Author Topic: Unusual method of measuring PV system EMI  (Read 3460 times)

0 Members and 1 Guest are viewing this topic.

Offline fourtytwo42Topic starter

  • Super Contributor
  • ***
  • Posts: 1183
  • Country: gb
  • Interested in all things green/ECO NOT political
Unusual method of measuring PV system EMI
« on: August 24, 2017, 12:59:08 pm »
I have a long telephone line to the exchange (53dB attenuation) so my broadband (ADSL) is always hard to maintain. Having checked out my PV system with the radio searching for excessive EMI I thought it was OK. But today I happened to run a broadband SNR plot (enclosed) and low and behold it is being modulated ~1db according to whether the sun is bright or diffused, this in turn determines if the converter runs in buck or boost mode.

So it seems the buck mode of the converter has an EMI problem and needs attention!

I am not entirely surprised, I call it buck as it reduces the PV input voltage to a lower voltage across the fixed heater load to allow the panels to maintain MPPT in low light, but it's actually a PWM, hence the long (maybe 10 Mtrs) heater connecting cable carries the switching current  :o but I thought the common mode choke and filtering components would be sufficient (slow down the edges).

Either I have to build a real buck converter (big job, extra inductor etc) or another idea is to reduce the PWM frequency as that will reduce the density of the fast edges.

Anyway if anybody is interested in a different kind of quantitative EMI sensor there it is :)
« Last Edit: August 24, 2017, 01:00:47 pm by fourtytwo42 »
 

Offline Inverted18650

  • Frequent Contributor
  • **
  • Posts: 875
  • Country: us
    • Test Gear Addicts
Re: Unusual method of measuring PV system EMI
« Reply #1 on: August 25, 2017, 02:06:53 am »
Are you also using this fixed heater load to heat your water in the house? If not, is there another reason not to build the buck?

Offline fourtytwo42Topic starter

  • Super Contributor
  • ***
  • Posts: 1183
  • Country: gb
  • Interested in all things green/ECO NOT political
Re: Unusual method of measuring PV system EMI
« Reply #2 on: August 25, 2017, 04:55:25 am »
Yes it is the water heater. The only reasons I don't want to build the buck are economy and space. According to simulation I should be able to reduce EMI substantially by reducing frequency and remain within the ripple current capability of the other components so I am going to try that first especially as it is a simple software change.
 

Offline fourtytwo42Topic starter

  • Super Contributor
  • ***
  • Posts: 1183
  • Country: gb
  • Interested in all things green/ECO NOT political
Re: Unusual method of measuring PV system EMI
« Reply #3 on: August 26, 2017, 12:00:23 pm »
Well I tried different PWM frequencies and it had no effect whatsoever (maybe unsurprising) from 0741 onward in the plot several different frequencies were tried ranging from 16Khz to 64Khz. The only reason for a change in level was the increasing power (as it was early morning) over the duration of the test.

Another strategy I had thought of was to remote the PWM stage itself to immediately next to the heater thus dramatically shortening the loop length of the modulated signal however this has quite a few implementation difficulties.

In simulation a modest LC filter added to the PWM output has a dramatic effect (see fft but ignore title) so I will pursue this path with the contents of the junkbox and see what transpires with a real field trial :)

Frequencies of interest as far as my broadband goes are approx 138-860Khz.
« Last Edit: August 26, 2017, 12:02:26 pm by fourtytwo42 »
 

Offline fourtytwo42Topic starter

  • Super Contributor
  • ***
  • Posts: 1183
  • Country: gb
  • Interested in all things green/ECO NOT political
Re: Unusual method of measuring PV system EMI
« Reply #4 on: August 27, 2017, 10:10:55 am »
Started building the LC filter today, junk box gifted four 2.2uF/250V poly capacitors (brown jobs in picture) and an ETD34 gapped at 2mm. Wound this with Litz made up of 3*1mm, got 13 turns on and put it on my inductor tester.
2nd pic is scope shot at 8A/cm vertical showing saturation at ~28A & inductance as 18uH (V is ~14.5).

This will do the job perfectly. The design equations gave a somewhat higher inductance and that's why I prefer to measure it :)
 

Online T3sl4co1l

  • Super Contributor
  • ***
  • Posts: 21608
  • Country: us
  • Expert, Analog Electronics, PCB Layout, EMC
    • Seven Transistor Labs
Re: Unusual method of measuring PV system EMI
« Reply #5 on: August 27, 2017, 08:40:30 pm »
Mind that it's probably common mode noise, not differential.

Tim
Seven Transistor Labs, LLC
Electronic design, from concept to prototype.
Bringing a project to life?  Send me a message!
 

Offline fourtytwo42Topic starter

  • Super Contributor
  • ***
  • Posts: 1183
  • Country: gb
  • Interested in all things green/ECO NOT political
Re: Unusual method of measuring PV system EMI
« Reply #6 on: August 28, 2017, 05:55:58 am »
Yup you may be right but I hope it turns out to be the easy option :) There is some evidence for it as the broadband only gets severely hit when the pwm is chopping the heater current, when the boost stage is just boosting the voltage (>600W) there is no noticeable broadband snr loss. Only a trial will tell :)
 

Offline fourtytwo42Topic starter

  • Super Contributor
  • ***
  • Posts: 1183
  • Country: gb
  • Interested in all things green/ECO NOT political
Re: Unusual method of measuring PV system EMI
« Reply #7 on: September 01, 2017, 06:10:35 pm »
Well I been quiet for a while due to various states of confusion and frustration!

Firstly trying to re-use components long forgotten in the bottom of the junkbox can be very misleading, in my case some ETD34 ish cores!! ish because unlike the modern variety they have grooves up the sides of the core (see 1st pic) I think these are Mullard/Phillips era (1970's) and there is no data anymore BUT at the time I was working on around 15-20Khz bipolar smps's so I dont reckon the material is that great and OMG did I get some core heating at a measly 32Khz. Anybody like to hazard a guess ?

Well then there's the scope (OWON VDSI1022I) on the whole it's not bad for ~£100 BUT for some reason when its got a 32Khz signal with lot's of ringing it swamps the USB so bad it cut's out the mouse, only way to restore mouse for a few 10ths of a sec is to unplug it! Eventually only scheme I had to get shots recorded was to unplug usb of scope that freezes display then restore mouse to do screenshot, very frustrating my laptop may be old but not that bad! (T2300 @ 1.66Ghz +2Gb ram), so down mark for OWON and good reason to have standalone scope.

Then there's the LTspice model of a lossy transmission line, it doesnt work at DC, seems to connect input - to 0V internally, very confusing I had to make my own transmission line model.

So we have bad components, bad scope and bad models, kinda sticks you up good and proper! So I resorted to empirical design (never one to give up). This is a very peculiar filter as it is neither a straight EMI filter in a DC line nor a buck converter output filter but somewhere in between so the published maths for design don't work.

The filter is to round off the edges of a 600W pwm where the load is resistive but remote along a transmission line!

I tried a ring core (classic green/blue) instead of the ETD34 see second pic but again the core heating was blistering and again god knows what the material or data is! although the same core performs well in common mode (3rd pic).

However the noise problem is not improved by common mode filtering, only differential, this being not entirely surprising given the circuit carries 600W at ~120V peak!

Eventually I removed an ETD54 inductor from my sleeping GTI project and threw that at it, success at last!! This guy is Litz (home made) wound with a 2.5mm gap, success, cool running and broadband restored to normal!!

BUT I want this inductor back in my GTI project..... I have a nice pair of ETD44/N87/2mm cores and I know there genuine having been bought new as spares for another project, I reckon they will do nicely as all the data collected from the ETD54 solution shows they should albeit at a reduced inductance according to my revised LTspice simulations so the bobbin was ordered today and I have the material to make the litz.

So to continue with the simulations a critical factor is the noise in the broadband downlink frequency range (in my case 132-860Khz. The 4th pic shows the effect of changing the pwm modulation frequency from 32 to 125Khz and surprisingly the lowest frequency provides the least noise and at the same time reduces the skin effect problems (meaning the litz build is simpler). This business of litz in an inductor is not something people are familiar with as normally an smps output inductor does not handle significant power at high frequencies, well this inductor does!

The 5th pic shows the difference between just a simple common mode choke and two values of big inductor, the simple common mode choke just doesn't cut any ice!

Will update hopefully next week if there is enough time for the build before I go away on holiday....hooray!!
« Last Edit: September 01, 2017, 06:36:43 pm by fourtytwo42 »
 

Online T3sl4co1l

  • Super Contributor
  • ***
  • Posts: 21608
  • Country: us
  • Expert, Analog Electronics, PCB Layout, EMC
    • Seven Transistor Labs
Re: Unusual method of measuring PV system EMI
« Reply #8 on: September 01, 2017, 10:09:20 pm »
Well then there's the scope (OWON VDSI1022I) on the whole it's not bad for ~£100 BUT for some reason when its got a 32Khz signal with lot's of ringing it swamps the USB so bad it cut's out the mouse, only way to restore mouse for a few 10ths of a sec is to unplug it! Eventually only scheme I had to get shots recorded was to unplug usb of scope that freezes display then restore mouse to do screenshot, very frustrating my laptop may be old but not that bad! (T2300 @ 1.66Ghz +2Gb ram), so down mark for OWON and good reason to have standalone scope.

You've given very little detail indeed about your setup, but I will ask one question:

Connect the probe tip, then the probe ground, one at a time, separately.  Which connection causes the USB to drop out?

Insulting a manufacturer for the shortcomings of an industry standard, that's unsuitable for the conditions you've put it under, is a very dubious statement to make.

Quote
However the noise problem is not improved by common mode filtering, only differential, this being not entirely surprising given the circuit carries 600W at ~120V peak!

This statement has no supporting evidence, and the power level is a non sequitur.

Quote
The 5th pic shows the difference between just a simple common mode choke and two values of big inductor, the simple common mode choke just doesn't cut any ice!

Again, you've shown a few pictures of components, and made a very bold, unrelated claim.

It is very possible that your diff noise is so bad, that it knocks out everything to begin with.  That doesn't mean it's not a CM problem, that just means the offending source needs to be treated accordingly.

The process of noise elimination is hierarchical, working from the worst offender, to the least.  You must first identify the loudest source, then treat it, then move to the next, and so on.  (In the process, it may happen that you've reduced all other noisy sources, and one of the original noisy sources needs further treatment.)

Tim
Seven Transistor Labs, LLC
Electronic design, from concept to prototype.
Bringing a project to life?  Send me a message!
 

Offline fourtytwo42Topic starter

  • Super Contributor
  • ***
  • Posts: 1183
  • Country: gb
  • Interested in all things green/ECO NOT political
Re: Unusual method of measuring PV system EMI
« Reply #9 on: September 02, 2017, 07:00:11 am »
Well then there's the scope (OWON VDSI1022I) on the whole it's not bad for ~£100 BUT for some reason when its got a 32Khz signal with lot's of ringing it swamps the USB so bad it cut's out the mouse, only way to restore mouse for a few 10ths of a sec is to unplug it! Eventually only scheme I had to get shots recorded was to unplug usb of scope that freezes display then restore mouse to do screenshot, very frustrating my laptop may be old but not that bad! (T2300 @ 1.66Ghz +2Gb ram), so down mark for OWON and good reason to have standalone scope.

You've given very little detail indeed about your setup, but I will ask one question:

Connect the probe tip, then the probe ground, one at a time, separately.  Which connection causes the USB to drop out?

Insulting a manufacturer for the shortcomings of an industry standard, that's unsuitable for the conditions you've put it under, is a very dubious statement to make.
I wasn't aware I was insulting anyone and I am also not aware that an OWON USB scope is an industry standard but there are certain conditions under which it does not behave very nicely! You may or may not be aware this scopes USB connection is galvanicaly isolated.

Quote
However the noise problem is not improved by common mode filtering, only differential, this being not entirely surprising given the circuit carries 600W at ~120V peak!

This statement has no supporting evidence, and the power level is a non sequitur.
This is a simple blog, not a legal submission and it is intended to be reasonably SHORT!

Quote
The 5th pic shows the difference between just a simple common mode choke and two values of big inductor, the simple common mode choke just doesn't cut any ice!

Again, you've shown a few pictures of components, and made a very bold, unrelated claim.

It is very possible that your diff noise is so bad, that it knocks out everything to begin with.  That doesn't mean it's not a CM problem, that just means the offending source needs to be treated accordingly.
Again you seek to criticize the work of others for some purpose that is entirely unclear to me.
Quote
The process of noise elimination is hierarchical, working from the worst offender, to the least.  You must first identify the loudest source, then treat it, then move to the next, and so on.  (In the process, it may happen that you've reduced all other noisy sources, and one of the original noisy sources needs further treatment.)
I think we all learn more and more about EMI everytime we encounter a severe problem.

« Last Edit: September 04, 2017, 03:52:03 pm by fourtytwo42 »
 

Offline fourtytwo42Topic starter

  • Super Contributor
  • ***
  • Posts: 1183
  • Country: gb
  • Interested in all things green/ECO NOT political
Re: Unusual method of measuring PV system EMI
« Reply #10 on: September 04, 2017, 04:16:36 pm »
Hohum what can I say I think I may be choking on humble pie! BUT sometimes when the fat's catching fire it's hard to always react appropriately!

Anyway there is definitely a differential and common mode problem, I am sorry to be short on detail but I don't have much in the way of diagnosis tools other than using LTspice to model possibilities and field trials to try solutions. Like I said at the beginning my best tool for measuring results is actually my broadband modem in conjunction with the dslstats software package.

I built the new differential inductor as the bobbin arrived today, I chose to use five strands of 0.6mm wire as the best compromise between current density at various frequencies and the complexity of making it (plus of course what sizes I had available). As I think I mentioned before as far as my broadband is concerned it seemed my lowest easily available frequency of 32Khz is actually better than the higher alternatives.

First three pics show winding, I use the vice in the garage as my 3rd hand holding the other end of the wire. You can see the core gap in the 3rd pic. 4th pic is connected to the inductor tester, result in 6th 4A/cm vertical shows saturation knee at ~13A and an inductance of ~290uH given a 14.5V supply. I should have said somewhere the inductor handles <=7.5A DC or <=6A pk-pk at 32Khz, design flux density is <=200mT.

Installing this inductor on it's own (no common mode) was a disaster, but also re-instating the common mode wound on a blue/green toroid worked wonders, no detectable loss of broadband snr when the solar is in pwm mode at last :):):)

The 6th pic shows the installation with the common mode wound with black & green wires to the right immediately at the output terminal block, the new ETD44 is balanced to it's right waiting final physical fixing.
The other inductor on the veroboard with a flux strap is the boost mode output inductor.

Last picture is nice peaceful broadband snr margin, at least as good as it ever gets for me :)

Guess I should stop blogging now, don't want to annoy anyone........
« Last Edit: September 04, 2017, 04:54:53 pm by fourtytwo42 »
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf