Some time ago I've looked for a similar solution that Jan has designed. I looked at some dual polarity isolated DC/DC converters like these. I would've put LDO regs from the TPS7Axx family after them just like Jan did. The only problem is that I wasn't sure about the radiated noise/interference. The metal shield looks like a good idea but how effective it actually is? Does anybody have an experience with such products?
Regards,
Oleg
Regards,
Oleg
The noise comes mainly in a conducted way. You must attenuate conducted noise sufficiently before applying a shielding. Attenuating conducted noise requires a definition of a reference potential. The shielding must be tied to this reference. All of the wires coming out from the shielding (case) must be ensured to be free of noise. If these are done, then the remaining noise emittion (radiation) is handled by the metal case. But an improperly connected metal case can increase radiation very much. An unconnected case can't make so much harm, but also rarely helps.
The above is true for up to n*100 MHz. Microwave is different.
The above is true for up to n*100 MHz. Microwave is different.
Hi,
Think about the amount of riple and noise within the audio band.
DC/DC-Converter: none || Transformer supply: right in the most sensitive hearing range >100Hz.
Think about the area of current loops and their sensitvity to send or receive EMV
DC/DC-Converter: small (complete Modules comply with EMC/EMV regulations) || Transformer supply: high
Think about regulation-precison and -speed:
DC/DC-Converter: very constant output voltage, very fast || Transformer supply: varying with load and input voltage, high load-induced riple.
Think about parts quality and cost:
DC/DC-Converter: small sized components of high quality but quite cheap || Transformer supply: bulky components with high tolerances and cost.
Think of convenience:
DC/DC-Converter: complete modules available with numerous protection features built-in (over-/under-voltage, overcurrent/shortcircuit-protect, inrush-current ramping, temperature, etc, etc. || Transformer supply: hardly anything apart from series Fuses and maybe a rush-In-protection.
Think about safety:
DC/DC-Converter: wide range of low-voltage input. Supply from easy to source, cheap wallwarts. Safety regulations less stringent || Transformer-supply: 115/230Vac with associated safety regulations.
Even designing a DC-DC-converter isn´t black art anymore as all Chip-Manufactureres provide for great support starting from Datasheets over app-notes to complete designs with layout recommendations.
As post filters the afore mentioned modern linear regulators guarantee high ripple rejection into the MHz-range and very low noise figures.
Due to their small size and cost one may design the supplies right at the PoL (Point of Load) resp. for each subassembly, thereby generating the power lines right where they are needed.
I wouldn´t want to go back to transformers anymore
jauu
Calvin
That's simply not correct.
Think about the amount of riple and noise within the audio band.
DC/DC-Converter: none || Transformer supply: right in the most sensitive hearing range >100Hz.
Think about the area of current loops and their sensitvity to send or receive EMV
DC/DC-Converter: small (complete Modules comply with EMC/EMV regulations) || Transformer supply: high
Think about regulation-precison and -speed:
DC/DC-Converter: very constant output voltage, very fast || Transformer supply: varying with load and input voltage, high load-induced riple.
Think about parts quality and cost:
DC/DC-Converter: small sized components of high quality but quite cheap || Transformer supply: bulky components with high tolerances and cost.
Think of convenience:
DC/DC-Converter: complete modules available with numerous protection features built-in (over-/under-voltage, overcurrent/shortcircuit-protect, inrush-current ramping, temperature, etc, etc. || Transformer supply: hardly anything apart from series Fuses and maybe a rush-In-protection.
Think about safety:
DC/DC-Converter: wide range of low-voltage input. Supply from easy to source, cheap wallwarts. Safety regulations less stringent || Transformer-supply: 115/230Vac with associated safety regulations.
Even designing a DC-DC-converter isn´t black art anymore as all Chip-Manufactureres provide for great support starting from Datasheets over app-notes to complete designs with layout recommendations.
As post filters the afore mentioned modern linear regulators guarantee high ripple rejection into the MHz-range and very low noise figures.
Due to their small size and cost one may design the supplies right at the PoL (Point of Load) resp. for each subassembly, thereby generating the power lines right where they are needed.
I wouldn´t want to go back to transformers anymore
jauu
Calvin
More than 10 years ago Linear Tech rolled out a bunch of gate drivers and switchers which used controlled slew rate to push the switching transistor/mosfet through its linear range -- this obviously causes some heat but it knocks down the switching transients. The older devices are available in SO-16 and SSOP-20 packages.
So the devices start off with good noise performance and excellent PSRR.
The crush of cell-phone technology, the need for ever-lower noise oscillator power supplies has pushed the margins.
Calvin -- i think that humans are most sensitive to noise from 3 to 8kHz
So the devices start off with good noise performance and excellent PSRR.
The crush of cell-phone technology, the need for ever-lower noise oscillator power supplies has pushed the margins.
Calvin -- i think that humans are most sensitive to noise from 3 to 8kHz
At some MHz filters can be incredibly effective. Over 100 MHz the situation is much worse. Parasitic effects ruin many things, and crappy 1 layer PCBs of cheap chinese PSU start to inject incredibly high RF noise into the whole mains network and everywhere. Fly-back converters are the ugliest.
Ripple must never be an issue. It's the easiest thing to get rid. (OK, not in a 1 cubic centimeter circuit...)
Ripple must never be an issue. It's the easiest thing to get rid. (OK, not in a 1 cubic centimeter circuit...)
What about PWM control? I guess it can generate harmonics below 20kHz following the load condition. Many cheap SMPS power supplies emit sound due to mechanical vibrations which change with load, so such audible frequency modulation should be present even if the carrier frequency is in the hundreds of kHz range.
The controle switching is very effective. For instance, the ripple of some of the latest converters is almost pure sine wave, at say 1MHz, just a few mV peak. Couple that with a linear post reg with 60dB+ PSRR at 1MHz and you can't find anything of it back at the output.
The ripple has no harmonics, and smart PCB design makes sure that even a fraction of an inch away is enough to push any radiation below the radar.
These new designs are to be experienced to believe.
Jan
I wonder if any one of you have tried this?;
Dual 12V Power Supply Mini Board
Part Code: MINIPOWERDUAL12V
Features
Incorporates the High Efficiency LM2575T Switching Regulator
Suitable for 14-24VDC input with PCB Terminals
Power LED for both Positive and Negative Outputs
Multiple +12V Output Termination Points
Suitable for use where dual polarity supplies are needed
Board Dimensions: 45 x 55 mm
Dual 12V Power Supply Mini Board
Part Code: MINIPOWERDUAL12V
Features
Incorporates the High Efficiency LM2575T Switching Regulator
Suitable for 14-24VDC input with PCB Terminals
Power LED for both Positive and Negative Outputs
Multiple +12V Output Termination Points
Suitable for use where dual polarity supplies are needed
Board Dimensions: 45 x 55 mm
Attachments
the trouble with AC-DC sw mode converters that they are based on the flyback principle. the transformers need a tight coupling between primary and secondary, resulting in capacitive coupling between primary and secondary with a strong noise source pumping noise into your system ground. try to connect your laptop headphone out while powered to a sound system. that is pump out noise..
Calvin,
You are right. But chinese crap SMPSs are only indicators showing the problem exists. In a good SMPS the problem is lower, but still there, and measurable. 1000...10000 times higher freq means 1000..10000 times higher capacitive coupling between primary and secondary (common mode current!) And also 1000..10000 times higher di/dt (magnetic coupling). This high attenuation of the noises is very, very hard to achieve. Typically it requires several nF Y capacitor between mains side and output. And this is not a good news for a Class II audio system connected to it.
BTW: can you design a decent SMPS, and willing to share the design here, or can you show a commercially (or DIY) available one that is at least as good and as cheap as a decent transformer based PSU? Because if not, then you make a comparison between an available and an unavailable option.
(I'm working on an SMPS especially designed for audio that overcome the high frequency noise issue without high capacitance, but it will not be as cheap as a traditional PSU, and it is not a public project, and I don't know any other design with the same characteristics.)
You are right. But chinese crap SMPSs are only indicators showing the problem exists. In a good SMPS the problem is lower, but still there, and measurable. 1000...10000 times higher freq means 1000..10000 times higher capacitive coupling between primary and secondary (common mode current!) And also 1000..10000 times higher di/dt (magnetic coupling). This high attenuation of the noises is very, very hard to achieve. Typically it requires several nF Y capacitor between mains side and output. And this is not a good news for a Class II audio system connected to it.
BTW: can you design a decent SMPS, and willing to share the design here, or can you show a commercially (or DIY) available one that is at least as good and as cheap as a decent transformer based PSU? Because if not, then you make a comparison between an available and an unavailable option.
(I'm working on an SMPS especially designed for audio that overcome the high frequency noise issue without high capacitance, but it will not be as cheap as a traditional PSU, and it is not a public project, and I don't know any other design with the same characteristics.)
Sorry, I thought I was in the other thread about PSU for power amp:
http://www.diyaudio.com/forums/power-supplies/51749-ultimate-psu-power-amps-2.html#post4838793
If the source is a floating Power Bank, then no problem with common mode noise. But as soon as it is operated from a charger, it will inherit the whole CM noise of the charger.
However I don't think this thread is limited to battery input PSUs neither.
So Jan, if not AC/DC, than at least indicate what else you are talking about, and don't generalise! Most people do use AC input.
http://www.diyaudio.com/forums/power-supplies/51749-ultimate-psu-power-amps-2.html#post4838793
If the source is a floating Power Bank, then no problem with common mode noise. But as soon as it is operated from a charger, it will inherit the whole CM noise of the charger.
However I don't think this thread is limited to battery input PSUs neither.
So Jan, if not AC/DC, than at least indicate what else you are talking about, and don't generalise! Most people do use AC input.
Last edited:
Yes you make a good point. SilentSwitcher can be run from a USB charger (and obviously it is then not totally separate from the mains). Or from a PowerBank, which normally is also 5VDC. That IS totally free from the mains.
In my use I have a PowerBank with the charge input connected to a USB charger, and the SilentSwitcher connected to the PowerBank output. For sensitive measurements or critical listening I just unplug the PowerBank from the charger.
You can also of course use a battery or a walwart. There's an advantage to doing that because the allowed output power from SilentSwitcher is higher with higher input voltage. A 9V 2A walwart would be nice for mains-connected use, or a 7.2V LiPo for disconnected use. Absolute max input (limited by one of the switchers) is 12VDC; to be on the safe side I would stay below 10V or so.
And, it's alive now on Kickstarter! I thought about a Group Buy but I have seen that many of them drag on months and months with a lot of problems and changes and bookkeeping. Kickstarter is just like a Group Buy but without the hassle!
https://www.kickstarter.com/project...her-mains-free-low-noise-15v-and-5v?ref=email
Jan
In my use I have a PowerBank with the charge input connected to a USB charger, and the SilentSwitcher connected to the PowerBank output. For sensitive measurements or critical listening I just unplug the PowerBank from the charger.
You can also of course use a battery or a walwart. There's an advantage to doing that because the allowed output power from SilentSwitcher is higher with higher input voltage. A 9V 2A walwart would be nice for mains-connected use, or a 7.2V LiPo for disconnected use. Absolute max input (limited by one of the switchers) is 12VDC; to be on the safe side I would stay below 10V or so.
And, it's alive now on Kickstarter! I thought about a Group Buy but I have seen that many of them drag on months and months with a lot of problems and changes and bookkeeping. Kickstarter is just like a Group Buy but without the hassle!
https://www.kickstarter.com/project...her-mains-free-low-noise-15v-and-5v?ref=email
Jan
Noise spectra defined in uV without measurement bandwidth or resolution indicated is an error. The correct dimension should be uV/sqrt(Hz), only at a fixed and specified resolution (for example at 1 Hz) it can be changed to uV.
And the real challange, the real problem with SMPSs is the common mode noise current, not within the audio band, but at high freq (from switching freq to at least 200 MHz).
Basically all manufacturers specify output noise in differential mode, and with 20 MHz Bandwidth. No wonder why it is "almost pure sine wave". Measure it yourself with more than 200 MHz bandwidth, and you will see nasty spikes at every switching event for sure!
And the real challange, the real problem with SMPSs is the common mode noise current, not within the audio band, but at high freq (from switching freq to at least 200 MHz).
Basically all manufacturers specify output noise in differential mode, and with 20 MHz Bandwidth. No wonder why it is "almost pure sine wave". Measure it yourself with more than 200 MHz bandwidth, and you will see nasty spikes at every switching event for sure!
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.