Crystal Filters

[AnalogSteph]

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)?

No, it's just a lowpass that you need (what else did you expect in an anti-alias filter for audio?). What I mentioned was the design considerations for it (passband, stopband and stopband rejection, you can add passband flatness to the list).

Your signal chain would generally be looking something like this:
anti-alias LPF --> ADC --> decimating digital filter --> DSP
At least your average DSP is not going to have the grunt to process several MHz worth of 16-bit samples (FPGA power may be required). Or what kind of SAR ADC was it that you were thinking of?

 

[DonH56]

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

High-speed ADCs rarely have the best performance in the audio band. Noise in particular may be higher, and a high sampling rate often means higher distortion than a typical audio converter. If you oversample, you must also be prepared to handle the higher digital bit stream and rate conversion if need be (e.g. to match typical audio-rate CODECs at 44~196 kS/s). I am not sure why you want to use a SAR architecture, nor at what level you plan to design the ADC (buy off-the shelf, designing an IC, ???)

Oversampling using a conventional ADC makes the anti-alias filter easier but you gain only 1/2-bit for each doubling of the sampling rate. Delta-sigma designs use an internal modulator, often coupled with multiple stages and multi-bit internal stages, to provide much greater noise reduction (several bits or more for each doubling of the sampling rate).

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

The problem is a crystal (or any high-Q) resonator is not the right component to create a wideband filter. The anti-alias filter must pass all the frequencies below cutoff, then sharply attenuate above. A high-Q resonator creates a very narrow bandpass filter, not the desired low-pass filter desired for anti-aliasing. To revert to a bad analogy, if you want to sprinkle a wide area of your lawn, you don't want a water nozzle that is fixed to a tight stream and that points in only one direction.

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

Many high-performance oscillators, crystal or otherwise, function just fine in high-vibration environments. But it takes special design techniques, both electrical and mechanical, to control their noise.
---
I have barely skimmed this thread but it seems like you need a more fundamental understanding of crystals, filters, and ADCs than you are likely to get from an Internet thread. People devote careers to learning how to design ADCs, oscillators, DSPs, and so forth. Reading manufacturer's application notes may be a better way to delve into the subjects. A college course in mixed-signal circuits could be beneficial.

 

[BKr0n]

At least your average DSP is not going to have the grunt to process several MHz worth of 16-bit samples (FPGA power may be required). Or what kind of SAR ADC was it that you were thinking of?

So the base design at the moment is a powerful SBC, and a FPGA with DMA access. Still shopping around on the FPGA (and me being stubborn about coding ) but as far as adc I was looking at this guy here (https://www.analog.com/media/en/technical-documentation/data-sheets/238718fa.pdf)

I am not sure why you want to use a SAR architecture

I'm going to try my hand at real time signal processing. I have enough compute for it so I figure why not?

(buy off-the shelf, designing an IC, ???)

This made me chuckle. You are silly. All kidding aside I've seen what it takes to learn about lithography... and here I thought this is tough... making tiny things is very difficult.

To revert to a bad analogy, if you want to sprinkle a wide area of your lawn, you don't want a water nozzle that is fixed to a tight stream and that points in only one direction.

I got what you meant

I have barely skimmed this thread but it seems like you need a more fundamental understanding of crystals, filters, and ADCs than you are likely to get from an Internet thread. People devote careers to learning how to design ADCs, oscillators, DSPs, and so forth. Reading manufacturer's application notes may be a better way to delve into the subjects. A college course in mixed-signal circuits could be beneficial.

I've been reading almost nonstop and it still isn't enough lol (if I'm being completely honest here, had to make time for baldur's gate 3). I do want to go back to school, but life is a little chaotic at the moment so not the best time unfortunately

 

[mhardy6647]

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

Given that "we" used to have crystal mics.
Some folks still do.

The venerable Astatic D-104 mic. I'd have cleaned the fingerprints off, were it mine and/or if it were I takin' that photo!

History of the Astatic D-104 Silver Eagle

<p><em>Excerpted from the 1946 Astatic Catalogue.</em></p> <p>The Astatic D-104 Silver Eagle is now considered an antique and collector item by many radio enthusiasts. A lot of the mics are still in good working condition today and are mainly used by CB collectors. The history of the Astatic...

www.roadprobrands.com

 

[DonH56]

So the base design at the moment is a powerful SBC, and a FPGA with DMA access. Still shopping around on the FPGA (and me being stubborn about coding ) but as far as adc I was looking at this guy here (https://www.analog.com/media/en/technical-documentation/data-sheets/238718fa.pdf)

I don't really know anything about FPGAs. I had several grad courses in DSP, so used to know a little, but not in years. My career took a mostly analog path designing (transistor-level IC) data converters.

I know the guy who designed that particular ADC (we used to work together). Not sure why you want want sampling that fast for audio but whatever.

I'm going to try my hand at real time signal processing. I have enough compute for it so I figure why not?

You can do that with any audio ADC, does not have to be a SAR (which also means synthetic aperture radar to me; my career began doing converters -- including successive approximation types -- for radar systems).

This made me chuckle. You are silly. All kidding aside I've seen what it takes to learn about lithography... and here I thought this is tough... making tiny things is very difficult.

You are asking questions a designer might ask, but with less knowledge base, and do not seem to understand the context of the responses. I have not been following this thread so had no idea exactly what level of design you were attempting. Application engineers and the application notes wouild be a good starting point IME/IMO.

I got what you meant

OK, sorry, in the previous few posts you still seemed to be targeting a crystal filter... I have used a number of those in RF systems but they usually end up in the IF strip or clock circuit, not audio band. There are some interesting SAW filter and mechanical filter designs that achieve extremely high Q (~10,000) but again very (very!) narrowband applications.

I've been reading almost nonstop and it still isn't enough lol (if I'm being completely honest here, had to make time for baldur's gate 3). I do want to go back to school, but life is a little chaotic at the moment so not the best time unfortunately

OK, we all have to live in the real world.

 

[DonH56]

Given that "we" used to have crystal mics.
Some folks still do.
View attachment 324830

The venerable Astatic D-104 mic. I'd have cleaned the fingerprints off, were it mine and/or if it were I takin' that photo!

History of the Astatic D-104 Silver Eagle

<p><em>Excerpted from the 1946 Astatic Catalogue.</em></p> <p>The Astatic D-104 Silver Eagle is now considered an antique and collector item by many radio enthusiasts. A lot of the mics are still in good working condition today and are mainly used by CB collectors. The history of the Astatic...

www.roadprobrands.com

Now, that's a blast from the past! Wish I had kept my HAM gear, which included a D-104, and a nice National receiver that had one of those mechanical filters with ridiculously high selectivity.

 

[mhardy6647]

Now, that's a blast from the past! Wish I had kept my HAM gear, which included a D-104, and a nice National receiver that had one of those mechanical filters with ridiculously high selectivity.

I've never had a D-104... and I've always wanted one. Nice "tone" on AM.
I have a National Rx here that needs a little TLC.



... and a much prettier (and fully functional), but much newer, and much less interesting one...



OK, I'll stop, sorry!

 

[BKr0n]

I know the guy who designed that particular ADC (we used to work together). Not sure why you want want sampling that fast for audio but whatever.

You can do that with any audio ADC, does not have to be a SAR

I believe it was some of the people here who pushed me closer to the "need better than an audio adc" trend. Originally I was just going to go with an audio adc, but was told delay would be an issue, and if I have any intent on real time anything processing, would need to look into something with less delay. SAR was a good place to start since SD add a lot of delay, and pipeline for this application make no sense. As for why that sampling rate, had the best noise figures for around where I wanted to sample (like @AnalogSteph said about 6.144MHz).

(which also means synthetic aperture radar to me; my career began doing converters -- including successive approximation types -- for radar systems).

Small world! I got my start repairing radar aboard Navy ships when I was enlisted a lot of the beforehand training was pretty much cramming a 4 year degree into about a year and a half. I actually paid attention which is probably how I've even gotten THIS far lol.

OK, sorry, in the previous few posts you still seemed to be targeting a crystal filter... I have used a number of those in RF systems but they usually end up in the IF strip or clock circuit, not audio band. There are some interesting SAW filter and mechanical filter designs that achieve extremely high Q (~10,000) but again very (very!) narrowband applications.

Not going to use them, but just as a point of curiosity, would they make good notch filters then? Or is it the same issue with microphonics that comes up?

OK, I'll stop, sorry!

Pfft. Dude. Post more lol. You reminded me of the first time I ever saw teletype and patch panels

 

[DonH56]

I've never had a D-104... and I've always wanted one. Nice "tone" on AM.
I have a National Rx here that needs a little TLC.



... and a much prettier (and fully functional), but much newer, and much less interesting one...



OK, I'll stop, sorry!

Mine was an NC-300, I think... Been a long time. Had a Heathkit Apache transmitter and SSB adapter mated to it, and some sort of 2 m rig I do not recall now (I was mostly HF). I had to sell my entire station, including a nice 70' tower with several antennas and a nice CDE rotator, when my family moved during high school.

 

[BKr0n]

So while you guys are here (and some of you will most likely already know this) I may have been using "real-time" in the wrong sense. I was thinking to myself when ask to define the specs, I wasn't sure what they meant. Apparently an industry all of its own cropped up around the idea of real time systems to the point of great granularity. Honestly, I should have known better given that real time anything could apply to so many other industries. But hey, we all need to learn somehow. Although I do apologize to those who's patience I've been testing

Mine was an NC-300, I think... Been a long time. I had to sell my entire station, including a nice 70' tower with several antennas and a nice CDE rotator, when my family moved during high school.

That kind of reminds me of my dad's old marantz he used to have. Had a set of ESS speakers to go with.

 

[DonH56]

I believe it was some of the people here who pushed me closer to the "need better than an audio adc" trend. Originally I was just going to go with an audio adc, but was told delay would be an issue, and if I have any intent on real time anything processing, would need to look into something with less delay. SAR was a good place to start since SD add a lot of delay, and pipeline for this application make no sense. As for why that sampling rate, had the best noise figures for around where I wanted to sample (like @AnalogSteph said about 6.144MHz).

OK. SAR also have latency, usually comparable to pipelined... "Successive" means you have to do a sequential series of comparisons. Depends upon what your application can tolerate. Delta-sigma designs may not have that much higher latency since processing can happen at high speed; they are used in a number of interfaces designed to operate with minimal latency. Other architectures, such as SAR, can "hide" the latency using special techniques inside the chips so there is essentially "no" latency to the outside world. The LTC design you linked is one of those.

For audio recording and post-processing latency is not usually an issue since much of the mixing and mastering is after the initial capture. It can be an issue for live sound or monitoring but there are ways around that if really needed.

Small world! I got my start repairing radar aboard Navy ships when I was enlisted a lot of the beforehand training was pretty much cramming a 4 year degree into about a year and a half. I actually paid attention which is probably how I've even gotten THIS far lol.

I have little experience with seaborne radars; the ones I worked on were mostly airborne or space based. Sea is tough due to reflections from the water.

Not going to use them, but just as a point of curiosity, would they make good notch filters then? Or is it the same issue with microphonics that comes up?

For some applications. Microphonics and other mechanical issues come up with many filters when you are needing extreme stopband attenuation.

 

[DonH56]

So while you guys are here (and some of you will most likely already know this) I may have been using "real-time" in the wrong sense. I was thinking to myself when ask to define the specs, I wasn't sure what they meant. Apparently an industry all of its own cropped up around the idea of real time systems to the point of great granularity. Honestly, I should have known better given that real time anything could apply to so many other industries. But hey, we all need to learn somehow. Although I do apologize to those who's patience I've been testing

Real-time in audio, or computing, means processing and responding fast enough that the lag is unnoticeable. Monitoring during recording is a common example, as is gaming applications. Technically it means response must occur within a specified time window. I do not know what you mean by it.

That kind of reminds me of my dad's old marantz he used to have. Had a set of ESS speakers to go with.

No, I am talking of a HAM (amateur radio) receiver. Something like this:

 

[BKr0n]

OK. SAR also have latency, usually comparable to pipelined... "Successive" means you have to do a sequential series of comparisons. Depends upon what your application can tolerate. Delta-sigma designs may not have that much higher latency since processing can happen at high speed; they are used in a number of interfaces designed to operate with minimal latency. Other architectures, such as SAR, can "hide" the latency using special techniques inside the chips so there is essentially "no" latency to the outside world. The LTC design you linked is one of those.

That's a great explanation thank you

For audio recording and post-processing latency is not usually an issue since much of the mixing and mastering is after the initial capture. It can be an issue for live sound or monitoring but there are ways around that if really needed.

Real-time in audio, or computing, means processing and responding fast enough that the lag is unnoticeable. Monitoring during recording is a common example, as is gaming applications. I do not know what you mean by it.

My idea was altogether was real time processing of a signal. Signal would go out the dacs, through the amps, and then the speakers. Speakers would have a return line where the signal is attenuated and then sampled by the adcs to the processor to get processed. I'm going to control said processor with a SBC and load it up with CamillaDSP and REW. Maybe some other stuff down the road. I already have the SBC and OS picked out. Just need to figure out and build the processor board (dacs, adcs, whatever microcontroller or dsp etc).

Sea is tough

Yes I know. Apparently some of the operators in the ATC couldn't tell a plane from a cloud

No, I am talking of a HAM (amateur radio) receiver. Something like this:

Interesting how a lot of audio equipment of that era used the analog meters they did. Is that just what was the best tech available at the time, or was it for a specific reason like parallax?

 

[voodooless]

I believe it was some of the people here who pushed me closer to the "need better than an audio adc" trend. Originally I was just going to go with an audio adc, but was told delay would be an issue, and if I have any intent on real time anything processing, would need to look into something with less delay.

Who were these people?

The big issue here is that you don’t lay down any requirements. Why do you need real-time processing? How much delay can you tolerate? You’ll need to define these and many more parameters before you start choosing components. If you want to do digital motion feedback, do the math and figure out how much delay you can work with.

Note that an ADC is just on of many components in the system. They all have propagation delay. Also note that the ADC chip you choose does not have a digital filter, so you’ll need to add that in DSP to lower the sample rate to something manageable. This will add delay as well, and probably you’ll end up with about the same delay as a normal ADC. From what I can see a typical ADC has a delay of between 5 and 12 samples depending on filter choice. That isn’t a lot..

 

[DonH56]

That's a great explanation thank you

NP

My idea was altogether was real time processing of a signal. Signal would go out the dacs, through the amps, and then the speakers. Speakers would have a return line where the signal is attenuated and then sampled by the adcs to the processor to get processed. I'm going to control said processor with a SBC and load it up with CamillaDSP and REW. Maybe some other stuff down the road. I already have the SBC and OS picked out. Just need to figure out and build the processor board (dacs, adcs, whatever microcontroller or dsp etc).

I have no idea what you are trying to do, but in addition to conversion latency, processing lag may be an issue for any sort of feedback system. Hopefully you've had some exposure to control theory. An FIR filter can add hundreds to thousands of clock cycles of delay just for converting the 15 MS/s output to the audio band, and then you whatever processing you're doing (more delay), and lag through the DAC. Plus the analog delay through various amplifiers, speaker components, mic or other sensors, buffers, etc. I may have just missed it, but as @voodooless said you need to clearly define your end-to-end requirements including resolution and latency.

Yes I know. Apparently some of the operators in the ATC couldn't tell a plane from a cloud

Water is a huge reflective surface that can overwhelm radar receivers, making it difficult to distinguish small objects like aircraft, missiles, or small boats from the "noise". Think of trying to take a picture of a bird or plane flying into the sun.

Interesting how a lot of audio equipment of that era used the analog meters they did. Is that just what was the best tech available at the time, or was it for a specific reason like parallax?

Again, that is an amateur radio receiver, not an audio component, but back then we did not have LED or LCD displays, DSP (or even computer) chips, and so forth. That receiver (National NC-300) was built and sold new in the 1950's. Analog meters were popular in stereo components to provide visual feedback to consumers, not just for levels but also for things like radio tuning (seeing if the tuner is centered on the station) and so forth.