EEVblog #689 - Back To The Future Time Circuits Troubleshooting

[EEVblog]

Dave troubleshoots the Back To The Future Time Circuits in 'real-time", warts and all.
A journey of red herrings, garden paths, the bleeding obvious, and the inevitability of Murphy's Law.
Can you discover the fault before Dave does?

Discussion and comparison of the Tagarno microscope vs the Mantis 3D stereo microscope.

Dave troubleshoots the Back To The Future Time Circuits in 'real-time", warts and all.
A journey of red herrings, garden paths, the bleeding obvious, and the inevitability of Murphy's Law.
Can you discover the fault before Dave does?

Discussion and comparison of the Tagarno microscope vs the Mantis 3D stereo microscope.

TLC59282 Datasheet: http://www.ti.com/lit/ds/symlink/tlc59282.pdf
The live camera feed: http://www.eevblog.com/live

[Towger]

The curse of the EEVblog defeated...

[nitro2k01]

Here's a simple idea to improve the contrast on the yellow display: Put some Kapton tape over it if you have a suitable width roll.

[EvilGeniusSkis]

Dave good to see you fix the time circuits, Doc Brown would be proud!!! I really liked this http://youtu.be/4-PhbPS5LQU?t=34m9s view for the soldering and would like to see it used more.

[bktemp]

I have reworked many PCBs with soldering errors, brokes traces and other common faults. One thing I have learned quickly was: When probing signals, you can use vias, SMD pads or other easily reachable points as an initial test. But always check each signal right at the end, directly at the point where it goes into the plastic of the ic. This is the only way to be sure the signals reach their destination.

At 13:00 when Dave probes the COLO signal, the frequency counter in the upper right corner of the scope shows about 5kHz, but the frequency counter at the bottom shows 122Hz. Can anybody explain what causes this difference?

[JackOfVA]

In ye olde days when DIP TTL or RTL chips were installed in sockets, it was certainly not unknown for the pin-to-socket connection to fail or, more often, be intermittent.

[eV1Te]

At 13:00 when Dave probes the COLO signal, the frequency counter in the upper right corner of the scope shows about 5kHz, but the frequency counter at the bottom shows 122Hz. Can anybody explain what causes this difference?

The top one is a hardware counter, while the bottom one only calculate the frequency in software from the data in the display buffer.

[Supercharged]

This footprint error is nothing against something i did  i had to acctualy bend the pins 

[vlad777]

This video is amazing.

When you touched the chip and contents changed , THAT was the main clue.
(Dave thought it was capacitive signal integrity problem but it was cold solder.)
Yet everyone would have missed it.

[Tek_TDS220]

Nice work, Dave.  I've learned that when I hand solder IC SMT parts, that I need to do a continuity check on each pin.  It's a pain, but it saves headaches later.

You did a careful job of soldering and used plenty of flux, but still had a problem with poor wetting between the pad and the pin.  Poor joints are very difficult to see, even with the excellent microscopes that you have.

[rs20]

Yikes, one chip multiplexed over 8 displays, no resistors to spare the TLC chip the full brunt of the LED's excess voltage drop; no wonder this thing is either hot or dim...

[marcan]

Heh, when Dave touched the chip with his finger I immediately though that it must be an issue with a control input being floating - clock or some kind of clock gate or latch.

What bugs me though is that he thought that the other two boards were fine to begin with. They're three identical boards and they were all connected together when the connector SNAFU happened; identical chips tend to blow identically, and just because the magic smoke didn't escape doesn't mean the silicon wasn't thoroughly cooked. After reworking the first board, I would've completely disconnected the other two boards and worked on the first one alone until it worked - that ensures that neither are the other boards negatively affecting the first, nor that you might be accidentally mislead by assuming anything about their state.

Admittedly though, the IC failure mode happening to have almost the same symptoms as the bad soldering was one heck of a confusing coincidence.

[rs20]

I wonder if the chip that failed most spectacularly actually shunted the damaging voltage away and "protected" the other chips from failing quite as spectacularly (even though they still got damaged beyond repair)?

What bugs me though is that he thought that the other two boards were fine to begin with. They're three identical boards and they were all connected together when the connector SNAFU happened; identical chips tend to blow identically, ...

But these chips didn't blow identically. The one chip he initially replaced failed short, which is a very obvious fault. It's the fact that the others also failed, but in a more subtle way, that understandably led him astray.

[marcan]

But these chips didn't blow identically. The one chip he initially replaced failed short, which is a very obvious fault. It's the fact that the others also failed, but in a more subtle way, that understandably led him astray.

What I meant is that identical chips tend to blow under the same conditions; they were exposed to the same (grossly outside absolute max ratings) environment. I wouldn't trust any chip that was exposed to something like that, particularly not when at least one unit has failed. That doesn't mean they're guaranteed dead, of course, but at least I'd isolate them until they can be independently tested.

I do like your theory of one chip failing short and preventing the others from blowing up too though.

[hans12]

I used to have an old HP laserprinter which printed everything about 1% smaller. I found out after I had etched a board with a DIP40 socket on it, it did not fit anymore.

[Rigby]

I used to have an old HP laserprinter which printed everything about 1% smaller. I found out after I had etched a board with a DIP40 socket on it, it did not fit anymore.

Be careful when you print; this is often a setting in the driver or the program doing the printing.  For example, PDFs always seem to like to "fit to page" which means they get shrunk, most often.

[Monkeh]

I used to have an old HP laserprinter which printed everything about 1% smaller. I found out after I had etched a board with a DIP40 socket on it, it did not fit anymore.

Be careful when you print; this is often a setting in the driver or the program doing the printing.  For example, PDFs always seem to like to "fit to page" which means they get shrunk, most often.

Software in general likes that as a default because it avoids clueless users not grasping why things don't fit on the page.

[hans12]

I used to have an old HP laserprinter which printed everything about 1% smaller. I found out after I had etched a board with a DIP40 socket on it, it did not fit anymore.

Be careful when you print; this is often a setting in the driver or the program doing the printing.  For example, PDFs always seem to like to "fit to page" which means they get shrunk, most often.

It did that too if you scaled to i.e. 80%. The thing just always printed 1% smaller. Ended up patching the software to render 1% bigger and that solved it.

[Fungus]

Yikes, one chip multiplexed over 8 displays, no resistors to spare the TLC chip the full brunt of the LED's excess voltage drop; no wonder this thing is either hot or dim...

You mean the TLC Constant Current LED Driver...?

[snipersquad100]

how do you download the eagle files from github. when I click on it a just get a page of code.

[grumpydoc]

I wonder if the chip that failed most spectacularly actually shunted the damaging voltage away and "protected" the other chips from failing quite as spectacularly (even though they still got damaged beyond repair)?

To provide some counter-examples

When I was, oh 15 or 16, I repaired a transistor radio for the brother of someone I knew.

He (the owner of the radio) had decided to save himself the cost of fresh batteries by running the radio from the mains but didn't quite grasp the difference between the two sources of power and connected 240V AC straight to the 9V battery terminals.

Amazingly the only thing which was fried was the audio amp output transistors which had clearly melted quickly enough to protect the rest of the radio.

A bit later when I was doing more repairs someone brought in a guitar amp - the sort of thing with a few inputs connected via a basic mixer feeding a power amplifier all basically mounted in a 'speaker enclosure. The story was that a disgruntled band member with an axe to grind with the owner of the amp had wired a BS 1363 to 1/4" jack lead - plugged one end into the mains and the other into the amp and turned the power on.

The inputs were fed into op amps - TL071 or similar - the one which bore the full brunt of 240V AC was literally blown apart, the next mixer channel along had the op amp shorted across the power rails but physically looked OK. Every other passive and active device survived intact.

Sometimes things survive extreme abuse better that you'd guess they would.

[thm_w]

how do you download the eagle files from github. when I click on it a just get a page of code.

Right click on Raw on that page, top right, and save link as.
Since its XML there is some browser confusion on how to handle the file. Also GIT probably thinks you will use their gui/command line tool to sync files.

[nitro2k01]

But these chips didn't blow identically. The one chip he initially replaced failed short, which is a very obvious fault. It's the fact that the others also failed, but in a more subtle way, that understandably led him astray.

Though arguably, he should maybe have understood that when one chip died with a magic smoke vent from having the polarity of the power supply reversed, the rest of the chips of the same kind couldn't have fared well either. On the other hand the purpose of the video was of course not simply fixing the sign, but doing some show and tell troubleshooting. From that angle it was a good video.

And also, this is a good example of when a pin header could have been made fail safe when reversed but wasn't. Especially since multiple pins are used for ground and led V+ respectively. As many pins as possible could have been matched to connect to the same pin when connected either way, and the remaining two pins (since the number of pins for each rail is odd in this case) could have been connected to an IO where this wouldn't have caused damage.

[rs20]

Yikes, one chip multiplexed over 8 displays, no resistors to spare the TLC chip the full brunt of the LED's excess voltage drop; no wonder this thing is either hot or dim...

You mean the TLC Constant Current LED Driver...?

Yes. Now I'm not sure if you're just clarifying what I mean, or trying to hint that I don't understand what constant current means; but in case it's the latter: Let's just assume the LED drops 2V @ 20mA (probably nowhere near right but not the point). The way this circuit was implemented:

2V drop + 3V drop across TLC chip = 5V supply. 3V drop across TLC chip * 20 mA = 60mW dissipation per line.

My proposal:

2V drop + 2V drop across fixed 100ohm resistors + 1V drop across TLC chip = 5V supply. Yada yada, 20mW dissipation per line. Much cooler running. The same thermal load is now spread between the TLC chip and the 16 resistors.

And also, this is a good example of when a pin header could have been made fail safe when reversed when wasn't. Especially since multiple pins are used for ground and led V+ respectively. As many pins as possible could have been matched to connect to the same pin when connected either way, and the remaining two pins (since the number of pins for each rail is odd in this case) could have been connected to an IO where this wouldn't have caused damage.

Agreed. Or use real polarised sockets in the PCB!

[Quarlo Klobrigney]

I use Git GUI. It took me some time to figure out just what to do before I found this program. Google is your friend. Search Git GUI. Or, try this: http://git-scm.com/downloads Then use the Clone existing repository, Enter the url and do the rest. It'll save the WHOLE repository in a folder of your choice & location.
Some help: http://code.tutsplus.com/tutorials/git-on-windows-for-newbs--net-25847

how do you download the eagle files from github. when I click on it a just get a page of code.

[rs20]

When using lots of high current LEDs like this I prefer to use switching power supplies to generate a voltage at or close to the forward voltage drop. That minimizes losses and tends to be more efficient.

Yeah, this is a more expensive approach, but easy to do (especially if you use an off-the-shelf complete switching DC-DC module) and definitely yields a superior end product.

Wouldn't it be cool if the switching reg could look at the excess voltage on the TLC pins, and use that as feedback (instead of its own output)? Then you wouldn't need to carefully tune the switching reg output voltage to the optimal setting; the circuit would be self tuning in that respect. It's tricky though, because measuring that excess voltage would involve an array of diodes at the very least, so whether it's worth the complexity is debatable. Would have been cool if the TLC chip could have done all this internally and provided a feedback signal for regulators.