On July 25, 2015 09:01:45 PM Andrew Deryabin wrote:
> Hi Tim!
>
> 25.07.2015 00:41, Tim E. Real пишет:
> > On July 22, 2015 03:26:48 PM Andrew Deryabin wrote:
> >> I think it's over-voltage (power supply was also broken).
> >
> > Hiya!
> >
> > Hopefully it was not AC input over-voltage (storm surge etc.)
>
> It can be so, because before broken ATX power supply there was an PCM
> UPS which was also damaged. I took transformer from this UPS, connected
> it to AC Line in series with 100 watt lamp and got a shiny light :).
>
> > Otherwise if it is a regular box type power supply, and sometimes even
> >
> > server supplies, if you open the power supply cover there is a good
> > chance that you will find bulged capacitors - the number one failure of
> > power supplies everywhere.
> >
> > Although it is somewhat rare to have catastrophic over voltage /output/,
> >
> > I've seen it happen with ugly blackened charred circuit results.
> >
> > Usually as capacitors slowly go bad, a supply simply becomes weaker.
> >
> > But a supply typically uses one of the output lines (say the +5V line) to
> > tap>
> > voltage feedback which self adjusts the regulation circuits.
> >
> > Sometimes the capacitors on that line go bad such that they cause too
> >
> > much load on the line or else very weak filtering and high ripple
> > voltage.
> >
> > Then the regulation attempts to compensate by driving harder to put more
> >
> > voltage on that line to make it increase to the correct voltage (say
> > +5V).
> >
> > But in doing so, the supply unwittingly /increases/ all the other supply's
> >
> > output voltages beyond their rating, since all output voltages come from
> > the same output transformer.
> >
> > With catastrophic results.
>
> I know only basic principles of Impulse power supplies and this intro
> was rather interesting, thanks :).
>
> > Next supply you buy, if you want to trust it, ask a qualified technician
> >
> > (well, me for example), to check the brands of capacitors in there.
> >
> > If they are crappy brands such as Ltec, OST etc. do not trust it.
> > Ask the tech to install only trusted Japanese brands.
>
> I took a new 450 watt FSP from my work, so there was no choice at this
> time, but I hope that it will serve for long time (at least for 2-3 years).
> > My order of preference:
> > 1: Panasonic. The (my) absolute go-to number-one trusted brand, ever!
> > 2: Nichicon, United Chemicon etc. all very fine manufacturers.
>
> Opened this FSP. There are caps labeled 'CAPXON'. What can you say about
> this brand?
I would throw in Panasonic FK, or FM/FC series caps.
>
> P.S. offtopic. I like electronic circuits and have lots of questions.
> Tim, if you don't mind, can you answer one of them, it's very simple for
> you, I think:
>
> Some time ago I made a JLH 1969 class A amplifier and it sounds great.
I know a guy specializes in assembling/customizing amps like this
for customers.
> I tested it with siglent sdg 1020 signal generator and atten ads 1102
> oscilloscope before connecting to my yamaha ns-333 speakers. Results
> were very good - noise floor was as little as 5 millivolts. When I lean
> my ear directly to speaker, I hear nothing (volume regulator of my
> amplifier is always at 100%, because I use volume knob on my sound card).
> The main drawback is that I had to make air cooling for the amp because
> radiators for output transistors are rather small (I took them from
> processor fans). But I like silence - all my computers reside in store
> room, so in my working room there are only monitors, hubs, keyboards and
> audio equipment.
>
> Here is my question. Is it worth to make a class-D amplifier? I have
> TDA8920CJ chip and it has near 90% efficiency (comparable to 25% of my
> current amp), so this means, that I need only one small radiator without
> any fans. But what about sound quality?
Whew, that's a pretty high power chip.
This chip exceeds the distortion and other specs of the JLH:
"The [JLH] amplifier has reasonable specifications; nothing special a
$3 chip doesn’t wildly exceed"
[1]
If I understand the datasheet correctly, the HSOP24 version has
a heat sink area on the bottom, while the upright DBS23P's heatsink
area is on the back.
"• Connect the external heatsink to the ground plane
• Use a thermally conductive, electrically non-conductive, Sil-Pad between
the backside of the TDA8920C and a small external heatsink"
[2]
But usually I see a HSOP chip's heat pad soldered right to the copper area
on a circuit board. So it may be harder to affix to the board without the use
of hot air or IR soldering.
But... the datasheet doesn't appear to mention this.
So maybe simple heatsink paste or pad will do ...
[1] http://6moons.com/industryfeatures/zen/plh.pdf
[2] www.nxp.com/documents/data_sheet/TDA8920C.pdf
Follow the directions on that datasheet carefully, regarding
grounding and conductor length. And especially shielding.
There may be very noisy high frequency high voltage signals in there.
See if you can purchase a kit with a board.
I've been fascinated with class D (and PWM) for a long time.
Being digital PWM of course it relates easily to computers.
I once made a Windows 3.1 PC speaker audio driver.
Remember, the PC speaker is driven with 'square' waves (beeps).
Drive it with carefully crafted PWM and you have analog sound.
The speaker itself acts as the low pass filter.
Wanna amplify it further without electrical connections?
Place a 'telephone recording earpiece pickup coil' in front of the
PC speaker, and connect that to an external amp.
I 'know' a certain mass market brand's AV receivers.
They have a line completely devoted to class D output.
Beautiful, cool amps.
Compare with other makers' five/seven/nine channel class A amps:
You can fry an egg on some of them - in idle mode, with no sound !
Still, I admit there's something... unsettling about driving 200 watts
into a speaker with a freakin' square wave !
T.
------------------------------------------------------------------------------
_______________________________________________
Lmuse-developer mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/lmuse-developer
