xman737 said:
The main advantage to a SMPS is the cost for mass production.
This is why Philips would be using SMPS'.
J-P said around a million UCD's were made.
Let's say that amount of amps fit into 100 000 cases and has 100 000 power supplies.
Let's say since they're all OEM they're all running off a single transformer (cheap as possible) and additionaly it's underated by the usual 70%.
What would each one weigh? I'm going to guess 10lbs.
What would shipping cost for a million pound transformer. Forget shipping, how much would it cost to buy it? They're a _very_ expensive component.
It costs them more to design it as it's more complex, but they find it well worth their while in the long run, as each supply winds up costing mere pennies.
The fact it is lightweight and smaller are what's told to consumers of the products, to some those are strong selling points, but there's another side to the story as well. For instance, no matter how good one can be made (sonically), it will always be far less reliable than a typical unregulated raw DC supply, and typically they're made so cheap that they aren't very good sonically either.
To answer to what I've quoted:
Caps too small? I wouldn't try to get the max out of the modules with those caps put it that way. I'm sure they're fine with the given transformer, but I'd be really careful trying to use my own with it, because even though the new voltage rating of the modules is 67V, you still can't have it exceed 63V without those caps going POOF.
Jan-Peter, a reliable /robust soft start would be a strong selling point, lots of people are nervous about those and can't find a reliable one to buy, and dont' want to make their own, or dont' know how to make a good one.
Not having one is forcing them to buy from your competitors, likely made with some other product in mind, they cross their fingers and hope it works. I didn't see it mentioned with the other features, have you any plans for a seperate soft start module or anything?
Would you consider releasing the designs of your supplies to further help out those who'd rather build their own, surely that's not confidential? For instance they might like to use your DC protection with their own super expensive dual bridge rectifiers etc. or just to have their own custom one which fits their needs?
Would it be feasible to have a sort of "watchdog" circuit sans supply, that just takes charge of start up/shutdown events? For example, soft start with a power cycle delay, DC protection, country music protection (just explode it), overtemp (optional), and maybe a remote turn on as well?
Thanks
ackcheng said:
I wanted to try a regulated supply since I want to feed 4 UcDs from one supply and want to avoid cross contamination via the supply rails as much as possible. It is not really small and not really cheap. The kit cost $329 and it is about 24x18cm and has a height of about 5.5cm. The supply is supposed to be able to deliver 1000W and 2000W peak, so it is a bit overspecified for my application, but I don`t mind that of course. I considered a conventional regulated supply using the THEL series regulators. These are also expensive and not easy to order from here. That`s why I took the "easy" way to order these SMPS kits in the US. Besides that, the THEL series regulators would burn more power, would need more heat sinking etc.
The 100Hz ripple had not much harmonics, so the supply output waveform may actually look better than that of a conventional supply (I know everybody will kill me for this remark). By optimizing the feedback network, I`m pretty sure I can get the 100Hz ripple down.
Besides that, I may use a capacitance multiplier like circuit after the SMPS to get rid of the 100Hz and 75kHz ripple. I have the white noise active supply already assembled at home. Just want to do this all nicely step by step since this is my first exercise with SMPS. I know this may not be a solution that other people prefer, I just have the stuff at home and may use it in this combination. Could use the active supplies also with a conventional transformer setup if the conclusion of the whole story is that SMPS is not the way to go.
Why is an SMPS by default bad??? A beautifully sounding UcD module uses the same switching principle. If it is good for an amp, why would it be by definition bad for a power supply?
Anyway, everybody is free to think what he thinks, I just want to experiment with this and get the experience/info first hand. I will admit it if it sounds bad or if I can`t get it working the way I want.
Best regards
Gertjan
Hi ghemink,
I wish you the best of luck with your SMPS. Your example will help blaze the trail others will surely follow once the way has been made clear.
I have designed SMPS's for use in pro audio products and found common mode noise much more problematic than a little bit of ripple or even normal mode spikes across the dc output terminals. Of secondary concern when sharing a common supply among outputs is cross talk via the common ground, but you shouldn't have audio grounding problems if you make wise use of the UcD's fully differential inputs.
For a DIYer the easiest thing to block common mode noise is to twist each of the UcD's power supply cables together as a separate triple and run each bundle several turns through its own ferrite toroid core. Getting core size, shape and material slightly wrong shouldn't matter as any will be orders of magnitude better than none. (This is something that also should be done on the UcD's outputs unless the UcDs are right in the speaker box, IMO.)
A slightly more difficult DIY mod would be to add small (1nF to 10nF) common mode bypass capacitors directly from each pin of the power supply output connector to power supply chassis. This could be done via a small bar of copper to which all the caps are soldered in a row. The bar must then be fastened to chassis on both ends in order to make a good RF ground.
By the way, common mode noise is insidious because it is phase aligned impulse noise (i.e. spikes) that get unevenly rectified by the high gain, high impedance inputs of the audio gear. One digital pro preamp product for which I designed the SMPS actually passed all the agency and product design specs for noise floor and EMI at the first pass prototype, but didn't pass the worst case listening test (all inputs cabled up but with no signal and gain set to max). Any beat signals or other artifacts that could be picked out from behind the full gain hiss was a failure (the next iteration passed).
Regards -- analog(spiceman)
I wish you the best of luck with your SMPS. Your example will help blaze the trail others will surely follow once the way has been made clear.
I have designed SMPS's for use in pro audio products and found common mode noise much more problematic than a little bit of ripple or even normal mode spikes across the dc output terminals. Of secondary concern when sharing a common supply among outputs is cross talk via the common ground, but you shouldn't have audio grounding problems if you make wise use of the UcD's fully differential inputs.
For a DIYer the easiest thing to block common mode noise is to twist each of the UcD's power supply cables together as a separate triple and run each bundle several turns through its own ferrite toroid core. Getting core size, shape and material slightly wrong shouldn't matter as any will be orders of magnitude better than none. (This is something that also should be done on the UcD's outputs unless the UcDs are right in the speaker box, IMO.)
A slightly more difficult DIY mod would be to add small (1nF to 10nF) common mode bypass capacitors directly from each pin of the power supply output connector to power supply chassis. This could be done via a small bar of copper to which all the caps are soldered in a row. The bar must then be fastened to chassis on both ends in order to make a good RF ground.
By the way, common mode noise is insidious because it is phase aligned impulse noise (i.e. spikes) that get unevenly rectified by the high gain, high impedance inputs of the audio gear. One digital pro preamp product for which I designed the SMPS actually passed all the agency and product design specs for noise floor and EMI at the first pass prototype, but didn't pass the worst case listening test (all inputs cabled up but with no signal and gain set to max). Any beat signals or other artifacts that could be picked out from behind the full gain hiss was a failure (the next iteration passed).
Regards -- analog(spiceman)
analogspiceman said:
ackcheng said:
Thanks for the support. I'll continue my experiments with the SMPS. Even if it will in the end not be used, it would be an interesting learning experience. I take utmost care that I stay away from the high voltage parts on the board, make sure the high voltage caps are decharged etc. I think experimenting with an SMPS can be quite dangerous if you don't know what you are doing with those high voltage. I have a PC based scope that I hookup on a notebook, that for the occasion works on batery power (so isolated from the mains) and I use a wireless mouse to control it. So I'm not going to be electrocuted via the scope for sure
Best regards
Gertjan
Gertjan et al,
A SMPS is no magical cure, it is plagued by a whole new set of issues. Common operating frequencies of a SMPS are in the 10...20KHz range which puts their output signal well into the audible realm, more audible than the 50Hz hum of a conventional PSU.
But adding to that is the switch noise of the electronics and the fact that you'll need output filtering to prevent the carrier frequency from entering the amplifier. Power reserve is another issue as well, most SMPS are meant to operate at a constant load, the use of it in an amplifier is far from that. Buffering, both primary and secondary, will be key and will need to be matched to the maximum power drain.
All in all my opinion is that a SMPS is not worth the trouble, only when you have hands-on experience with designing SMPS and a good grasp of what works and what doesn't and access to development tools and measurement equipment you'll be able to make this work properly.
There's a reason why you don't see much high-powered SMPS designs in commercial amplifiers, for many of the reasons described above, and adding to that is reliability.
One of my personal mantras is 'keep it simple stupid', which is clearly applicable here.
Best regards,
Sander Sassen
http://www.hardwareanalysis.com
A SMPS is no magical cure, it is plagued by a whole new set of issues. Common operating frequencies of a SMPS are in the 10...20KHz range which puts their output signal well into the audible realm, more audible than the 50Hz hum of a conventional PSU.
But adding to that is the switch noise of the electronics and the fact that you'll need output filtering to prevent the carrier frequency from entering the amplifier. Power reserve is another issue as well, most SMPS are meant to operate at a constant load, the use of it in an amplifier is far from that. Buffering, both primary and secondary, will be key and will need to be matched to the maximum power drain.
All in all my opinion is that a SMPS is not worth the trouble, only when you have hands-on experience with designing SMPS and a good grasp of what works and what doesn't and access to development tools and measurement equipment you'll be able to make this work properly.
There's a reason why you don't see much high-powered SMPS designs in commercial amplifiers, for many of the reasons described above, and adding to that is reliability.
One of my personal mantras is 'keep it simple stupid', which is clearly applicable here.
Best regards,
Sander Sassen
http://www.hardwareanalysis.com
AC regeneration
While we are at the subject of power supply, I recently come across some articles on AC regeneration. The theory is that the street electricity is noisy and to clean it up, the best way is to convert it to DC and then regenerate AC out of it.
They use a Wein bridge to generate an oscillation of 50Hz and then an opamp for amplification. I wonder if we can use UCD400 for the amplication part?
While we are at the subject of power supply, I recently come across some articles on AC regeneration. The theory is that the street electricity is noisy and to clean it up, the best way is to convert it to DC and then regenerate AC out of it.
They use a Wein bridge to generate an oscillation of 50Hz and then an opamp for amplification. I wonder if we can use UCD400 for the amplication part?
Bgt said:
SSassen said:
Thanks for the remarks/comments. Well, Linn uses SMPS in their flagship amps, so it is possible. Anyway, yes, it is not simple plain straight forward technology, I agree. But every now and then one needs some challenge
I will not recommend this project to people that do not have the technical background to experiment with an SMPS since the high voltages are very dangerous (read deadly) and one has to understand the feedback theory when one wants to modify the feedback loop.The SMPS is from:
http://www.a-and-t-labs.com/K6_Sw_Amp/index.htm
This SMPS switches at 75kHz, so well above the audio band. The switching noise at the outputs at about 100W load is about 30mV RMS. The 100Hz mains related noise is 0.6V peak/peak, so about 200mV RMS with a load of about 100W. When I switch on a 2nd lightbulb, the voltage briefly drops a few volt for a few ms, but quickly recovers while the 100Hz component only slightly increases. The bulb has a cold resistance of about 12 Ohm, so pretty high currents flow when I switch on that 2nd bulb. I will later try to measure those lightbulb switch on currents. I will also hook up more lightbulbs later this week to torture this supply further. I have increased the softstart capacitor to a value of about 54uF (actually 10uF and 2x22uF in parallel), this results in a ramping up output voltage from 0 to about 63V in about 1 second with less overshoot than previously, but still 5-6V overshoot. I'm still using it without heatsink, just with the aluminum L profile. After burning continously 100W in that lightbulb for half an hour or so, the power mosfets reach a temperature of about 50C, so not too bad, of course some form of heatsinking is preferred to keep them running cooler.
Best regards
Gertjan
Ackcheng,
If you live in Hong Kong I'd indeed worry about the amount of 'noise' coming in through the power lines. I've been to Hong Kong a few times and I know how things work there, in Gertjan's case this isn't an issue, he lives in Holland, where clean power is usually guaranteed.
Best regards,
Sander Sassen
http://www.hardwareanalysis.com
If you live in Hong Kong I'd indeed worry about the amount of 'noise' coming in through the power lines. I've been to Hong Kong a few times and I know how things work there, in Gertjan's case this isn't an issue, he lives in Holland, where clean power is usually guaranteed.
Best regards,
Sander Sassen
http://www.hardwareanalysis.com
I just received my transormers and am pretty impressed- big, resin center, toroid monsters. However, being corporate surplus, they have more options on them than I need and I cannot figure out which are the correct wires to use or how to hook them up. Here is a quote from the auction:
On one side, there are (left to right) black, white, purple and dark grey wires coming out- all about the same gauge and double insulated (white insulation under the colored jacket). I am guessing that these wires are inputs (primaries) and (wild guess) black and white would be for US 120vac? Is this right? Which is for + and which -?
On the other side are 2 red heavy gauge wires that I would guess are for the main 43v output (secondaries). Also, there are 2 blue wires, 2 yellow, and 1 orange wire. I am guessing that the other pairs are for the 14v and 53v outputs, would the orange one be for some kind of shield ground? If I am not using the other wires, should I just cap them? If I use the 14v output for a lighting supply, will this affect the sound at all?
More info off the side of the transformer:
Toroid International
P/N:TI-023439 AU Iss 3
PS Audio: 13-023-11-1 rev g
(I would guess that PS Audio was the original customer)
Once I have figured out which wires are which, how do I wire the pair to get the 43VAC+/- I need for a stereo UCD 400 amp?
I also have a lot of capacitance (4x30,000uF Sprague- 3.5" wide by 7" tall each), but with caps this size, should I bypass with smaller ones?
I know that these probably sound like total novice questions- but that's what I am. If the information I've given doesn't tell enough to get me an answer, what would be the safest way to use a multimeter to find out?
Thanks to everyone,
Larry
On one side, there are (left to right) black, white, purple and dark grey wires coming out- all about the same gauge and double insulated (white insulation under the colored jacket). I am guessing that these wires are inputs (primaries) and (wild guess) black and white would be for US 120vac? Is this right? Which is for + and which -?
On the other side are 2 red heavy gauge wires that I would guess are for the main 43v output (secondaries). Also, there are 2 blue wires, 2 yellow, and 1 orange wire. I am guessing that the other pairs are for the 14v and 53v outputs, would the orange one be for some kind of shield ground? If I am not using the other wires, should I just cap them? If I use the 14v output for a lighting supply, will this affect the sound at all?
More info off the side of the transformer:
Toroid International
P/N:TI-023439 AU Iss 3
PS Audio: 13-023-11-1 rev g
(I would guess that PS Audio was the original customer)
Once I have figured out which wires are which, how do I wire the pair to get the 43VAC+/- I need for a stereo UCD 400 amp?
I also have a lot of capacitance (4x30,000uF Sprague- 3.5" wide by 7" tall each), but with caps this size, should I bypass with smaller ones?
I know that these probably sound like total novice questions- but that's what I am. If the information I've given doesn't tell enough to get me an answer, what would be the safest way to use a multimeter to find out?
Thanks to everyone,
Larry
Nice, J-P. I wonder also how Lars Clausen's, B&O's ICE and the Tripath modules compare to yours.Jan-Peter said:An EMI measurement on the latest version of the UcD400. FYI 30dBuV/M is the official limit.
Jan-Peter
Tja, altijd hé
I know also that this official limit is one thing and real life is an another.
Hi Larry,
I couldn't find anything worthy on the Toroid International website, but I am under the impression the toroid makers tend to follow the same wiring convention, surprisingly enough.
Therefore, maybe you can use this information from Plitron, as a guide to get you started:
http://www.plitron.com/pages/Products/Std/schemati.htm
Doubt they're indentical, but if you spend some time with that and a continuity tester/ohm meter you should be able to sort out most of it with some confidence.
Then maybe you can wire a 100W light bulb in series with that you found to be the primary, to limit the current to safer levels.
Hook some jumpers (alligator clips) up to your volt meter and each secondary lead one at a time, safest to connect the jumpers while the toroid is unpowered, then power it up, take an AC voltage reading, write down your reading on some masking tape, unplug the power, and label the wire you just tested with the tape, move the jumper to the next wire and repeat.
You can do this with no load or anything attached, I'd expect slightly higher (~10, 20%) readings than you're expecting.
I think with that and the schematic above you should be able to figure out your transformer leads OK. Just use the probing to verify what you've worked out by the color codes. Be smart though and tape your wires in such a way so that they can't short with each other while you plug it in for testing. Maybe fold them at different lengths and tape them off.
After that you should be able to use the diagrams at plitron in the link I gave you to figure out how you need to wire it. You can find a few others here too:
http://www.plitron.com/pages/technote.htm
I think you'd be fine using the auxiliary 15V supply for an LED or whatever else you want, but I wouldn't bother using it just for a light. Maybe it would be best used to power a soft start circuit.. I'll leave that up to you and google.
Hm, oh, there's no need at all to bypass your reservoir caps. They're already included on the module where they're most effective.
You can do it anyway if you want to experiment.
You should certainly bypass your bridge rectifiers though.
So, hope that gives you something to start with.
Regards,
Chris
PS: Unrelated topic:
Hi Sander Sassen
Nice site you have there, great forums, nice seeing you here. I look forward to your review of the UCD.
I couldn't find anything worthy on the Toroid International website, but I am under the impression the toroid makers tend to follow the same wiring convention, surprisingly enough.
Therefore, maybe you can use this information from Plitron, as a guide to get you started:
http://www.plitron.com/pages/Products/Std/schemati.htm
Doubt they're indentical, but if you spend some time with that and a continuity tester/ohm meter you should be able to sort out most of it with some confidence.
Then maybe you can wire a 100W light bulb in series with that you found to be the primary, to limit the current to safer levels.
Hook some jumpers (alligator clips) up to your volt meter and each secondary lead one at a time, safest to connect the jumpers while the toroid is unpowered, then power it up, take an AC voltage reading, write down your reading on some masking tape, unplug the power, and label the wire you just tested with the tape, move the jumper to the next wire and repeat.
You can do this with no load or anything attached, I'd expect slightly higher (~10, 20%) readings than you're expecting.
I think with that and the schematic above you should be able to figure out your transformer leads OK. Just use the probing to verify what you've worked out by the color codes. Be smart though and tape your wires in such a way so that they can't short with each other while you plug it in for testing. Maybe fold them at different lengths and tape them off.
After that you should be able to use the diagrams at plitron in the link I gave you to figure out how you need to wire it. You can find a few others here too:
http://www.plitron.com/pages/technote.htm
I think you'd be fine using the auxiliary 15V supply for an LED or whatever else you want, but I wouldn't bother using it just for a light. Maybe it would be best used to power a soft start circuit.. I'll leave that up to you and google.
Hm, oh, there's no need at all to bypass your reservoir caps. They're already included on the module where they're most effective.
You can do it anyway if you want to experiment.
You should certainly bypass your bridge rectifiers though.
So, hope that gives you something to start with.
Regards,
Chris
PS: Unrelated topic:
Hi Sander Sassen
Nice site you have there, great forums, nice seeing you here. I look forward to your review of the UCD.
ghemink said:
I have tried to increase the output caps after simplified simulations with switchcad indicated I could do that without modifying the feedback circuit (although phase margin decreases and the step response showed some ringing in the simulations).
So to avoid problems I first hooked up a 8200uF cap with a 0.2 Ohm resistor in series since that will improve the phase margin in comparison with no resistor (only serie resistance of the cap). This works fine. 100Hz ripple is reduced to about 0.2V peak/peak, so about 70mV RMS with a load of about 200W, not bad I would say.
When I start it up with only one light bulb, I get a 100Hz ripple that is only 100mV peak/peak, so about 35mV RMS, that is with a 100W load, try that with a conventional supply. The high frequency harmonics, mainly 75kHz based are still about 100mV peak/peak, looks like a 75kHz triangle that is strongly bandwidth limited.
Without the 0.2 Ohm resistors, the SMPS became slightly unstable with 1 lightbulb but it was stable and very low ripple with two bulbs (forgot the ripple level). Obviously I have to optimize the feedback network for use with the bigger caps without series resistor. I'll probably will do that since these results with bigger caps look very promising.
Cheers (I'll have a beer on the encouraging results obtained so far)
Gertjan
matjans said:schouderklopje!
Proost.
Anyway, this is by far not the end of the story. Of course I will continue. Plan to use my 10.000uF ELNA Cerafine caps at the output and of course without series resistance, so I have to modify the feedback loop.
Besides that, I still don`t like the overshoot during start-up, I can reduce it by making the softstart slower, but the actual problem is that the feedback loop (which has an integrator) is fighting against the softstart circuit that forces the output down while the feedback loop wants to force the output up. When the softstart stops working, the feedback loop integrator is still giving too much output and it takes time before that feedback loop is stabilized. I`m thinking of a solution for that, one solution would be to have a reference voltage for the error amp that does not immediately go to 5V but that is delayed itself by an RC network. Then by properly choosing the time constant for the softstart and the reference voltage (making the reference voltage slower than the softstart), one should be able to avoid these problems and should get a very nice controlled ramping up output voltage. I`m thinking of some other method as well, will try something.
This project is fun, although a bit tricky because of the high voltages/power. If something goes really wrong with the feedback loop, I could get about +-150V at the outputs, this will destroy the caps at the output stage (and my lightbulbs too of course). I guess that blowing up a couple of 8200uF caps gives a pretty bad mess. So I try to simulate things first and make small changes step-by-step.
Don`t want to recommend this type of tuning to people with a non-electrical background and lack of test equipment because of the high-voltages and power involved. Be really carefull, I don`t want people to become enthusiastic with my results and start experimenting and hurting themselves (very possibly fatal). I use a scope all the time to verify at each stage what is going on.
Gertjan