[Ryan] wanted a spectrum analyzer for his audio equipment. Rather than grab a micro, he did it the analog way. [Ryan] designed a 10 band audio spectrum analyzer. This means that he needs 10 band-pass filters. As the name implies, a band-pass filter will only allow signals with frequency of a selected band to pass. Signals with frequency above or below the filter’s passband will be attenuated. The band-pass itself is constructed from a high pass and a low pass filter. [Ryan] used simple resistor capacitor (RC) filters to implement his design.
All those discrete components would quickly attenuate [Ryan’s] input signal, so each stage uses two op-amps. The first stage is a buffer for each band. The second op-amp, located after the band-pass filters, is configured as a non-inverting amplifier. These amplifiers boost the individual band signals before they leave the board. [Ryan] even added an “energy filler” mode. In normal mode, the analyzer’s output will exactly follow the input signal. In “energy filler” (AKA peak detect) mode, the output will display the signal peaks, with a slow decay down to the input signal. The energy filler mode is created by using an n-channel FET to store charge in an electrolytic capacitor.
Have we mentioned that for 10 bands, all this circuitry had to be built 10 times? Not to mention input buffering circuitry. With all this done, [Ryan] still has to build the output portion of the analyzer: 160 blue LEDs and their associated drive circuitry. Going “all analog” may seem crazy in this day and age of high-speed micro controllers and FFTs, but the simple fact is that these circuits work, and work well. The only thing to fear is perf board solder shorts. We think debugging those is half the fun.
I wonder if this is the best (cheap also) way to do this. because I need to rework his notes (all images) to make a schematic again to make my own. anyone suggestions ?
While this is how a sophomore electrical engineering student would probably build it, modern analog designers would use a switched-capacitor bandpass filter, which provides high Q, low component count, and reconfigurability. If I were to do it, I’d get a single LTC1059 (or something of that sort) and switch it rapidly between each of the 10 bands you want to sample. The output would be precision-rectified and then sent to a 1-to-10 analog mux, which would drive 10 different electrolytic capacitors. By switching the frequency and the mux at the same time, each electrolytic capacitor would store a voltage representing the energy in that particular band. A 40-cent MCU (or much more expensive analog timer circuit) would be used to drive the switching action.
Since you already have the scanning logic, you can also mux the analog voltage from the individual caps to a single VU driver chip. The VU chip drives the common rows while the scanning logic drives the column on a multiplexed LED matrix.
Instead of individual band-pass filters, I would have considered a design that amplified the difference between successive low-pass filter stages. Assuming phase delays between filters are well matched, this would cut the number of filter components in half. (or allow for sharper filters) The biggest cost, would be the extra resistors in the output difference amplifiers.
when I say “successive filter stages” I’m thinking of 9 parallel low pass filters, with successively higher bandwidths.
What do you mean. Could you elaborate?
I would go buy the cheapest ARM chip and run FFT on it. :)
(AVR ADC is too slow)
That’s the easiest/cheapest way, but it’s not analog :-)
There used to be a kit from Velleman (the K4300) that did this identically; a 10-band audio spectrum analyzer done entirely with bandpass filters and LED bar drivers. I built one in 1999 and the display board was indeed the crappiest part of the build. Two hundred through-hole solder joints in a 10×20 grid on 0.1 spacing, plus a ribbon cable. AND you really wanted to make sure that the flat face of each LED was at the exact same level (the holes were large enough that the ‘wide’ spot on the lead would fit through)…
Klasse!
I’d like to see “page 61” (http://i.imgur.com/RNgpRTg.jpg) to find out what the actual bands are but it does look line an interesting project.
For something a bit more advanced, but analog:
http://sound.westhost.com/project136.htm
I really enjoyed reading about this project…… It’s nice to see some analog design without anything blinking, arduinoing, or servoing.
Find an old n-band Graphic Equaliser.
Remove sliders and replace with LED rows.
Add n-signal rectifiers,
and n-way commutated log law Line-of-Light chip (or just n-off).
My colleague build 31 band (10 LEDs on each band) spectrum analiser on 80′. Peak and average indication+noise generator for speaker aligment. He told that he soldered it 4 or 5 months. Want some photos?
It would be too much work to hack a GEQ. The gyrators are all common to the input-output. Cutting and rerouting the ckt-bd, yech. Gulbransen Organ flute filters are one octave, there are 7 in an organ usually times 2. They have separate inputs and common out but that’s easy to change. Lowery has half octave filters but they wank.
Commutation is an analogue no no. In a word,flicker! N-display chips or stacks of quad comparators. All display all frequencies all the time.
my led spectrum analyzes diy. some led are partially lit. how can i remove that? its like a shadow from other group of led in sync but partially lit.
Would love to see it in action … that IS why this was built right ?