preselector schematics/design for conducted emissions measurements?

[CopperCone]

Preselectors are expensive and I have seen people build home made filters that are quite good or even beat out stuff thats for sale.

Are there any good designs out there? Not sure how you would synch it to the sweep on the SA

As far as specs go, I think the bare minimum for sale in EMI receivers is like 2 bands between 30MHz, 100KHz and 30MHz filtered, but I believe I have seen some that break up the 30MHz bandwidth into 8 or more bands.

Would have to be built with beefy magnetics to discourage distortion. I would be interested in a 3-400MHz conducted bandwidth though.  Commercially it seems that LISN go up to around 400MHz or so. Not sure how many bands it would need, more is better I guess but it becomes unreasonable.

Though I think how things are, a 30MHz one and a 300MHz one with wider bands would be more reasonable since you would not go past 30MHz normally and the signal power past 30MHz conducted would be low, but it might be useful for broadband antenna testing.

What filter order and topology would be a good start for this design? Not sure if you would do butterworth since you would probobly use tables on a  computer to modify the collected data, so you can account for ripple..

Appendix C gives an idea of commercial specifications on filter performance but does not go into specifics related to the questions posed:
https://www.keysight.com/upload/cmc_upload/All/11960A.pdf

[German_EE]

Start your preselector with a wideband transformer that has a 9:1 impedance ratio, it's easier to do a sharp bandpass filter at 5.5 ohms than at 50. You then have your inductor and capacitor in series and finish off with a second 9:1 to bring things back to 50 ohms. The bigger you can make the magnetics the better, my HF preselector covers from 200 KHz to 30 MHz in six bands and it uses 3 x two inch diameter toroids for the two transformers and the tuned element.

Then again I'm not doing conducted emissions measurements, I'm trying to pick ham radio signals out of the noise in the middle of a small city.

[CopperCone]

What do you think is a reasonable pass band to set for designing such a thing?

If you want 40dB ultimate attenuation, and you set 4MHz bands, where should the 40dB hit? How did you structure yours?

[CopperCone]

also why is it easier to build filters around a lower impedance?

[T3sl4co1l]

Well, it's easier to do series bandpass tanks at 5.5 ohms, but quadratically harder to do parallel tanks.  You need both to do more than 1st order with a canonical ladder topology.

Best is the geometric mean of whatever resonator impedances are easy to make.  Capacitors are limited by ESL, and inductors are limited by Q, EPC and proximity.  Not hard limits, but at the very least you need to take account of these properties in the design and adjust it accordingly (parasitics act as frequency-dependent impedance transformations, changing the required resonator and port impedances on either side of a given resonator -- and changing the frequency response so other parts need to be tweaked).

Tim

[German_EE]

Well, the question about the 5.5 ohm impedance has already been answered.

Elsewhere in my transceiver I have six low-pass filters at the following frequencies:

1 MHz
2 MHz
4 MHz
8 MHz
16 MHz
30 MHz

These appear in both the transmit and the receive path. After the low-pass filters the preselector covers the following bands using three values of inductor and a 'digital' capacitor that covers from 1pF to 1024pF. Depending on the band various capacitor values are switched in series and in parallel to give a minimum and maximum capacitance.

The schematic for the whole thing covers 1.5 sheets of A3, it's quite a complex project.

[CopperCone]

Hmm, that is similar to the agilent one in terms of bands. I don't know if I want to use a variable capacitor or tuning, I think I just want relays or PIN diodes or whatever is appropriate at those frequency ranges to switch.

But what are your filter orders and passband widths? Thats the crticial data I need to establish what I need to simulate and start experimenting with what filter toplogies are good.

Since you use a variable capacitor I assume you have a first order filter? My simulations indicate to have a decent rejection in the nearby bands you need something like a 6th order elliptic filter, or some crazy order butterworth filter (+20). I thought that it should fall to 40dB within like 5 MHz of the center of the 4MHz pass band, but I suspect that If I try to build this filter it will misbehave like crazy, to drop 40dB in 1MHz step. You get something like 5th order requirement to do 40dB / 2MHz with 4MHz pass band with ~1dB ripple

The so called parametric filter in NuHertz is rather interesting. You can get parts match nH / pF this way. But it is lossy to 6dB. Might have real nice high frequency performance since you can use tiny air coil though. The commercial filter is only 4dB of loss. Tons of parts too, but nothing crazy like 1mH inductors.

Also, what would be a good fast RF switch to use if I wanted to connect it to the sweep signal?

I wonder what kind of 'slew' is possible. I need to build a few to find out.

Do you think that trying to make torroidal inductors would help?

Gain of 2 parallel LTC1028's would work as a preamplifier for the band up to 30MHz to counter effect this loss, but that is cutting it close to the GBWof 70MHz. And probably not the best part. Use a transistor I guess? Or should I look for a opamp?

Also, what is a good fast RF switch to use if I wanted to connect it to the sweep signal?

[CopperCone]

Thought that was interesting. 5% tolerances on the inductors and the capacitors. I need to do monte carlo to see what drift will look like. I put 100 as a capacitor Q. Not sure. You can buy inductors that have a Q of something like 400 in those sizes.

100 passes 10% component tolerance.
*this is with uniform Q  I don't think it changes during the simulation iterations, you would need to remake the schematic in a diff program to do it

Wanted to see what it would look like with garbage parts.

With Inductor Q of 120 and Capacitor Q of 1000 (not sure how optimistic or pessimistic this is, I read a quality one will have something like 10,000 at these frequencies)



Do you think that is reasonable for a preselector band?

I figured I could wind all the coils using air coils, on a strip of copper, then test with torroids. It should work nicely with my VNA because its 50 ohms and I can put BNC sockets on it so its not some hodge podge alligator clip shit

If you want the 40dB in 500KHz,

I thought 40dB to be reasonable in order to prevent gain compression and mixing. That is what the commercial instrument is specified to.

Given the cost of these things is low (Though I don't know about the switch), it would seem like a better idea to have two filter sets that are spaced differently rather then trying for higher orders , then switch between them to see what happens?

If you set a tighter pass band and lower block band you can get ugly shit


I do think you would need variable inductors to tune this thing though....

Seems OK with 1% tolerances and 1% monte carlo though. my waste of time experimentation will tell



Anyone wanna place bets on the highest order I can decently replicate?

[T3sl4co1l]

Sure as hell not getting a capacitor with << 15nH ESL, nor an inductor with a high Q in that value.

Tim

[CopperCone]

https://wcmagnetics.com/product/high-q-rf-inductor-202-series/

It looks like a reasonable Q spread is between 90 to 180

Values bit too high for most uses

https://www.global.tdk.com/corp/en/news_center/press/aah38900.htm

Lower values, claim Q of 100 but at high frequencies. I don't know how Q fluctuates over frequency as a device parameter though.

As for capacitors, this website seems to illuminate it a little

http://www.seed-solutions.com/gregordy/Amateur%20Radio/Experimentation/HexArray/ComponentQ.htm

it seems very variable based on frequency. One of them has a measured Q of almost 10k at 1.5MHz which drops to like 300 at 9 MHz.

The filter does seem to work with dirty Q factors so long you amplifier it afterwords though. I will try some Qspread manual monte carlo simulations