this will probably stop Power Good from being asserted. It could also be that the chip on the PSU that is responsible for generating this signal is dead.

1b) Even once you've confirmed that the main PSU is okay, that doesn't mean that there aren't other power supply components on the motherboard that have been killed. See Dave's recent DSA repair videos on youtube for a good example of what I'm talking about. You should check that all the voltage regulators etc on the motherboard are providing the output that they ought to be. Check also any electrolytics on the motherboard, especially those near heat-prone or power supply areas.

2) Once you're sure that all the power supply bits are working as they should, you can try connecting to the console on your device to see if you can see anything. It's possible that the FLASH memory has gone bad. If you get nothing you can try to reflash the FLASH from blank if you've got the image and the programmer, or you can move on to checking things like the CPU clocks (you will definitely need a scope for this). If the clocks are good and you still get nothing then this is probably beyond economical repair, at least in regards to time spent.

In regards to your actual question, yes! It is entirely possible to pick up a decent amount of knowledge by working on equipment like this. Many of the failures you'll come across are actually something very simple, and the fact that this equipment was going to be binned anyway means that you don't lose anything if you can't get things working. There's also the bonus of being able to keep whatever you can repair to spur you on. You'll also find that most of them follow the same sort of design pattern - companies like to stick with what has been proven to work. I worked on a variety of digital projection gear for 6 months this year, and trust me, after you've seen one you've seen em all. This applies with most electronic gear to some extent. What you see might be daunting now, but once you've figured it out you'll find that what you've learnt can usually be applied elsewhere.

HTH

EDIT: I forgot to mention that it will help you to do a little research about what you're facing first. Your current candidate has a switch mode power supply that might be causing issues, so it's a good idea to run a Google search like "how does a switch mode power supply work?". Understanding how things are supposed to work will greatly aid in you in finding out why yours does not.

[tcv]

What does an oscilloscope cost? I've looked them up before, but I wouldn't know what to buy. I do have an analog multimeter. Maybe I'll see if I can see peaks and troughs with it.

You know: I would be happy to pay for shipping to send this PSU to someone here who is in the U.S. and has a scope. Maybe I can learn something that way, too.

[AG6QR]

What does an oscilloscope cost?

The Rigol DS1052E sells for around $330.00 at various websites.  A couple of years ago, that was the standard recommendation for hobbyists (and it was more expensive back then).  Now it's still a reasonable recommendation, but there are scopes that are very significantly better at only slightly higher prices.  There are also some scopes available, particularly used ones, including analog ones, that are significantly cheaper.

Anway, move over to the "test equipment" forum and read all of the "which oscilloscope should I buy" threads.  If you don't find the info you need, then start another thread.

Two words of warning: 1) Not long after you get an oscilloscope, you'll start wanting a signal generator, so remember this as you plan your budget. 2) No hobbyist class oscilloscope will allow you to probe the signal lines of an ethernet switch - the data goes by too fast.  Any reasonable oscilloscope should be able to look at its power supply, though.

[tcv]

Thanks. Certainly a scope is quite a mystery to me, probably beyond my depth at the moment.


I think I will take up garak's suggestion and google "how does a switch mode power supply work". Seems like a great idea.

[cs.dk]

Sorry for breaking in - Is there some typical ripple-values/max. values one could look for? Ie. on 5V/USB chargers, notebook-chargers, PC PSU's, and other SMPS stuff?

Just for the question, i did find an old HDD-connector-kit, one of these things with a powersupply and a USB-device.
The +12V rail shows ~150 mV (triangle-wave) ripple at ~130 Hz. Is this normal, or am i using the scope wrong?

[janoc]

If the PCB doesn't show any obvious damage and the thing doesn't seem to be powering up, the first thing to check are the electrolytic caps in the PSUs  - bulging, leaking, exploded ones are obvious, but the caps could be dead even without any external symptoms. Get a cheap ESR meter from eBay - capacitance measurement on your multimeter is not enough! (bad cap can still have enough capacitance but the ESR is too high for the circuit to work).

Just had an ATX PSU that fails to power up when under load, but it works when powered up standalone. I have opened it and there are a few bulging caps inside ... My Samsung LCD monitor died after 6 years, caps in the PSU looked OK, but when I measured them, most of the electrolytics were dead. Replaced them with new Nichicon ones and the screen works like new again.

Bad capacitors are by far the most common problem with switchmode PSUs.


Jan

[abyrvalg]

Don't buy a scope, just do a quick test - disconnect the original power supply and power the board from some known good 5V supply (i.e. red and black wires of HDD power cable inside any desktop PC).