Homebrew Digital Sampling Audio spectrum analyser.

[GK]

A few years ago I designed a fully fledged digital sampling spectrum analyser sweep generator / controller board with double buffered screen memory that used a conventional analogue oscilloscope with X/Y inputs for the display (which I also designed around a CRT pulled from a wrecked Telequiptment oscilloscope).

This circuit board was for a HF spectrum analyser with a narrow dispersion bandwidth and a high resolution IF filter to examine the audio side bands of SSB and AM transmissions.
Digital sampling was used to accommodate the low sweep rates that are required for the high resolution (narrow) IF bandwidth, while double buffered screen memory, (flipped during the fixed 25Hz horizontal re-trace interval) provides a static visual display of the previously measured/recorded low frequency sweep on the oscilloscope CRT during each sampling sweep interval.
Preceding the ADC is a highly accurate logarithmic detector with high dynamic range (an Analog Devices part, can’t remember the part number ATM)

I’ve decided to reengineer the design for audio, with an input bandwidth of 20Hz to 200kHz. There should be a lot of interest in such a design here, so I’m entertaining the possibly of having a batch of professionally made PCB’s for the complete design made for group buy.

Now, I have elected 1MHz to be the first IF frequency and the first mixer will be a MiniCircuits doubly balanced diode ring mixer (low noise + high dynamic range).
For the first local oscillator I will use a hybrid PLL with a DDS (direct digital synthesis) chip for the divider.
The DDS allows precise microprocessor control of the frequency sweep in minute increments, while the frequency output taken from the PLL’s VCO avoids the close-to-carrier spurious frequency outputs of the DDS.


All I have to do now is to find a decent low phase noise VCO IC to cover the relatively low frequency range of 800,000 Hz – 999,980 Hz.

I do not prefer to fiddle around with a discrete varactor tuned VCO at these frequencies. Harmonic distortion doesn't matter (an LPF will fix that), but a linear response would be desirable as would a fair degree of inherent stability.

Any suggestions?

[Jan Dupont]

There should be a lot of interest in such a design here, so I’m entertaining the possibly of having a batch of professionally made PCB’s for the complete design made for group buy.

I'm in

Sadly I can't help you with the parts

[h_a]

A distortion analyzer for x/y input of the oscilloscope

Whoay, that would be fun to have!

All the best, Hannes

[Giaime]

I'm in too, that's very interesting and I also have got 2 or 3 old CRT scopes that can be used (and finally my avatar means something )

Sorry for your parts, I can't help. I don't remember I've ever seen a VCO designed for such low frequency use: don't you think you can use a good one and then divide the frequency with flip-flops?

[syn08]

A low cost solution: 74HCT4046 the high speed version of the CD4046 CMOS IC.

A high performance solution: TLC2933 from TI plus dividers, etc...

Would this project be (performance, cost and effort wise) competitive with a high end sound card? I can get here a brand new Lynx L22 for $600 all inclusive. Or with a second hand pro spectrum analyzer?

[GK]

Originally posted by syn08
A low cost solution: 74HCT4046 the high speed version of the CD4046 CMOS IC.

A high performance solution: TLC2933 from TI plus dividers, etc...

Would this project be (performance, cost and effort wise) competitive with a high end sound card? I can get here a brand new Lynx L22 for $600 all inclusive. Or with a second hand pro spectrum analyzer?

I actually currently intend to use the phase/frequency detector part of the 74HC4046 (which is quite good) in the PLL, but the VCO section of that IC is horrible (fortunately the VCO section has an inhibit input).

I'll have a look at the TLC2933.

All up cost should be ~AU$200.

EDIT:
TLC2933A is a 30MHz+ part, no good.

[syn08]

Originally posted by G.Kleinschmidt


All up cost should be ~AU$200.

TLC2933A is a 30MHz+ part, no good.

$200? Hmmmm...

Just divide that thing by 30. One more chip to add, anyway less board space than a Colpitts + varactor.

[djk]

Swept filters do not work for low frequency audio.

Real pink noise is only flat after a long integration time, even pseudo-random cylical digital noise has a very long integration time.

A) I have designed traditional audio spectrum analyers with individual filters, precision rectifiers, and integrators with guard rings on Teflon boards.

B) I have participated in the re-design of above for lower cost using digitally generated noise and computer integration.

C) I have participated in the re-design of swept-filter designs such as you are suggesting, requires lots of software to be useable.

My experience has been that A is best, B is OK for the money involved, and C is OK for controling a vibe-table, but I wouldn't want to use it for audio analysis.

I still use my production prototype of B for my audio needs.

[GK]

Originally posted by djk
Swept filters do not work for low frequency audio.

Real pink noise is only flat after a long integration time, even pseudo-random cylical digital noise has a very long integration time.

A) I have designed traditional audio spectrum analyers with individual filters, precision rectifiers, and integrators with guard rings on Teflon boards.

B) I have participated in the re-design of above for lower cost using digitally generated noise and computer integration.

C) I have participated in the re-design of swept-filter designs such as you are suggesting, requires lots of software to be useable.

My experience has been that A is best, B is OK for the money involved, and C is OK for controling a vibe-table, but I wouldn't want to use it for audio analysis.

I still use my production prototype of B for my audio needs.

?????

This instrument is intened to examine the harmonic content of steady state audio waveforms such as the distortion residual from a THD analyzer.

If you use digital sampling you can have a sweep time of 100 seconds if you want.
This allows the examination of low frequency audio signals (not possible with a 'traditional' analogue swept frequency spectrum anaylzer) using a high resoultion (narrow bandwidth) IF, as dictated by the formula.

Ro=SQRT(D/T)

Ro = maximum resolution without distortion (guassian filter response)
D = Dispersion
T = Sweep time

[Bonsai]

Glen,

take a look at this

http://uk.geocities.com/cyrilb2@btinternet.com/

Might give you some inspiration/ideas with your project.

Bottom line is most people on the forum would love an easy to build, dedicated distortion analyzer with a 80KHz bandwidth i.e. a poor mans AP.

I looked at Bob Cordells distortion analyzer but it would be a bitch to build - too many boards/interconnects etc and a lot of th e componenets are unavailable now.

Some kind of pre-processor that fed into a PC or scope is probably the right way to go in '08.