5uH LISN for Spectrum Analyzer EMC/EMI work

[georges80]

Hi Folks, I recently purchased a Rigol DSA815-TG primarily for EMI debugging/snooping. I've purchased and made a few near field probes and current clamps and been busy crimping coax connectors to RG174/LMR100 and generally getting the RF side of things put together.

So, next on the list was a LISN to check conducted noise issues from DC power inputs. Saw the Tekbox unit and since it is nominally open hardware with schematics/bom and layout available I figured it would be cheaper* (and more instructive) to build one up versus purchasing it ready made.

First step was to verify component availability from the tekbox bom (mostly digikey stuff) and then do my PCB layout since I prefer to have my own database versus someone elses gerbers. Created a few footprints and then layed the board out being reasonably faithful to the original layout. I used a gerber viewer to measure some of the key dimensions (especially the inductor spacing) to replicated it into my pcb cad file.

To get a netlist I entered (from scratch) the schematic into Orcad and here's a screen capture,

essentially the equivalent of the tekbox schematic (can be found at www.tekbox.net):



Once I had the netlist I could import it into the pcb cad program and verify all the footprints and connectivity. A couple of hours later and here's a screen capture of the PCB layout I put together:



The layout was sent off to PCB fab about a week ago, so I should receive boards back this coming week. Simple 2 layer board, soldermask/silkscreen etc. Ordered 20 boards since their individual cost is somewhat minimal compared to setup/shipping. More about that later. I ordered components from Digikey & Mouser and most of them are pretty cheap so 20 sets of parts are in my hands. The only parts I didn't order up in volume were the box, bnc connector, GDT etc. I order from Digikey often (too often), so getting more of those parts later is a non-issue.

With one box in my hands I decided to move ahead (since boards are in fab and dimensions are cast in stone) and start drilling some of the holes - for the PCB and BNC to start with.

So, here's the box with a few holes, small drill press makes life easy and a step drill does a neat job on the larger BNC hole. Next to the box is a 1:1 print of the layout, makes it easier to visualize and verify mounting location and hole placement.



To verify the air wound inductors were the values claimed in the original schematic I wound a test inductor with some scrap wire onto the former and then checked it with my DE-5000 and sure enough inductance was in the 1uH range. I'm still waiting on the nominal new 1.4mm diameter (16awg) enameled wire to arrive - should turn up Monday. So, I turned a 17mm former on my minilathe in preparation for winding the actual inductors.

Here's a picture of the former machined from some scrap delrin with the inductor wound on it:



This was wound from scrap wire (16 awg as it turns out) that I recycled from a power inductor. I had to straighten the wire so it still looks a bit rough. The results should be nicer when I get the new wire that comes on a spool.

Based on the physical dimensions of the inductor, I entered the parameters into an online calculator:

http://wcalc.sourceforge.net/cgi-bin/air_coil.cgi

The coil former is 17mm diameter, so the effective coil diameter (at 1/2 the thickness of the wire) is (17mm + 0.7mm x 2) = 18.4mm. The coil measured length is approximately 12.4mm. Wire diameter is 1.4mm and 7.5 turns. Calculated frequency independent inductance is about 1.02uH



and here is the DE-5000 reading of the inductor. Certainly close enough. I did calibrate the DE-5000 with the croc clip adapter plugged in.



Pic of some of the parts waiting for the PCB's to turn up:



That's where I'm up to at this point. Hopefully later in the week I'll have the PCB's in hand and can assemble up my unit. I'll post up more pics as the build proceeds.

*So, what about the other 19 boards?? I figure if some folk on this forum want to buy a kit I can purchase the remaining components and ship them out as kits. I'll include the PCB, all the components, banana jacks and the BNC that need to be soldered to the board and the wire that you'll need to wind on a former. I won't include the box. Cost to cover the PCB/components (likely $30 per 'kit') + mail/packing (which will depend on US/international).

Most of the discrete components are SMD, so you need to be comfortable soldering those parts - though on this board they are quite large (0805 is the smallest part)... Anyhow, first let me get the PCB's in and build one unit up.

cheers,
george.

[Circuitous]

Neat!  I look forward to updates on your project.  I expect to need a LISN in a few months for another project.

[georges80]

The new 16g enameled wire turned up this morning.



I made a mod to my former, drilled a hole in the side to hold one end of the wire and also made a bit of a trench (hot soldering tip) for the wire taper. Makes winding the coils so much easier.

Winding a coil on the former:



4 coils done, just need the PCB's to turn up:



cheers,
george.

[Rufus]

Saw the Tekbox unit and since it is nominally open hardware with schematics/bom and layout available I figured it would be cheaper* (and more instructive) to build one up versus purchasing it ready made.

As I pointed out in the blog thread the Tekbox design and the way Dave used it is garbage. I was hoping someone more expert would comment but I guess we haven't got any here or they couldn't be bothered.

[georges80]

Feel free to explain where the issues with the unit are. It is meant to be used with a ground plate and screwed/bolted to it as per their manual. The ground posts on the unit are the grounding points. Dave's video of course side steps the full setup procedure presumably to simplify his task in the demo.

I hope you can educate us with why it won't work as shown in the tekbox manual for DC power supply conducted emissions. I'm NOT planning to use the unit for accuracy, it's use in my case it to identify the conducted emissions (frequency/harmonics etc) and then use that information to resolve issues that are in the design and/or front end filter circuitry.

cheers,
george.

[Rufus]

I hope you can educate us with why it won't work as shown in the tekbox manual for DC power supply conducted emissions. I'm NOT planning to use the unit for accuracy, it's use in my case it to identify the conducted emissions (frequency/harmonics etc) and then use that information to resolve issues that are in the design and/or front end filter circuitry.

I already did but I will say it again

As said in the mail bag thread it is arsed up anyway. You can't properly measure anything on a power supply pair with a single channel LISN. If you stuck a capacitor across the UUT side of that LISN you would short out most of what you are measuring while such a capacitor would have no effect at all on common mode signals.

That LISN is crap being manufactured and documented to support the way Dave used it.

You absolutely don't connect equipment ground wires to the ground plane. A single channel LISN is only useful for testing equipment which has a single wire.

[georges80]

You seem a little 'excited'... Anyhow, I'd be glad to read some explanations of how it is all meant to be designed and used - got any good links for me to read up on?

cheers,
george.

[T3sl4co1l]

Simple remedy: make two!  Which fortunately you have the boards for

Remember of course, it's for automotive work or whatever, whereas an AC line grade unit needs 150uH.

I made one of those.  It's untested as yet, but evaluating the inductors seems to suggest it won't be grossly (like >10dB) out of whack.



Tim

[georges80]

Rufus: Well, I've done some reading and searching on the web and I've found several descriptions of a 5uH scheme with the same basic topology as the tekbox LISN. They are indeed used to measure conducted EMI in vehicle type applications, which is exactly what I need to use it for (DC power source/battery feeding the DUT which in my case is a LED driver).

So, for my use, the tekbox design certainly appears to be applicable and not 'garbage'.

cheers,
george.

[Lunasix]

I also made one, because it was too expensive to order it for France. I made inductors with a lathe, very easy and perfect result, a copper spring ! I've checked my boards and improved power supply with a better filter design. But I'm not an expert...

[georges80]

Blank boards arrived today from the PCB house:



Unfortunately some shelving arrived today too and reorganising the mess I call my work bench takes priority. So, it'll be tomorrow before I start to assemble my LISN. More pics to come then.

cheers,
george.

[EEVblog]

Rufus: Well, I've done some reading and searching on the web and I've found several descriptions of a 5uH scheme with the same basic topology as the tekbox LISN. They are indeed used to measure conducted EMI in vehicle type applications, which is exactly what I need to use it for (DC power source/battery feeding the DUT which in my case is a LED driver).
So, for my use, the tekbox design certainly appears to be applicable and not 'garbage'.

That is correct as I understand it too. I may have failed to mention in the video that there are many testing scenarios, and the Tekbox is suitable for just a subset of them.

[Jay_Diddy_B]

Hi,
I am going to share my implementation of a 5uH LISN. I used carefully selected SMD inductors instead of the traditional method of winding air-cored inductors. The inductors are Wurth 744314110.

I also incorporated a combination 10dB attenuator, limiter and 9kHz High Pass filter into the design. The attenuator / filter is to protect the spectrum analyzer from damage.

The style of construction was chosen to make the LISN performance very predictable.


Schematic
This is the schematic diagram:



Attenuator / Limiter / Filter Design

LTspice model:



Modeling Results:



Board Design



Construction
The board is designed to replace the lid of a Hammond 1590 diecast box.



Testing
This graph from a HP 3753C Network Analyzer shows the attenuation from Load Connection to the Spectrum Analyzer connection. It is flat within 0.5dB over the range 2MHz to 100MHz.



This Smith chart shows the input impedance from the load connection. After 2.6MHz the input impedance is pretty close to 50 Ohms. This matches the requirements for a CISPR 25 LISN.



The performance of this LISN has been compared to a commercial LISN and found to agree very closely.


Jay_Diddy_B

[T3sl4co1l]

About the only thing better you could do -- I would guess -- is staggered inductor values, or maybe some C or R+C on some of the inductor nodes (taps).  Exactly how much (of either case!) would depend.

Either way it's quite flat, as it should be!

Tim

[Lunasix]

Here is my lisn made according Tekbox schematic. Probably not perfect, but helped me to improve switching power supply on my boards.

[georges80]

Finished the initial tidy up of my work bench:



and soldered some of the parts down on the new board:



and the back



I have the banana jacks turning up hopefully Sat, Mon worst case and then will drill their holes  wire things up. Also a couple of components that I forgot to order before will turn up.

Here's a trial fit into the box



and the BNC (have a locknut coming just to make it more structurally sound)



I'll have to remove the enamel coating on the ends of the 4 coils I wound and solder them in place too.

More updates in a few days. I've also ordered enough parts to kit up 10 units for folk that have PM/Email me with an interest to buying a kit or two.

cheers,
george.

[georges80]

Remaining few parts turned up and I finished assembling the board. Drilled the holes for the banana jacks and installed the finished board into the box:



First test was to play with a LED driver (the DUT) to measure the conducted noise back to the power supply. Just a rough test to get a feel with how things work:



So, with the DUT connected directly to the LISN I scanned and stored the yellow trace. Lots of conducted noise up to around 300MHz. Then I wired inline (you can see the red/black croc clips and blue/yellow wires) a simple LC pi filter built rough as guts christmas tree style. The purple scan shows the noise level with the pi filter in place. The noise floor is at -68dbm with the DSA settings.



Very pleased with the instant gratification. More time to learn/educate myself with the capabilities, but for a quick inspection and assessment of problems and possible solutions it seems to be a great little addition.

I have parts on hand (all except the box). $30 per kit + postage (based on US or international) will cover costs. I'll provide the schematics & gerbers (including Orcad & PADS datafiles, I'll put them up online).

cheers,
george.

[Jay_Diddy_B]

Hi,

It looks good.

It would be interesting to see a sweep from 150kHz to 5MHz with an RBW setting of 10kHz. You should be able to see the switching frequency and its harmonics.


I would recommend using a 10dB attenuator on the Spectrum analyzer to protect the input. It is much easier to replace than get the analyzer fixed. We an HP SA you set a pre-amplifier gain. When I use an external attenuator I set the SA pre-amplifier gain to -10dB so the display on the screen is corrected for the external attenuator.

You can also use the tracking generator in your Spectrum Analyzer to check the LISNs performance. Feed the Tracking generator to the load side of the LISN and the BNC connector to the SA input.

Regards,

Jay_Diddy_B

[georges80]

Hi Jay,
 yes, I have a 10db attenuator in my "bag of tricks" - but thanks for reminding me to USE it

The performance should match the texbox data (since it's a "clone"), but I do plan to play with a lot more and sweep it etc.

Lots more testing to do as I get time (precious commodity!).

I've decided the easiest way to kit these - after sticking components down on a schematic printout is to just solder the R's & C's down and provide semi-built kits. Fast to just put some solder paste down and reflow than tape them down on a sheet of paper and less chance of folk loosing parts or mixing up the C's... Folk will need to solder down the mov, gdt and wind their own air coil inductors etc.

cheers,
george.

[georges80]

Had a bit of time to do a scan of the LISN. First step was to get the signal through the coax and to the LISN DUT connectors. Obviously signal integrity is key to getting a good scan, so I ended up stripping just the end of the coax and enough ground braid to touch the case (bare metal) and the center conductor to touch the banana jack pin.

I did normalise the TG with the same coax making contact with the BNC center pin of the input cable. Obviously the ideal case would be to have a coax BNC soldered right onto the PCB. As is, the TG signal is traveling through the banana jack and the wire that solders down to the PCB. Not the greatest impedance match for doing a scan.

Nevertheless the scan looks pretty damn good out to 180MHz. If I can be bothered I'll see about soldering a coax connector onto the PCB to get a better scan of the LISN characteristics, though it seems to perform great all the way to the 100MHz 'spec' and pretty good even way beyond 180MHz if you can live with 2db 'ripple' (some of which may just be an artifact of the coax/banana jack connection.



cheers,
george.

[georges80]

So, I soldered the coax directly to the PCB (no BNC), just direct soldering to the LISN DUT input and the internal BNC ground.



Doing that of course improved the signal integrity of the measurement and the LISN measured performance is much better than the previous scan.

Here's the full 1.5GHz scan to see how it behaves:



Then a scan to 500MHz.



A scan up to 200MHz:



and finally the low freq (up to 5MHz) performance:



So, from about 2.5MHz to 200Mhz it is flat to within about +/- 0.25db. I consider that pretty good performance and will now move on to actually using the unit for its intended application.

cheers,
george.

[T3sl4co1l]

Willing to bet the junk over 300MHz or so is due to parasitic resonances of the coils (helical waveguide modes) and board.  Maintaining a 50 ohm characteristic impedance from DUT connectors to the BNC would help (of course to be fair, the DUT side needs to be a clean 50 ohm termination as well).  Graded choke sizes, rather than equal dimensions stacked, would probably help.  (Really wideband bias tees use ferrite cored, conical inductors!)

Just for S&G, not really useful... I wonder, do you have any ferrite rods, or shapes that would fit inside the coil?  A piece added to the supply end might absorb some of the lower frequency energy bouncing around.  However, placed in the DUT end, it would probably attenuate too much.

Tim

[georges80]

The tekbox design is meant to be for up to 100MHz, so the performance is 'good enough' given it seems very flat all the way to 200MHz. My application is to ensure that my DC : DC LED drivers are 'quiet' through the FM band and this will deal with that. Even to 475MHz is only deviates about 1db - I'm happy enough.

It's primary use (for me) is to evaluate input power filtering to my drivers and this unit appears to easily be able to meet that requirement.

If you want to experiment - feel free to buy a kit  , but thanks for the suggestions. When I have some free time I may play with it some more, but at this point I need to USE it.

cheers,
george.

[EMC]

Hello,

I have been involved in automotive EMC for some time. I also have a DSA815 and was thinking of making some 5uH V Networks, or LISNs.   These look good.   A couple of comments:

> air core inductors are definitely best
> housing should be diecast but will impact on the air core inductors
> phase is also specified in CISPR 16
> impedance & phase must be verified, including housing influence, with a VNWA (see DG8SAQ https://www.sdr-kits.net/)
> impedance & phase verification done at the DUT side with PS side shorted
> the network becomes more & more useful the higher it goes at 50 ohms; you start to get confidence in radiated performance
> only one input and one output required; everything else is connected to the diecast housing
> they must be used in pairs, one for positive one for negative; best bonded to a small ground plane
> only one at a time connected to the DSA815; the other must be terminated with 50 ohms on the RX output
> 1uF required across the power supply input (option to switch out for transient immunity testing)
> the MOV needs to be switched out for transient immunity testing
(the noise under test develops voltage over the specified inductive reactance; the 1uF gives low impedance and is critical/fundamental to a good result)
> best operated from a battery (R_int typically 0.01 ohm) to ensure low impedance at the input with a PS (R_inttypically 0.10 ohms) in parallel for charging between tests

When used in pairs the test setup emulates intended environment; i.e. power cable to the DUT.   If the positive result is higher than the negative result then differential suppression, cap to ground.   If positive and negative results are the same then common mode suppression.   If negative result higher than positive result then DUT ground scheme is wrong.   This is all just very general and typical.   Result will have two types of 'character'; broadband and narrowband.   Broadband to be measured with quasi peak (9kHz BW <30MHz & 120kHz BW > 30MHz); 'brush on rotor' type noise.   Narrowband to be measured with average detector; uP XTAL and harmonics.

Typical Limits in dBuV:
LW 0.15 to 0.30 MHz QP 77 AV 70
MW 0.53 to 1.8 MHz QP 57 AV 50
SW 5.9 to 6.2 MHZ QP 52 AV 45
FM 76 to 108 MHZ QP 37 AV 30

Steve

[didjado]

i have had some Pcbs made and have a quite a few left over if any one wants the smt stuff is all placed by machine and all you have to do is put the  inductors and the in put out put wires on...

i needed one and put it on a panel that was being made now i  have heaps of them... i will put some images up  so that the response can be viewed

if any one is keen