I have seen that yes 
With regards to why the AC modules aren't working.
If the ASC modules still work when the AC modules are plugged in then it would seem like none of the AC modules power supply inputs have shorted to ground. If they had then it would stop the ASC from working too.
It is possible that fusing, of some sort, has been included, either on the ASC or the AC board, to protect the ASC in the even of an AC failure.
Given what the AC modules are, I would expect the 4 output MOSFETs to be the most likely to break. These are some of the only active components that actually carry power, they undergo the most thermal cycles and are only heat sinked via the PCB. You could measure these and see if any of them have gone short circuit from the drain to the sink, but if they have I would expect the high voltage power supply input to the AC board to have gone short circuit. If that had happened then the ASC wouldn't work with the AC plugged in unless fusing was included.
If the output MOSFETs are fine then it's possible that the output capacitor has failed. This sits across the two outputs. If it has failed short circuit then it would short the outputs together and cause the amplifier to shut down.
With regards to why the AC modules aren't working.
If the ASC modules still work when the AC modules are plugged in then it would seem like none of the AC modules power supply inputs have shorted to ground. If they had then it would stop the ASC from working too.
It is possible that fusing, of some sort, has been included, either on the ASC or the AC board, to protect the ASC in the even of an AC failure.
Given what the AC modules are, I would expect the 4 output MOSFETs to be the most likely to break. These are some of the only active components that actually carry power, they undergo the most thermal cycles and are only heat sinked via the PCB. You could measure these and see if any of them have gone short circuit from the drain to the sink, but if they have I would expect the high voltage power supply input to the AC board to have gone short circuit. If that had happened then the ASC wouldn't work with the AC plugged in unless fusing was included.
If the output MOSFETs are fine then it's possible that the output capacitor has failed. This sits across the two outputs. If it has failed short circuit then it would short the outputs together and cause the amplifier to shut down.
If you get desparate, there's this as an option --
Power Amplifier Board ICEPOWER200ASC 200W Digital Audio AMP ICEpower | eBay
Looks to be genuine rather than clone, though can't be sure.
Power Amplifier Board ICEPOWER200ASC 200W Digital Audio AMP ICEpower | eBay
Looks to be genuine rather than clone, though can't be sure.
In a class D amplifier there are essentially a few failure paths:
- Noise margin problems in switching. Oscillation at too-high frequency. Equivalent to overload. Something has changed in the circuit w.r.t. original design (I have seen this happen in old ICEpower depending on batch of parts used and temperature).
- Stability problems. Intermittent or permanent oscillation at too-low frequency. Equivalent to overload. Something has changed in the circuit too, or there was some kind of feedback from output to input.
- User dependent overload (abuse). Overheating. Either input section (signal), or output section (load). For cases not covered by limiting circuits included.
- Programmed obsolescence, like mounting main power semiconductors (subject to marked thermal cycling) directly to PCB, also subject to marked thermal cycling.
- Degradation of materials used for heat transfer. Design techniques from computer mainboard have to be borrowed with care, as these mainboards were intended to last just a bit longer than what the computers took to become obsolete, although that trend is progressively changing (theoretical silicon processing and magnetic storage limits reached).
- PSU/mains overvoltage.
- Problems related to SMD manufacturing (cold solder joints, parts failing due to thermo-mechanical stress).
When main power transistors fail, depending on the quality of the design, there may be a chain reaction of destruction and non recognizable parts, or there may be just a couple of "firewall" parts blown. Same happens for input section when it fails, the power stage may shutdown or may blow. A good design has some tolerance to powering up with faults for measuring, without blowing anything.
- Noise margin problems in switching. Oscillation at too-high frequency. Equivalent to overload. Something has changed in the circuit w.r.t. original design (I have seen this happen in old ICEpower depending on batch of parts used and temperature).
- Stability problems. Intermittent or permanent oscillation at too-low frequency. Equivalent to overload. Something has changed in the circuit too, or there was some kind of feedback from output to input.
- User dependent overload (abuse). Overheating. Either input section (signal), or output section (load). For cases not covered by limiting circuits included.
- Programmed obsolescence, like mounting main power semiconductors (subject to marked thermal cycling) directly to PCB, also subject to marked thermal cycling.
- Degradation of materials used for heat transfer. Design techniques from computer mainboard have to be borrowed with care, as these mainboards were intended to last just a bit longer than what the computers took to become obsolete, although that trend is progressively changing (theoretical silicon processing and magnetic storage limits reached).
- PSU/mains overvoltage.
- Problems related to SMD manufacturing (cold solder joints, parts failing due to thermo-mechanical stress).
When main power transistors fail, depending on the quality of the design, there may be a chain reaction of destruction and non recognizable parts, or there may be just a couple of "firewall" parts blown. Same happens for input section when it fails, the power stage may shutdown or may blow. A good design has some tolerance to powering up with faults for measuring, without blowing anything.
Regarding PSU feedback circuit complexity:
A quick look at the pictures reveals 2 facts.
- ASC modules use a 2nd SMPS for obtaining low voltage rails, including standby rail, powered from main SMPS.
- Main SMPS is a peak current control forward converter.
These 2 facts probably imply that a dual output voltage scheme is used for main SMPS, achieved by switching something in feedback network for standby, thus the extra complexity. This approach may seem familiar from CRT TV flyback SMPS, where typically the "+B for deflection" output is switched to the "+5V logic" output for standby, to achieve idle power consumption below 1W (due to very low duty cycle, and of course regulation threshold is switched in feedback network).
Has the root cause of the failure been figured out?
A quick look at the pictures reveals 2 facts.
- ASC modules use a 2nd SMPS for obtaining low voltage rails, including standby rail, powered from main SMPS.
- Main SMPS is a peak current control forward converter.
These 2 facts probably imply that a dual output voltage scheme is used for main SMPS, achieved by switching something in feedback network for standby, thus the extra complexity. This approach may seem familiar from CRT TV flyback SMPS, where typically the "+B for deflection" output is switched to the "+5V logic" output for standby, to achieve idle power consumption below 1W (due to very low duty cycle, and of course regulation threshold is switched in feedback network).
Has the root cause of the failure been figured out?
Each daily power cycle qualifies as a thermal cycle. But keeping modules powered on permanently consumes faster the life hours of electrolytic capacitors used for low-voltage rail decoupling. These caps may be oversized or undersized, cheapo ones rated 1000h, normal 2000h, or hi-reliability rated about 10000h, or even aluminum polymer types which in same HF circuits last like 10 times longer than standard 2000h types due to ~3 times higher current ripple capability at HF.
No Eva the problem hasn't been fixed as I've been working on other things. The root cause of my failure would appear to be the main FET switch in the SMPS failed and took with it a number of components. I've so far replaced what I can with the help of the internet plus hokas help here.
What you are saying has maybe given me some food for thought.
What you are saying has maybe given me some food for thought.
The chance of a FET operating under its ratings failing is way lower than the chance of something else failing before the FET. SMD mounting of power devices without heatsink, relying only in audio crest factor for protection, has a trend towards leveling this out, but it would also demonstrate the modules have programmed obsolescence in them.
I find the PCBs by B&O quite cryptic, I can't just deduce the schematic from the pictures in this case, maybe you could draw an schematic of the PSU (as someone else did for the class D section) and we can interpret it, including solving unknowns for 3~6 lead SOT parts that could be dual diodes, dual transistors, biased or not, or triple/quad diodes, or configurable logic gates.
Understanding the circuit it should be a matter of hours to get it working. As I mentioned before, complexity of feedback network comes from the need for a low (standby) and a high (run) output voltage settings.
I find the PCBs by B&O quite cryptic, I can't just deduce the schematic from the pictures in this case, maybe you could draw an schematic of the PSU (as someone else did for the class D section) and we can interpret it, including solving unknowns for 3~6 lead SOT parts that could be dual diodes, dual transistors, biased or not, or triple/quad diodes, or configurable logic gates.
Understanding the circuit it should be a matter of hours to get it working. As I mentioned before, complexity of feedback network comes from the need for a low (standby) and a high (run) output voltage settings.
Last edited:
With my board I know that the amplifier section works because I've powered it independently from the main SMPS, the issue is getting the main power supply running.
Once again I've been busy with other things but I'm working on a decent schematic. One of the first things I did was draw one up on paper, so I could separate off and visualise the functioning blocks, but these are without part numbers or component values. The thing I am absolutely terrible at is discreet analogue signal design so this doesn't help matters.
The main SMPS FET is mounted to the PCB for cooling and the main PCB was installed in a small closed cabinet subwoofer. From what I understand low frequency (ie below the line frequency) reproduction would be quite stressful, in terms of current draw, for the PSU.
From what I can see is that when the device is in standby mode pin 2 (feedback pin) of the NCP1203 is possibly clamped to ground (or close to it?). This prevents the NCP1203 from switching properly but does allow a small voltage to develop on the secondary side. Once the standby mode is exited the optocoupler changes mode of operation and feeds back to regulate the high voltage line.
If I power the amplifier side of the board alone this appears to be how things operate. When powered on, but in standby, the opto is turned on (this would clamp pin2). Exit standby and if the high voltage is too low the opto turns off, presumably feeding back to the NCP1203 that the output voltage needs to increase. If I increase the voltage on the secondary side, as soon as the regulation voltage of 47V is exceeded, the opto turns on fully again, presumably to tell the NCP1203 that the output voltage needs to decrease.
Having tested that it would appear that the secondary side of the opto coupler is functioning as it should. It displays a different mode of operation during standby vs normal operation with the opto activating and deactivating around the regulation voltage. This theoretically should provide enough feedback for the various modes of operation.
If pin 2 is completely clamped to ground then absolutely no switching occurs and no secondary voltage develops. One part of the feedback circuit (besides the optocoupler) is designed to clamp pin 2 to ground via a low side NPN switch. Now pin 2 is connected to the collector of the NPN switch via a resistor. Sadly I have no idea what value this resistor is supposed to be as it was originally destroyed. Hoka's version does not contain this resistor but for the clamp to be effective (if that's what it's actually trying to do) then it needs to be of sufficiently low value to actually pull pin 2 to ground.
The trouble is that the clamp above is actually pulling pin 2 to ground and is, in and of itself, stopping any switching from occurring. If I remove the clamp (remove the collector resistor to pin 2) then voltages can develop on the secondary side and within the aux windings. This isn't enough to start the power supply. If I jump start the NPC1203 with a separate power supply then things start and also remain started when the separate power supply is removed. A high voltage is developed but it isn't regulated.
Once again I've been busy with other things but I'm working on a decent schematic. One of the first things I did was draw one up on paper, so I could separate off and visualise the functioning blocks, but these are without part numbers or component values. The thing I am absolutely terrible at is discreet analogue signal design so this doesn't help matters.
The main SMPS FET is mounted to the PCB for cooling and the main PCB was installed in a small closed cabinet subwoofer. From what I understand low frequency (ie below the line frequency) reproduction would be quite stressful, in terms of current draw, for the PSU.
From what I can see is that when the device is in standby mode pin 2 (feedback pin) of the NCP1203 is possibly clamped to ground (or close to it?). This prevents the NCP1203 from switching properly but does allow a small voltage to develop on the secondary side. Once the standby mode is exited the optocoupler changes mode of operation and feeds back to regulate the high voltage line.
If I power the amplifier side of the board alone this appears to be how things operate. When powered on, but in standby, the opto is turned on (this would clamp pin2). Exit standby and if the high voltage is too low the opto turns off, presumably feeding back to the NCP1203 that the output voltage needs to increase. If I increase the voltage on the secondary side, as soon as the regulation voltage of 47V is exceeded, the opto turns on fully again, presumably to tell the NCP1203 that the output voltage needs to decrease.
Having tested that it would appear that the secondary side of the opto coupler is functioning as it should. It displays a different mode of operation during standby vs normal operation with the opto activating and deactivating around the regulation voltage. This theoretically should provide enough feedback for the various modes of operation.
If pin 2 is completely clamped to ground then absolutely no switching occurs and no secondary voltage develops. One part of the feedback circuit (besides the optocoupler) is designed to clamp pin 2 to ground via a low side NPN switch. Now pin 2 is connected to the collector of the NPN switch via a resistor. Sadly I have no idea what value this resistor is supposed to be as it was originally destroyed. Hoka's version does not contain this resistor but for the clamp to be effective (if that's what it's actually trying to do) then it needs to be of sufficiently low value to actually pull pin 2 to ground.
The trouble is that the clamp above is actually pulling pin 2 to ground and is, in and of itself, stopping any switching from occurring. If I remove the clamp (remove the collector resistor to pin 2) then voltages can develop on the secondary side and within the aux windings. This isn't enough to start the power supply. If I jump start the NPC1203 with a separate power supply then things start and also remain started when the separate power supply is removed. A high voltage is developed but it isn't regulated.
Yesterday I repaired a 36V 2A offline lead-acid battery charger, in laptop PSU format. You may find the experience useful.
It was the usual peak current control flyback, 70khz, with usual SO-8 control IC, with startup resistor to +HV and bias winding, with single optocoupler feedback. Actual no-load ouput voltage was 44V. A 14V zener was used to reduce that to 30V to power feedback loop.
Voltage feeback loop was comprised of a TL431 with R divider and RC compensation, driving the optocoupler.
Current feedback loop was comprised of a 2nd TL431 for 2.5V reference, shunts for sensing output current, and a LM258 (ground sense op-amp), with dividers and RC, for activating opto, to transition from voltage control to current control at LF.
The unit would produce only <=6V output, cycling between 6V and 5.6V faster and faster as mains power was removed. Upon opening and taking a look with oscilloscope I found primary bias voltage cycling between about 16V and 11V, and during that small time of operation switching waveform was looking OK. I zoomed further to inspect the progress of duty cycle during the operation time, and it was full at beginning, reducing towards zero at end. (It was secondary side not allowing startup).
Upon further inspection the 14V zener was semi-shorted, the LM258 had the output driving the "charge" LED semi-shorted to +VCC (op-amp current limiting was relied on for LED drive, no series resistor). I replaced the zener by a 13V zener with series diode, replaced LM258, replaced open "charge" LED (actually dual reg/green LED), and added series resistor. Worked fine.
I also revised the output connector, it had signs of being reworked by low-skilled person with soldering iron, not caring much about things such as polarity (maybe 1st attempt was inverted) or shorts from the wires to the metal case of the connector (3-pin XLR type as used in Chinese low cost misc stuff). The adapter became tolerant to negative voltage transients with the addition of the diode in series with the zener, as rest of potentially damaging current paths have high enough series resistors.
It was the usual peak current control flyback, 70khz, with usual SO-8 control IC, with startup resistor to +HV and bias winding, with single optocoupler feedback. Actual no-load ouput voltage was 44V. A 14V zener was used to reduce that to 30V to power feedback loop.
Voltage feeback loop was comprised of a TL431 with R divider and RC compensation, driving the optocoupler.
Current feedback loop was comprised of a 2nd TL431 for 2.5V reference, shunts for sensing output current, and a LM258 (ground sense op-amp), with dividers and RC, for activating opto, to transition from voltage control to current control at LF.
The unit would produce only <=6V output, cycling between 6V and 5.6V faster and faster as mains power was removed. Upon opening and taking a look with oscilloscope I found primary bias voltage cycling between about 16V and 11V, and during that small time of operation switching waveform was looking OK. I zoomed further to inspect the progress of duty cycle during the operation time, and it was full at beginning, reducing towards zero at end. (It was secondary side not allowing startup).
Upon further inspection the 14V zener was semi-shorted, the LM258 had the output driving the "charge" LED semi-shorted to +VCC (op-amp current limiting was relied on for LED drive, no series resistor). I replaced the zener by a 13V zener with series diode, replaced LM258, replaced open "charge" LED (actually dual reg/green LED), and added series resistor. Worked fine.
I also revised the output connector, it had signs of being reworked by low-skilled person with soldering iron, not caring much about things such as polarity (maybe 1st attempt was inverted) or shorts from the wires to the metal case of the connector (3-pin XLR type as used in Chinese low cost misc stuff). The adapter became tolerant to negative voltage transients with the addition of the diode in series with the zener, as rest of potentially damaging current paths have high enough series resistors.
Last edited:
I have a feeling that in my case something might be interrupting the device from switching correctly.
I have observed voltages between 0 and 1.78V present on the feedback pin but have never observed any changes in the duty cycle on the switching waveform to the MOSFET.
I'll have to double check this but as nothing seems to be changing that could indicate that the problem is coming from elsewhere. One thing that I do know (I think) is that part of the circuit is designed to slow down the rise time switching of the MOSFET as a way of reducing EMI. From what I could gather this is implemented by sending a short pulse of overcurrent to pin three. Every time the MOSFET switches I do indeed get an associated spike on pin 3 but this could be doing something besides slew rate control.
One thing to mention is that whenever anything switches it's never a PWM train or train of pulses just a single event.
I have observed voltages between 0 and 1.78V present on the feedback pin but have never observed any changes in the duty cycle on the switching waveform to the MOSFET.
I'll have to double check this but as nothing seems to be changing that could indicate that the problem is coming from elsewhere. One thing that I do know (I think) is that part of the circuit is designed to slow down the rise time switching of the MOSFET as a way of reducing EMI. From what I could gather this is implemented by sending a short pulse of overcurrent to pin three. Every time the MOSFET switches I do indeed get an associated spike on pin 3 but this could be doing something besides slew rate control.
One thing to mention is that whenever anything switches it's never a PWM train or train of pulses just a single event.
Interesting. I was doing some testing earlier on and the board transitioned into cycle skipping mode and much to my surprise it began regulating properly and without all the evil spikes on the over current pin.
Proper switching commenced and with a change in the load the duty cycle appropriately changed with the supply voltage remaining stable.
Proper switching commenced and with a change in the load the duty cycle appropriately changed with the supply voltage remaining stable.
Last edited:
Good. Did you find something else broken? Or was it an observation bias problem? Or cold solder joint? No more options.
Extra parts in primary side of a simple "plain switcher IC" SMPS can serve 5 purposes:
- Open feedback shutdown. Output voltage would rise uncontrollably otherwise.
- Input under/overvoltage shutdown. Undervoltage prevents overheating at full load.
- Establishing the limits of operation of bias supply (like duty cycle range or input/output voltage range), as run out of bias supply is a way of protective shutdown.
- Tailoring of ON/OFF times in low load burst mode.
- Latching over-current shutdown.
Extra parts in primary side of a simple "plain switcher IC" SMPS can serve 5 purposes:
- Open feedback shutdown. Output voltage would rise uncontrollably otherwise.
- Input under/overvoltage shutdown. Undervoltage prevents overheating at full load.
- Establishing the limits of operation of bias supply (like duty cycle range or input/output voltage range), as run out of bias supply is a way of protective shutdown.
- Tailoring of ON/OFF times in low load burst mode.
- Latching over-current shutdown.
Now I need help, After taking pictures my 200asc is completely dead, the led goes nicely from red to green so it should be OK but there is no sound. I thought these where pretty sturdy modules but now I have three of them dead without any cause or visible damage.
Who can steer me in the right direction to getting these fixed!
Who can steer me in the right direction to getting these fixed!
I've been very busy with other stuff and haven't had a chance to look any further into my problem although as to Eva's question it just started operating correctly without rhyme nor reason. I still have to find out why.
Hoka you haven't really provided enough information towards understanding where the problems lie with your modules.
Hoka you haven't really provided enough information towards understanding where the problems lie with your modules.