On Wednesday, September 7, 2016 at 3:46:28 PM UTC-4, hon...@yahoo.com wrote:> Thanks for everyone's responses. I tried some LC filtering and saw no obvious improvement. At that point it became obvious to me that I have to use a linear power supply since a switcher will never be quiet enough. This device is used to calibrate resistance measuring devices, so a switcher probably should have never been in there in the first place.Is the smps on the pcb or an external thing? I've had great luck with external smps's. (I make power supplies with a few nv/rtHz with them... there is both an LC filter and cap multiplier.) In some way I think the high frequency is easier to filter than the 60 Hz. (and linear supplies are heavy and spit 60 Hz magnetic fields all over.) You might have common mode noise as Tim W suggested. Do you have any voltage head room, such that you could put a linear regulator after the smps? George H.
Reducing SMPS noise
Started by ●September 6, 2016
Reply by ●September 7, 20162016-09-07
Reply by ●September 7, 20162016-09-07
On Wednesday, September 7, 2016 at 5:34:44 PM UTC-4, John Larkin wrote:> On Wed, 7 Sep 2016 05:35:23 -0700 (PDT), George Herold > <gherold@teachspin.com> wrote: > > >On Tuesday, September 6, 2016 at 4:11:46 PM UTC-4, John Larkin wrote: > >> On Tue, 6 Sep 2016 09:46:13 -0700 (PDT), hondgm@yahoo.com wrote: > >> > >> >I'm working on a PCB that contains nearly 30 resistors in a ladder configuration, each step switched in or out by a SMT solid state relay. It's essentially a decade box for a specific application. The relays are controlled by a micro and all is powered by a boardmount enclosed CUI AC/DC switching converter, supplying up to 400mA. > >> > > >> >The resistors are electrically isolated from the SMPS, however I see a lot of noise on the resistance output that can be from nothing other than the SMPS. I'm assuming the power planes on the board are inducing all kinds of noise into the traces that connect the resistors and solid state relays. > >> > > >> >Replacing the SMPS with a linear would be ideal, but the amount of space required is a big problem. So I'm looking at filtering the output of the SMPS. I'm a digital and DC analog guy, so this switching noise scares me. Some research and an educated guess tells me to try an LC filter on the SMPS output. I'm looking for any first hand experience with this and dumbed-down advice that will hopefully get me a solution without days of headaches. > >> > > >> >The SMPS claims a 100kHz switching frequency, and spikes on the scope confirm that. I've already tried putting several MLCC 1uF caps across the SMPS output but that didn't change anything. > >> > >> If the noise is in-plain-sight conducted, adding a ferrite bead, or a > >> small inductor, and a cap, can help a lot. I just tried that this > >> morning. > >> > >> https://dl.dropboxusercontent.com/u/53724080/Parts/VREGS/CUI_P7805.zip > > > >What's 600 ohms on the FB. Is that it's impedance at 10 or 100 MHz? > > The convention is Z at 100 MHz. That bead has a low-frequency > inductance of a few uH.Thanks, is a bead better than an inductor?.. lower Q or something. (I always worry about making a low pass with a high Q. Resistors are safer that way.)> > > > > >I figured out yesterday how to do the ground for a board with multiple > >CUI smps's. Shunting everyone's noise to ground, made the local gnd > >"bad". A star gnd should work but I want to test that today. > > One can cut a ground peninsula out of the PCB ground plane and locate > things such as to keep switcher ground loop currents out of the rest > of the board.OK I could try that... I've got 5 supplies, five peninsula's looks like a star. :^) The HF crud in my lab seems to get nothing but worse. ~20 mV @ ~100 MHz. (with a 60 MHz 'scope, x1 probe) I made a little pomona box LP filter to stick on 'scope. A little active box with some gain would be nice. George H.> > > -- > > John Larkin Highland Technology, Inc > picosecond timing precision measurement > > jlarkin att highlandtechnology dott com > http://www.highlandtechnology.com
Reply by ●September 7, 20162016-09-07
"John Larkin" <jjlarkin@highlandtechnology.com> wrote in message news:gp11tbt58a7n4hjd8gc3kglcugqf7pgnah@4ax.com...> The convention is Z at 100 MHz. That bead has a low-frequency > inductance of a few uH.Unfortunately it saturates by 20-50mA or thereabouts (a few 100s mA for the biggest multilayer beads), so it's useless for power applications. A ferrite bead that's rated for DC bias is called an inductor. Yes, they do actually make that distinction. Even though the ferrite bead technically has higher quality material (high purity = high mu), and they're both inductive components. Inductors have higher Q, so you usually need to dampen the filter resonance with an R+C. (Or something with ESR, like tantalum, but really, who puts tants on supply rails? Use a ceramic plus resistor for goodness sake.) Tim -- Seven Transistor Labs, LLC Electrical Engineering Consultation and Contract Design Website: http://seventransistorlabs.com
Reply by ●September 7, 20162016-09-07
On 9/6/2016 2:24 PM, Dave Platt wrote:> In article <83484c52-da61-4712-98b9-16c01338091d@googlegroups.com>, > George Herold <gherold@teachspin.com> wrote: > >>> Replacing the SMPS with a linear would be ideal, but the amount of space required is a big problem. So I'm looking at >>> filtering the output of the SMPS. I'm a digital and DC analog guy, so this switching noise scares me. Some research and >>> an educated guess tells me to try an LC filter on the SMPS output. I'm looking for any first hand experience with this and >>> dumbed-down advice that will hopefully get me a solution without days of headaches. >>> >>> The SMPS claims a 100kHz switching frequency, and spikes on the scope confirm that. I've already tried putting several >>> MLCC 1uF caps across the SMPS output but that didn't change >>> anything. > >> You'll want an indcutor in series with the cap... making a low pass. >> I think I used 10 uF and 22 uH... (You want the LC frequency to be below 100 kHz.) > > If you've got a volt or so to spare, you could use a transistor-based > capacitance multiplier, to give you an effective capacitance value > quite a bit larger than its physical value. That would make it easier > to push the corner frequency of the low-pass filtering down even > further.I think this is not likely to do anything effective. The problem is not lower frequency components getting past the filter. The problem is parasitic effects making the filter a pass through for high frequency noise.> Check the layout of your PC board, and your grounding arrangements. > Try to make sure that current pulses from the SMPS (either the input > side or the output side) and voltage jumps aren't being coupled into > the output circuitry via either capacitive coupling, or > inductive-loop coupling.Excellent advice. The noise is being coupled across (or around) the isolation by some means. Rather than adding bandaids, better to understand where the problem is coming from. When the OP says, "I'm assuming the power planes on the board are inducing all kinds of noise into the traces that connect the resistors and solid state relays." it makes me suspect poor layout. The ground plane should not be anywhere near the isolated circuits.> In an extreme case you might need to have a shield around one side or > the other, broken only for the relays themselves. > > Thinking outside the (shielded) box a bit: Using magnetically-loaded > "latching" relays might be another helpful option, as their coils > could be left un-energized (either open-circuit, or both sides > grounded) except for a short pulse to turn them on or off. This would > also cut your power supply current demand, and might let you use a > smaller/quieter switcher, or a small linear supply with a big output > filter/reserve capacitor.He's using a solid state switch, so I wonder if those are available in a latching form? Still, if it requires a connection to the power supply it might still be coupling noise across the barrier. -- Rick C
Reply by ●September 7, 20162016-09-07
On 9/7/2016 9:40 PM, George Herold wrote:> On Wednesday, September 7, 2016 at 5:34:44 PM UTC-4, John Larkin wrote: >> On Wed, 7 Sep 2016 05:35:23 -0700 (PDT), George Herold >> <gherold@teachspin.com> wrote: >> >>> On Tuesday, September 6, 2016 at 4:11:46 PM UTC-4, John Larkin wrote: >>>> On Tue, 6 Sep 2016 09:46:13 -0700 (PDT), hondgm@yahoo.com wrote: >>>> >>>>> I'm working on a PCB that contains nearly 30 resistors in a ladder configuration, each step switched in or out by a SMT solid state relay. It's essentially a decade box for a specific application. The relays are controlled by a micro and all is powered by a boardmount enclosed CUI AC/DC switching converter, supplying up to 400mA. >>>>> >>>>> The resistors are electrically isolated from the SMPS, however I see a lot of noise on the resistance output that can be from nothing other than the SMPS. I'm assuming the power planes on the board are inducing all kinds of noise into the traces that connect the resistors and solid state relays. >>>>> >>>>> Replacing the SMPS with a linear would be ideal, but the amount of space required is a big problem. So I'm looking at filtering the output of the SMPS. I'm a digital and DC analog guy, so this switching noise scares me. Some research and an educated guess tells me to try an LC filter on the SMPS output. I'm looking for any first hand experience with this and dumbed-down advice that will hopefully get me a solution without days of headaches. >>>>> >>>>> The SMPS claims a 100kHz switching frequency, and spikes on the scope confirm that. I've already tried putting several MLCC 1uF caps across the SMPS output but that didn't change anything. >>>> >>>> If the noise is in-plain-sight conducted, adding a ferrite bead, or a >>>> small inductor, and a cap, can help a lot. I just tried that this >>>> morning. >>>> >>>> https://dl.dropboxusercontent.com/u/53724080/Parts/VREGS/CUI_P7805.zip >>> >>> What's 600 ohms on the FB. Is that it's impedance at 10 or 100 MHz? >> >> The convention is Z at 100 MHz. That bead has a low-frequency >> inductance of a few uH. > Thanks, is a bead better than an inductor?.. lower Q or something. > (I always worry about making a low pass with a high Q. Resistors > are safer that way.)I don't think the issue is Q. An inductor has some amount of self capacitance which essentially parallels the inductor. A turn or two or possibly none (just slip the bead over a wire) can add inductance with very little capacitance. If using SMT components they make ferrite filters with a small amount of capacitance to ground and ferrite to add inductance.>>> I figured out yesterday how to do the ground for a board with multiple >>> CUI smps's. Shunting everyone's noise to ground, made the local gnd >>> "bad". A star gnd should work but I want to test that today. >> >> One can cut a ground peninsula out of the PCB ground plane and locate >> things such as to keep switcher ground loop currents out of the rest >> of the board. > OK I could try that... I've got 5 supplies, five peninsula's looks > like a star. :^) > > The HF crud in my lab seems to get nothing but worse. > ~20 mV @ ~100 MHz. (with a 60 MHz 'scope, x1 probe) I made a little > pomona box LP filter to stick on 'scope. A little active > box with some gain would be nice.-- Rick C
Reply by ●September 8, 20162016-09-08
On Wed, 7 Sep 2016 21:12:42 -0500, "Tim Williams" <tiwill@seventransistorlabs.com> wrote:>"John Larkin" <jjlarkin@highlandtechnology.com> wrote in message >news:gp11tbt58a7n4hjd8gc3kglcugqf7pgnah@4ax.com... >> The convention is Z at 100 MHz. That bead has a low-frequency >> inductance of a few uH. > >Unfortunately it saturates by 20-50mA or thereabouts (a few 100s mA for the >biggest multilayer beads), so it's useless for power applications.We have some 1206 power beads in stock. The chunkiest one is 33 ohms, 6 amps. -- John Larkin Highland Technology, Inc lunatic fringe electronics
Reply by ●September 8, 20162016-09-08
On Wed, 7 Sep 2016 18:40:14 -0700 (PDT), George Herold <gherold@teachspin.com> wrote:>On Wednesday, September 7, 2016 at 5:34:44 PM UTC-4, John Larkin wrote: >> On Wed, 7 Sep 2016 05:35:23 -0700 (PDT), George Herold >> <gherold@teachspin.com> wrote: >> >> >On Tuesday, September 6, 2016 at 4:11:46 PM UTC-4, John Larkin wrote: >> >> On Tue, 6 Sep 2016 09:46:13 -0700 (PDT), hondgm@yahoo.com wrote: >> >> >> >> >I'm working on a PCB that contains nearly 30 resistors in a ladder configuration, each step switched in or out by a SMT solid state relay. It's essentially a decade box for a specific application. The relays are controlled by a micro and all is powered by a boardmount enclosed CUI AC/DC switching converter, supplying up to 400mA. >> >> > >> >> >The resistors are electrically isolated from the SMPS, however I see a lot of noise on the resistance output that can be from nothing other than the SMPS. I'm assuming the power planes on the board are inducing all kinds of noise into the traces that connect the resistors and solid state relays. >> >> > >> >> >Replacing the SMPS with a linear would be ideal, but the amount of space required is a big problem. So I'm looking at filtering the output of the SMPS. I'm a digital and DC analog guy, so this switching noise scares me. Some research and an educated guess tells me to try an LC filter on the SMPS output. I'm looking for any first hand experience with this and dumbed-down advice that will hopefully get me a solution without days of headaches. >> >> > >> >> >The SMPS claims a 100kHz switching frequency, and spikes on the scope confirm that. I've already tried putting several MLCC 1uF caps across the SMPS output but that didn't change anything. >> >> >> >> If the noise is in-plain-sight conducted, adding a ferrite bead, or a >> >> small inductor, and a cap, can help a lot. I just tried that this >> >> morning. >> >> >> >> https://dl.dropboxusercontent.com/u/53724080/Parts/VREGS/CUI_P7805.zip >> > >> >What's 600 ohms on the FB. Is that it's impedance at 10 or 100 MHz? >> >> The convention is Z at 100 MHz. That bead has a low-frequency >> inductance of a few uH. >Thanks, is a bead better than an inductor?.. lower Q or something. >(I always worry about making a low pass with a high Q. Resistors >are safer that way.)Ferrite beads are low-Q by design. Still, the inductance would resonate with a biggish cap at low frequencies. LT Spice includes a bunch of Wurth bead models, for playing with.>> >> >> > >> >I figured out yesterday how to do the ground for a board with multiple >> >CUI smps's. Shunting everyone's noise to ground, made the local gnd >> >"bad". A star gnd should work but I want to test that today. >> >> One can cut a ground peninsula out of the PCB ground plane and locate >> things such as to keep switcher ground loop currents out of the rest >> of the board. >OK I could try that... I've got 5 supplies, five peninsula's looks >like a star. :^) > >The HF crud in my lab seems to get nothing but worse. >~20 mV @ ~100 MHz. (with a 60 MHz 'scope, x1 probe) I made a little >pomona box LP filter to stick on 'scope. A little active >box with some gain would be nice.We're in a wooden building in plain sight of this: https://dl.dropboxusercontent.com/u/53724080/SF/Sutrotower1.jpg so we are sort of an EMI horror lab. -- John Larkin Highland Technology, Inc lunatic fringe electronics
Reply by ●September 8, 20162016-09-08
On Wed, 07 Sep 2016 20:53:51 -0700, John Larkin wrote:> On Wed, 7 Sep 2016 21:12:42 -0500, "Tim Williams" > <tiwill@seventransistorlabs.com> wrote: > >>"John Larkin" <jjlarkin@highlandtechnology.com> wrote in message >>news:gp11tbt58a7n4hjd8gc3kglcugqf7pgnah@4ax.com... >>> The convention is Z at 100 MHz. That bead has a low-frequency >>> inductance of a few uH. >> >>Unfortunately it saturates by 20-50mA or thereabouts (a few 100s mA for >>the biggest multilayer beads), so it's useless for power applications. > > We have some 1206 power beads in stock. The chunkiest one is 33 ohms, > 6 amps.I'm going to bet that 6A is the thermal rating, not the saturation current. Allan
Reply by ●September 8, 20162016-09-08
On Wednesday, September 7, 2016 at 12:46:28 PM UTC-7, hon...@yahoo.com wrote:> Thanks for everyone's responses. I tried some LC filtering and saw no obvious improvement. At that point it became obvious to me that I have to use a linear power supply since a switcher will never be quiet enough. This device is used to calibrate resistance measuring devices, so a switcher probably should have never been in there in the first place.I'm not following that reasoning. A switcher can be a source of power supply ripple, or magnetic induced currents (interference), but neither of those are absolute disqualifications. Generating a reference voltage with a linear regulator makes a lot of sense, but 'linear' doesn't automatically reduce your ripple or induced currents. There's a lot of induction around a line-frequency transformer. RF goes through a linear regulator and past a big filter capacitor like it has a VIP card.
Reply by ●September 8, 20162016-09-08
On 08 Sep 2016 10:29:39 GMT, Allan Herriman <allanherriman@hotmail.com> wrote:>On Wed, 07 Sep 2016 20:53:51 -0700, John Larkin wrote: > >> On Wed, 7 Sep 2016 21:12:42 -0500, "Tim Williams" >> <tiwill@seventransistorlabs.com> wrote: >> >>>"John Larkin" <jjlarkin@highlandtechnology.com> wrote in message >>>news:gp11tbt58a7n4hjd8gc3kglcugqf7pgnah@4ax.com... >>>> The convention is Z at 100 MHz. That bead has a low-frequency >>>> inductance of a few uH. >>> >>>Unfortunately it saturates by 20-50mA or thereabouts (a few 100s mA for >>>the biggest multilayer beads), so it's useless for power applications. >> >> We have some 1206 power beads in stock. The chunkiest one is 33 ohms, >> 6 amps. > > >I'm going to bet that 6A is the thermal rating, not the saturation >current. > > >AllanThe data sheet doesn't say anything about saturation. We used it to reduce EMI hazards in the drive to an NMR sample heater. Beads work on RFI the same sort of way that garlic keeps vampires away. -- John Larkin Highland Technology, Inc picosecond timing precision measurement jlarkin att highlandtechnology dott com http://www.highlandtechnology.com
