Any high resolution diy spectrum analyzers ?

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
With many 24bit/192Khz soundcards and audio interfaces on the market these days and lots of freeware programs you dont even need to do anything special for signal analysis upto that frequency and dynamic range resolution. What I am curious is why arn't there any (or atleast I haven't managed to find any) pc based diy spectrum analyzers, especially the hardware, for higher sampling rates. I see that there are a few 16+ bit (enob) ADC chips from TI and AD that have sample rates of several (2-10) MSPS. I mean if anyone can build a 12bit 1MSPS "pc scope" then why not a 24bit 4MSPS spectrum analyzer ? What difficulty could be preventing this from happening ?
 
If you google for "build spectrum analyzer" you will find many:

Google

But a lot of them are analog and meant for RF (Radio Frequency) work. But even some of those would be nice to have, if they go down to around DC, as some do.

Below are a few links, but not necessarily the best ones, since I didn't look at most of the search results.

Our big Aeroflex analyzers at work are based on digital sampling and playback hardware (512 MHz recorded bandwidth that you can put anywhere from 0-6 GHz!), with everything else done in software. It seems like all you would need would be an off-the-shelf data acquisition module and some software, especially if the display didn't need to be "real time". I'd guess there's got to be some already-done stuff out there, like that. 50 to 100 MHz should be plenty high-enough for most audio work, although 200 MHz would be nicer. Anyway, there are some below that go to 1 GHz and higher.

Or find a working HP 141T system for around $200 which will get you to 110 MHz with the lowest frequency plug-in. Even a cheap non worker is easy to fix as long as the CRT is OK.

A Network Analyzer and Bode Plotter sure would be nice to have, too.

Spectrum Lab by dxzone.com

Build a 1000 MHz RF Spectrum Analyzer Inexpensively

http://www.signalhound.com/

BITSCOPE = PC OSCILLOSCOPES AND ANALYZERS (link to construction article at bottom)

Homebrew spectrum analyser

http://www.google.com/url?sa=t&rct=...l9mLAg&usg=AFQjCNGXdSBszbl6g7zCOIFk0Huy0uy8CQ

Create a FFT Analyzer part I: prerequisites, concerns and setup Sample & Hold

Lots of SA links:

Spectrum Analyzer - Technical Reference: Spectrum Analyzers

Google search for "spectrum analyzer ADC OR sampling OR "data acqisition"":

http://www.google.com/#hl=en&gs_nf=....,cf.osb&fp=b69c23b9f9168a7a&biw=1280&bih=818
 
Last edited:
The words "high resolution" in the subject were intentional. By high resolution I meant anything over true (enob) 16 bits, as I would imagine to correctly analyze a 16bit(audio) signal the analyzer would have to be more than 16bits.
Can you tell me if any of these spectrum analyzers have more than 16bit (-96db) resolution AND more than atleast 2Mhz spectrum range ? Imagine if you'd like to investigate -80db 500khz smps noise/ripple riding on the signal, or if you'd like to investiage amp oscillation into a few Mhz range. How would you be able to do that with a spectrum analyzer that has 12 or 8 bit resolution ?
 
Last edited:
It seems like it would be a small market, as someone famous might have said, "24/192 should be more than anyone could ever need for audio."

Since the A/D chips are available, it would just be a matter of making the interface with enough buffering, bandwidth, the latest USB or other high-speed interface, etc. I smell a raspberry pie here.

A near-off-the-shelf solution would be to use a standard interface for the audio band, and also send the signal through a high-pass filter set around 20kHz to 100kHz with that going to a 12 to 16 bit interface that goes up to several MHz. The filter would get rid of most of the "large" audio-band signals, and you could crank up the gain to get better resolution. Of course if an amplifier is oscillating rail-to-rail at 1MHz, you'd want the option of full attenuation on the the higher band as well.

And it would only take a "simple matter of programming" for the PC to read both of these interfaces and put the composite signal on one spectrum display.
 
Member
Joined 2008
Paid Member
640k is all anyone will ever need

:p
 

Attachments

  • BG.jpg
    BG.jpg
    17.7 KB · Views: 680
The words "high resolution" in the subject were intentional. By high resolution I meant anything over true (enob) 16 bits, as I would imagine to correctly analyze a 16bit(audio) signal the analyzer would have to be more than 16bits.
Can you tell me if any of these spectrum analyzers have more than 16bit (-96db) resolution AND more than atleast 2Mhz spectrum range ? Imagine if you'd like to investigate -80db 500khz smps noise/ripple riding on the signal, or if you'd like to investiage amp oscillation into a few Mhz range. How would you be able to do that with a spectrum analyzer that has 12 or 8 bit resolution ?

I already gave you sufficient links. Even a 10-minute perusal tells us that yes, many of them do _WAY_ better than -96 dB, and yes, many of them do from DC to several GHz.

This type of stuff just isn't that hard, any more. I saw one PDF for a single college undergrad Lab course session where the students were supposed to do all of that and more during just one week of the lab course, having designed and implemented the RTOS (real-time operating system) for their embedded processor in the two prior weeks!

And if you look at the links for some of the many that are labeled as "sound card" spectrum analyzers, you will see that many or maybe most of them can instead use other ADC data acquisition hardware, which is most-definitely not restricted to audio-type sample rates or word sizes.

So maybe you should go look at the range of data acquisition/sampling/ADC (analog to digital converter) solutions that are out there. THAT is what might limit the specs. But we already know that 1 Hz to multi-GHz and -150 dB are pretty easy and are available ready-to-use, with software, for under $1000, with possibly-lesser capabilities (but still FAR above audio-type sampling) available for much less.

I guess it will come down to how much you want to spend and/or how much DIYing you are able or willing to do.
 
If it's DIY you want, consider getting one of the Texas Instruments development boards and outboard the signal conditioning/protection and amplification front end -- you can do the programming yourself or use that which comes with them -- I think that TI is one company which likes struttin' their stuff. It isn't always necessary to re-invent the wheel.
 
Hello percy,

To follow jackinj, you can show this ADC from TI.
It's a 24bits 4Msps ADC (it's the faster 24bits ADC i know).
You can purchase the eval board of this IC from TI, that include a software allowing
to perform FFT,THD measurement and much more.

As you, i'm very interested to get high resolution system with better bandwidth than traditional sound card.
Regards.

Frex

You can also read this web site,
It's a very interesting project of 12bits 100MHz acquisition system.
 
yep, the ADS1675 was the reason behind the origin of this thread. I only wish there was a chip for 17 or 18 or anything over 16 bit but a little more bandwidth. I’d gladly trade in some vertical resolution for horizontal bandwidth.
Considering an EVM was going to be my next logical step if I didn’t find any diy projects. However I see that for this chip they don’t have a traditional EVM but rather a REF (“Reference Design”) and its $500. At that price it makes you want to consider just buying used equipment or even just renting one when needed. Or who knows maybe its only a matter of time before someone here builds one around this chip, and costs far less than the evm/ref. Jack, what are your thoughts on that ? I haven't looked at that implementation in detail but do you think it can be done for much less than $500 - say $250-300 ?

Gootee, I did check some of those links and didn’t find what I was looking for so stopped there. But if you’re saying there are some in there I will go check them all. I am sure it will be worth the time and effort.
 
Last edited:
Hi percy,

Designing an hardware interface for the ADS1675 is a big job but that would be very pleasant for me...
The main issues (for me) to get a functional system is the needed software to speak with !
If you purchase the EVM for 500$, you have the hardware and the software that allow to use it directly as a FFT spectrum analyzer.
I really doubt that you will find used hardware that could be comparable for that price.
I don't have the EVM, but i'm very tempted !

Jackinj, could speak a little about it ? How it work in real world ?
What type of analysis have you already perform with ?
Thank you !
Regards.

Frex
 
I just did buy on eBAy a card (E-MU1820) that is capable of relevant ADC performance and use it to take some measurements. Now I wil try to get the "m" version of the dock that has even better specs for ADC. My measurements with the present dock (at 24bit/96kHz) provide some 17-18bit of real resolution up to 50kHz. I can try to get 192kHz SR measurements, but I didn't think I needed them.
http://www.diyaudio.com/forums/digital-source/208134-measurements.html
 
Last edited:
yep, the ADS1675 was the reason behind the origin of this thread. I only wish there was a chip for 17 or 18 or anything over 16 bit but a little more bandwidth. I’d gladly trade in some vertical resolution for horizontal bandwidth.
Considering an EVM was going to be my next logical step if I didn’t find any diy projects. However I see that for this chip they don’t have a traditional EVM but rather a REF (“Reference Design”) and its $500. At that price it makes you want to consider just buying used equipment or even just renting one when needed. Or who knows maybe its only a matter of time before someone here builds one around this chip, and costs far less than the evm/ref. Jack, what are your thoughts on that ? I haven't looked at that implementation in detail but do you think it can be done for much less than $500 - say $250-300 ?

What you are asking for, you won't find, at least not for cheap. This is pushing the envelope for chip design, i.e. way over 16 ENOB (i.e. 20-24bit) AND over 50MHz sampling / frequency. If it does exist, expect to pay dearly as it will be in great demand by those who will.

As mentioned, you can get what you want, and people do, with bandpass filtering and adjustable front end end gain / attenuation using solutions with less resolution. You just can't see the entire BW "all at once".

Also, as folks mentioned, the hardware is easy compared to the software and calibration.

Getting a display is one thing - getting an amplitude and time / frequency calibrated display is another. This has to tie in with the front end amplitude gain / attenuation ranging, and so on, and is not trivial, as this part is not built into the A/D chip ;)
Building practical test gear is not an easy thing, especially for people that don't design and build it for their real job.

I'd suggest used, off the shelf test gear, even for more $, like a decent digital scope w/ FFT and filtering in front of it. Or a low frequency SA (40-200MHz). HP made quite a few, and there were others brands too.
Very easy to resell and likely break even once you are done playing.
 
As mentioned, you can get what you want, and people do, with bandpass filtering and adjustable front end end gain / attenuation using solutions with less resolution. You just can't see the entire BW "all at once".

I didn't quite get the 'with bandpass filtering and adjustable front end end gain / attenuation using solutions with less resolution' part. Can you please explain that me in a little more detail ? Perhaps with an example ?
Thanks.
 
I presume he was talking about my "near-off-the-shelf solution."

Furthermore, it wouldn't really be lower resolution. If the device you're testing is any good at all, ultrasonic distortion products will be (pulling out a rough figure) 60dB or more below full audio-band signal output. Once the audio band is filtered out, the remaining ultrasonic component can be amplified by 60dB and analyzed with a "standard" DSO with as little as 8 bits of resolution that then shows the range from 60dB to 110dB below the audio-band signal.
 
A typical scope has about ballpark 45dB dB range on the display screen. So this means at the current full scale input setting, you can see about -45dB from that number.

But, you can adjust the full scale input (peak to peak voltage) on most scopes from ballpark 1mV/division to 5V/ division. It's assumed there are 10 divisions, so typical full scale range is 10mV to 50V. The ratio of these 2 numbers is 74dB.
Since 10mV is full scale, it's assumed you can see 45dB below this, so in essence, by adjusting ranges, you can see 74dB + 45dB = 119dB.

As mentioned above, if you excite in the audio range at say 2Vrms (= 5.65V pk-pk), and then high pass filter with a 4 pole filter (24dB attenuation/octave) above audio (80kHz) you'll have 48dB attenuation at 20kHz. You can then set the scope range to say 20mV full scale and effectively look at the bottom 45dB of the (45+48dB=93dB) range above 100kHz.

For coverage up to 96kHz, you can use a 24/192 sound card, which is probably not good to 96kHz, depending on how the input is alias filtered, but will be usable well above 20kHz.
 
Last edited:
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.