DC-DC vs DC-AC

[lagoausente]

Tipical DC-DC converters have an inherent R&N due to the harmonics of the swithing, isn´t? For example, this one, 80mV
http://www.switching-powersupplies.com/shop/dc-dc-converter/module-type/dke15.html
Linear PSU, are usually more quiet or not?

Let´s think now, take a battery to a pure sine wave inverter, to get AC, and then use the current linear PSU.

What woud be the advantages/disvantages of each? What one do you think would be quieter?

A quote from other forum:

I said that because I have been considering such a sine-output inverter design recently and there are a lot of chances to generate gentle RFI with such a circuit. Note that the most harmless part is the DC-DC converter, since the output PWM section is a high voltage class-D amplifier that is forced to perform hard switching of 200V to 400V over a *conducting* diode in each clock cycle (If two phases are used to obtain 230V AC, then hard switching happens twice in each cycle)

If understand it ok, the inverter has the same problem since it need a DC-DC section. My question is, if in the AC to DC process, at the linear PSU, the posible switching noise gets slower, for the proportion voltage/noise. I mean, once the 230 VAC is converted to for example 40VAC the swithing noise would decrease in the same proportion? and maybe so be neglectable, or not? To simplify the question, is any advantage to go to 230VAC first, then again to +/- 24 on a preamp, or would be just the same, as going directly from DC to +/- 24V?
I´m quite confuse with this.

 

[mwkeene]

Pure sine wave inverters are generally hard to design, and would probably generate as much noise, if not more than, a well designed switcher would.

Most around here hate switchers. :roll:

A simple linear supply is easier to implement, but will take up more space with its bulky 60Hz transformer. But forget that, I wouldn't DC:AC at 60Hz, id do it at 100kHz, or something out of the audible spectrum. Dont forget, you'll get noise from the bridge rectifier diodes turning on and off, so the linear supply aint all pure. Plus, a linear supply is generally inefficient, burning up power in a regulating pass transistor.

Switchers are harder to implement, but have the advantage of being very tight and efficient supplies that generally take up less space and dont radiate 60Hz field all over the place (You'll have to place your line transformer carefully if you go linear). With a switcher you can put the fundamental frequency all the way out at 250kHz to a couple MHz which is well beyond the audible range. Noise can definitely be an issue, and sometimes linear post-regulating schemes are used to clean up power. Swichthers are also a bit of a challenge to lay out on PCB, and you cannot bread board them (trust me, please).

I'd do just a DC/DC. DC/AC/DC is a lot of conversion. Linear post regulate if you have to.

-Mike

 

[lagoausente]

What about a DC to AC converter without swithing, is no posible?
I mean:
Take the 12VDC, then convert to 12VAC, with an oscilator, then transfomer, convert to 40VAC for example, then enter in the linear PSU to the rectifier section..

What do you think about this with 80mV R&N http://www.switching-powersupplies.com/shop/dc-dc-converter/module-type/dke15.html
Maybe a practical solution if I put a linear post regulator after?

 

[lagoausente]

whats the MG set?

 

[mikep]

Ive been thinking about sinewave versus switching DC-DC converters

if you think about it, sine might be a good idea in applications where the circuitry is dealing with frequencies HIGHER than the switching frequency. for example: low noise RF amplifier. a 200kHz switching DC-DC converter is generating harmonics way up into RF which have to be aggressively filtered.

conversely, for audio, where the frequencies of interest are all below the fundamental swithing frequency, those harmonics will matter much less, or I should say, are easier to deal with. so for audio maybe you should concentrate on conventional swithing technology and just try to implement it right. it is doable, there are loads of examples out there.

if you want to do sine anyway, heres some food for thought: a light bulb stabilized oscillator, a VCA and a class AB driver stage feeding a small ferrite core transformer to do the step up. then rectify, filter and regulate. I was thinking about using feedback to the VCA to keep constant drop across the regulators. I *think* you could find appropriate magnetics cheaply that are intended for the mainstream switchymode industry.

mike

 

[lagoausente]

What would happen if take a Tone Generator, and plug a 50 hz tone to a hifi amp, then take the output and go to a transformer. Setting the correct Z, would not work? maybe a stupid idea, but I´m curious.

 

[JohnRoberts]

Yes, it would work. I've done this to simulate 50 Hz mains for product testing before (you need a decent power amp though).

I'm inclined to ask why go to so much trouble and wasted energy? (Al Gore is watching you know). Even 50/60 Hz sinewaves must be rectified to convert to DC and that conversion from AC to DC also causes noise.

It basically comes down to managing LF or HF noise. Since good operation requires PS that is clean and low impedance at HF too why not just figure out how to properly manage HF noise?

JR

 

[bcarso]

As I may have mentioned in another thread, the latest Audio Precision units use a switcher. I spoke to Bruce Hofer about it when he was in town presenting, and he said he had been dubious to begin with that they would ever get the noise down enough. He added that they did have to carefully select a good vendor, encase the unit in a good shielding package, and do some very good common-mode chokes on the mains input tuned for excellent balanced inpedance. The result was not cheap, but within reason.

The output noise is a few millivolts, which works all right because there is always some local filter/decoupling, and reasonable power-supply rejection, in the rest of the subsystems of the instrument.

As others point out, a linear supply always has a big magnetic field generator with the transformer, and low frequency fields are hard to shield. Hofer said that there was essentially no trace of such normally ubiquitous fields from the switcher.

I believe Madrigal (Mark Levinson brand) did a high-end power amp once that had a lot of power regeneration stuff going on, along the lines discussed earlier in this thread. It was inefficient and bulky and I suspect as much done as a marketing gimmick as anything else, but this is just a suspicion.

 

[lagoausente]

(Al Gore is watching you know). Embarassed

what do you mean? Don´t understand.
The reason for starting this question, is not to me going this way. I would prefer a set of batteries that going DC-AC-DC. I just wanted to know if some "pro" guys out there that are going so happy with a determinated brand inverter (I don´t consider important to mention the brand), are having maybe a extra noise, at least, they didn´t notice it.
I confess I like to research some times for curious, and of course I think is a good way to discover things. In this case I´m not researching on this as a practical option for the Supergreen, just I´m guessing if options other respetable people are taken, are well based or not. In any case is fine to know, for if one day for any reason, must use a AC preamp, to know if will have any con using the inverter.
Sometimes some doubts takes you to others.

As I may have mentioned in another thread, the latest Audio Precision units use a switcher. I spoke to Bruce Hofer about it when he was in town presenting, and he said he had been dubious to begin with that they would ever get the noise down enough. He added that they did have to carefully select a good vendor, encase the unit in a good shielding package, and do some very good common-mode chokes on the mains input tuned for excellent balanced inpedance. The result was not cheap, but within reason.

The output noise is a few millivolts, which works all right because there is always some local filter/decoupling, and reasonable power-supply rejection, in the rest of the subsystems of the instrument.

As others point out, a linear supply always has a big magnetic field generator with the transformer, and low frequency fields are hard to shield. Hofer said that there was essentially no trace of such normally ubiquitous fields from the switcher.

I believe Madrigal (Mark Levinson brand) did a high-end power amp once that had a lot of power regeneration stuff going on, along the lines discussed earlier in this thread. It was inefficient and bulky and I suspect as much done as a marketing gimmick as anything else, but this is just a suspicion.

I´m not sure what´s your conclusion in relation with preamps.

Yes, it would work. I've done this to simulate 50 Hz mains for product testing before (you need a decent power amp though).

I find some contradictions about linear/switching, in general. In one side, seems swithing noise is dificult fo filter, but out of audio band ,isn´t so why so many people hate switchers?
And if linear noise, from rectifiers is more present in audio band, why some people claim is more easy to implement with a good result?

 

[mwkeene]

[quote author="lagoausente"]

(Al Gore is watching you know). Embarassed

what do you mean? Don´t understand. [/quote]

He is saying that a conversion from DC to AC to DC is a two step, and thus, inefficient process. Your DC/AC inverter will maybe be 80 to 90% efficient if you're lucky, and your linear recficiation, filtering, and regulating will be another step following that that could be up to 65% efficient. (best case you have (.9) * (.65) = 59% efficieny). That DC/DC you're looking at there is 75% efficient. If you had a switcher designed specifically for your application using a transformer (say, a forward converter), you could have up to 80 - 85% efficiency. LC LC LC filtering .1uF, .01uF caps, snubbers, and good layout with low parasitics would get you pretty damn clean. Throw a linear post regulator on there if you really wanna get crazy, but that will cost you some efficiency, cost, and space. You probably dont need it. You may however need it if you go with an off the shelf DC/DC.

The guy from seventh circle audio has a switcher power supply for his preamp kits and i havent heard any complaints about it (I havent been listening very hard for complaints however).

Dan Kennedy (great river) uses an off-line switcher in one or two of his boxes i think.

[quote author="lagoausente"]
I find some contradictions about linear/switching, in general. In one side, seems swithing noise is dificult fo filter, but out of audio band ,isn´t so why so many people hate switchers?
And if linear noise, from rectifiers is more present in audio band, why some people claim is more easy to implement with a good result?[/quote]

Switching noise is difficult to filter, but it can be done. interwinding inductance on filter chokes allows the higher freq stuff to pass, and ESR on electrolytics limits their ability to filter the higher freqs, and this is where the challenge comes in.

Don't forget, switchers are really DC:ACC, its just that some topologies produce square waves. There are resonant DCC converters, which will generate a near perfect sine wave for filtering and rectification. If you have to go with a sine wave, i'd go with one of these. They're difficult to design however. You have to design your transformer to have just the right amount of leakage inductance such that it forms a hi-q LC bandpass filter in conjunction with caps strapped across or in series with the primary and secondary. They have poor line rejection (they have the best rejection at the switching frequency) but that doesnt matter because youve got a nice DC source.

Do you have a scope with decent bandwidth? Buy a DC/DC and look at the noise. Add a linear regulator after it, (remember, the dropout on a linear reg is about 1.7 volts, so your DC/DC should probably be putting out about 3 - 5 more than you need for your supply). Do they make a 28 volt DC/DC, or a 30 volt one? one of those would be good. Add the regulator (with proper decoupling) and look at the noise. what does it look like now? i bet its pretty clean. Just grabbing the datasheet for any old reg, i see the L7824 series has about 60dB line rejection at 100kHz, and i bet you could do better with another reg out there. You're really gonna need to know the switching frequency of that DC/DC to know what kind of supply rejection you need, and theres always gonnna be harmonics generated well into the MHz.

Give it a try, the best, and most fun way to do it is just buy up some stuff and try it out.

-Mike

 

[mwkeene]

[quote author="bcarso"]As I may have mentioned in another thread, the latest Audio Precision units use a switcher. I spoke to Bruce Hofer about it when he was in town presenting, and he said he had been dubious to begin with that they would ever get the noise down enough. He added that they did have to carefully select a good vendor, encase the unit in a good shielding package, and do some very good common-mode chokes on the mains input tuned for excellent balanced inpedance. The result was not cheap, but within reason.[/quote]

That common mode choke may have been for FCC requirements, and im assuming this was an off-line, isolated converter. He doesn't need something that extravagant. Non-isolated is fine since he's using a battery, and the expensive bulky common mode choke wont be necessary. Opto-couplers or magnetic isolation amplifiers wont be necessary either. Those are two of the most expensive parts in a switcher anyway (The common mode choke costing upwards of $20/piece sometimes). You could, however get into games of using a common mode choke as the filter L in your secondaries (+24, -24V) to improve cross regulation, which is poorer in forward converters. That would be an expensive (by commercial standards) solution, but it would probably be more than fine for a one off.

-Mike

 

[lagoausente]

What sense have to remove noise from 150kz to Mhz, if those frequencies will not be heared?
would not be more logical a Low pass filter from 20khz down to 20hz?

That common mode choke may have been for FCC requirements, and im assuming this was an off-line, isolated converter. He doesn't need something that extravagant

Did you refer to this ones,?
http://www.audioprecision.com/products/option_filters.htm

 

[bcarso]

[quote author="lagoausente"]What sense have to remove noise from 150kz to Mhz, if those frequencies will not be heared?
would not be more logical a Low pass filter from 20khz down to 20hz?

[/quote]

What is not being discussed so far is the propensity of circuits to detect the high frequency energy envelope and reproduce this as an audible artifact. All it requires is a suitable nonlinearity somewhere in your circuitry and high enough gain. A base-emitter junction of a bipolar is quite effective, a JFET typically much less so.

In addition, if there is more than one such source of high frequency energy, and if the intensities and frequencies meet certain conditions, then you are likely to have your nonlinear circuits detect heterodyne ("beat-frequency") artifacts. To have a complete discussion of all of these effects is more than I have time for at the moment, but you could do some googling. Or maybe someone else can chime in.

 

[JohnRoberts]

Indeed, while perhaps not immediately obvious, ICs are typically 4 input devices; +, -, and both supplies. Bad stuff on the rails, can show up at reduced levels in the inputs, and then in the output depending upon closed loop gain of feedback networks.

For a perhaps not so interesting story about noise. I was helping a junior engineer chase down a 1 Hz noise in a powered mixer. Since 1 Hz is not a typical noise frequency this was a head scratcher. Turns out it was a beat frequency between the computer monitor in his office, and the engineer's office next to his. Noise went away if either one was turned off.

Power supplies, IMO should be clean and low impedance at all frequencies.


JR

 

[jdbakker]

[quote author="lagoausente"]What sense have to remove noise from 150kz to Mhz, if those frequencies will not be heared?[/quote]
What bcarso said. Just like any amplifier aspires to be an oscillator, pretty much any circuit will do its best to act as a radio (RF) receiver.

To add to JR's point, it is very common for circuitry to have power supply rejection getting worse with increasing frequency, as does common-mode rejection for more than a few topologies. All this means that the higher the frequency of the spurious signals on your power lines, the more will find a way to your device's output (up to a certain point -- no device has infinite bandwidth).

An extra factor to consider in this digital day and age is aliasing by ADCs. Although all audio ADCs have anti-aliasing filters on their inputs, these filters are usually no better than 12dB/octave. If you present an audio ADC with a strong signal near (a multiple of) its modulator frequency, you will end up with audible artifacts in your digital stream.

JDB.

 

[mwkeene]

In terms of the common mode choke reqirements, I was purely talking about FCC requirements pertaining to conducted emissions back down the power lines. If you are doing a one-off design here, you wont absolutely have to have it because you probably won't plan on selling your supply. That doesnt mean I wouldnt suggest one, as you want to be a friendly "neighbor" to all your other equipment.

Where is svart? He knows a thing or two about switchers. He could add some good info to this conversation.

Check out the app note by linear tech on their site, it was mentioned to you in another discussion...AN101. Like I said, linear regs only have 40dB attenuation out to 100kHz, and past 1MHz they are essentially useless, so you need to add ferrite beads to kill the stuff above that. Switching noise (rectifiers, xfmr leakage spikes, etc.) will be well above 1MHz, and a ferrite bead or two will help you attenuate that noise quite well.

-Mike

 

[lagoausente]

Check out the app note by linear tech on their site, it was mentioned to you in another discussion...AN101. Like I said, linear regs only have 40dB attenuation out to 100kHz, and past 1MHz they are essentially useless, so you need to add ferrite beads to kill the stuff above that. Switching noise (rectifiers, xfmr leakage spikes, etc.) will be well above 1MHz, and a ferrite bead or two will help you attenuate that noise quite well.

I´m learning much things here. I will need some time to get into details, and read some theory.
So, I think I will need an osciloscope to analyze the results of the filtering.
Maybe would be more practical to buy a LPF, and test and modifie it?
What about this ones? http://www.audioprecision.com/products/option_filters.htm

I´m not having too much time just now, although I expect have more time in the future, and hope will start studing at university, if I can.
So, to be running the preamp, do you think those filters would do the job, or better desing the entire filter myself?

 

[bcarso]

The Audio Precision lowpass filters are specifically designed for and compatible with AP equipment. They are measurement filters, and they are not suitable for power supply filtering.

 

[mwkeene]

Check these guys out, they're the main player if you want a common mode power filter:
http://www.corcom.com/Series/PowerLine/default.asp

This is something that goes on the "wall" side of the power supply. Its connected directly to the AC coming in from the outside world. If you're using a battery as a source, you wont need it. If you're using AC as a source, id suggest it.

In terms of the common mode choke on the output of a dc/dc module, I wouldnt suggest it, unless you were able to break into the module and insert one in instead of your filter inductors inside the control loop. The last thing you wanna do is try to dig into that potted part and create a huge mess, and trust me, potted modules are a huge mess. When the technicians at work do failure analysis, they basically have to put a stencil down on a potted part and drill where they know certian test pads are to probe them, but going beyond that is ridiculous. Removing parts is a nightmare. And, unfortunately, adding a common mode choke outside your module will not be as effective.

-Mike

 

[lagoausente]

[quote author="mwkeene"]Check these guys out, they're the main player if you want a common mode power filter:
http://www.corcom.com/Series/PowerLine/default.asp

This is something that goes on the "wall" side of the power supply. Its connected directly to the AC coming in from the outside world. If you're using a battery as a source, you wont need it. If you're using AC as a source, id suggest it.

In terms of the common mode choke on the output of a dc/dc module, I wouldnt suggest it, unless you were able to break into the module and insert one in instead of your filter inductors inside the control loop. The last thing you wanna do is try to dig into that potted part and create a huge mess, and trust me, potted modules are a huge mess. When the technicians at work do failure analysis, they basically have to put a stencil down on a potted part and drill where they know certian test pads are to probe them, but going beyond that is ridiculous. Removing parts is a nightmare. And, unfortunately, adding a common mode choke outside your module will not be as effective.

-Mike[/quote]

And so? Best option is to add the beads and caps as recomended on AN101 and go?
I´m quite surprised with a DC-.Dc converter from Bellnix that claims only 3mVpp of Ripple an noise, while all other brands have around 75mpp or more.
Contacted them, but seems they sell 10 units minimum. Still waiting for they confirm it.
Although Bellnix converter is +/- 15Volt, and I´m not sure if could run the supergreen. http://www.akustische-kunst.org/audio/supergreen.html
Any help with this? Must I consider the Op-amps datashem for this, or is there any other things to have in account in the whole circuit?