I asked this in the power supply section and didn't get any response, so I thought I would ask here. I'm looking to build a 3 channel 200 watt per channel amp, and wanted to design a nice quiet power supply for it. Typically in such a design people seem to only use a standerd CCC power supply, with somewhere between 20,000 and 60,000uf of total capacitance. I modeled with PSUD2 what the 20,000uf per rail supply looked like with a series resistor, 4700uf parallel, series resistor, 4700ud parallel capacitor. This is per rail again. Compared with even a 100,000uf power supply, this design seemed to offer much lower ripply. I can get away with the voltage loss no problem as I plan to use a 60-0-60v transformer, instead of the 50-0-50 the design asks for, using two 250-300ohm resistors will let me take it down to around 68-70 volts, which is fine. I can even go to 500 ohms and get it down to closer to 65 volts. Anyway, I have asked the people at Aussie amps and Amps lab about using their modules in such a setup, and they are yet to answer me about that specificly, they keep sayiing that the power supplies they offer is adequate.
I'm even wondering if in general, can you greatly reduce the noise of a power supply by added a series resistor followed by a smallish capacitor. For instance to modify the power supply in my solid state amp, or in my tube amp. I also noticed that reducing the esr of the caps reduces ripple, so I was wondering if bypassing the power supply caps with film caps can help reduce esr a bit, I'm guessing mostly only in the high frequencies of the supply. Given that I'm trying to reduce ripply at 60hz and its harmonics, does that mean that film caps wont do anything useful in reducing ripple where I need to?
I'm even wondering if in general, can you greatly reduce the noise of a power supply by added a series resistor followed by a smallish capacitor. For instance to modify the power supply in my solid state amp, or in my tube amp. I also noticed that reducing the esr of the caps reduces ripple, so I was wondering if bypassing the power supply caps with film caps can help reduce esr a bit, I'm guessing mostly only in the high frequencies of the supply. Given that I'm trying to reduce ripply at 60hz and its harmonics, does that mean that film caps wont do anything useful in reducing ripple where I need to?
3 channels = left, right and sub, right? (I'm building a similar setup as well, but 150 W. per channel).
Of interest: http://www.vicr.com/products/configurable/flat_pac/ ... note that these industrial switching (take two flat paks for + 50 VDC and - 50 VDC rails w/ ground being the series connection between 'em.) I take the edge off of these as the residual switching noise is quite low ... by feeding the smps outputs through a "regular" filter bridge board ... the DC output passes directly through the diode bridge and the 10,000 uF caps on the filter board cut the noise down to a very acceptable level = 100 db down or more. Anyway these run very cool and there is no resistor heat to deal with ...
Of interest: http://www.vicr.com/products/configurable/flat_pac/ ... note that these industrial switching (take two flat paks for + 50 VDC and - 50 VDC rails w/ ground being the series connection between 'em.) I take the edge off of these as the residual switching noise is quite low ... by feeding the smps outputs through a "regular" filter bridge board ... the DC output passes directly through the diode bridge and the 10,000 uF caps on the filter board cut the noise down to a very acceptable level = 100 db down or more. Anyway these run very cool and there is no resistor heat to deal with ...
I appreciate the feedback, but I wasn't really looking at doing a switching power supplu, and that setup is quite a bit more expensive than I was looking at. Even going with a 2400VA transformer doesn't put me into the cost of those units.
I figured I could mount the aluminum housed 50watt power resistors either on the heatsinks, or I have a smallish heatsink that would probably work fine. Also a block of aluminum on the base of the amp might be another way to deal with this.
I figured I could mount the aluminum housed 50watt power resistors either on the heatsinks, or I have a smallish heatsink that would probably work fine. Also a block of aluminum on the base of the amp might be another way to deal with this.
adding series resistors, even though it reduces ripple, will reduce the current available from the power supply. if you have an amp design with good PSRR (power supply rejection ratio), a little ripple shouldn't be a problem. the idea of using a bipolar switching supply is a good solution, and in some cases can be less expensive than a transformer supply. if you are handy at winding toroids, you could try to find an old PC supply that uses toroids, and wind your own secondary for your 60v rails. you will need to use series strings of schottky rectifiers to get the required PIV (peak inverse voltage) ratings for the rectifier, which should be at least twice the rail voltage. most schottky rectifiers hav a PIV of 40V or less. the PC supply should also give you +/-12V as well, so that can be used for op amps and such. make sure you terminate the +5 rail with a 200 ohm resistor, because many of those switching supplies won't operate without a load on the +5V rail.
pjpoes said:
CRC supplies may work great for preamps, but the series resistance you're talking about is going to kill you in a power amp. You're talking about 200W x 3 into 8 Ohms. Thats 5Arms per channel, or 15Arms total. 15Arms across 500 Ohms sounds like you'd need 7500 Volts, and you'd be dissipating ~56kW per resistor. Are you sure you don't mean 250 milli-Ohm? Even then, you're still wasting ~56W per resistor, 112W total per rail. If you're that concerned about ripple, go with a high-current regulated supply, and you can probably get away with wasting only 45-60W (3-4V dropout * 15A) per rail. And with a 60-0-60 transformer, I'm seeing ~+/-61V max, and not the 68-70 you're talking about.
Just my $0.02
--Greg
Hi,
a normal supply is RC, a single pole filter.
You are proposing a cascaded two pole filter RCRC.
The current to supply the amp is two parts the DC and the pulse current.
The DC comes from the first C and is kept recharged by the transformer.
The ClassAB pulse current comes from the last C.
In the single pole, the single C does both duties.
In the RCRC, the ClassAB pulse currents are fundamental to the correct operation of the amplifier.
The first C tries to turn the charging pulses into DC. It suffers large ripple currents and must be selected for this duty. An expensive high ripple current cap can be used or a bank of small cheap normal ripple duty caps in parallel, They have little influence on sound quality.
The last C must supply all the pulse current and must be of a similar size as used in an RC supply.
For 8ohm duty that is about +-10mF to +-25mF. For a three channel amp the PSU sees three 8ohm loads in parallel.
The last C should be in the range +-30mF to +-75mF. They have the most influence on sound quality and some benefit may come from adding substantial plastic film caps for fast response.
If you use an R that drops significant voltage then you need higher voltage capacitors. This is due to the voltage rise that gets stored in the caps when current demand is low. A 60Vac transformer when fed maximum mains voltage and when regulation is also high will push over 80V into the caps. You will need lots of 100V caps that are going to cost plenty.
a normal supply is RC, a single pole filter.
You are proposing a cascaded two pole filter RCRC.
The current to supply the amp is two parts the DC and the pulse current.
The DC comes from the first C and is kept recharged by the transformer.
The ClassAB pulse current comes from the last C.
In the single pole, the single C does both duties.
In the RCRC, the ClassAB pulse currents are fundamental to the correct operation of the amplifier.
The first C tries to turn the charging pulses into DC. It suffers large ripple currents and must be selected for this duty. An expensive high ripple current cap can be used or a bank of small cheap normal ripple duty caps in parallel, They have little influence on sound quality.
The last C must supply all the pulse current and must be of a similar size as used in an RC supply.
For 8ohm duty that is about +-10mF to +-25mF. For a three channel amp the PSU sees three 8ohm loads in parallel.
The last C should be in the range +-30mF to +-75mF. They have the most influence on sound quality and some benefit may come from adding substantial plastic film caps for fast response.
If you use an R that drops significant voltage then you need higher voltage capacitors. This is due to the voltage rise that gets stored in the caps when current demand is low. A 60Vac transformer when fed maximum mains voltage and when regulation is also high will push over 80V into the caps. You will need lots of 100V caps that are going to cost plenty.
In other words this is an expensive idea with little benefit.
Is the reason CRC's are so popular in Class A amps because the current draw is fixed, instead of variable?
There is also the design on Rod Elliot's site for a capitive multiplying cuircit, however he also talks about issues with such a supply on a Class AB amp, requirering a good sized set of caps. I guess I hadn't planned on that, but maybe a reverse of my design, and only one series resistor would fix that.
So for instance, a 4700uf cap, then a series resistor, and then a 30-60,000uf set. That also shows very low ripple, but I understand the issues with the resistor dissipating a lot of current. I didn't realize that it would be that high, but in PSUD2 I didn't put in a modifier for the value, just 250, so I assumed it was 250ohms. Milliohm doesn't seem to even be a reasonable value for large power resistors, unless I'm confused.
Is the reason CRC's are so popular in Class A amps because the current draw is fixed, instead of variable?
There is also the design on Rod Elliot's site for a capitive multiplying cuircit, however he also talks about issues with such a supply on a Class AB amp, requirering a good sized set of caps. I guess I hadn't planned on that, but maybe a reverse of my design, and only one series resistor would fix that.
So for instance, a 4700uf cap, then a series resistor, and then a 30-60,000uf set. That also shows very low ripple, but I understand the issues with the resistor dissipating a lot of current. I didn't realize that it would be that high, but in PSUD2 I didn't put in a modifier for the value, just 250, so I assumed it was 250ohms. Milliohm doesn't seem to even be a reasonable value for large power resistors, unless I'm confused.
most ClassA amplifiers do not have a fixed current draw when delivering output current. The rail currents follow the output current.
The need to reduce ripple is in part due to hum appearing on the output when there is no signal present and this is due to the larger quiescent draw on the smoothing caps. Even high bias ClassAB suffers from this.
RCRC and RCLC is very common in valve designs due to the much lower ciurrents involved.
better than using resistors would be series chokes. you would get more current and less ripple. a "pi" filter would give you less ripple and more current under load than using resistors. remember that for 60V rails, you need at least 15A current for a solid state amp, and a resistor will severely limit your current. even inductors will (because of internal dc resistance) limit your current somewhat. the best solution is to use a high current voltage regulator, the best for power amp use being a series pass MOSFET driven by an op amp and zener. this will give you a somewhat stiff supply rail with almost nonexistant ripple. the only problem with a stiff supply rail is that you have no dynamic headroom. definitely stay away from the resistors..... at 15A a 100 ohm resistor would be dropping 1500V which is impossible with 60V rails(in other words, you would have no rail voltage once you began drawing current). solid state amps require a completely different approach to power supply design. when you test your power supply designs with SPICE, you need to have an 8 or 4 ohm load on the output, because you need to see what it does when the amp is driving a load. with tubes, 500 ohm resistors in the power supply are OK, because you are only supplying a few hundred milliamps, but with solid state you need multiple amperes, and even a few tenths of an ohm is going to adversely effect the power supply performance.
Thanks for the input, I appreciate it. I had considered chokes, but they are very expensive. In order to get one large enough to be effective, and to also handle the current. How do you pick a choke for such an application, are we talking about custom made only? Seemed like most of them were only for tube amps and thus rated at a few hundred milliamps. Do they have to dissipate the full load of the amp or no?
can you give me more information on the Pi filter, I believe it would be a CRC, correct, however how do I determine the size and value of the choke. How large a capacitor is needed to handle the current, etc. I assume a larger capacitor should go after the choke in this situation? I am starting to wonder if it would be cheaper to try a capacitor multiplying cuircit as suggested on the ESP labs site, given how cheap transistors are, and how simple it could be. Anyone have any thoughts on the positive and negative of the plausible methods. I have also considered saving my money on such extravigances as chokes, and simply using better capacitors such as the Jensen 4 poles, but still question the difference they will make over a better power supply.
If I understand correctly, you are referring to a series resistor between the first cap and the second, to isolate the pulse currents of the transformer from the pulse currents of the amplifier, which are different. Since these currents are large, I would think something on the order of 0.1 Ohms, no?
Hi pjpoes,
An old fashioned way of smoothing amplifier rails was to split the Cs insert a low value resistor selected to drop about 1V under quiescent conditions, then parallel with a power bridge to limit the voltage drop under load. As here.
Since then however, amplifiers have been designed for decent immunity to rail fluctuation and noise.
Cheers ........ Graham.
An old fashioned way of smoothing amplifier rails was to split the Cs insert a low value resistor selected to drop about 1V under quiescent conditions, then parallel with a power bridge to limit the voltage drop under load. As here.
Since then however, amplifiers have been designed for decent immunity to rail fluctuation and noise.
Cheers ........ Graham.
Attachments
Graham & pjpoes: That's a very neat trick ... diode forward voltage is quite predictable.
Silicon diodes = about 0.6 volts (typical of 1N4001)
... X 2 in series = 1.2 volts +/- 0.05 volts
... X 4 in series = 2.4 volts +/- 0.10 volts
... dependant on current.
Germanium diodes = about 0.2 volts, likewise as above.
Schottky diodes = about 0.4 volts, but these have upper limits on reverse voltage & current (breakdown).
"Fast recovery" diodes = about 0.4 to 0.5 volts, depending on ...
Anyway, this is often used to get solar cell voltage regulation under control = double or quad or more, parallel diodes = high current, low voltage drop, etc. Diode pop corn noise can be greatly reduced with an inductor in parallel with or in place of the low value resistance, creating an avalanche effect = sudden release of electrons across the diodes rather than a trickle at the transition on the current curve = "reverse spring loading" the diode threshold as it were ... I love the hydraulic model of electron flow ...
Silicon diodes = about 0.6 volts (typical of 1N4001)
... X 2 in series = 1.2 volts +/- 0.05 volts
... X 4 in series = 2.4 volts +/- 0.10 volts
... dependant on current.
Germanium diodes = about 0.2 volts, likewise as above.
Schottky diodes = about 0.4 volts, but these have upper limits on reverse voltage & current (breakdown).
"Fast recovery" diodes = about 0.4 to 0.5 volts, depending on ...
Anyway, this is often used to get solar cell voltage regulation under control = double or quad or more, parallel diodes = high current, low voltage drop, etc. Diode pop corn noise can be greatly reduced with an inductor in parallel with or in place of the low value resistance, creating an avalanche effect = sudden release of electrons across the diodes rather than a trickle at the transition on the current curve = "reverse spring loading" the diode threshold as it were ... I love the hydraulic model of electron flow ...
Hi pooge,
You can use two single diodes in series, but a 35A bridge is a very convenient way of ensuring low voltage drop under load.
For instance with R= 2R2, second C= 22mF and quiescent current= 400mA, a 100Hz psu ripple can be reduced by 20dB !!!
The only sacrifice being two diode volt drops under full power.
Cheers ......... Graham
You can use two single diodes in series, but a 35A bridge is a very convenient way of ensuring low voltage drop under load.
For instance with R= 2R2, second C= 22mF and quiescent current= 400mA, a 100Hz psu ripple can be reduced by 20dB !!!
The only sacrifice being two diode volt drops under full power.
Cheers ......... Graham
The first album would be called, " Down a Few More db"...damn that's magical!
"Squishing Ripple" ... could be a new soft drink ... or a new milder detergent.
See the image = http://www.diyaudio.com/forums/attachment.php?s=&postid=1163885&stamp=1174470303 ... That "Low R" resistor could just as easily be a wire wound resistor = it has a slight inductance. Otherwise you might wind your own resistor / inductor and put 'er there.
See the image = http://www.diyaudio.com/forums/attachment.php?s=&postid=1163885&stamp=1174470303 ... That "Low R" resistor could just as easily be a wire wound resistor = it has a slight inductance. Otherwise you might wind your own resistor / inductor and put 'er there.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.