Exact match. Not showing close matches.
'[EE] an astonishingly high no-load loss'
|part 1 2295 bytes content-type:text/plain; charset="gb2312" (decoded quoted-printable)
I know it's hardly a hardware guys' forum, forgive me for being desperate.
I designed a non-power factor corrected flyback converter type battery charger before, it works not bad, about 80% efficiency, 90-240VAC in, 24VDC, 2.5A out. 0.8W no-load loss.
Recently I added a PFC boost stage in the front of the flyback converter with the same transformer, the new charger also works, gives the same load, power factor as high as 0.99, total efficiency 78%.
Its no-load behavior is sort of weird, the no-load loss is about 0.8W at first.then after a few seconds, it's oscillating between 0.2W to a few watts, I cannot explain, but still not harmful since I don't need to meet the energy star standard.
Then I designed a new transformer with higher primary turns, ( to cut down on core loss). So with the old transformer, the circuit works only at 17% duty cycle, while with the new one, the duty cycle can be as high as 50%. What strange is: both the transformer winding and flyback FET temperature are lowered by 20-30C, but the total charger efficiency is still similar. That's unknown #1.
The only changes I made to the board other than the transformer are: change current sense resistor values, add a low pass filter at the primary feedback path (since with higher primary inductance, the feedback voltage spike is higher, thus it will force the charger into overvoltage protection), also a higher voltage FET was used.
The no-load behavior is totally weird this time. The no-load loss is as high as 10W! If it sees load within this period then the charger will work as normal. However, if after a few seconds without seeing the load, the input power can suddenly shoot to 40-50W, which you may assume the transformer soon gets saturated and adding load to it won't pull it back.
Then if you turn off the input and turn on again, the charger repeats this process, it's not dead at all, it always work if a load is applied to the output in time.
I was confused a lot, how to explain that?
I'm using a UC2843B chip from Onsemi for the flyback stage, with only a primary feedback.
part 2 35 bytes content-type:text/plain; charset="us-ascii"
2009/4/17 gardenyu <hotmail.com>:gardenyu2004
> I know it's hardly a hardware guys' forum, forgive me for being desperate.
Why? There are a lot of hardware engineers here even though some
may not deal with the power supply design you are doing. I myself
do not do much work on off-line power supply but I am a hardware
design engineer doing a lot of work related to DC/DC, MCU, analog, etc.
> I was confused a lot, how to explain that?
> I'm using a UC2843B chip from Onsemi for the flyback stage,
> with only a primary feedback.
Nothing can beat a schematic. It is very difficult to have a discussion
without schematics and waveforms for hardware engineers.
To me this sounds a lot like your feedback loops from the pfc and from the main
switcher are locking into each other and setting up some kind of oscillation
when there is no load. One way to have power being consumed and 'disappear' (no
heat), is by transmitting it as RF elsewhere. Radiating 2-3 Watts at HF from a
usual size PSU brick using its wiring as antennas is no problem. Try to use a RF
field strength meter near the PSU when it consumes a few Watts in idle and does
not heat up. Also suspect that the RF can cause the digital meters you are using
for instrumentation to read unusually high or lie altogether.
I think that I posted here about the tiny oscillator with the badly soldered
decoupling cap that caused the bench PSU's digital meters to display ridiculous
> To me this sounds a lot like your feedback loops from the pfc and
> from the main switcher are locking into each other and setting up
> some kind of oscillation when there is no load. One way to have power
> being consumed and 'disappear' (no heat), is by transmitting it as RF
He's loosing a few 10s of watts if I recall correctly, which seems very very
unlikely to be radiated.
> Radiating 2-3 Watts at HF from a usual size PSU brick
> using its wiring as antennas is no problem.
Oh? Let's do a little math instead of hand waving. What may be "high"
frequency for a power supply is relatively low for RF. Let's be generous
and say this thing is a power oscillator at 1MHz. The power supply may have
some harmonics higher than this, but not ones that produce 10s of watts at
no load. The wavelength at 1MHz is 300m. That's a lot for a small board
that sits on your desk. That means that any stray wiring isn't going to
make for a good antenna. Put another way, the impedence of the wires acting
as a antenna at 1MHz is going to be quite high. Now consider the voltage
that would have to go into this antenna to radiate even just 10W. Being
very generous and saying the antenna impedence will be as low a 1Kohm, it
would take 100V to radiate 10W. If there was really a 100V RMS signal at
1MHz, I think he'd know about it.
I originally didn't comment because the OP did a bunch of hand waving
without a schematic. I can't say where the power is going, but a far more
likely answer with the little information we have is that the voltage and
current from the input are out of phase. This causes losses due to the
current, but not as much heat as looking at the voltage or current in
isolation would lead you to believe.
Embed Inc, Littleton Massachusetts, http://www.embedinc.com/products
(978) 742-9014. Gold level PIC consultants since 2000.
Olin Lathrop <olin_piclist <at> embedinc.com> writes:
> Peter wrote:
> > To me this sounds a lot like your feedback loops from the pfc and
> > from the main switcher are locking into each other and setting up
> > some kind of oscillation when there is no load. One way to have power
> > being consumed and 'disappear' (no heat), is by transmitting it as RF
> > elsewhere.
> He's loosing a few 10s of watts if I recall correctly, which seems very very
> unlikely to be radiated.
True, but it mostly goes into the attached instrumentation and supplies, not out
in the air. Also, the instruments are not to be trusted if they 'see' even a few
hundreds of mW of HF in the leads. There are very few digital instruments that
are not disturbed by RF/HF injection, and most benchtop supplies with digital
readout will display total BS when subjected to unfiltered switching noise from
the load. Reading high in this case is rather normal, but some instruments read
low or even negative. Estimating input power based on such readings is not a
good idea. Some good lab supplies do have a RF warning light on them, though, so
the problem is known.
After wasting a number of hours on related problems I did bother to do some
informal testing with a 10W RF test generator and all the meters and testers I
had in the lab at the time. At about 2W output (the instruments were on the
wooden floor, in a square series tuned wire loop loaded with a 50R load) *none*
of them was reading right. The only one that was somewhat ok was an old
d'Arsonval multimeter, but that showed some interesting effects too, likely due
to the moving coil protection diodes. FWIW even a usually reliable Fluke meter
beeped desperately and showed high volts on an Ohms(!) continuity scale when
subjected to this test. At least it did show that something was not ok. Later I
got used to the Fluke (not connected to anything on the table) beeping off when
the switching circuits I was working on let loose some 'leakage', and used that
as a reasonable reliable warning that odd things are happening ...
You have to admit that losing a few 10s of Watts is *bound* to heat up something
rather quickly, no matter what. Given the power ratings of today's miniaturized
parts it would cause some kind of noticeable problem rather soon, if the leakage
would be real.
More... (looser matching)
- Last day of these posts
- In 2009
, 2010 only
- New search...