Electrical noise-free LED workbench lighting

[5U4GB]

This came up indirectly in another thread, and there's been plenty of discussions of LED noise getting injected into measurements in other threads, does anyone have any thoughts on using a well-filtered linear supply for their LED lighting?  I was thinking of something like this (there are plenty of similar clones) which, if you're actually getting what the photos show, seems like a fairly solidly-designed low-noise DC source for powering an LED strip.  The one notable issue I can see on that one is the missing earth wire to the case, I'm assuming the metal plug body provides the necessary connection.

[jonpaul]

Use high quality EU approved drivers, comply with EMI cond/radiated, CISPR, CE etc.

We specify TCI, Italy, no EMI issues, pure filtered DC out, variable current.

Avoid the low cost Chinese junk

J

[5U4GB]

I'm driving a $10 LED strip that lights the workbench from it, I realise I'm not getting anything CE (or whatever) certified or similar, but all I need is something that's good enough, i.e. can drive the LEDs without radiating a mess of switching hash into the surrounding area.

[Gyro]

If noise is a real issue for you then you could always build yourself a linear power supply to drive the strip, either with a 3 terminal regulator or a simple resistor after the transformer, bridge rectifier and reservoir cap. An added bonus would be total absence of flicker (if you are sensitive).

[Edison]

I made my work table lighting from industrial 24V led strips 2700Lm/m, I used a Mean Well industrial switching power supply for power supply and I couldn't stop being surprised. The light is absolutely flicker-free, but the interference was so intense that the BT mouse stopped working - when the transponder was brought closer, the mouse functionality problem disappeared, but the interference remained. The led strips were mounted in massive aluminum profiles and temporarily connected with an ordinary double line, I made a modification and replaced the double line with a shielded cable - the interference decreased by half and the problem with the mouse disappeared completely, then I added ferrite filters (also on the supply cable) and the interference is negligible - it will only show up with very weak radio stations of the test receiver.

the front light has 1.5m of strip in it and the rear 1m

[CosteC]

LED strip consisting only LEDs and resistors does not produce noise, yet it can radiate it. Elimination of this radiation is very expensive. It is much easier not to "pump noise" into LED strips.
I think, as mentioned earlier, that use of certified, high quality power supplies is key point. Chinese "low noise PSU" will not do it. In PSU linked in first post everybody can see how input and output wires from transformer are tied together to ALLOW noise coupling... It is not good design, despite lot of inductors and capacitors visible everywhere.

Another key factor is wiring. Good, tight wiring will just emit less noise. Large loops, separating "+12V" wire from "GND" wire create such loops acting as antennas.
Next cheap and efficient method is grounding. GND of PSU can be connected to PE (Protective Earth). It however opens large discussions about efficiency of such solution. There is however no definite answer as it is PSU dependent and location dependent. It is easy to make however.

Last interesting point is additional filtering. But is it needed if good MeanWell PSU will have 100 kHz 200 mV ripple at output?

Side note:
Most of cheap Chinese PSUs "for LEDs" does not have PFC, despite it is required for "lighting applications". This is yet another way to limit cost of PSU at expense of quality. Among of many other methods, where filter removal is probably first what comes to mind of owner of company competing on this market.

[coromonadalix]

same here,    dc 24v led strip  they are dual row 1 inch wide  (neutral white), it doesn't change the colors reading if needed 

give plenty (maybe too much) of light when put to 24vdc  loll

on a meanwell smps, and a led dimmer + remote,   and i don't have problems  .... if any   i have an adjustable linear version


and yes avoid china  crap ........ far from good

[Xena E]

Just don't use switchers.

Linear is OK, but a simple solution would be capacitive reactive dropper.

[CosteC]

Dimmer... I would thing twice before installing any dimmer between PSU and LED strip. Most of dimmers use PWM - naturally rich with harmonics and noise.
Constant Current (CC) instead Constant Voltage (12 V or 24 V in most of cases) is better choice to regulate light output if really needed. Yet this is different design than classical LED strip.

[graybeard]

I just had two Lindgren shielded rooms upgraded to LED lighting because the incandescent bulbs were no longer available.  They installed a shielded DC power supply extensive RF filtering on the inputs and outputs.  Only clean well filtered DC comes into the room.  I have detected no noise from the lights,  I do very low level 1/f and phase noise measurements.

[rhb]

I am installing 48x 8' LED lamps in my 1530 sq ft shop building which will get a DIY 11' x 16' screen room with advice from a friend who held a TEMPEST level 2 cert.

I had bought a pair of Corcom 20VB1s on ebay, but was not happy with the performance below 1 MHz.  He suggested using a 6ESB1 in series with a 20ESB6.  That worked much better, but still not to my satisfaction.  As I had the 20VB1s I tried adding one of those.  Placed between ESB series gives really good results up to 5 MHz.  Each string of 3 lamps will have its own switch with 2 strings per 6ESB1 fed via the 20VB1 and 20EBS6 from a master switch so I can choose which lamps come on when I throw the master swtch.

I was using an Agilent 4395A VNA to measure the response of various arrangements and took a *lot* of screen shots.  I'll go through the photos and post a summary showing each filter alone, just the EBS series filters and then the effect of putting the 20VB1 in  between the two.  Interestingly, to be effective the VB must be between the ESB filters.  The VB has a 1.5 M ohm shunt whereas the ESBs are 330 k ohms.  Apparently the impedance mismatch suppresses coupling between the 2 ESB filters.

This is *just* for suppressing conducted mode EMI from the building lights.  The lamps will all have 1/8" hardware cloth screens enclosing them.  The screen room lighting will be a separate effort.

Have Fun!
Reg

[coromonadalix]

yes for previous thread, i use dimmer  and yes they are pwm,  in my case,  i dont have any problems, i do know their working frequencies  ...  so far all is good,  and the smps and dimmer are enclosed in metal frame

but for sure   it could become a problem is some cases ...

finger crossed

[Kim Christensen]

I worked in a place where they had two switches/circuits for the lights. One was incandescent and the other fluorescent for day to day use when low noise was not a requirement.

[5U4GB]

I was using an Agilent 4395A VNA to measure the response of various arrangements and took a *lot* of screen shots.  I'll go through the photos and post a summary showing each filter alone, just the EBS series filters and then the effect of putting the 20VB1 in  between the two.  Interestingly, to be effective the VB must be between the ESB filters.  The VB has a 1.5 M ohm shunt whereas the ESBs are 330 k ohms.  Apparently the impedance mismatch suppresses coupling between the 2 ESB filters.

It'd be interesting to see a few sample shots when you get the time.  I was so concerned about switching noise being radiated from the transmission lines (LED strips) directly above the workspace that I overlooked the fact that EMI hash would be sent back out the mains wires as well.  Are you filtering the output in any way?  A while back I bought a few cheapie "DC power filter" boards off Aliexpress just out of morbid curiosity but they seem to be just low-voltage mains EMI filters (common-mode choke and an LC filter built with cheap low-voltage components) rather than a two-stage LC filter, but this is definitely not my area beyond basic exposure to the theory many years ago.

[rhb]

I described both the mains filter and the shielding of the lamps.  What "output" are you referring to?

Edit:

Here are photos showing the pass band of the filter I am using.  The bottom trace is with the 20VB1 between the 2 EBS filters and the upper trace is with the VB followed by the 2 EBS units.  In 4009 you can see that putting the VB between them knocks down the peak at 500 kHz but at the cost of 5 dB at 100 kHz.

I've typically seen the most noise below 200 kHz, so I'm going to be looking to knock down the lower frequencies a bit more.  The best result for either sequence of filters at 100 kHz is only -30 dB, but at this point I need to examine the raw spectrum of the lamps first.  I wish I'd thought to do that before I hung them on the ceiling as the shop is currently unheated and very cold right now.

[Randy222]

What type of noise?
Noise you see on the wires, like via FFT?
Or noise like EMI?

Won't even the diodes of a linear PSU make noise as they turn on and off?

For the noise free LED light strip, perhaps used on a workbench when doing certain types of "sensitive" tests measuring, power the LED strip from a battery. When you power off the LED some other charging ckt adds more joules back into the batt, etc.

Another option is, skip DC PSU altogether, just wire two sets of LED's 180deg from each other, then run them using AC (simple stepdown xfrmer to the strip voltage needed). Not sure how much noise they make turning on and off, but when one set goes off the other set goes on. LED rise and fall times should be very fast, so you should have good continuous light, but the on/off flicker from each set does still exist.

Edit: duly noted, in a series of LED's ran by AC, there's still a dead band (albeit small) where the AC voltage needs to rise high enough to fully fwd bias the LED's.

[5U4GB]

I described both the mains filter and the shielding of the lamps.  What "output" are you referring to?

The output of the power supply.  You mentioned Corcom filters which are mains (input) filters, I was more concerned about what was coming out of the power supply and into the LED strips.

[rhb]

I described both the mains filter and the shielding of the lamps.  What "output" are you referring to?

The output of the power supply.  You mentioned Corcom filters which are mains (input) filters, I was more concerned about what was coming out of the power supply and into the LED strips.

Did you not read "I am installing 48x 8' LED lamps in my 1530 sq ft shop building which will get a DIY 11' x 16' screen room with advice from a friend who held a TEMPEST level 2 cert."?

The EMI filters are to keep noise generated by the LED lamps from getting on the AC power lines in the shop.  That simplifies keeping them out of the screen room.  Each of those lamps has its own SMPS.

I have 1 pair of LED lamps overhead with a single Corcom filter and hardware cloth screening.  Not as good as I'd like, but a huge improvement over not having them.

@Randy222 you need to examine the construction of LED lamps.   EMI is both conducted and radiated.  The Corcom filters are there to suppress the conducted modes on the wire and hardware cloth and EMT to suppress the radiated EMI.

All the AC power to my current bench is fed from a 2.5 kVA Topaz isolation transformer via EMT and MX all the way to the IEC plug at the back of the instrument.  That might seem excessive, but if you probe the EMI coming out the back and bottom of an Instek MSO2204EA you will discover that it is appalling.  I don't see how they ever got it approved  other than  by paying bribes.  The photo shows the general style of the bench AC feeds which are switched or not depending upon whether it should not be shutdown such as a GPSDO or is an instrument which doesn't need power 24/7. So a single switch powers down the bench without having to turn off each instrument one at a time.  With ultimately a pair of 6 bay, 6' high racks filled with T&M instruments, that is a lot of instruments.

As good as the current EMI filter stack is, I don't consider it adequate below 100 kHz which is where most of the SMPS noise is located.  I've contacted Corcom for advice on what to use below 100 kHz, but not gotten an answer yet.  As my goal is -60 dB above 1 kHz, I may well have to design and build custom filters.

The real challenge will be suppressing noise when arc welding as I will have a number of experiments which will run 24/7 for years (OXCO and Vref aging).  10-15 kVA isolation transformers are not common 2nd hand and *very* expensive new.Likely I'll be forced to record the noise when welding and then use DSP to remove it.  The OXCO experiments will be very sensitive to the change in voltage caused by the 50 A welding current.  If necessary, I'll construct a rotary motor and alternator with a heavy flywheel to protect the screen room from the voltage drop.  I installed 165 ft of 4/0 Al for the meter to entrance panel run. That limits the drop at 200 A to 1-2% IIRC.  I'll never pull more than 100 A at any time, so it's a very conservative wire sizing.

Have Fun!
Reg

[5U4GB]

Did you not read "I am installing 48x 8' LED lamps in my 1530 sq ft shop building which will get a DIY 11' x 16' screen room with advice from a friend who held a TEMPEST level 2 cert."?

Yup, which I read as "I'm doing something weird and exotic that doesn't apply for 99.99% of the people reading this", which is why I asked about distinctly non-exotic filtering of the output of a standard SMPS.

[rhb]

It applies to anyone using LED room lighting and doing electronic measurements.  With the campaign against incandescents that is becoming *everyone*.  I've got 48 SMPSs running if I turn on all the lights. 

The EMI filter in the mains input to an SMPS is not there to protect the SMPS output from being noisy.  It's there to keep the conducted mode EMI from getting into other equipment via radiation from the mains wiring and conducted mode EMI from entering via the mains.

The only exotic part is the screen room.  I mentioned that to convey *why* I am going to the level of effort which I am.  I mentioned that I had advice from a retired TEMPEST technician in an attempt to head off the "you don't know what you're doing" crowd.

Have Fun!
Reg

[Echo88]

A suggestion: Linear LED panel power supplies instead of SMPS (why produce trouble if you can avoid it), a simple UPS as AC-source during varying mains when arc welding and suitable voltage regulators to power the DUTs: LT3042/3045 instead of series connected selected LM317.

[rhb]

I was primarily commenting about the building lighting.  It's quite impractical to replace the OEM SMPS in 48x $18 LED lamp fixtures for a 1530 sq ft building.  Lights in the screen room are another topic, though very closely related.

Arc welding induced voltage drop is an order of magnitude more difficult.  I simply thought I'd share the issues I have and my solutions.  Still waiting for a response from Corcom for which filter to use below 1 MHz.  Might have to make my own.

Have Fun!
Reg

[5U4GB]

I think, as mentioned earlier, that use of certified, high quality power supplies is key point. Chinese "low noise PSU" will not do it. In PSU linked in first post everybody can see how input and output wires from transformer are tied together to ALLOW noise coupling... It is not good design, despite lot of inductors and capacitors visible everywhere.

Don't worry, that's just the demo version for the sales photos, in the version they'll ship the wiring isn't anything like that neat :-).  This one seems to be the original that they're copying.

My motivation for getting prebuilt was, apart from the fact that I'm lazy, or at least have a ton of other stuff that I'd rather be focusing on, is that I can't even buy the parts for the price they're charging for the whole item.  I'm really just after good-enough, not perfect.

For people interested in all the mistakes you can make in low-noise power supply design, PA0NHC has a on this, using Aliexpress units as examples.

[Edison]

LED strip consisting only LEDs and resistors does not produce noise, yet it can radiate it.

The problem is that my LED strip is not just diodes and resistors, it has blocks of twelve LEDs and each block has its own linear current source, dimming is not possible due to this design.
The light is completely uniform and I don't see a decrease in brightness during use (approx. 10 years).

[5U4GB]

The problem is that my LED strip is not just diodes and resistors, it has blocks of twelve LEDs and each block has its own linear current source, dimming is not possible due to this design.

Could you post more details on this?  That sounds pretty... unusual.

[cosmicray]

Just don't use switchers.

Linear is OK, but a simple solution would be capacitive reactive dropper.

Yes, a linear regulator (e.g. LM317 as a two-terminal current regulator) does not generate the EMI that some regulators do. But you need a (relatively clean) DC source to feed it from. I'm currently looking at using a USB-C power source with a PPS controller to achieve the voltage/current I need (~18v).

As to dimming ... Dimming can be achieved by building a board with multiple interleaved (or dithered) arrays of LEDs. Simply turn on/off as many of those as needed to achieve the desired level of light. All of those LEDs will be run down the curve so as to control heat dissipation and current loading. It may even be possible to tweak the adjust voltage resistor (on the LDO) value in real time, to achieve different amounts of current/light thru the LEDs, but that becomes fiddly with heat dissipation values (primarily on the LEDs, but possibly also the adjust resistor). With the low current values (to each branch of the dithered LEDs) I might even toss in a few ferrite beads to see of they suppress any additional circuit noise.

[Edison]

The problem is that my LED strip is not just diodes and resistors, it has blocks of twelve LEDs and each block has its own linear current source, dimming is not possible due to this design.

Could you post more details on this?  That sounds pretty... unusual.

The LED strip from which I made the work table lighting was not bought in a normal store, but from a company that produces industrial lighting. I have been alerted to the fact that due to this LED connection, the strip cannot be dimmed, but has a completely constant brightness. I didn't open the light all the way, it's hard to get into, but here you have at least a piece as proof that the tape isn't just balance resistors.

[5U4GB]

Interesting, here's my one, also not from a generic supplier, but with the standard resistor-only configuration.



Edited to add: This was lit with a flash to get light into the alu channel which was otherwise partly shaded, thus the uneven colour appearance of the LEDs.

[Edison]

Interesting, here's my one, also not from a generic supplier, but with the standard resistor-only configuration.

I don't know what chips you have installed on your LED strip, but the photo clearly shows the degradation of the phosphor and thus overheating of the chips.
The color of the phosphor determines the temperature of the color of the emitted light, the brighter the yellow, the colder the light, some LED strips have mixed chips, but mostly they alternate regularly. In your photo, you can see a series of bright yellows (according to the color, this would correspond to 5500 K and above - both the camera and the monitor can be deceiving) but then towards the right the color of the phosphor changes and it seems to gradually darken - this would indicate the use of an unsuitable source - thanks due to deviations in resistance values, the chips may be overloaded in individual sections, or due to uneven cooling - although it looks like the tape is stuck in an aluminum strip, but the chips themselves are mounted in width and because they are cooled to the wider side, in the place where there are resistors which also emit heat, they overheat and thus degrade the phosphor and the chip itself.
On my strip, you can notice the longitudinal assembly of the chips and the uniform color of the phosphor - the strip has been used every day for several hours for about 10 years, because the LED chips are cooled to the sides, there is no mutual transfer of temperatures between the components, and thanks to the approx. 5 mm thick cooler, will cool down.

[Randy222]

PWM is one way to have "dimmable" light. But that's very digital and can introduce EMI / noise.
One other way to dim is to keep it analog. As example, lower the diode amps (lower voltage), and you can do that w/o wasting power. There are those addressable volt regulators, many made with i2c interface. You can essentially dim the led by lowering voltage by way of i2c signal to the regulator. An MC that has i2c works best for this. Yep, more complicated, but less wasted power and "all" linear.

I guess one could also make a linear reg setup with a dial pot for adjustment.

[Edison]

A source with current regulation, not voltage regulation, must be used to power the LED. When using a voltage-regulated source, the lifetime of the LED can be greatly reduced

[Randy222]

How do you "regulate" current if voltage is fixed? Even PWM is not constant current, and PWM slams the led junction open and closed at whatever frequency is used. Good thing with PWM is you can actually drive the led a bit harder at lower duty cycle.
If you mean like fixed 12v (as example) and then add resistance between source and device, that's still dropping voltage on the device (also wasted power).

[cosmicray]

How do you "regulate" current if voltage is fixed? Even PWM is not constant current, and PWM slams the led junction open and closed at whatever frequency is used. Good thing with PWM is you can actually drive the led a bit harder at lower duty cycle.
If you mean like fixed 12v (as example) and then add resistance between source and device, that's still dropping voltage on the device (also wasted power).

The short answer is you are going to drop voltage somewhere (and produce some heat as a by-product).

The longer answer is that you design so that the input voltage is ~2v above what your calculations say the LED string (series or parallel) will need, and then you run with it. Dropping volts on an LDO (assuming it's not too many volts) is likely preferable to heating up a bunch of LEDs. In a Cree thermal management guide, it even says to plan for (up to) 70% of the LED current being dissipated as heat. You really don't want that if you can design around/away from it. In some specific use cases, you have to accept all that heat, and plan accordingly.

[Edison]

How do you "regulate" current if voltage is fixed?

Current regulation by resistance only is possible only with signaling LEDs with a relatively low current consumption of approx. up to 20mA, this can be applied up to approximately 20V differential voltage, then the losses are already too big.
First of all, we have to realize how the LED behaves with the increase in temperature, after reaching a certain temperature point of the chip itself, there will be a sharp change in the internal resistance and a logarithmic increase in the current - if the current maximum is not set on the source below this breaking point, degradation and possible destruction of the chip will occur very quickly .
The general rule says - with a really well-designed switching power source, we can load the LED to almost 100% of its nominal value, if we use a linear current source, the load should not exceed the limit of 75%

Industrial current-regulated sources are produced directly for specific lamps with a fixed I/U ratio or universal ones for which the ratio curve can be set, for example, using DIP or using an IR controller (my 200W LED source). The source primarily regulates to the maximum set current and the voltage follows the regulation curve.
A simple example is an ordinary laboratory power supply in CC mode.
Probably the simplest current source is a current mirror made of two transistors and a couple of resistors to set the operating points - this circuit controlled the interior lighting of my previous car for 12 years without any problem.
Thanks to the expansion of LED lights, specially manufactured integrated circuits for these purposes can be found today.
However, heat dissipation must always be ensured, preferably according to the specifications of the LED manufacturer

[beanflying]

A little late to the party but there is a some thoughts and issues here  in a thread of mine from a few years ago. Several years later and the Linear supplies over the bench are still doing what I need.

https://www.eevblog.com/forum/metrology/metrology-grade-lighting-i-am-in-the-dark/msg2225853/#msg2225853

[5U4GB]

I don't know what chips you have installed on your LED strip, but the photo clearly shows the degradation of the phosphor and thus overheating of the chips.

Actually it shows that a flash was used to get light into the alu channel :-).  That strip has been lit for maybe a minute in total, and never even got to the point of being noticeably above ambient temperature.

[5U4GB]

If you mean like fixed 12v (as example) and then add resistance between source and device, that's still dropping voltage on the device (also wasted power).

From the very basic test I ran there's no need to externally limit or otherwise try to control current, it varies based on voltage, or at least decreases with voltage drop.  This isn't meant to be any high-tech measurement so don't rely on it, I just wanted to find out what level to undervolt at to get the amount of light I wanted.

Oh, and one of the photos shows it illuminated rather than externally lit via flash, so you can see the colour is consistent in that one.

[Edison]

Actually it shows that a flash was used to get light into the alu channel :-).  That strip has been lit for maybe a minute in total, and never even got to the point of being noticeably above ambient temperature.

I took a photo with a flash and the luminophore on my LED strip is the same color everywhere, but as I wrote - both the camera and the monitor can be deceiving

[Terry Bites]

I'd use an enclosed smps and be careful about the wiring and filtering.
Your digitial scope has an smps inside it so its not really the issue of psu type.
Use star grounds and star supply wiring. Use CM and DM filters in the dc line.
I had to build a low noise lighing scheme for a metrology lab. I used a a Traco (nice) enclosed switching supply and added post regulation of the LED current with LM317HV's as the sources. Twisted pair everywhere I could. It worked out really well.
At the end of the day modern smps are fine if you take care to keep radiating loops to a minimum.

I have found that linear supplies can be harder to tame. Even noiser than a decent smps.
The charging current pulses in the rectifier circuit can create naasty EMI problems.
If you got down the linear path make sure to use snubbing caps the fullwave bridge.

[Xena E]

I have found that linear supplies can be harder to tame. Even noiser than a decent smps.
The charging current pulses in the rectifier circuit can create naasty EMI problems.
If you got down the linear path make sure to use snubbing caps the fullwave bridge.

This.
The biggest problem is this instance on the use of slow rectifiers, many, notables are the 1N 400x/1N540x, they are hardly fast enough for 50/60 Hz.

Not being able to switch fast enough, reverse current builds until the junction effectively "snaps" shut causing the RFI.

Made worse if the current gulps are overly narrow by use of gross over capacity in the smoothing. The old RoT here is 1microfarad per milliamp, (50Hz supplies).

A side project for us was the LED lighting in a development lab that had to be EMI quiet. It was found there was still a need to tame the 100Hz switching noise in the chosen linear supplies: it was just a case of swapping to UF series rectifiers in the PSUs in each panel.

Saved face for the guy who specified the linear supplies in the first place, his not apparently knowing they had their own vices.

It did ultimately prove to be the best choice to use linear with current regulation.

Though not the most efficient.

[5U4GB]

The biggest problem is this instance on the use of slow rectifiers, many, notables are the 1N 400x/1N540x, they are hardly fast enough for 50/60 Hz.

Not being able to switch fast enough, reverse current builds until the junction effectively "snaps" shut causing the RFI.

Ah, that would explain the HER 60x's in a power supply I looked at recently.  I didn't realise you needed a fast-recovery diode at 50/60Hz.

[Edison]

It did ultimately prove to be the best choice to use linear with current regulation.

Though not the most efficient.

Yes, this is what I mentioned - the maximum current load of the LED is 75% (because if we exceed this limit, there will be rapid degradation due to overheating even with sufficient cooling) and greater losses on the source - the reward is completely silent operation

[5U4GB]

A little late to the party but there is a some thoughts and issues here  in a thread of mine from a few years ago. Several years later and the Linear supplies over the bench are still doing what I need.

https://www.eevblog.com/forum/metrology/metrology-grade-lighting-i-am-in-the-dark/msg2225853/#msg2225853

I've just posted a note to my very occasional blog about a similar problem, but this time with a linear power supply where the cables were picking up noise from a nearby SMPS.  So the important thing is, no matter what you're using, run some measurements to make sure you're getting what you're supposed to be getting.