I would like to learn about this.
Right now, I have 3v extra and the only source for this is noise. Noise rectified by caps isn't strong enough to reinforce bass but will certainly do something awful to treble. Its those darned MUR860's causing a heck of a noise into my caps. Will a "shuntie" leave me with strong power as the only power source?
Please illustrate care and feeding of a "shuntie" as used in practical application? 😉
Last edited:
Well, the linear voltage regulators can be either series or parallel (or shunt regulators, nickname shuntie). Shunt regulators have much better RF behaviour and they keep a low impedance even at high frequency where the PSRR is lower and all the noise passes through the amplifier without any problems. In simple unregulated power supplies I woudn't use fast recovey diodes and low ESR capacitors, the diode ringing will cause high voltage peaks even with snubbed diodes.
One example would be this one:
http://www.diyaudio.com/forums/powe...hunt-w-ccs-vs-jsr03-05-other-series-regs.html
One example would be this one:
http://www.diyaudio.com/forums/powe...hunt-w-ccs-vs-jsr03-05-other-series-regs.html
voltage peaks. . . fast recovery diodes. . . Capacitors will charge up to the peak. Perhaps the problem has a simple solution. 😉
Thank you for the shuntie article.
Also, thank you for the confirmation about my suspicions on the diodes. I was told that MUR860's are the best thing since sliced bread. I guess that really loud diodes are one possible way to get extra dynamics, but that really is noise, isn't it?
Unfortunately it's only noise. The peaks I'm talking about are seen after the filtering and decoupling caps. The waveform after this should be a nice triangle like shape. Diode ringing will cause voltage bursts which appear over this and they're not filtered at all. You can reduce them, but the power supply gets complicated when it was alot easier and cheaper to use regular rectification diodes. Engineering = art of compromise.
Last edited:
Good idea! I like the 1n504 diode a lot (because of results), but, unfortunately, its not quite strong enough for this application. What are some seemly substitutes?
The diode bridges you showed some time ago are excellent. I think they were KBPC2506 or something like that.
EDIT: I just remembered a great article regarding power supplies. There you go:
http://www.calex.com/pdf/3power_impedance.pdf
EDIT: I just remembered a great article regarding power supplies. There you go:
http://www.calex.com/pdf/3power_impedance.pdf
Last edited:
I do have a KBPC2504 and a KBPC1506 handy, right here in the house. 😉 Anyway, which of the 1-piece bridges is best for this application?
Either of them. The only difference is the current and inverse voltage they withstand. If I remember correctly, KBPC2504 is rated at 25A and 400V and the KBPC1506 is rated at 15A and 600V. You can still google it though 😀
It would have been great for you if you had an oscilloscope to literally see the differences.
It would have been great for you if you had an oscilloscope to literally see the differences.
OH, I know. I need to check in somewhere for some training. Like a "boot camp" sort of thing. Then I might have some idea of what sort of scope to buy and perhaps how to use it.
Question:
I have, per each rail, at the power supply board:
diodes, 3300uF, 10,000uF, output jack
Would it be a lot better to stick a resistor in-between the 3300uF and 10,000uF?
The resistor won't help too much, but you can give it a try. Make sure it will cope with the current at full load (Pdis = R*I^2, R in ohm, I in A and Pdis in W). How do you feel now that you started to understand from where all the sound changes are coming from? 😉
Started? That's a question I can't answer. However long that I knew where the sound changes are located. . . seems to be less relevant than being empowered to make those changes on purpose. 🙂
Coping with the current? Oh easily. So, I spent extra to get a transformer with thick enough wire. A similar resistor would be a parallel pair of 10w (per each rail) in order to avoid installing extraneous inductance, unintentionally. Sure is overkill on power handling. Maybe it will be useful when someone connects a parallel LM3886 to 4 ohms speakers.
But, those maths were impenetrable, and I can't guess the figure for ohms.
Actually an inductor used instead of resistor will have way better beahviour. That math is very simple. Let's assume 5 amperes and a 2.2 ohm resistor. The power dissipation on the resistor would be 2.2 * 5^2 = 2.2 * 25 = 55W. That's kind of big, ins't it?
The filtering must reduce the 100/120Hz waveform resulted after the diode bridge. Depending on the desired ripple you have to design a low pass filter with a specific slope. RC filtering isn't as good as LC filtering. At low frequencies cutoff you need either big R or big C. C wouldn't be a problem, but after a limit it doesn't matter ho much you increse it. R it will also dissiapte lots of heat and it will dry out the caps in time. L on the other hand can be made pretty big while keeping a low parasitic resistance so the power dissipation on it it will be less than on the resistor. The downside is the cost, as copper is quite expensive nowadays.
Above that, even for 2Vpp ripple will get to the speakers as 0.1mVpp at a PSRR of 85dB. Let's assume that you have that ripple at your 13,300uF filtering. This means that the current drawn continuously from capacitors and transformer is 1.5A. On a 4 ohm load this translates in 9W output power, therefore the voltage swing on the load is 17Vpp. Comparing 0.1mVpp (generated by the 2Vpp ripple) with 17Vpp on the load (the actual signal) it is quite obvious that the difference is so small that you might barely hear it. On 8 ohm load, the things are changing even better. Output power is 18W,
voltage swing is 24Vpp.
Quite interesting, isn't it?
The filtering must reduce the 100/120Hz waveform resulted after the diode bridge. Depending on the desired ripple you have to design a low pass filter with a specific slope. RC filtering isn't as good as LC filtering. At low frequencies cutoff you need either big R or big C. C wouldn't be a problem, but after a limit it doesn't matter ho much you increse it. R it will also dissiapte lots of heat and it will dry out the caps in time. L on the other hand can be made pretty big while keeping a low parasitic resistance so the power dissipation on it it will be less than on the resistor. The downside is the cost, as copper is quite expensive nowadays.
Above that, even for 2Vpp ripple will get to the speakers as 0.1mVpp at a PSRR of 85dB. Let's assume that you have that ripple at your 13,300uF filtering. This means that the current drawn continuously from capacitors and transformer is 1.5A. On a 4 ohm load this translates in 9W output power, therefore the voltage swing on the load is 17Vpp. Comparing 0.1mVpp (generated by the 2Vpp ripple) with 17Vpp on the load (the actual signal) it is quite obvious that the difference is so small that you might barely hear it. On 8 ohm load, the things are changing even better. Output power is 18W,
voltage swing is 24Vpp.
Quite interesting, isn't it?
Oh, its very interesting indeed!
This time, the point isn't if I can hear it.
The amplifier can hear it, and then make more heat.
4 ohm speakers, cut-down gain, and solo LM3886. Oh, this must have clean power.
Too little voltage and its own spike circuit clipping will overheat it, too much voltage and it'll blow one output transistor because of the 4 ohm load, too noisy voltage and it will overheat with any load regardless of the amount of voltage. Nice Job NatSemi!!! This will make an excellent high-efficiency space heater.
So clean power, with a shuntie, uses enough parts to almost build an entire discrete amplifier. I'm dizzy now.
Well, I'm going to go have another look at Mark Houston's amp and see what there is to see.
edit: Aha! It wasn't the baking pan. Its two whole dc power supplies cabled separately to just one bridge rectifier--a "dark voice" treatment. Those tube amp guys know stuff. 🙂
Last edited:
It's alot harder to work with tubes because the majority uses class A and the PSRR is way low when compared even with an ordinary chip. And yet the results are surprisingly good with a normal CRC filtering and voltage divider. The ripple from the power supply has nothing to do with chip's temperature. Ripple frequency is also low. 50 or 60Hz mains will result in 100 or 120Hz ripple.
Something simple then. . .
Sure. For the chip, where MUR860 = heat, then the solution is plain--just swap those for something more ordinary.
Before that, I'll try a 3.3nF polyester at the transformer's ac output (aka at the bridge rectifier's ac input) and look for a temp drop/volt drop. When that works, its really obvious and when it doesn't work, that's also easy to spot. 😉
P.S. Just guessed my way to fixing a zyxel router's power circuit. She's holding steady. 😉 If only audio were that easy. 😉
Last edited:
Diode ringing is reaching MHz area and RF interferences caused by this are not that easy to deal with. I really really doubt that you can do something without an oscilloscope. Now for the router, that's just digital stuff, they tend to be immune to noise.
Well, the timeframe for doing it with no scope could be quite long; therefore, do you have any power supply design recommends? Something nice and simple perhaps?
Oh! I just realized that my question was just like: "Can you make a crossover for a woofer without using the woofer's specs?" 😉
Still, why not ask and see what happens? 😀
Hmm. The 3.3nf cap increased the voltage.
Now the power supply converts 41vac to 63.5vdc No kidding.
It was converting 41vac to 61vdc.
Its not too much voltage, its just that peaks are such a weak source of power that the voltage isn't steady. I think, either remove the MUR860's or add more 10,000uF caps for stability. Is that about right?
Now the power supply converts 41vac to 63.5vdc No kidding.
It was converting 41vac to 61vdc.
Its not too much voltage, its just that peaks are such a weak source of power that the voltage isn't steady. I think, either remove the MUR860's or add more 10,000uF caps for stability. Is that about right?
Well... it's trial and error without measurement equipment, so test and choose what you like the best.
Need to drop the rails with something other than the amplifier.
I've successfully made lots of DC, but about 3v of it is at insufficient amperage, and I need something to eat that up. I think? Everything I've tried for filter makes yet more DC at even less amperage capacity. omg! Blunder! 😉
I've successfully made lots of DC, but about 3v of it is at insufficient amperage, and I need something to eat that up. I think? Everything I've tried for filter makes yet more DC at even less amperage capacity. omg! Blunder! 😉
Last edited:
- Status
- Not open for further replies.