Thursday, April 17, 2014

A Crossword Just for EMC


ACROSS

3) and 6) down. If the walls have ears in this judge’s room, don’t expect a reply
7) and 8) across. EHT can cause this to light up the dark
8) See 7) across
10) A heck of a lot of power, apparently
12) The theme of this crossword
13) Germany’s own version of 12) across
14) A few billion of these involved in 7) across
16) With 15 down the bible of EMC since 1970
18) Abbreviated unit of measurement for a French natural philosopher
20) This antenna sounds like it has plane section curves with a focus
21) 1 million Coulombs per second
22) and 23) across. Samuel Plimsoll was way ahead of the game on these constraints
23) See 22) across
25) With 28 across, checking the horizon for the tallest emitter?
28) See 25 across
30) Military purchasers will buy these items if they suit, but expect them to be in the stock room

DOWN

1) Sounds like this piece of lumber only turns up now and then
2) When measuring, dBms. Watts, Amps, and Volts are all one of these
4) This is the mark for Europe
5) German scientist George involved in the resistance?
6) See 3) across
9) Electric guitar control used in compliance testing?
11) This doesn’t look good in the test report
13) Energy per unit charge produced by a dynamo
15) See 16 across
17) Never direct and undecided on polarity
19) A horse should never go hungry in this field set up
24) Times are a changing to disseminate
26) The center of attention initially
27) and 29) down. This container contains fields, not food
29) See 27 down

Answers below.












ANSWERS


-Tom Mullineaux

Wednesday, April 16, 2014

Elephant in the Test Room #3 – Fixing the Broken Automotive Emissions Test Fixture

Please note, regarding the fate of Elephant #2 – ‘Disharmony in Harmonic Limits’, it has now fallen under the umbrella of ‘Linearization of EMC Amplifiers’ and will be alluded to again during the part of the design exercise that determines practical harmonic levels.

And so, with no further ado, onto a whole new elephant…………

Elephant #3 – Fixing the Broken Automotive Emissions Test Fixture

The Room: Automotive RF Emissions Testing

The Elephant: Hard to justify systemic uncertainties

The Culprit: An over-reliance on weakly related historical methods compounded by the turning of a blind eye to known serious flaws

Note: It is important that we are all at the same place when we discuss this topic or you may be left behind, or find yourself pondering needlessly over points already covered in the recently published article ‘A Design Review of the Automotive Radiated Emissions Test Fixture’, so it will pay to read this first.

We start by describing a means of getting a feel for the present level of error when measuring the fields emitted by the wiring harness. The test currently relies on the 1m calibration data supplied with the biconical antenna.

Very simply, ten single frequency fields are emitted from the harness one at a time over 20-200MHz and a field probe is used to measure the field strength 1m away. The article already describes how the test fixture can be characterized with a signal generator providing a known spot frequency at a known injected power level.

The measurements are then repeated with the biconical antenna in place of the field probe. It is easy to convert the dBuV reading on the receiver back to the field the antenna ‘sees’ using the supplied calibration data.

Anyone out there have the facilities, time, and inclination to try this? For the purposes of this experiment we can probably devise a way of initially avoiding the need for a 50-300 Ohm transformer.

Before signing off until next time, I would estimate the figure for getting an independent RF house to review and fix the system as $30k-$50k. Even if it turned out to be twice or even three times that estimate, surely this pales into insignificance when compared to the cost, (financial and injury to reputation), of just one vehicle recall. You might think that the automotive manufacturers’ EMC departments would get their heads together and fund the project jointly. In my view, the various groups that presided over this debacle of a test set-up to date have demonstrated their ability already, and should be excluded from the project until the new fixture design has been proven.

To be continued…


Linearization of EMC Amplifiers

Homing in on the Phase Control

For optimum control of the harmonic cancellation process we need to be able to adjust the phase and amplitude of the cancelling signal. The amplitude control is comparatively simple as we can get the PC to change the output power of the signal generator as required. So for now we will concentrate on the phase control, but should bear in mind that any phase shifter we choose will be frequency dependent. That is the insertion loss of the shifter will increase with frequency, causing a reduction in signal amplitude. However this is of no great matter as we are designing a self monitoring / correcting system.

To make the objective we are seeking here perfectly clear, we will divorce our thinking from the job in hand and simply look at the output of two signal generators.

The generators are locked together by connecting the ‘sync’ connectors on the back panels and so will start in phase as shown in the first figure. The amplitudes are different only so we can see both signals when they coincide.


The second figure shows what happens when we introduce a phase shift (lengthen the electrical length of the second generator cable connection).


The phase shifter can only add electrical length to the output connection from the second signal generator and so, as shown, the black trace can only move to the right. This is a nuisance when to phase align the two signals we only need to move the black trace slightly to the left. So we will deliberately put a phase lag into the red trace. This way we can always get the two to align when we phase shift the black trace to the right. This is shown in the next figure.


The final figure shows the resultant when we press the ‘invert’ button on the second signal generator and use the phase shifter to align the two signals.


The resultant is the only signal that will appear when the two outputs are combined.
Getting back to the job in hand, we must remember that the two signals we are talking about are signals at the output of the TWT amplifier. And so we must bear in mind that we do not as yet know the insertion phase of the TWT amplifier, so the offset lag (read the length of the line stretcher we will install) and which arm it needs to go in will need to be determined when we characterize the TWT amplifier.

To be continued….

The Cellphone Threat

In this post we complete the description on the 4G orthogonality concept, then in the next post we will move onto the 4G and 3G transmitted waveform characteristics.

4G (OFDMA) Orthogonal Codes Continued……

It was explained previously that OFDMA splits the very fast serial data to be sent into many parallel, far slower, frequency separated channels sent as a tightly packed spectrum. At the receive end the slow parallel data channels are converted back into the original fast serial data stream. The separation of each parallel channel from the ‘pack’ is achieved through the fact that a signal frequency multiplied by itself and integrated produces a non-zero result, whereas when this frequency is multiplied by any other frequency in the pack a zero result is produced. The figure below shows the concept.

The variable signal generator presents each ‘bin’ frequency in turn (f1 then f2 etc) and this is multiplied with the ‘pack’. To see why only the wanted signal produces a result when mixed with the correct bin frequency we need to revisit the high school trigonometric identities we were all forced to learn by rote.

Proof of a zero result when sinusoidal waveforms of different frequency are multiplied and integrated

From the trigonometric identity cos(u)cos(v) = ½[ cos(u-v) + cos(u+v)]

∫cos(u)cos(v)dt = ½ [∫cos(u-v)dt + ∫ cos(u+v)dt]

But the algebraic sum of the area under the curve of the cosine function over one cycle is 0, so

½[ ∫ cos(u-v)dt + ∫ cos(u+v)dt] = ½ [(0) + (0)] = 0

That was easy. Now for the next.

Proof of a non-zero result when a sinusoidal waveform is multiplied by itself and integrated

From the trigonometric identity cos(u)cos(v) = ½[ cos(u-v) + cos(u+v)]

And letting v = u

cos(u)cos(u) = ½[cos(u-u) + cos(u+u)]

cos2(u) = ½[cos(0) + cos(2u)]

cos2(u) = ½[1 + cos(2u)]

Again, the algebraic sum of the area of the cosine function over one cycle is 0, so

∫ cos2(u)dt = ½ ∫ [1dt + cos(2u)dt] = ½T + ½ (0)

= ½T, where T is the time to complete one cycle

And that completes the basic concept behind 4G OFDMA.

Next time we will look at the transmitted waveform characteristic.

To be continued ……

-Tom Mullineaux

Thursday, April 3, 2014

Elephant in the Test Room #2 Continued and The Design Guide Article

Elephant #2 - Disharmony in Harmonic Limits

The Room: RF immunity testing

The Elephant: The curbing of the contribution made by harmonics to a calibrated test-field varies wildly from standard to standard, and within standards

The Culprit: Harmonic limits that are seemingly not well thought through, and / or are open to interpretation

The consequence: A customer’s product sent to two different test houses for the same RF immunity test can be subjected to test fields of very different harmonic content

Continuing…

The dilemma faced by test houses when trying to abide by the automotive -20dBc harmonic limit is that it is hard to figure out how to interpret this. As mentioned previously a test house may have been bitten when buying a 1kW amplifier after it is realized that the amplifier -20dBc specification is at less than the amplifier rated power. What if the amplifier needs to be run at rated power to achieve 200v/m? Is the test house in compliance with the automotive limit or not?

And it only gets worse. Sellers of amplifiers provide data on the amplifier performance under ideal conditions. Is the data still valid under real test conditions? You will be hard pressed to find an amplifier supplier willing to guarantee the same performance as that specified on the data sheet.

I do not envy the procurement guy that has to buy an amplifier costing in excess of $100k who through due diligence discovers he needs to buy one at double the power (and double the price) in order to guarantee -20dBc harmonics when generating 200v/m under all test conditions. That will make for a fun conversation when he talks to the company accountant.

In my view the commercial sector approach to harmonic limits is eminently more sensible where you can work backwards to the amplifier and deal with the two issues mentioned above.

And here is an entirely realistic situation that may impact on a customer’s product going through a test house. The test house offers automotive and MIL-STD testing. One of the TWT amplifiers in the MIL-STD test suite goes down right in the middle of the test run. The customer’s delivery deadline is looming and the MIL-STD chamber is booked for another customer next week. However the commercial/automotive chamber is free at present. There is every chance that the test manager will borrow a filtered TWT amplifier from the automotive suite to replace the failed amplifier. There is nothing ‘illegal’ about this but now the test applied has changed substantially. If corrected, instead of 120v/m at the intended frequency and 80v/m at the harmonic, the product is now exposed to almost 200v/m at the intended frequency. If the EUT fails, would it have passed if the original amplifier was available? And if left uncorrected, is the 200v/m achieved?

To be continued……

The Linearization of EMC Amplifiers

Now we have the intern up to speed in terms of what to expect when cancelling the harmonic noise we can start to create a proof of concept instruction set for single test frequency 1GHz.


INSTRUCTION SET

Set F2 to 2GHz
Set F2 amplitude to look-up table entry
Set F2 phase to look-up table entry
Set F1 to 1GHz
Set F1 amplitude to look-up table entry
Set F2 to ON
Set F1 to ON
Check harmonic level is as in look-up table
If not then run subroutine 1
Check Pi is as in look-up table
If not then run subroutine 2
Loop through subroutine 1 and 2 until both requirements are met
Dwell
Set F1 to OFF
Set F2 to OFF

The subroutines are required to cater for the TWT amplifier quirks we educated the intern about (non-monotonic behavior, possible changes in the power of the fundamental as the harmonic is cancelled), but more than this, we want the system to be self-correcting when under the control of the PC. Too many RF immunity tests are completed ‘flying blind’, where the GO button is pressed and everyone hopes for the best. We will get the PC to ramp up the power and fine tune the cancelling as it does so, before it dwells.

Looking ahead it occurred to me that we could use some of these low priced USB controlled ‘half brick’ instruments that rely on being connected to a PC (Sig Gens, spectrum analyzers, power meters, etc). Would be great fun, but for now let’s get the intern up to speed and get the system automated using ‘real’ instruments.

Sanity Check

Before we spend any further time and effort we should do a sanity check regarding the practicalities of the exercise.

Let’s play devil’s advocate and list two possible objections to the implementation of the new system:

1. The test time will be doubled compared to the present system arrangement.
  • Only the first 20% (possibly less) of the band is involved
  • No time is wasted investigating whether the intended test frequency or its harmonic caused an EUT failure
2. The amplifier behavior may change under real test conditions.
  • This question is a bit rich given the present ‘flying blind’ mentality
  • Once the concept is proven, the self-correction capability can be expanded to cater for further scenarios
Another feature is that the amplifier/linearizer harmonic level can be changed to suit different standards

To be continued…..

The Cellphone Threat

In this post we complete the description on the 3G orthogonality concept. It is fairly long so we will continue with the 4G orthogonality explanation next time.

3G (WCDMA) Orthogonal Codes Continued…..

If you had a shot at auto-correlating one of the special orthogonal codes (EXNORed it with itself) then you established the output of the EXNOR gate was permanently at +1 volts. It was also easy to establish that when two of the special codes are cross-correlated the average output is zero volts.

I completed the other exercises needed to complete the picture in this concept.

The first diagram ( The effect of ‘1’ and ‘0’ …) shows how a user’s data modulates the special code allocated to the user, where basically a ‘1’ leaves the code as it is and a ‘0’ inverts the code. 


The next diagram (Recovering the data …..) shows how demodulation is accomplished by simply EXNORing it with the user’s special code again (in actual fact the decoding code is the self-same allocated special code).


The last diagram (Data not recovered…) shows how another special code does not recover the data (sum of output = zero volts).


In summary, we can state the following:

On the transmit end:
  • When the user’s data is at 1, the special code is unaltered
  • When the user’s data is at 0, the special code is inverted
  • These are the only two circumstances as regards the sent modulated data
 On the receive end:
  • When the unaltered code is EXNORed with itself the gate produces all ‘1’s
  • When the inverted code is EXNORed with the special code the gate produces all ‘0’s
  • When the unaltered code is EXNORed with a different special code the average output is zero volts
  • When the inverted special code is EXNORed with a different special code the average output is still zero volts (this is because inverting a special code creates another orthogonal code)
Next time we will look at the characteristics of the transmitted WCDMA waveform spectrum.

EMC Directory and Design Guide Article

I recently wrote an article that questioned the operation of the automotive RF emissions test set-up. The article, titled, "A Design Review of the Automotive Radiated Emissions Test Fixture," appeared in Interference Technology’s EMC Directory and Design Guide magazine. I could put only a fraction of what I wanted to say within an article of reasonable length, so I will expand on this here. We will start formally next week by christening it ‘Elephant in the Test Room #3’, but for now, here is one of my thoughts on ‘testing’ this test set-up. If we were to inject known single frequencies of known power level into the test set-up wiring harness, we could use a field probe to measure the field at 1m and compare the measurement with the field strength picked up by the nearby antenna. Under normal circumstances we cannot use a field probe as it sums all the noise fields into one single output power level. However, when the noise field is one single frequency the field probe works perfectly in measuring the single emitted field.

To be continued…..

-Tom Mullineaux