# Substitute for 455 kHz Ceramic filters in new designs

I am looking to design a dual conversion FM receiver/power meter for t-hunting similar to the MK4 sniffer using 10.5 MHz and 455 kHz IF. In my research it appears that manufactures have stopped making 455 kHz ceramic filters. I am in the U.S.A. and the only thing I can find still in production is in England.

My question is what can/is used for a very high Q if filter (Q of about 45 = 455 kHz / 10 kHz) at 455 kHz in the absence of ceramic filter/discriminators.

Since this platform considers questions asking for products as off-topic, I'll interpret your question as:

I have an IF receiver, but there's no good filters for the rather low IF it uses (455 kHz). What to do?

If you're doing a new design, simply pick an IF that works with the filters you can get.

That, and pretty much nothing else, is what defines what IF is chosen in commercial detectors. Your high-sensitivity Spectrum Analyzer does that. Your cheap FM Radio receiver IC does that.

On the other hand: If you're just reworking a superhet where the final receiver stage failed, then consider why you need the IF filter:

Its purpose is to isolate the wanted spectrum from all the unwanted stuff, so that noise and interference has lesser influence on the receiver. That's why you want that IF filter to be sharp; and that's why you'd use an IF to fit your filter (and not the other way around). Adjusting an oscillator to a different LO frequency is easy, building a good filter for an arbitrary frequency band is not.

Now, what if I tell you that 455 kHz is "practically baseband" in 2019. If you, say, build a very relaxed low-pass filter that lets through 0 to 550 kHz and a very relaxed high pass that lets through 400 to $$\infty$$ kHz, and then digitized at more than 300 kS/s, then you could do the IF processing with digital filters (in software!), and these can be pretty much arbitrarily steep. You can modify them to your hearts delight (for example, to account for actual signal spectrum shape or LO inaccuracies), and that doesn't cost anything and has no risk of not working.

People's classical argument is that a good IF filter is paramount to receiving weak signals between and especially near to strong ones – but that's no problem for a digital receiver, that is, as long as the dynamic range of the ADC can still "see" the weaker signal without clipping the stronger one.

So, say you want to be able to, without even using oversampling¹, detect a signal that is say 80 dB weaker than its neighbor. Your Superhet receiver with nice IF filter would require an IF filter with a suppression of at least 83 dB if you want your SNR to be still 3 dB, i.e. crosstalk to only be half as powerful as the signal of interest. In reality, you'd probably need more filter attenuation.

I doubt the MK4 achieves that selectivity (I've not found two-tone test charts for it), but let's assume it did and you'd want to achieve the same.

Your ADC would thus need to be able to not overflow on the 80 dB stronger signal, and still have at 3dB dynamic range to spare on resolving the weaker one; a 18 bit ADC gladly does that (and far more) and comes at less than 10 USD.

If you can accept only a DR of a little more than 60 dB (which will, for fox hunting and the like, more than suffice), a < 20 USD Microcontroller eval board's on-chip ADC will simply do. That microcontroller itself is so plenty fast enough at doing math that applying even a mean filter on a couple kHz of signal is no big deal.

Notice that I'm an SDR guy – I really know that the time for narrowband filters under benign circumstances implemented as analog hardware components is over. You still need to do your anti-aliasing filtering, but seriously, instead of hoping you can calibrate your analog filtering to be acceptably steep and stable, digitally filter. Microcontrollers are plenty fast enough for the bandwidths you use.

¹ You can increase the SNR of signals by sampling faster than you need to. And this will have happened dramatically after your digital filter.

• To avoid aliasing wouldn't you need at least 920 kS/s for 455 kHz+-5 kHz instead of 300 kS/s? Also I was trying to keep it in the analogue domain to allow for an automatic gain control loop to extend the dynamic range and avoid having to design a DSP subsystem as my embedded DSP skills are negligible. – Jason R. Moore Mar 16 at 2:05
• Band pass undersampling! So: the idea is that while, yes, you alias, that's no problem: you only got less than half of the sampling rate in bandwidth, so nothing "folds over" any other signal. So, 400 – 450 kHz end up in baseband's 100–150 kHz (namely, by being mathematically aliased down by $-f_\text{sample}$) and 450 – 550 kHz end up at 0 – 100 kHz (namely, by being mathematically aliased down to -150 kHz through being shifted by $-2f_\text{sample}$, and then realising that for real-valued signals, spectrum is 0 Hz-symmetrical). – Marcus Müller Mar 16 at 10:26
• That at first looks ugly (because it is ugly – a circularly shifted version of the spectrum you want to observe), but it can be corrected in digital by multiplying the signal with a complex sinusoid $e^{j2\pi\frac{\Delta f}{f_\text{sample}}t}$ – mixing it right so that the middle ends smack at 0Hz, the lower half at $[-\frac{f_\text{sample}}2;0]$ and the upper half at $[0;\frac{f_\text{sample}}2]$ in complex baseband. – Marcus Müller Mar 16 at 10:31
• However, you'd be free to pick a higher sampling rate than 300 kHz – in this case, 400 kHz would be pretty clever, because that aliases the lower end of our 400 kHz–550 kHz band to 0 Hz, and you could at the expense of handling a slightly larger sampling rate work without any digital frequency shifting :) – Marcus Müller Mar 16 at 10:33
• So, thing about gain control is that you don't really need it all that much if your ADC has > 60 dB dynamic range to begin with, and your software could very well adjust the gain of an analog amplifier, (or, like many devices do it, a fixed high-gain, low noise figure amplifier, followed by an adjustable attenuator) based on what it sees :) – Marcus Müller Mar 16 at 10:34

Murata makes the CERAFIL(R) line of ceramic filters, some of which are designed for use at 455kHz and 450kHz. For example, the SFPLA450KD1A-B0 has a -6dB bandwidth of 10kHz, designed for use with AM stereo receivers. A search of eBay for "filter if 455" turns up units like the AVX Kyocera KBF455R, which may be suitable for your project.

Depending on your goals, you might consider building the Elenco AM/FM-108CK receiver kit. It is a superhet design, though it does not achieve the high selectivity of a ceramic or mechanical filter. This kit is an excellent hands-on educational project - being the focus of Elenco products - and you could learn a great deal by simulating the design with the free LTSpice from Analog Devices.

If your goal is a portable, low-drain FM receiver - as opposed to reproducing an arcane architecture for historical or nostalgia value - I agree with an earlier answer that recommends using a modern integrated circuit approach for a new design. There are many how-to articles online showing everything from standalone battery-powered units to "shields" for the current generation of single-board computers.

• I find my answer misrepresented: The main point I made was: you choose your IF to match the filters you need, not the other way around. If you can't choose, don't filter in analogue. – Marcus Müller Mar 15 at 18:51