I want to extend the bandwidth of a PCIe or PCI soundcard to accommodate more spectrum for my SDR radio. I'll start with one that has a great noise floor spec. and no ripple in the passband. All I've found so far is the M-Audio Audiophile 192 and it's an old design. I'm guessing that they all have an anti-aliasing filter. I'm wondering if it's possible to bypass that and work on the amplifiers to make the passband flat to 100kHz?
3
-
\$\begingroup\$ Even if you could remove the antialiasing filter, it wouldn't help. If you sample without the filter, then all you get is a mess. You could put a bandpass filter in front of the ADC and then you would have at least a predictable mess, but even then you won't get more bandwidth. Read up on Shannon and Nyquist. \$\endgroup\$– JRECommented Nov 24, 2017 at 5:49
-
\$\begingroup\$ I want to move the corner frequency of the anti-aliasing filter up to 100khz... They implement that filter digitally from what I read so I have little hope of modifying anything to what I want but I could bypass it maybe, then add in my own filter... \$\endgroup\$– MaxWebXperienZCommented Nov 25, 2017 at 0:59
-
\$\begingroup\$ Seriously, look up Shannon's theorem, and the Nyquist limit. There is a reason why the anti-aliasing frequency is set lower than 100kHz. It is probably a bit under 96kHz. That value wasn't chosen at random. \$\endgroup\$– JRECommented Nov 25, 2017 at 9:09
Add a comment
|
2 Answers
\$\begingroup\$
\$\endgroup\$
1
Audiophile 192 features high-definition 192kHz sampling rate.
How to search for simple answers? like NO.
Search Engine: M-Audio Audiophile 192 specifications
- find out the ADC chip, then search for the specs and get your answer in seconds.
https://www.akm.com/akm/en/file/datasheet/AK5385BVF.pdf
The decimation filter is integral part of the ADC chip. Therefore "no"
https://www.soundonsound.com/reviews/m-audio-audiophile-192#para5
-
\$\begingroup\$ Yeah, it's pretty much cast in concrete.. there have been two projects of custom designed soundcards that achieved what I want, I'll emulate their designs... \$\endgroup\$ Commented Nov 25, 2017 at 1:02
\$\begingroup\$
\$\endgroup\$
1
Even if you could change the cutoff frequency of the antialiasing filter, you cannot change the bandwidth of the card.
It samples at 192kHz. It therefore has a bandwidth of 96khz. You cannot get more bandwidth through it without changing the sampling rate.
If you check the datasheet that Tony Stewart provided, you will see that the built in anti-aliasing filter allows you to use up to around 89kHz out of the theoretically available 96kHz when using the 192kHz sampling rate. That's pretty good, and I don't think you will do better with anything you implement yourself.
If you want to sample a signal that has a frequency higher than the sampling rate, you can run your signal through a bandpass filter and then down convert it so that its frequency range is within the bandwidth of the ADC.
Say you want to sample a signal containing frequencies from 150kHz to 200kHz.
You pass it through a bandpass filter that removes everything below 140kHz and 210kHz. Now, you mix it with 140kHz, and run that through a low pass filter at 96kHz. The result is a signal with 70kHz bandwidth, with your signal of interest between 10kHz and 60kHz. You can then sample that with your ADC at 196kHz. You do have to keep the down conversion in mind when doing FFTs and things if you need to calculate frequencies from the sampled data.
I'm not going to try to write an explanation of the physical limitations of sampling, and why you only get a bandwidth of half the sampling rate. There's an explanation of the Nyquist theory on Wikipedia. I suggest you read that. It has a clear explanation, and links to other parts of sampling theory.
-
\$\begingroup\$ You don't even need to use mixing to down-convert if the input bandwidth of the ADC is high enough (I'm not taking about Nyquist limited bandwidth). If you bandpass filter something between fs/2 and fs, then sample it at fs, the resulting signal gets down-converted by aliasing to 0 to fs/2. See undersampling. Although it's true that wouldn't help for this particular ADC due to the built in filtering. \$\endgroup\$ Commented Nov 25, 2017 at 12:41