5uH LISN for Spectrum Analyzer EMC/EMI work

[NikWing]

alright, I assembled it and just made a real quick test and now I'm not sure if everything is alright
(SA setup for this quick test: preamp on, attn: 10dB, 150kHz-30MHz, 100k/1k for BW)

1st thing I notice: the RG58 BNC cable is about 30 cm. I see peaks on the SA without anything connected to the LISN when preamp on the SA is on. They disappear if I press the LISN case against the input connector GND (and probably remove the GND loop of the cable I guess)
2nd: I get random "ADC overload" warnings, for example when I connect the PSU to the LISN, if I short circuit the PSU after the LISN etc (there's no voltage on the BNC if I use a multimeter to measure)

what exactly should I set on the SA? I did not have much time to play around today, but I did not yet find out how to compensate the LISN's attenuation in the SA
I guess I should still enable the SA's attenuation and set it to 10dB
I see/measure peaks if I enable the SA's preamp but nearly nothing if it's off. But these peaks clearly come from the DUT (and change if I make changes on the DUT)

The LISN does not have the 4148 mounted yet.
Now I'm not sure where to start improving the whole setup to have something I can use to make significant and repeatable measurements.

(oh, just in case: these are my 1st steps using SAs )

[Jay_Diddy_B]

Hi,

You may see some local AM radio stations if you connect the LISN to a spectrum analyzer. You can identify if they are local stations from the frequencies.
Here is a picture from my lab.



i

So long as the interference from the local stations are around 6dB below the limits you don't have to worry about them.

Check your setting against the ones shown on the screen:

Vertical unit dBuV
Vertical reference: 100 dBuV (top of the screen)
Pre-amp gain: -10 dB (to compensate for 10dB attenuation in the LISN)

RBW 9 kHz ( use 9 or 10 kHz)

Frequency: 150 kHz to 5 MHz.



Can you post a picture of your spectrum analyzer screen?

Regards,

Jay_Diddy_B

[NikWing]

alright, let's see ...
I suppose preamp gain is Ref Offset on the Siglent

pic 1: preamp is off, ref lvl 100 dBµV

pic 2: preamp is on, ref lvl 87 dBµV is the maximum I can set

[Jay_Diddy_B]

Hi,

It looks like it is working correctly. The noise level with the pre-amp off is around 25dBuV, the same as the noise floor on my measurement.

Trying hooking up a switching power supply.

Regards,

Jay_Diddy_B

[NikWing]

I was a bit unsure about the noise, it looks less noisy in your screenshot but I did not see any reference at the 1st glance

ok, pic 1: no LISN connected
pic 2: LISN connected, DUT is a 5" Nextion Enhanced ... PreAmp is off

again, I'm a bit nervous: the moment I connect the PSU, the SA warns about ADC overload, not everytime, but 30-50%

[rauol]

Very Nice work, are the gerbers available.
Regards

[ferdinandkeil]

I've just finished my own copy of the LISN. So far it seems to work properly, however using my own equipment I could only check it for flatness from 500 kHz to 25 MHz. Will post some pictures once I've measured it with a proper network analyzer.

The PCB was designed in KiCad. Being my first layout in KiCad it might not look super pretty. It is designed to fit a Hammond 1590B diecast aluminium box. The holes I've placed did not fit the box, but luckily I could fix it using a PCB mill. Please find the KiCad and GERBER files attached to this post. The second version of the PCB should have the drills in the correct location.

Thank you Jay for making this design available 

[slug13]

I am a little behind the rest of the folks here in this blog.
I am just beginning to build the 5uH LISN per the Tekbox plans.
I have almost all the components now but am dismayed by the tiny size of the TVS diodes specified. They are only 0.8 x 0.6 mm and I don't think I will be able to hand solder them.
I wonder why the designers chose such tiny devices for ESD protection, and which are rated only at 150 mW ??
Then they add the 60 volt gas tube in parallel which seems a little on the over-kill side.
Has anyone used alternatives, like a pair of 1N4148 diodes back to back, or as I was thinking 1N5711 schottky diodes back to back?

[Ranganok]

Hello everyone,

I'm trying to build an LISN of 5uH but I run into several doubts:

- How can I measure impedance? I have a VNA (SVA1015x) and I think it has to be connected to the power ports (like attachment picture) of the LISN but I'm not sure.

- The metal box is necessary, but how much is necessary? I want to use an extruded aluminum box type (https://www.banggood.com/en/15010555mm-Aluminum-Instrument-Box-PCB-Enclosure-DIY-Electronic-Case-p-1080020.html) is it possible? Or should the enclosure of the box be without grooves?

Regards,

[ferdinandkeil]

Attached the measurements.

DUT_PORT_MEAS_PORT shows S12 from DUT port to measurement port.
DUT_PORT_SMITH shows S11 at the DUT port.
MEAS_PORT_SMITH shows S22 at the measurement port.
POW_PORT-DUT_PORT shows S21 from power input port to DUT port.

The LISN is very nice and flat up to more than 100 MHz. Should be usable at higher frequencies as well.
Impedance at the DUT port is well behaved. Same is true for the measurement port.
Isolation between power input and DUT port is high.

Measurement was done using an Agilent E5062A.

[r0d3z1]

anyone have a couple of this homemade LISN for sale ?

[jonpaul]

Hello all

The industry standard is Solar Electronics

This link has all the types and many useful app notes.

https://www.solar-emc.com/LISN.html

In our lab, we used the 50 uH 10 A version 7225-1

You can find them on eBay and at Ham Radio Fleas.

Do not count on the transient protection Zeners  to save your Spectrum Analyzer front end!

In any 2 sided LISN NEVER switch the line/neutral with analyzer attached!

Just the ramblings of an old retired EE

Enjoy

Jon

[r0d3z1]

Georges80 could you please indicate the model number and manufacturer of the metal case that you used ? Thanks

[georges80]

BOX ALUM 4.51X2.52X2.17" Hammond   1550D   Digikey   HM1184-ND


cheers,
george.

[r0d3z1]

I have finally been able to build the LISN and test it.
There are 4 test condition

• LISN connect & PREamp off
• LISN connect & PREamp on
• SA background noise & PREamp off
• SA background noise & PREamp on

the SA is a new rigol rsa3015e, the test setup is with 2 LISN and metallic reference plane. For sure I still don't have

• a good low impedance wiring between battery negative terminal and metallic plane
• a good low impedance wiring between negative lisn terminal and metallic plane. Some suggestion on how to make it good with the banana connectors ?
• the right 50ohm termination on not used lisn (I have incorrectly bought the 75ohm )
• 1uF and 1000uF capacitor on the supply side of lisn as dictated by CISPR25

What do you think about ? is it an overall good result ? I am very uncertain about the peak at 15MHz visible in the image LISN-noise-PA-0-30

Here is some screenshot, the filename comment itself.



this tool is very useful to compare the pictures.
https://www.diffchecker.com/image-diff

[prasimix]

Just sent into production a LISN PCB built using the schematics that @Jay_Diddy_B sent me . I arranged it so that it goes directly into the Bud industries EX-4521 enclosure. The front and rear panel are also PCBs, something I have long wanted to make .

[ferdinandkeil]

Just sent into production a LISN PCB built using the schematics that @Jay_Diddy_B sent me . I arranged it so that it goes directly into the Bud industries EX-4521 enclosure. The front and rear panel are also PCBs, something I have long wanted to make .

Awesome, very slick layout. Please post pictures once it's finished 

Are you planning to sell / open source the design?

[doktor pyta]

It seems that we have follower of idea by Jay_Diddy_B.
https://www.ebay.co.uk/sch/herczegz992/m.html?

[Noy]

 Die somebody build the Lisn from here:
https://github.com/LibreSolar/5uH-DC-LISN

Little differences in schematic / layout. Is it good and can be copied?

[T3sl4co1l]

Looks fine.  Give or take exact component values, or if you want limiters or attenuation or what, which is all up for grabs on final assembly.  I'd cut out the ground under the inductors, and add a few spots for optional damping resistors.  And obviously you need two for most purposes, or higher inductance and some voltage clearance for mains (probably including bigger capacitors for the voltage rating).  Not a big deal.

Tim

[Noy]

THX, i'm new to these things. Still waiting for arrival of my SA/VNA.
I'm only working on DC <60V so do i need 2 devices for DC ?

Why is this design using 2 serial caps with instead of 1 in the other?

You mean cut out on "top" on the opposite side from the inductors? I thought it is for shielding because there is no additional case top like the tekbox approach.

Why 10db? I mean 3.61 is uncommon? Better 6db (2) or 20db (10) ? I don't understand why 10db is used in this designs.

Edit: Can answer myself. Forgot power db its 10 not 20 (Volt) for 10x..

[Jay_Diddy_B]

Hi Noy,

I know as much about this circuit as anybody 

THX, i'm new to these things. Still waiting for arrival of my SA/VNA.
I'm only working on DC <60V so do i need 2 devices for DC ?

Why is this design using 2 serial caps with instead of 1 in the other?

You mean cut out on "top" on the opposite side from the inductors? I thought it is for shielding because there is no additional case top like the tekbox approach.

Why 10db? I mean 3.61 is uncommon? Better 6db (2) or 20db (10) ? I don't understand why 10db is used in this designs.

Edit: Can answer myself. Forgot power db its 10 not 20 (Volt) for 10x..

The design shown in this post:

https://www.eevblog.com/forum/projects/5uh-lisn-for-spectrum-analyzer-emcemi-work/msg404662/#msg404662

is my original design.

I gave the design to Linear Technology and they developed the DC2130A demo board that can be found here:



The Linear Technology design was copied and presented here:

https://github.com/LibreSolar/5uH-DC-LISN

When Linear Technology made their version, they made three changes:

1) They increased the thickness of the traces on the inductor wiring:




This makes the RF performance worse.
This is addressed in their demo board manual. They suggest lifting the inductors off the board to minimize the capacitance to the ground plane.

Look at page 15 of this document:

https://www.analog.com/media/en/dsp-documentation/evaluation-kit-manuals/DC2130AF.PDF

(I have since built a few versions of the my original LISN using a single-sided PCB, no ground plane, and they work even better.)

2) They added a 22uF 100V Electrolytic across the input. The useful part of the electrolytic is the ESR. It is providing damping. This can be analyzed like this:




These models look at the impedance of the LISN for different conditions on the input.
There is OPEN, SHORT and inductance. The inductance represents the wiring between the LISN and the power source.




The results show that if the wiring and the LISN is in the approximate range 0.1uH to 0.5uH the LISN doesn't meet the impedance specification.
If the 22uF capacitor, including 1 Ohm ESR, is added it does.

This is a good change.

3) Two capacitor in series.

This is done to as an extra safety measure to protect the spectrum analyzer input in the event that the capacitor fails short.
The change in effective capacitance from 0.1uF to 0.05uF has very little impact on the RF performance.

Since there already is the 9kHz HP filter, built into the LISN, there is little benefit to this change. It does not make the performance worse.


I consider this a neutral change, neither good or bad.

10dB

10dB is a very easy number to work with on a Spectrum analyzer, since they typically display amplitude on a log scale. On my spectrum analyzer, there is a setting for an external pre-amp. I set the pre-amp gain to -10dB. This corrects the display for the attenuation in the LISN.

Regards,
Jay_Diddy_B

[Noy]

 Many thx for the detailed discription.
Only question left:
Do i need 2 if im using DC only?

And some former designs from me radiated emission oder GND.
Is the shown lisn also then useable or do i need 2 then 1 for. + one for und?
Or can i simply change the in/out polarity? Or something else?
Or is it enough to put inductors into GND path?
But then which GND is used for SA connection / attentuator? Dut or supply?




Edit: GND Planes are probably added to prevent bending during reflow process.

[Noy]

Did some tweaks on the github design.
Would you suggest to cut out the GND plane on top right above the inductors?
Thought GND Plane is for shielding.

Tweaked the resistor valus to the LT ones and added a resistor in series to the 22uF cap (don't like to use a electrolytic caused by polarity and dampening can be adjusted). Changed to SMA connector (use it more than BNC).

Will order it from JLCPCB in Black.

Im still not sure if i need 2 ones for DC supply.. And like i asked if i have a noise / radiationed GND. How to measure? Additional Inductors into GND? Or how?

[TimNJ]

You basically always need two LISNs, unless the DC input wires are very short (<200mm per CISPR25), or if you have a "single wire" DC system, that is, the return current is allowed to conduct through the metal chassis of a vehicle or aircraft (DO-160, for example).

The need for two LISNs is not inherently related to the fact that AC switches polarity. In an ideal world, a device's input current is purely differential-mode. But, in the real world, there will almost always been common-mode currents on both input wires. Common-mode current must return externally, through earth, or at least the surroundings. These currents tend to cause EM radiation. The justification for why short wires, or "single wire" systems are exempt is, I believe, that those configurations are less prone to radiation in this way.

For a "normal" two wire input, the LISN should provide a symmetrical impedance for common-mode currents on both wires. I suppose there could be issues with common-mode-to-differential-mode conversion (or vice versa?) if the impedances do not match.

More generally, the impedance presented to EMI currents will be different if you leave one side out. The equivalent circuit of a dual LISN presents 25 ohms to common-mode currents, and 100 ohms to differential-mode currents. By changing the "load" seen by these signals, you may change the amplitude detected at the port on the LISN. The limits in CISPR11, etc. are defined with the expectation that the CM and DM currents see these impedances.

Something I don't understand, and maybe someone can help to clarify, is why single LISNs are allowed (for the above mentioned exemptions), even though the DM impedance will be 50 ohms instead of 100. Won't this affect the results?