Hardware engineers everywhere in the world all have to deal with that black art know as Electromagnetic Immunity or Compliance (EMI / EMC). These are the nasty radio frequencies (RF) that blast and affect other equipment or your circuits. So in this blog I look at one time when the going got tough, we just added a little wiggle.
The project I have in mind is a product that drives ‘something’ with a 400V DC square wave at around 200+ watts. The product contains a Power Factor Correction section (PFC) that takes in our normal supply voltage and using a continuous mode PFC circuit, boosts the voltage and regulates it to 400V DC. Now I’ll not go into how a PFC works, but the key point here is that the circuit has a large inductor used to store and boost the supply voltage. This inductor is cycled via a device like a power FET at some frequency – for example 100kHz.
One of the rules of avoiding EMC problems is to not switch high voltages and currents and in so doing, avoid switching frequencies – Duh! Oh well, this means we are going to have problems then me thinks!
So with all this power switching through the inductor, we gets nice EMC problems at the first few harmonics of the switching frequency. This is expected and a good EMC filter fitted in the supply lines can solve this issue. However, the real problem comes from elsewhere.
For nearly every PFC I have looked at, you will find a nice radio transmitter at around 10MHz. This is coming from the windings of the coil which are singing at this frequency. There are two ways to stop this problem: spread spectrum and/or random winding.
Spread spectrum is a commonly used term. The trick is to move the frequency around a lot so you can pass EMC. What actually happens is that the noise being generated is still there, however spectrum analyzers look at a set frequency (and small band width) for a short moment and then move to the next frequency. By wiggling your frequency around you never spend long enough at any one frequency for the analyzer to see it. Its still there, just you have cheated by bouncing around everywhere. The trick with our PFC circuit is to feed in a pseudo-random generator into the frequency control of the main PFC IC. This is normally a capacitor and or resistor that sets the timing. You can either inject a small voltage into this or place a NPN across the cap and then as the title suggests, wiggle it – just a little bit.
OK, so this will help hide emissions from your product and may help the effect on other products, but in my view you are still making the noise in the first place. So how else can we get round this problem without adding extra circuit devices. Well, the reason your inductor is singing at one frequency is because the inductance is pure – one inductance, one resonance frequency. So if you had two inductors at similar values they would resonant at slightly different frequencies. So the trick is to randomly wind your inductor to generate lots of small inductors that all add up to your total inductance required. You make lots of messy windings that crisscross and overlap backwards and forwards – the more messy the windings the better. Then in place of getting a single inductance that rings at 10MHz with lots of energy, you get a very weak one that centres on 10MHz but has a wide spread either side of it.
Making the inductor by hand, its easy to generate a messy winding, but in mass production it is hard to tell a machine to make a mess. So once again, the key word is ‘wiggle’. Set the machine to make a winding that wiggles backwards and forwards. You will get a less random winding but the overall effect is still very good.
In the product that I was talking about at the beginning, we did start with a pseudo-random generator injecting into our PFC circuit, but in the end, we found that one of our output waves forms could be coupled back and used as a wobble more than a wiggle. This, along with my messy windings, meant we whizzed through EMC with no problems and all without making massive circuit changes and without strapping on lots of EMC filters circuits.
One key phrase stays with me when it comes to EMC and getting a pass is; “first stop generating the noise before adding a band-aid”.
So have you ever added a wiggle to your circuit? Post a comment below and let me know.
I like your “messy windings” idea. It would be nice if you could buy off the shelf inductors made like that.
My only wiggles have been traces, for length matching, though I try to avoid it if I can. I it seems like electronic hardware design constantly breaks the “looks good works good” rule, as many “works good” solutions are pretty ugly.
One silicon, we may wiggle the layout of a wire to match timing delays, but that’s usually frowned upon as a kludge rather than a robust solution. I also like your messy winding idea.
I’ve never been a fan of the spread spectrum approach, as invariably the frequency of interest seek out a victim somewhere and cause no small amount of field headaches at some point. Have you ever played around with old school pi winding? I’ve often thought it might be worth investigation, albeit the efficiency loss might well shoot one in the foot as far as other standards go.