|
I’ve come to the realization that every time I write a repair story something has broken! This time it’s a Freeview set top box, aerialBox model T1000n with an inbuilt universal power supply.
Figure 1. aerialBox T1000n Freeview Set Top Box
The Fault
This particular unit is connected to the roof mounted UHF antenna in the master bedroom (the most direct coaxial cable routing into the house) and drives a circa 2008 Sony LED flat panel dumb TV. The loop output from the unit provides the RF gain to drive an additional 20 m of coaxial cable to the lounge TV. The unit is normally plugged in, but turned off with a red LED illuminated on the front panel. The loop RF amplifier still operates with the unit off, but with power plugged in.
The lounge television image suddenly started to become highly pixilated on all Freeview channels. It didn’t take long to identify that the red led on the T1000n was weak and flickering.
The T1000n wouldn’t turn on from the remote or the power switch. Unplugging the unit (power down reset) did not restore functionality.
Sometimes when stuff is out of warranty and breaks it just isn’t worth fixing. But I can’t get the same model and a new one (with all manner of functionality that I don’t need) will cost between about $180 and $500 plus freight. Okay, let’s proceed to the service bench.
Safety First
This unit is mains powered and almost certainly has a Switch Mode Power Supply (SMPS) because it is physically small and light weight. This likely means stored energy at rectified mains potential and this can cause serious harm (electric shock and fire are in this category).
As an interesting aside the T1000n isn’t actually grounded to mains earth. It floats and assumes the ground of connected equipment. This helps prevent ground loops which can be a significant cause of noise. It also means that under fault conditions the metal case could be at rectified mains potential!
The SMPS uses rectified mains. When taking measurements with mains powered instruments (like an oscilloscope) an isolating transformer is essential. Inadvertently connecting rectified mains to ground through a test instrument is simply not a good idea.
Attaching test equipment with power applied is potentially hazardous so take appropriate precautions such as powering down between test-point adjustments and thinking about what you are connecting to. Allow or force the discharge of stored energy after the power is removed (high voltage electrolytic capacitors for example) before physical handling.
When servicing you don’t want to make the situation worse. Prodding around randomly on a board with a multimeter, or failure to observe anti-static precautions can actually induce faults. Before taking a service measurement you truly need to think about why you are doing it, the expected outcome and the likely consequences.
Looking for the Obvious
A really good place to start is checking the obvious, trying to work systematically. Launching in with a soldering iron is seldom a good idea.
The unit demonstrates the same dim red LED flickering fault from another wall outlet. Conclusion: the problem is likely with the T1000n and not the mains.
Remove the cover (three screws on the rear panel). There’s a bit of dust but otherwise nothing looks untoward.
Figure 2. Case Open
Figure 3. Nothing Immediately Apparent
Check the fuse. All good, rated at 2 A. Conclusion: We haven’t had an overload fault.
Check the plug and power cable electrical connectivity right to the board. I can see you saying ‘you’re kidding me’, but I’m not. This unit has a cheap molded 230 V AC appliance plug and from time-to-time I’ve come across these plugs with a high resistance due to poor terminations in the molding, cable faults due to kinks and flexing, and incorrect phasing. We’re all good here.
Check the internal plugs and sockets (there are just three). Remove and re-seat them taking care not to pull on the wiring to prevent causing damage. Reapply power. Same fault. Conclusion: unlikely to be a connector issue.
Failed electronics tend to look or smell burnt. Have a look and a sniff around. Nothing looks burnt or overheated. But something smells cooked around the SMPS transformer. There are no visible signs of overheating. The transformer insulation (blue Mylar tape) looks fine. There is no distortion at the top of electrolytic capacitors, and no apparent heating damage or gunk ooze on the board. Note that while burnt stuff is clearly indicative of a problem this may not have been the root cause.
Conclusion: something has got hot in the SMPS based on the smell but it’s not clear if this is associated with the fault.
Figure 4. Something Smells Burnt Around Here
The signal processing and RF sections of the board contain a lot of big silicon chips. If any of these are toast then the board will be beyond economic repair. Further, this section of the board is not where you want to make random test measurements without risking further damage.
But here’s a handy trick.
The output of the SMPS is almost certainly 5 V which is a necessary rail for the USB host socket (confirmed from the silk screen mask on the underside of the board). This is supplied by a single through-hole rectifier diode at the output of the SMPS. So if we apply 5 V from an external supply between the cathode of the diode and floating ground (connected to the chassis) then we can test the functionality of the signal processing circuitry and measure operating current without energizing the SMPS.
With an external 5 V DC supply the unit is completely functional! The current is about 0.57 A on and 41 mA in standby. The front panel red LED is now bright and continuous. The current is reasonable (there is a lot of big silicon on the board and one IC has a large heat sink adhered to the top). Conclusion: the problem is definitely with the SMPS.
In retrospect this conclusion was wrong. If you read on you might be surprised to find that the SMPS was probably completely serviceable. Sure the SMPS wasn’t working but the cause was downstream. The use of the external 5 V DC supply actually masked the problem. This sort of issue can arise any time you’re working in a feedback loop.
This also suggests an alternative method of completing the repair by mounting a simple DC plug on the rear panel and using an external 5 V 1 A DC universal supply or wall wart. But let’s see if we can actually fix the SMPS before taking this course of action.
The heart of the SMPS is a Silan SD6830 IC in an 8 pin DIL socket. It has an integrated primary winding transistor switch. The data sheet indicates that the IC should be good for up to 12 W SMPS output power (that’s 2.4 A at 5 V). It has numerous functional elements including fault, Vcc and temperature protection. As a general rule semiconductors only break if stressed beyond their maximum ratings. We should look elsewhere before attributing the fault to this IC.
I found a draft data sheet with a typical application circuit. The T1000n application appears to be pretty close to the typical application, although I can’t fully trace the feedback section on the board without removing components.
The LM431 and the two resistors R8 and R9 determine the output voltage trip point of the SMPS. Resistor R6 provides about 1 mA through the anode of the LM431 when it is turned on, and resistor R5 limits the current through the optocoupler photo diode to provide active regulation of the feedback pin (Pin 5) of the SD6830.
Figure 5. Typical SD6830 Application Schematic (Preliminary)
A good thing about the SMPS is that we can test stuff with an Ohm meter without significant risk of causing damage. All of the SMPS resistors, diodes and choke check out good with simple Ohm meter measurements. The electrolytic capacitors can be a bit tricky to test in circuit because they are in parallel with other stuff. But they are easily removed and can be checked by measuring their charge time through a high value resistor and their leakage current at rated voltage. The two 10 uF 400 V electrolytic capacitors were confirmed at 10 uF with minimal leakage.
That leaves just the transformer and a small 47 uF 25 V electrolytic capacitor which is the half wave power supply filter for the SD6830. The transformer windings appear to be intact and the capacitor checks out just fine (just over 47 uF).
The next stage is to confirm that the rectified DC rail is actually about 340 V DC and that it is present at the IC MOSFET drain. A multimeter test indicates that we’re good here.
It appears as though the SD6830 is starting and shutting down. Every other individual component tests fine and there are no apparent dry solder joints or board cracks. The transformer windings appear to be intact. The only symptom of overheating (that burnt smell) is around the transformer and SD68730.
As much as I hate repair by component replacement (which can end up causing damage to replacement components if the problem isn’t actually identified) I have no service information and everything other than the IC and the transformer has been statically tested.
Actually not quite everything, but read on...
So I need a replacement SD6830. These aren’t available locally so I have ordered two from China at a cost of $0.31 each.
In the interim I will remove the IC. Darn, despite as much TLC as I could muster, one land on the board has lifted. This is entirely my fault because I don’t own a vacuum desoldering station or an 8 pin DIP desoldering block, and I wanted to get the IC out intact. Sometimes stuff like this happens, particularly with lead-free solder due to its relatively high melting point. I’ve repaired the damaged PCB land with a 0.8 mm ID copper rivet.
I now had an interlude of three whole months waiting for the replacement ICs. After 50 days I opened a dispute with the supplier and ordered some more ICs from another source. The dispute hasn’t been resolved, but the second order from the new supplier arrived in just a few weeks.
So I replaced the SD6830 but the SMPS still isn’t operating. The T1000n symptoms persist (dim red LED) so whatever the problem is, it probably isn’t the IC. The only other component in the SMPS that I haven’t checked is the transformer (actually not quite everything, but read on...). My previous tests indicated that the windings were intact but this doesn’t ensure that that the transformer is serviceable. There are a number of possible faults including leakage or shorts between turns and between windings. A regular step down transformer fault would likely result in significant over-heating and maybe a fuse failure - but this isn’t a regular step down transformer.
I don’t have any information on the transformer so I’ve decided to rewind it. Since my track damage disaster removing the SD6830 I have invested in a Duratool vacuum desoldering station. The 7 lead transformer came away from the board in just a couple of minutes with the Duratool with no track damage. The last time I had access to a vacuum desoldering station was about 40 years ago. I should have invested in my own one years ago.
Figure 6. Duratool Vacuum Desoldering Station
The tricky part of transformer disassembly is separating the core to get at the windings because it is glued together, and to the former. Sure enough I cracked the ferrite but this can easily be repaired using adhesive (Super Glue). With the core removed I carefully removed the Mylar insulation, photographing each layer, noting the winding direction, counted the turns in each layer, and measured the wire diameter.
Figure 7. Oops, I Broke the Core
(but easily fixed with Super Glue)
Figure 8. Transformer Unwinding Layer by Layer
Here’s the transformer schematic and my simplified rewind.
Figure 9. Transformer
(T = number of turns, L = Layer Number from bobbin)
Nothing seemed to be untoward on the transformer deconstruction so best I get on and reassemble it. But there is one unused winding and the primary doesn’t need to be tapped and split above and below the secondary. Rather than use Mylar for the layer insulation I used paper masking tape because it helps with maintaining even (unjumbled) windings and provides excellent insulation (16 KV/mm or, for 0.08 mm thick tape about 1.3 KV). And I used some Kapton tape to insulate the winding terminations (cheek insulation). The windings were tested for continuity and insulation before the core was re-adhered into the former. The rewound transformer windings were significantly more compact that the original, despite using thicker insulation and identical wire diameter.
But with the rewound transformer back in circuit the fault persisted!
I refuse to be defeated by a simple SMPS and consign the T1000n to recycling. It’s time to get serious and use an oscilloscope to see what is actually happening (all measurements to date have been with a multimeter). The SMPS runs off rectified mains so this requires the use of an isolating transformer. Let’s go back to the start of our servicing.
The rectified mains is at 340 V DC with no ripple. The SD6830 is switching the 340 V DC supply (Pins 7 and 8) through the primary transformer winding. Vcc to the IC (Pin 2) is at 11 V DC with no ripple. There is almost no DC voltage on the IS over-current pin (Pin 5). The DC output (after the rectifier diode) is about 2 V with significant positive and negative transient at switching. The feedback optocoupler photodiode has a saw tooth waveform with a significant transient at switching. In writing up this repair I am kicking myself for not recording these waveforms. They might have been helpful to others but the opportunity is gone!
The SD6830 data sheet is a poor translation into English, and is self-inconsistent so it’s difficult to work out what should be happening. For example let’s consider te various descriptions of Vcc:
Vcc max: 18 V
Supply Voltage Vcc: Minimum 5: Typical 6: Max: 12 V
Overvoltage Protection Threshold: Vovp 11.0 (min), 12.0 (typ), 13.0 V (max),
. when VCC Voltage reached 9V After the chip starts working, it enters the normal working state.
2. PWM control VCC Normal operating voltage range 5-9V . The peak current of the switch is determined by FB The voltage is determined.
3. Vcc Overvoltage protection Integrated inside the chip VCC Overvoltage protection circuit, when VCC Voltage is greater than 12V , Through the internal control circuit, put FB The voltage of the foot is pulled down, so as to turn off the output; VCC Voltage back 12V the following, The chip resumes output. This control method, the highest VCC Voltage clamped at 12V ,Guarantee IC Work reliably.
Vcc was measured at 11 V. This is higher than typical and implies that the over-voltage protection could be active but I don’t think it is because the FB pin (Pin 5) is not being pulled to ground.
As a quick check I powered the T1000n from ~120 V AC through my isolating transformer. Vcc had dropped to about 9 V but there was no change in the SMPS circuit performance.
The transient on the output (which is inverted and coupled through to the FB pin) is a problem. The positive transition is almost certainly causing the IC to shut down. This implies that either the diode, resistor and capacitor snubber clamp across the switch isn’t working effectively and/or the LC filter on the 5 V DC supply rail has lost capacitance. I’ve tested the snubber components and they’re fine. So this leaves the capacitors in the 5 V DC filter.
These assumed faults would not have been apparent when operating the T1000n using an external 5 V DC supply because there are no transients and the external supply is well filtered, masking a filter capacitor fault.
The nature of the onset of the original fault also suggests that the problem could be the filter capacitors on the 5 V DC supply. I removed the two 1,000 uF 10 V electrolytic capacitors and tested them. On removal I can see residue on the surface of the board. This implies that the first capacitor in the CLC filter has leaked electrolyte. The capacitance and leakage checked out just fine. Maybe the Equivalent Series Resistance (ESR) has gone high? Sure enough the first capacitor has an ESR of about 4.4 Ohms at 100kKz when it should be about 50 milli Ohms. The second capacitor is fine. The SMPS failure is directly attributable to the high ESR.
Note. A reasonable estimate of ESR can be obtained by connecting the capacitor directly to a signal generator (in my case the AFG on my oscilloscope) at a relatively high frequency (say 100 kHz) where the capacitive reactance is small. The ESR is then simply:
Resr ~ Rsig x Vout / (Vin-Vout) where Rsig is the output impedance of the signal generator.
The through-hole plating for the capacitors also looks pretty dodgy so I reworked these with copper rivets to make sure of a connection to between the solder pad on bottom of the board and the track on the top.
Figure 10. Gunk Ooze from the First Filter Capacitor
While removing the capacitors with the Duratool desoldering station I found a tombstoned SMD capacitor which looks like it was caused during manufacturer’s rework (not of my making). I resoldered the capacitor in place. This fault must have always been present hence it cannot be the cause of the current problem. But in general, well designed commercial boards don’t have redundant components. Fixing this fault is expected to improve some aspect of the T1000n performance.
Figure 11. Tombstoned SMD Capicitor due to Manufacturer Rework
(above SMD jumpers)
I don’t have any appropriately rated 8 mm diameter electrolytic capacitors in stock so I replaced the originals with 10 mm diameter components (1,000 uF 16 V). They’re a tight fit and sit atop several SMD passive components but they should be fine. With the new capacitors in place I applied 230 V AC power to the T1000n and the unit is operating again. Whew! Replacement of the SD6830 and the transformer haven’t been fatal. The 5 V DC supply rail at the second filter capacitor is flat (no transients) at about 5.1 V DC.
So I didn’t need to replace the SD6830 or rebuild the transformer. I might have avoided this if I had realized that the 5 V DC CLC filter was part of the SMPS feedback loop, and that using an external DC supply would mask problems with the filter. Servicing within feedback loops is seldom straight forward and I’ve experienced grief and frustration with this in the past. A fault can be expected to propagate throughout the loop and downstream so nothing is making sense. While breaking the loop may assist, this is not always possible and, in the case of SMPS, excess forward path gain could take out downstream components.
I left the T1000n powered up on the test bench for about 6 hours. Nothing is getting hot (although the big silicon chip with the heat sink does get warm) and the unit continues to be functional. I put the T1000n back into service with the UHF antenna, the output RF loop, and the television via the HDMI interface. The unit is operating just fine, but I’ll be waiting a few more days before announcing a successful repair.
Two days have passed and we’re still operational so job well done!
While I missed the opportunity to save the fault condition waveforms here are the serviceable measurements that may assist with your repair.
Figure 12. Rectified Mains at 340 V DC
Figure 13. Snubber Clamp RLC Junction
(Rectified mains reference - about 500 V DC)
Figure 14. 5 V CD at First Filter Capacitor
Figure 15. 5 V DC at Second Filter Capacitor
Figure 16. SD6830 Transformer Switch (Pins 7 and 8)
Figure 17. SD6830 Vcc (Pin 2) at ~ 7.4 V DC
Figure 18. SD6830 FB (Pin 4) at ~ 2 V DC with no Transients
Aside from my time the repair cost has been minimal. About $3 for the replacement SD6830 including postage, maybe $1 for the transformer rewind, $2 for the replacement capacitors, and a few cents for consumables. If I were billing the repair time then even replacing the capacitors, reworking the tombstoned SMD capacitor, fitting the through-hole rivets and soak testing would have consigned the T1000n to e-waste.
Regrettably this is the consumer society that we now live in.
|
