Crystal Filters

[bluefuzz]

So many rules and regulations

Just join the Boy Scouts ...

This 17 Y.O. Boy Scout Built A Nuclear Reactor In His Mom’s Backyard​

 

[BKr0n]

Just join the Boy Scouts ...

This 17 Y.O. Boy Scout Built A Nuclear Reactor In His Mom’s Backyard​

When I first read that when it first hit the news I was floored. How did his parents not even know!?

 

[mhardy6647]

You guys have got me thinkin' about Q multipliers, now... maybe it's just me.

Not my R-100A (unfortunately)

 

[Digby]

A pocket portable nuclear reactor has always been a dream of mine.

Even if it remains forever but a dream, at least Kraftwerk did it some justice in song form.

 

[BKr0n]

Anti-aliasing seems to be a "solved problem"

Where would be a good place to start in terms of design? I've seen a lot of articles on it in terms of the math, but not necessarily topology.

 

[AnalogSteph]

Where would be a good place to start in terms of design? I've seen a lot of articles on it in terms of the math, but not necessarily topology.

For starters, define your signal bandwidth and the ADC sample rate. That should give you some indications for what you need in terms of filter order.

For example, let's say you want a 50 kHz signal bandwidth and a modern-ish 128fs audio ADC (64fs at double speed etc., so 6.144 MHz for 48 kHz derivatives). That's a 50 kHz passband and 3.072 - 0.048ish MHz stopband, or 5.9 octaves or 1.8 decades. That's a stopband rejection in excess of 106 dB with a 3rd-order filter. In reality you can probably get away with substantially less, given that your input is not likely to be blasted with signal levels anywhere close to nominal audio input level in the MHz.

 

[voodooless]

modern-ish 128fs audio ADC

He wants to use a SAR ADC though…

 

[Philbo King]

Xtal filters were originally used for SSB (single side band) shortwave where only the lower or upper RF sideband was used to carry audio. They were developed at the company I worked at, long before I was born, around WWII. When I started at the company I was fixing airborne HF transceivers that used it. Interesting toys...

They're very narrowband, 5k-10k and generally cut to operate at an IF frquency.

 

[BKr0n]

For example, let's say you want a 50 kHz signal bandwidth and a modern-ish 128fs audio ADC (64fs at double speed etc., so 6.144 MHz for 48 kHz derivatives). That's a 50 kHz passband and 3.072 - 0.048ish MHz stopband, or 5.9 octaves or 1.8 decades. That's a stopband rejection in excess of 106 dB with a 3rd-order filter. In reality you can probably get away with substantially less, given that your input is not likely to be blasted with signal levels anywhere close to nominal audio input level in the MHz.

So then it's really something like a band pass and a low pass? If that's the case, would something like a svf be good here (having access to more than one filter band with variable gain and Q)?

Xtal filters were originally used for SSB (single side band) shortwave where only the lower or upper RF sideband was used to carry audio. They were developed at the company I worked at, long before I was born, around WWII. When I started at the company I was fixing airborne HF transceivers that used it. Interesting toys...

Have you seen them used for anything other than RF?

 

[Philbo King]

So then it's really something like a band pass and a low pass? If that's the case, would something like a svf be good here (having access to more than one filter band with variable gain and Q)?

Have you seen them used for anything other than RF?

Not in audio, except clock oscillators. It'd be hard, maybe impossible, to get one big enough to operate on raw audio. Perhaps a custom created one made with laser ion deposition, layer by layer. I'm not very familiar with that stuff; it's a bit 'exotic', White Room stuff.

In theory it could be done by audio modulating an RF carrier, do the filtering, then demodulate back to audio. But these days it's probably more practical to use a DSP filter.

 

[BKr0n]

Not in audio, except clock oscillators. It'd be hard, maybe impossible, to get one big enough to operate on raw audio. Perhaps a custom created one made with laser ion deposition, layer by layer. I'm not very familiar with that stuff; it's a bit 'exotic', White Room stuff.

In theory it could be done by audio modulating an RF carrier, do the filtering, then demodulate back to audio. But these days it's probably more practical to use a DSP filter.

Funny you should say that. I'm doing DSP. It was recommended to me that I put anti-aliasing before the ADCs. Now that you say that, why am I doing that if I'm using a DSP? Would a second layer of filtering even be necessary?

Also, what do you mean by "big enough" for audio?

 

[voodooless]

. It was recommended to me that I put anti-aliasing before the ADCs

Because you don’t want aliasing Even if it high up in frequency, it will fold down in the audible range. You use high sample rate so that the anti-aliasing filter can start outside of the audible range and you don’t need a very high order filter to get enough attenuation. This saves cost and complexity.

As for svf filters, you don’t need precise Q control, because you filter outside of the audible range anyway. Normal filters will do just fine. ADC datasheet usually include a reference.

 

[Philbo King]

Funny you should say that. I'm doing DSP. It was recommended to me that I put anti-aliasing before the ADCs. Now that you say that, why am I doing that if I'm using a DSP? Would a second layer of filtering even be necessary?

Also, what do you mean by "big enough" for audio?

It goes a bit deep. But it basically resonates based on the physical size and the type of cut the crystal has. Bigger = Lower freq.

As regards crystal cuts:

Quartz Crystal Cuts: AT, BT, SC, CT . . » Electronics Notes

The angle to the main crystal axes at which a crystal blank is cut has a major impact on its performance. Popular cuts like AT, BT, SC, CT give known performance . . find out more

www.electronics-notes.com

Note that one cut of crystals can be made to work down to 5 Khz. (toward the end of the article).

 

[Prana Ferox]

They were developed at the company I worked at, long before I was born.

Nowadays we have laws that you have to go to school first

 

[fpitas]

Just to add to the list of problems, they tend to be microphonic, too.

 

[DonH56]

Aliasing: https://www.audiosciencereview.com/forum/index.php?threads/digital-audio-aliasing.1920/

The anti-aliasing filter must be implemented in the analog domain before the ADC and digital conversion. It is required to ensure all out-of-band signals are not aliased into the desired signal band, or at least suppressed below the quantization noise floor. Once aliased, the ADC/DSP has no way of determining if the resultant is a desired signal or an aliased (undesired) signal. Oversampling (e.g. delta-sigma) converters move the Nyquist frequency well above the desired signal passband so implementing the anti-alias filter is simpler. Oversampling also means that, after the initial analog anti-aliasing filter, further band limiting can be performed in the digital domain. That is typically where one sees very high order "anti-alias" filters in delta-sigma designs (and similarly high-order DAC anti-image filters).

The size of a crystal determines its resonant frequency. Crystals are also usually high-Q resonators so very narrow bandwidth. The former means an audio-frequency crystal is generally impractically (physically) large. The latter means creating a wideband (multioctave) filter such as required for audio is also impractical.

 

[BKr0n]

Oversampling (e.g. delta-sigma) converters move the Nyquist frequency well above the desired signal passband so implementing the anti-alias filter is simpler. Oversampling also means that, after the initial analog anti-aliasing filter, further band limiting can be performed in the digital domain. That is typically where one sees very high order "anti-alias" filters in delta-sigma designs (and similarly high-order DAC anti-image filters).

So then would oversampling with a SAR be just as simple as using one in the MHz range like @AnalogSteph was suggesting?

The size of a crystal determines its resonant frequency. Crystals are also usually high-Q resonators so very narrow bandwidth. The former means an audio-frequency crystal is generally impractically (physically) large. The latter means creating a wideband (multioctave) filter such as required for audio is also impractical.

What about if it were beyond the audible range like for oversampling?

Just to add to the list of problems, they tend to be microphonic, too.

Would that include any kind of oscillator, or just quartz based?

 

[fpitas]

Would that include any kind of oscillator, or just quartz based?

Most piezoelectric elements are microphonic, for obvious reasons. So quartz, ceramic, SAW, etc.

 

[BKr0n]

I meant something like a timing IC/programmable oscillator. Or do all of them have the aforementioned materials to create the signal?

 

[fpitas]

I meant something like a timing IC. Or do all of them have the aforementioned materials to create the signal?

That depends on the timing IC. Some just have an RC or LC to set frequency. Some use a quartz crystal. You'll never get good phase noise or stability from an RC or LC, though.