How to select low distortion ferrite
So, I am designing a sub 500kHz filter with a difficult requirement for very high Q and very low intermod distortion. This seems like a juggling act at the moment as I can meet one or the other.
I have Epcos M33 core which gives me my required Q of 400 at 500kHz but when I increase the air gap (and hence also the number of windings accordingly) to give me my required intermod level the Q has drop significantly. At the moment, we are asking the customer if they will accept a performance compromise.
But anyway, what I wanted to ask was what factors to look for when selecting a ferrite for low distortion. I would imagine the following are the main ones: but I might be wrong:
1) Air gap - the more the better as you then move towards a pure air core in the limit.
2) Saturation Level -often quote for a material per volume rather than core size but important nonetheless
3) Core size - The more material you have for a particular material the high the saturation point
4) Hysteresis - Surely this is a source of distortion? And how is this specified? How can it been compared between datasheets? I have seen remnance specified which I would say was a measure of hysteresis.
5) Frequency response of ferrite - you want to be in the band of interest for a material I guess
Basically, as it's difficult to obtain samples of, and even buy at short notice, some materials and core sizes, I need to come up with a way of selecting the right materials. Does anyone have any experience in this? Distortion in ferrites seems much less documented that I would like!
Thanks
James
For the theory of inductor nonlinearity and intermodulation, I would again refer to E. C. Snelling, Soft ferrites properties and applications, 1st (1969) or 2nd (1988) edition. I know the 1st edition that has a paragraph Wave form distortion and intermodulation. The book should be present at profound technical university libraries and is recently available as a reprint. Currently Available from PSMA | Power Sources Manufacturers Association
You shouldn't be operating anywhere near the saturation level for the ferrite. Your flux excursion should be relatively small, so the actual saturation flux density shouldn't matter.
No, saturation flux density is just a function of the material. Increasing the core size will decrease the flux excursion (for a given voltage applied), which is good for low distortion though.
Also you should make sure that the cores are biased at B=0, meaning they're not conducting any DC current, and there is no remnant flux in the core. Maybe you should build a little degaussing circuit to make sure there's no Br.
The THD or whatever should also be a function of signal level. You could also tell the client that there is just a cap on signal power before distortion loses its spec.
The best way to know is to test it with various ferrite cores. Manufacturers often fail to address the distortion (no specifications are available).
Ok, thanks. It certainly seems this book is widely recommended. It might be a bit of a push to get the boss to okay this though. We're on a bit of a spending freeze at the moment and this is an expensive book! But yes, I should get it when I can.
---------- Post added at 12:20 ---------- Previous post was at 12:16 ----------
Ok, good to know I can disregard saturation level, so thanks for that. There shouldn't be any DC as I have input and output coupling capacitors to remove any potential DC.
Reducing the signal level would help but unfortunately the customer has specced the input power level with their THD. They want 0.5V rms per tone which is around 5dBm.
Thanks for your input
---------- Post added at 12:27 ---------- Previous post was at 12:20 ----------
I was worried this might be the case! I have had a suspicion that I was going to have to do that in the end. if or when I do go down this route I need to deduce a good repeatable test that can be repeated at different frequencies.
My intermod measuring apparatus that I made only give me fixed frequencies. It's also not so representative to measure an single inductor on it's own, when it is to be made into a high order filter. so I could really do with a measure of distortion other than my intermodulation measurement, that can be conducted as a standard test across cores. If it's true that I am nowhere near saturation and that isn't causing my intermodulation (IP3 specifically) then am I left with hysteresis as my cause? In that case do I make a test test up to measure my hysteresis and then quantify it in some way (e.g. measure the remnance)?
The fact, that manufacturers rarely specify these parameters doesn't necessarily mean, that they are unknown or arbitrarily different between manufacturers for otherwise similar ferrites. Ferrites can be distinguished in basic types depending on the material composition. Your previously used M33 is a common MnZn type and you'll find almost identical materials from other leading manufacturers.
I understand, that your present question is about the optimal combination of material, air gap and core size for the given problem. Unfortunately I don't have own experience related to low IM/distortion small signal design, so I can't give a well-founded answer. Literature suggests that large air gap is preferred (nearer to a fully distortionless "air" core), also most likely a larger core. Ferrite selection is more hard to guess. I would still opt for M33 like material, but higher μr ferrite should be worth a try.
Yeah, that sounds similar to what I have been deducing. Basically, in the limit, you have distortionless air core. So you move along the path between no gap and no ferrite until you reach your spec. Unfortunately at the same time your Q goes down. So there is an optimum point you need to fine.
The main problem in my case is a practical one then. I have a high order filter. I need to know what gap I need in each inductor such that the total distortion meets my spec. But, as I change the gap, I change my AL factor, which means I then have the wrong inductance, and need to change the number of turns on all of them! This then changes the flux through the ferrite, and will then change the distortion etc... So, it could end up being a long and iterative process that uses a load of litz wire along the way. But eventually, you'd reach the point where you were happy with the performance and confident it was the optimal solution. You also want it to be tunable, for production ease, but inserting a ferrite slug increases the distortion again, especially if you are on the edge of your spec. so all in all, not each to get right.
If I can find a higher μr material then that will increase my Q for sure. One intersting thing I found when trying different sized cores is that the AL factor decreases by less for larger cores. I had two different sized p-core assemblies and determined the AL factor with one half removed. My large core decreased less, which makes sense if you think about it. So, Q wise, I'm definitely better with a larger core, for this reason and because you have more area to wind on, which lowers losses in the windings too, as it decreases the number of layers you need.