Another PSU Question

[hongrn]

Hey All,

I want to build this unregulated PSU for a Class D amp I have in mind. I noticed there are no bleeders on the schematic, so should I add a couple of resistors in parallel before the bank of large caps (10,000uF)? If the bleeders are needed, what do you suggest in terms of values? Thanks.

Hong

[Wolf]

Re: Another PSU Question

The 333 ohms are bleeding already, Hong. No need to add them.

Personally, I would use 25-50k as a bleeder for a slower discharge rate.
Later,
Wolf

[hongrn]

Re: Another PSU Question

My inexperience. I thought bleeders are positioned right after the rectifier bridge. Thanks Wolf!

[Wolf]

Re: Another PSU Question

As long as the resistors are parallel, then it matters not location.
Later,
Wolf

[AJ]

Re: Another PSU Question

As Wolf mentioned, 333ohms is a really low value for a bleeder. Assuming a 50Vdc rail, that's 7.5W being dissipated continuously in that resistor as shown. Depending on the rail voltage, 2Kohm+ is more appropriate. The resistor is there solely so the caps will discharge for safety when the case is opened.

[hongrn]

Re: Another PSU Question

Thanks guys. I use 2k ohms in previous PSUs, so I'm glad I asked. Also, I want to use a NTE 5330 bridge rectifier instead of the 4 MUR860s in the schematic. Should I keep the four 100nF caps in parallel with the bridge legs to minimize switching noises? Thanks.

[jonpike]

Re: Another PSU Question

Now, if you want to get techier... (is that a word?) you could calculate how fast a resistor value will bleed down your supply by working out the time constant: RC. Take the total cap value of one supply (plus or minus, not both) in Farads, (i.e. 10,000uF = .010F) and multiply by the ohms of your bleeder resistor. That will equal the time it takes for the voltage to go down 63% of what it was. Also, you can see the power burned by the resistor with V^2/R.

You can see that it's a tradeoff between a smaller, faster discharging resistor that will need to be bigger and burn more excess power, and a larger value that will have less heat load but will take longer to bring down the caps. Typically, bleeder R's are spec'd to bring down the supply in 5-10min or so, if not longer, but it's really taste, and how much power you're willing to waste. (and money, to spend on high wattage resistors)

On the bridge rectifier... as long as it has similar voltage and current ratings, it should be a drop in with no problem.

[AJ]

Re: Another PSU Question

The RC time constant only tells part of the story, though. Even with no signal there will be quiescent current draw by the amplifier which will drain the caps quickly by itself. It's all academic though, as it only really comes down to how much power you want to waste. It's not a necessary item because any good tech will know to make sure the caps are discharged before sticking his hands inside the case.

[hongrn]

Re: Another PSU Question

On the other hand, it's very uncomfortable to see sparks flying when connecting or disconnecting from a PSU without bleeders...

The RC time constant only tells part of the story, though. Even with no signal there will be quiescent current draw by the amplifier which will drain the caps quickly by itself. It's all academic though, as it only really comes down to how much power you want to waste. It's not a necessary item because any good tech will know to make sure the caps are discharged before sticking his hands inside the case.

[SirNickity]

Re: Another PSU Question

I find a power LED on each rail is about right for me. I design them in for a power good indicator during testing anyway, having them pull duty as a bleeder (and bonus bleed status indicator!) is just icing on the cake.

Feeding the LED with maybe 10mA will bleed the caps down to the forward voltage of the LED, and at that point, I don't really mind if the last couple volts is parasitic only. If the LEDs are dark, you're safe.

[dthomas]

Re: Another PSU Question

On the other hand, it's very uncomfortable to see sparks flying when connecting or disconnecting from a PSU without bleeders...

333 ohms is to low at least from the standpoint of it wastes a lot of current that could be delivered to your speakers. It will bleed your caps faster. The rectifier change is fine. The MUR diodes are faster and quiter than the bridge you have but whether that equates to audible difference is hard to say. I have used cheap bridge rectifiers in well filtered supplies with very good results. I have used the MUR diodes also and I can't really say I heard an audible difference.

My luck with getting a quite low ripple supplies has always been to oversize the power transformer. There is no substitution for current delivery.

[lhwidget]

Re: Another PSU Question

I question the need for the capacitors across the bridge diodes. MUR's shouldn't ring the way big slow bridge diodes do, and shouldn't need snubbers at all.

Also, just adding a cap won't snub the spike on a regular bridge diode at all, it will usually make it worse. If snubbers are used on a regular bridge diode, a resistor should be placed in series with the cap to dissapate the energy.

Without doing the calculations, a 100 nF in series with a 20Ω 1/2W resistor across each diode in the bridge is a good place to start.

The attached PDF is a post about power supply design from the DIY Audio forum a few years back that really rang true.

[hongrn]

Re: Another PSU Question

Thanks Jay for the article. Now I'm totally confused about snubbers. There are pages and pages of the pros for snubbers, and pages and pages of cons. Jon Pike, I need that oscilloscope inservice asap my friend.

I question the need for the capacitors across the bridge diodes. MUR's shouldn't ring the way big slow bridge diodes do, and shouldn't need snubbers at all.

Also, just adding a cap won't snub the spike on a regular bridge diode at all, it will usually make it worse. If snubbers are used on a regular bridge diode, a resistor should be placed in series with the cap to dissapate the energy.

Without doing the calculations, a 100 nF in series with a 20Ω 1/2W resistor across each diode in the bridge is a good place to start.

The attached PDF is a post about power supply design from the DIY Audio forum a few years back that really rang true.

[SirNickity]

Re: Another PSU Question

Snubbers take a lot of effort to get right. Just tossing a cap across a diode can turn it into an oscillator. Correctly engineering a snubber network takes some effort, for questionable benefit. It's one of those things where there's usually more low-hanging fruit to pluck before you should be bothered taming noise in the rectifiers. There will always be those tweakers out there looking to squash each and every source of noise. Good for them. But it calls for a systemic approach, not a knee-jerk reaction to an idea someone posted on a forum once -- which is how a lot of DIY PSU builders begin looking into them.

As always, my perspective is this... Is there an audible problem? Yes? Find and solve it. No? Move on. For example, I'm currently rocking out to a home-made headphone amp. It takes a ~12V AC input from a wall wart I chose by randomly picking one from Digikey. I literally cannot tell it's on with no input connected, even with the volume up at max. Good enough for me.

[lhwidget]

Re: Another PSU Question

SirNickity has a good point. The real design articles are pretty deep. The web chatter presents a lot of fluff and some of it is just wrong.

The only case I can think of where you might get good payback, with no audible difference is if the high frequency noise (up around 400 - 500 kHz) excites the amp into oscillation. To be honest, I'm not sure if it's possible with the signal coming in on the poswer supply connections.

The guys with more experience than me will know if this is a worry or not.

Have fun with it