No commercial links

I will delete all comments, which contain links to commercial web sites. This blog is about sharing, not advertising.


The Andonstar ADSM302 microscope - more versatile than you think!

BGA reballing and main board repair means staring at very small components. I am 51 years old now and need some decent magnification to be able to see these little balls, chips, resistors and whatnot.

The high end microscopes by Amtech or Eakins, which you see in various repair videos, are prohibitively expensive for occasional use, and also very bulky for my restricted workbench.

The new Andonstar ADSM302 came at the right time for me. After checking some reviews on YouTube and reading all about it, I bought one on eBay from Germany for 200 EUR.

This thing is great! The picture quality is more than I ever need. I'll show you some modifications, which I haven't seen so far anywhere. Those make it even more useful!

Turn the base by 180°

The base is not terribly small, but still annoying. I want my stuff to lay flat on my bench. It is dead easy to mount the holder to point away from the base. An old disk drive attached as a counterweight and there you have it, a flat workspace underneath the microscope!

Mount in two positions

That's not all. The optics can be attached to the holder in two positions, thanks to a second groove below the focus ring. The manual doesn't reveal that.

Large "zoom" range!

And now look what a fantasic range you get from this little microscope. We are looking at a CPU taken from a Philips TV. Beautiful full HD images on a 22 inch monitor.

Top position:

Bottom position, still enough room to work under it:

I dare to say that this is the best microscope in the 200€ range today. Second to none.


IR6500 Rework Station Modification - Ineffective bottom heater due to overly thick glass

First of all: get the original manual for the controller HERE. The description in the IR6500 manual is an unreadable desaster.

It did not take me long to realize that the thick glass of this rework station is absorbing most of the energy from the bottom heater. The 800W of the heating element were basically ineffective.

Also, the station has trouble with the peak temperature of 230°C in the reflow program. I need to adjust the top heater very close to the chip if I like to reach it. Closer than I feel comfortable with. The manual says >2 centimeters. That's actually the maximum distance. Further away and the peak temp will never be reached. If a window is open or a fume extractor running, it is even worse. I think that a good heating support from the bottom will stabilize the temperature.

I got inspired by this YouTube video.

Screw the warranty, I want this thing to work! Let's remove the top:

Oh dear, I have used the station only a dozen times and the temperature sensor shield was already burned and coming off. The sensor had no contact with the heater anymore. This thing gets too hot for any glue. A stupid design.

This is garbage. I am not interested in the temperature of the heater, anyway. I want to know the lower temperature of the board, if anything.

The sensor has a rugged, metal sleeve. So let's pull it outside through one of the slits:

The downside of the top cover:

Unscrew the frame that holds the glass and remove the glass. This thing is really thick!

I got a metal grill from a hardware store. I cut it to the right size and screwed it to the cover:

The difference in heat radiation is remarkable. The glass takes everything away!

Sweet, now let's do a preheater test without running a program and measure a few things:

With the bottom temperature set to 100C, the board heated up nicely to the specified temp in a few minutes. Note that with the glass on, the board would go up to about 40C max after ages with the heater set to 250! 
The heating element has around 300C, which should be well inside its spec.

Well, it is no surprise that a relatively cheap chinese rework station is not quite finished and has a sloppy build quality. On the other hand, those things are simple and easily tweaked.


Philips 55PFL6158 - 2 blinks - Code 53 - QFU 1.2 - CPU reflow - IR6500 rework station

I am getting the hang of those QFU boards. :-)

Update: I have revived another two of those boards recently with the same symptom by reflowing the CPU.

This Philips did not show any sign of life except for the standby LED, which responded to an infrared remote. The 2-blink would show after a while and the level 2 code in SDM mode was 53.

Quick reminder: if you attempt to repair a QFU Philips, you must have a logging adapter, otherwise you are blind. See this blog post.

I plugged in the logging adapter and I got this in many repetitions until the two blinks came:
20:49:53.548 0x00000002 page is reading 
20:49:53.557 0x00000000 0x00000000 0x00000000 StartUnit/EndUnit/offset 
20:49:53.557 Reading out data: 
20:49:53.557 00 unit: 
20:49:53.567 waiting ECC result ready 
20:49:53.567 00 bits error in the unit. 
20:49:53.567 {preboot} 
20:49:53.567 <000> 
20:49:53.567 <010> 
20:49:53.567 <020>K

That was it. Abrubt end.

To learn what was going on in a good TV I logged my 40 inch QFU1.2. The section where the dead device looped should look like this:

20:49:53.567 {preboot}
20:49:53.567 <000>
20:49:53.567 <010>
20:49:53.567 <020>K
20:49:53.773 <030>DDDDDDDDDDDDDD 
20:49:53.831 <040>secure 
20:49:53.832 Load uboot 
20:49:53.884 U-Boot 2009.01_Production (Jun 11 2013 - 10:29:06) 
20:49:53.884 U�U 
20:49:53.911 NAND: NAND, size:1024MB, Micron(ID:0x2c,0x38), block size:512KB 
20:49:53.913 1024 MiB 
20:49:53.932 Env: NAND @ 0x01800000 
20:49:53.983 bootcmd1 
20:49:53.996 BOOTREASON=coldboot


So, the processor could not reach the point where it starts reading the main software, which is actually a Linux boot.

I suspected the SPI ROM first and swapped it. No change. Then I tried the NAND chip with an image from a 40PFL6008. No change.
The last option left was the CPU. I did not dare to reflow the whole board with its uncountable miniscule parts everywhere. I needed a tool with more precision.

I had my eyes on a BGA reflow workstation for a long time and this was the opportunity.
Here it is, the affordable IR6500 made in China (where else).

The preparation of the board was as follows:

  • Remove the heatsink from the CPU. Use a heat gun briefly and it will come off easily.
  • Apply flux under the chip. My favorite flux is Chipquick. I applied it at one edge, heated it to make it flow under the chip and shot some compressed air through the gap until it came out on the other side.
  • Cover everything around the CPU with tin foil. Some instruction videos use a tape. I don't think this is necessary as this device does not blow hot air.

Fluxing the chip:

Reflow in progress:

The first test was almost successful. It started booting but something was wrong with the NAND flash that I swapped. Linux refused it. Because the NAND was not the fault in the first place, I put back the original and voila! It started up fine.

Last task: put the heatsink back on. The original adhesive pad got ripped apart. To make the sink removable I decided to glue a thin thermal pad between the CPU and the sink:

The glue stays somewhat flexible, yet it might be strong enough to make removing a glued sink from the chip an unpleasant work. I'd rather destroy the pad instead.

This all worked well. Hallo Frau Johansson!

I do like this TV. The LG panel produces beautiful, realistic colors. My Panasonic Plasma is still the king of skin tone, but I could totally live with this one. With the sharpening reduced to a minimum, that is. I also like that it can be switched to computer mode, which bypasses all extra image processing and pixel by pixel is displayed as is. This gives you a realistic impression which information an image really contains and how much these devices invent by themselves!

The sound is not bad either. The 6000 series is still a bit on the budget side. Stereo is not taking place. The 8000 and 9000 series have a lot more to offer there.

Improving the cooling situation

Philips has undersized the CPU cooling on all QFU chassis. The QFU1.2 is even worse than the QFU1.1. The CPU won't die from that, but the solder balls will break. Interestingly, the 1.1 seems to have different typical faults than the 1.2, yet both are caused by overheating.

This is an image of the back cover from a white 40 inch:

There is some convection going on. The air stream hits the CI socket (bravo!) and exits through a small portion of the upper grill. The dark streaks are dust accumulations on the footprint of the heat sink. Hot plastic attracts dust. This solution relies exclusively on air convection. The infrared radiation hits the plastic and in turn, being a bad heat conductor, the plastic gets hot, radiating back onto the device.

Without the back cover, at 21°C room temperature, the CPU gets 55° hot. That's okay. But with the cover on and in summer I estimate another 20° or more on top of that. That's almost 80°. Too hot for sustained function, if you ask me.

I found no way to attach a larger heatsink to the CPU. There is just no free spot on the board where to fix it.

When I put the back cover on, I noticed that the thermal pad closed the gap between the sink and the cover. That was bad! Thermal death unavoidable and high danger of mechanical shock.

I cut out a rectangle at the sink position and glued a metal grid on it. Nothing beats a sink exposed to fresh air.  Infrared radiation and air convection in one. It is not exactly robust and does not protect the CPU too well, but unusual problems require unusual solutions :-D
There are more stable grids available in hardware stores, but those are are not fine enough.

One hour break-in. About 55°C constant. This is perfect. Identical thermal situation as with an open cover.

About the IR6500

This thing is a little rough around the edges as you might expect from a budget Chinese product.

The manual is funny Chinglish. I could figure out the meaning of most sentences. However, not all details are needed. There is one predefined program for leaded and one for unleaded solder. That's all I need to know. Press start and you are good.

The glass shield on top of the downside heater is bullshit. It reduces the function of the heater plate drastically. The preheating is practically ineffective. I ran the program with the top heater swiveled aside. The board made it to 45°C on the top side. That's a joke. Proper pre-heating begins with 80 or more! I think I am going to remove the glass.


Philips 42PFL7008 - QFU 1.2 - defect NVRAM - strange boot looping

Yet another Philips QFU case. This time, the TV got stuck in a boot loop. It would not reach the point where it displays anything.

The first thing to do is to start the emergency software installation. For that, you need an infrared remote, which does not come with the TV. You can get one for around 10€. The reason being that the software, which manages the original radio remote control will not be loaded under such circumstances. Silly Philips!

  • Disconnect the TV from the mains
  • Press OK or DOWN on the remote and hold it
  • Plug in the mains.

I did that and the recovery program came up. Good! This means that the screen, the backlight, and the main processor are doing fine. It can only be a problem with the software or the devices that store it, respectively.

The QFU main software, which I downloaded from Philips, installed fine. The TV went to standby - and did still not start.

Next up: logfile reading. I plugged in my adapter (see this post) and realized that the device is looping. It did not stop randomly while booting, rather it happened at more or less the same moment. The last log message was truncated, no indication that the loop was intended by the software.

This means that the processor always crashed for the same reason. Processors crash due to corrupted software. As I had installed the software previously, I came to the conclusion that the NVRAM chip must be faulty.

In the schematics, the chip is a MT29F8G08ABACAWP. In reality though, it is a MT29F8G08ABABAWP.

It is a 8GB NAND chip.

I was able to to find the binary image of the software. Now I needed a programmer with which to load the software onto the chip, and a supplier from where to buy it.

I got the chips from Aliexpress HERE. It worked fine.

The programmer I chose was THIS MODEL. It also worked flawlessly.

In case you search for programmes yourself, make sure the exact name of the chip is on the compatibility list. The letter soup is confusing and I got it wrong once and bought a non-suitable programmer first.

Swapping the chip was surprisingly easy with my preheater plate and a large nozzle on the hot air gun. In about 15 seconds the chip already floated.

With a new programmed chip soldered in (use lots of flux and a good magnifier glass!) the TV behaved differently. I was able to start it with the remote and everything seemed fine. However I was  not successful each time from standby. It would not listen to the remote everytime.

So I checked the log again and spotted yet another form of boot loop, this time initiated by the software in a consistent and regular manner. The crash was gone, but the TV would refuse to go to standby properly. Instead it first attempts to stop, ambilight goes off, standby LED switches off - and then it reboots into some semi-standby. This repeats forever.

The standby LED actually has three states, which it goes through:

- Dimmed. The TV is not listening. Any command from the remote gives me a quick flickering.
- Off. The TV is not listening. Any command from the remote gives a slow, bright blinking.
- On. The TV will only start in this state.
...repeat from top

A reinstall of the software did not change anything. I also read the boot EEPROM and it was 100% identical to the binary that I have on disc.

I have no idea what to try to fix the loop. The hardware is working. WiFi, ambilight, radio remote, all good.

My only guess is that there needs to be some extra software on the NVRAM chip. Yet, the binary image I got from the Russians was a dump of the chip. I am not even sure anymore whether it is necessary to burn the software onto the chip, because the emergency recovery software will program it, anyway.

So here I am. A TV which is basically working fine, but won't switch itself off to standby. At least I have learned yet another QFU fix, sort of.


I soldered in another NAND with the working software from a 55PFL6158 and now the thing does not even write a log anymore. The CPU gets a little warm and then cools off. This cannot have anything to do with the NAND. It is not even trying to boot. The SPI is good, too.
I tried a reflow of the CPU with no luck.
As it turns out, there is a short on one of the power supply lines for the Fusion processor. This CPU needs a reball and if that's not helping, it is probably toast.


Onkyo A8470 - speaker relais not clicking - "servo operation" lamp not coming on - degraded glue on protection IC's pins

That was an interesting repair. A friend brought me his Onkyo amplifier. It did not switch the speaker relais on and the servo operation lamp did not come on. Both are controlled by the integrated protection chip Toshiba TA7317 and after some tests and measurements, it became pretty clear that the error had to be there. The amp produced a signal just up to the relais and there was no DC on the output, either.

I remembered a YouTube video where a guy said that in old devices a certain type of glue would degrade and become ever so slightly conductive. Just enough to cause sensitive circuits to malfunction. And what have we got here? A big splash of that brownish gunk right over the pins of the protection IC! They had glued the patch wire to the board with it and splashed glue all over the place.

I cleaned it with acetone and the amp came back to life. I resoldered everything just to be safe that it wasn't a dodgy solder joint.

As a precaution I also took care of the other glue spots:


Reading Philips TV logs with an USB-UART adapter

With my Philips TVs I never had the requirement to read the log, as they all had measurable faults or the Service Default Mode revealed everything I needed to know. In case of a two blinks error code, which points to the mainboard, or when the TV won't boot at all, it can be beneficial to peek into the log.

You need

  • An USB-UART Adapter. This device maps a serial  (UART) connection across USB to a serial port (COMx on Windows). Device drivers are required.
  • A terminal program, which can handle serial ports.

The hardware

There are various types of USB-UART adapters on the market (eBay or AliExpress). First I tried this type:

Those are garbage. They contain an illegal copy of a Prolific PL2303 Revision A chip, which is discontinued since 2012. Read HERE. The problem is that the latest Windows 8 & 10 driver won't work with it anymore. Some articles in the net say that Prolific has changed the device signature in their later revisions to lock out the copies. You need to install an older version of the driver. I wasted way too much time with this rubbish.

I opened mine up and the chip had no marking on it. Sure sign of a copy. Also, the USB plug already started to come off the board.

So I tried another one with the Silicon Laboratories CP2104 chip:

This one's legit. No driver problems, Windows found the driver itself and the device worked.

The software

On the PC you can use PuTTY. A more sophisticated program is RealTerm. It can record sessions, which is quite useful, and has more features than you'll ever need. Both are free.

The Android app Serial USB Terminal by Kai Morich also works fine. You can read the log on your tablet or phone quickly without a bulky laptop. Just put a micro USB adapter in front of the UART device.

The connection

Now this kept me busy for a while due to my own incompetence.

The UART / service socket on the TVs is a stereo 3.5inch type like for headphones. The connections are as follows:

Here is the rub: you need to cross RXD and TXD. Don't connect the RXD on the adapter with the RXD on the TV. Makes total sense once you understand it :-)

A schematic from a ComPair device manual put me on track:

And that's how I built the thing. I attached a 3.5 inch stereo socket to the adapter and used a stereo cable I had lying around:

First tests

I had a working 42PFL9803 sitting in my living room and I tested the device with it. To my surprise I could not get any useful log. The service manual says 38400bps 8N1. I configured everything accordingly and all I got was garbage. The TV sent data but it wasn't readable. I tried many bps setting with no luck. This TV fooled me for quite some time. I thought something was wrong with the UART adapter :-/

Yesterday, I picked up a 32PFL9606 and with this one it worked flawlessly. I don't know what's wrong with the 9803. Very strange.

Sweet! I can add one more diagnostic tool to my repertoire. I currently have a 46PFL8007 with the dreaded QFU chassis, which doesn't show any signs of life even though the standby voltage is good. It's not writing a log either. But that's the subject of an upcoming blog post once I have reprogrammed its boot EEPROM, which I suspect.

In the meantime, I glued it into a nice blue box:


RUNTK5351 TCON - defect analysis - ISL98602

The TCONs with the ISL DC/DC chip notoriously go bad. Sometimes they are fixable by swapping the chip, sometimes they are not. I had the chance to play with four broken boards. Two of them got a new ISL and still didn't work. However, I present you a few tips how to avoid fruitless work, because the main video chip may be measurably dead. Also I think I have measured the reference voltages that the ISL should produce.

The next image shows the voltages of a good ISL chip:

If the voltages are all there and there is still no image, the main chip is dead.

Normally though, the TCONs come with an ISL, which produces no voltages at all. Here are the tests you can do to asses whether it is worth changing:

  • Test 1: In diode mode, measure the breakthrough voltage of the 1.2V trace. It should be around 0.5V. If it is 0, forget the board.
  • Test 2: Attach a lab power supply with 1.2V and current limit around 100mA to the 1.2V supply. The main chip should draw about 0.01A. If not, it is broken.

Alas, I did not yet have a working board in my hand to know how much current the board is supposed to draw from the 12V line. The ones with the fixed ISL and no shorted main chip both pulled 0.39A. The main chip got pretty hot quickly. I guess this is not normal.

Changing the ISL

This is very difficult. I never managed to solder it properly with hot air only. It always took me an extra step with the soldering iron to get the solder to flow at the pins. I failed with a needle tip. It doesn not have enough heat capacity. Spade tips neither worked, even small ones, because they all were too clunky to reach the pins. The only tip that worked was the horse shoe with an excess amount of solder on it. A perfectly rectangular tip would be best. And lots of flux is required, of course.

Be extremely careful with the microscopic SMD parts around it, especially on the upper right corner (previous image). That one 0 Ohm resistor close to the edge gets pushed away easily.