here is a spectrum analyser design that i would like the group to
comment upon.
1) we take the input via a low pass filter, up convert it to an IF of
100Mhz or so, and follow it up with a direct conversion receiver at
100 Mhz with 20 khz bandwidth.
2) the upconverting local oscillator is a VCO that is controlled by a
sweep generator. the sweep is controlled by a PWM modulated signal in
the audio range.
3) the sweep generator input is connected to the output of a PC sound
card. the output of the direct conversion receiver is connected to the
input of the PC sound card.

Now, by clever programming of the sound card on the PC, we can make
the VCO sweep our passband of interest. The sound is often digitsed at
16 bit levels (in the better systems at 32-bit level). This will
effectively give us 90db range. the lograithmic scale can be
implemented in software. DSP can be used to set the bandwidth to any
particular size.

the most important benefit of this design will be that even hams
without expensive oscilloscopes will be able to easily make a PC based
analyser that is easy to assemble and use.

if there isn't any glaring problem with this design, i would like to
pull out my soldering iron and take a go. is anyone here with spectrum
analyser experience willing to share knowledge?

"Ashhar Farhan" wrote in message
om...
here is a spectrum analyser design that i would like the group to
comment upon.
1) we take the input via a low pass filter, up convert it to an IF of
100Mhz or so, and follow it up with a direct conversion receiver at
100 Mhz with 20 khz bandwidth.
2) the upconverting local oscillator is a VCO that is controlled by a
sweep generator. the sweep is controlled by a PWM modulated signal in
the audio range.
3) the sweep generator input is connected to the output of a PC sound
card. the output of the direct conversion receiver is connected to the
input of the PC sound card.

Now, by clever programming of the sound card on the PC, we can make
the VCO sweep our passband of interest. The sound is often digitsed at
16 bit levels (in the better systems at 32-bit level). This will
effectively give us 90db range. the lograithmic scale can be
implemented in software. DSP can be used to set the bandwidth to any
particular size.

the most important benefit of this design will be that even hams
without expensive oscilloscopes will be able to easily make a PC based
analyser that is easy to assemble and use.

if there isn't any glaring problem with this design, i would like to
pull out my soldering iron and take a go. is anyone here with spectrum
analyser experience willing to share knowledge?

I have recently completed a spectrum analyser, see
http://www.hanssummers.com/electroni...yser/index.htm.
It is
awaiting possible magasine publication so there are not yet any circuit or
construction details on the page above. If you want the full details, email
me privately and I'll show them to you.

I also tried a direct conversion receiver initially. It doesn't work on in
analogue (i.e. non-PC) analyser, because there are all sorts of heterodynes
of the sweep frequency against the directly converted incoming signal. Of
course I kicked myself afterwards for not thinking of it in advance to save
myself the time of the experiment.

I think broadly speaking the final IF should be substantially higher than
the frequency of the sweep waveform, so that the final filtering works
faster than the sweep. There are lots of people in this forum far more
advanced than me who will probably be able to explain it better in terms of
filter response times or group delays or something.

Though there might be a way of untangling everything in software so it may
work. Being direct conversion you'll also have both sidebands present, which
will create further complications. Again, clever software might untangle it
but I think it's far from straightforward.

Another problem is the narrow bandwidth. 20KHz is a nice bandwidth to have
but I think in a spectrum analyser you also want wider bandwidths available.
In particular, if you are digitally generating your sweep voltage, and
trying to cover the whole 100MHz, you need of the order of 100,000 / 20 =
5,000 discrete measurement intervals. You can't display that many horizontal
pixels on screen. You could average them in software, but at the low 20KHz
bandwidth, you're going to need quite a slow sweep rate. 5,000 measurements
are a lot and will take a long time.

It's a nice idea but I don't think it will work as it stands.

My recommendation would be to add a 2nd IF to your design, 2nd IF amp and
logarithmic detector. In my design I used a 145MHz 1st IF, so the VCO sweeps
145 - 290MHz. The 2nd local oscillator is at 153MHz for an 8MHz 2nd IF,
amplified then passed into an AD8307 logarithmic amplifier. Anything similar
would work well. I used an SA602 front end for simplicity, but a diode ring
mixer would give potentially better performance than the 65-70dB dynamic
range I achieved.

You can still use the PC for a nice display, rather than an oscilloscope.
Just feed the log output into your PC sound card, and have the PC sound card
control the sweep as you suggest. I think you'll solve a lot of problems by
adding these few extra modules to the analogue front end before introducing
the PC. Incidentally, this is exactly what I'm doing with my Mk2 analyser,
see
http://www.hanssummers.com/electroni...ser2/index.htm.

Now I'm hoping for someone to put it all more clearly and professionally
than I have here ;-)

73
Hans G0UPL
http://www.HansSummers.com

"Hans Summers" wrote in message news:bmgam6
I also tried a direct conversion receiver initially. It doesn't work on in
analogue (i.e. non-PC) analyser, because there are all sorts of heterodynes
of the sweep frequency against the directly converted incoming signal. Of
course I kicked myself afterwards for not thinking of it in advance to save
myself the time of the experiment.


a) i am not proposing a direct direct conversion. i am proposing an up
conversion exactly like han's MK1. for the second conversion, i am
suggesting a direct conversion to base-band from the high IF. as you
rightly point out harmonic mixing is a problem with broad-band direct
conversion receivers, therefore, you need to have a low-pass or a
band-pass ahead of a diode mixer working as a product detector.
b) my second suggestion is to do the logarithmic calculations
digitally on the PC. that will simplify the design.
c) as for the granuality of the sweep, for finer resolution the sweep
range will be decreased. that is one way to get higher resolution, the
other is to slow down the sweep.

what kind of a VCO are you using? and how are you ensuring that the
vco output remains constant throughout the sweep? i expect that the
VCO's amplitude will also effect the first mixer gain.

- farhan

"Ashhar Farhan" wrote in message
om...
"Hans Summers" wrote in message news:bmgam6
I also tried a direct conversion receiver initially. It doesn't work on
in
analogue (i.e. non-PC) analyser, because there are all sorts of
heterodynes
of the sweep frequency against the directly converted incoming signal.
Of
course I kicked myself afterwards for not thinking of it in advance to
save
myself the time of the experiment.


a) i am not proposing a direct direct conversion. i am proposing an up
conversion exactly like han's MK1. for the second conversion, i am
suggesting a direct conversion to base-band from the high IF. as you
rightly point out harmonic mixing is a problem with broad-band direct
conversion receivers, therefore, you need to have a low-pass or a
band-pass ahead of a diode mixer working as a product detector.

That wasn't the effect I was worrying about... after all your up mixing and
direct conversion of the VHF IF down to baseband, and low pass filtered it,
you'll end up with say 20KHz of audio baseband to feed into your soundcard.
Unfortunately your sweep rate is also down in the same range. Which is a
problem because whatever method you use for envelope detection of the
baseband will have to react faster than the sweep rate. Though as I say, if
you do it all in software you might be able to untangle it.

b) my second suggestion is to do the logarithmic calculations
digitally on the PC. that will simplify the design.

It simplifies the analogue design and complicates the software. 16 bits of
resolution should give adequate dynamic range. Depends what you want to
spend time on... The AD8307 chip is very easy to use, if a little on the
expensive side.

If you have to use the PC to unwrap the envelope detection mess as mentioned
above then you have to do the logarithmic bit in PC anyway.

c) as for the granuality of the sweep, for finer resolution the sweep
range will be decreased. that is one way to get higher resolution, the
other is to slow down the sweep.

Agreed, but if the widest resolution is only 20KHz, then you have to do
5,000 samples to cover 0-100MHz input bandwidth. This fine resolution is
imposed by the relatively narrow bandwidth, otherwise you will suffer
significant loss of accuracy on your displayed result. The way I think of it
is in terms of frequency spikes falling into the holes between samples. In
the extreme imagine taking a VCO going in 100 steps, so making an amplitude
measurement at each 1MHz of the range 1-100MHz. Your bandwidth 20KHz.
Signals exactly on the MHz will be no problem. But what about a signal at
say 12.5MHz... what does it look like on the analyser? It's 500KHz away from
the centre of the 20KHz passband for both adjacent measurement points 12MHz
and 13MHz. It's fallen into the hole. What you'd see on your display depends
on the skirt selectivity of your filters. If you're using a 20KHz soundcard
as the filter, the skirts will be quite sharp, so it's likely you'd see
almost nothing at 500KHz baseband. This would mean the majority of
frequencies in your input spectrum would be absent from your display or at
entirely the wrong amplitude.


what kind of a VCO are you using? and how are you ensuring that the
vco output remains constant throughout the sweep? i expect that the
VCO's amplitude will also effect the first mixer gain.

I'm just using the internal oscillator of the SA602 mixer/oscillator chip.
As you suggest, probably not at all constant ;-)

Hans
http://www.HansSummers.com

i just ran a check with the sound card:

a) the sound card contributes noise. i shoved in a 50 ohms resistor
into the mic input, cranked up the mic level, save a 'silent' wave
file and saw it through a hex editor. There is a random noise of about
2-3 bits. That means, the effective range is not 16 bits but 12 bits.
12 bits can represent 4096-1 as the highest number. That means, the
dynamic range is 4096 squared: about 16 million or 72db. A little low
really for interesting work. there isnt enough 'roof' to resolve a
spur about 60db down without touching the grass

b)the sampling is at 44khz. some of the posts got confused between
20khz b/w and 20khz sampling rate.

hans is right about the PC being a kludge when compared to an
oscilloscope. I live in India and second-hand oscilloscopes are a
rarity. The new ones cost an engineer's whole year's salary. I have
purchased a tektronix 454 last month on ebay for $300, it is still on
its way to India. So, while i can imagine that in a number of more
developed countries oscilloscopes are not really a problem, they do
remain a problem in many parts of the world.

More often than not, an amateur already has a PC. Adding a simple
hardware to be able to do quick and dirty spectrum analysis might be
an interesting option. As Wes writes in SSD and EMRFD, the purpose of
test instruments is to help with the projects on hand, rather than be
projects themselves (then he went on to homebrewing a spectrum
analyser, hehe).

- farhan

Actually, I'd be tempted to rig a phasing type conversion to baseband from
the 300 MHz or whatever the first IF is....

At best it might get 40-50 dB of suppression of the other sideband, but it
would remove most of the undesired spurs and something like an AD8307 log
detector could be used to convert to a DC level.

After that, feed it into an A/D converter and into a PC using the printer
port.

Given a dual channel A/D converter, I'd also sample the tuning voltage for
the VCO and feed that into the PC as well.

All of this would move the programming from a possibly complex DSP problem
to a relatively simple Parallel port input and then number crunching to a
display.

There would be spur responses 40-50 dB down, but generally speaking, it
should give useful information.

===========

Granted, one problem is how fast the data transfer is through the parallel
port, that would limit how many samples/second the system could process.

There is the second issue, about maximum sweep speed Vs the IF bandwidth,
and the maximum useful bandwidth would be a function of the Phase shift
detector system. Still, I think this approach has some possibilities.

Almost forgot... WIN XP and WIN2000 make getting to the parallel port for
something like this MUCH more difficult then it was under WIN 98.

You could use RS232 and perhaps two COM ports to allow getting the AD8307
and the VCO sweep voltage in parallel, so to speak. Once again, it would be
a limit on how many samples per second the systme could provide.

If you want to be more complex, you could feed OUT commands to a
synthesizer to tune the VCO and this would allow really slow sweeps and very
precise accuracy, but it complicates the design.

I have to admit, I've been looking at the synthesizer design John Miles,
KE5FX, did which tunes from 1 GHz to 2 GHz with very fine tuning steps and
all osrts of ideas for various test gear built around it keep dancing in my
mind. I suppose I should see if I can build one and get some idea as to
the cost and so on. But it'd make a wonderful starting point for a
digitally tuned spectrum analyzer !!!

Admitted, the close in phase noise could limit dynamic range 'way below
the range pssible with an AD8307 detector, but even so the frequency
accuracy of an analyzer would be fantastic.

=========

Overall, Avery is probably correct that the result may be a lot of work an
less than perfect performance, but as a homebrewer it does seem like a fun
way to get something fairly useful and learn a lot as the project goes on.




Jim Pennell
N6BIU

hans is right about the PC being a kludge when compared to an
oscilloscope. I live in India and second-hand oscilloscopes are a
rarity. The new ones cost an engineer's whole year's salary. I have
purchased a tektronix 454 last month on ebay for $300, it is still on
its way to India. So, while i can imagine that in a number of more
developed countries oscilloscopes are not really a problem, they do
remain a problem in many parts of the world.

That puts another angle on it. In that case I'd say you could produce a nice
instrument by using a complete analogue analyser such as my design
http://www.hanssummers.com/electroni...yser/index.htm
or similar, and feeding the sweep (X-axis) and logarithmic output (Y-axis)
into the PC via ADC's on the parallel port. 8-bit ADC's would probably be
marginally good enough, 10 bit for sure.


More often than not, an amateur already has a PC. Adding a simple
hardware to be able to do quick and dirty spectrum analysis might be
an interesting option. As Wes writes in SSD and EMRFD, the purpose of
test instruments is to help with the projects on hand, rather than be
projects themselves (then he went on to homebrewing a spectrum
analyser, hehe).

It became a project in its own right for me. So much that it spawned a
second project
http://www.hanssummers.com/electroni...ser2/index.htm.
But not wasted time - I learnt so much making it and got my first experience
at VHF work etc, so it was time well spent.

Hans G0UPL
http://www.hanssummers.com

Actually, I'd be tempted to rig a phasing type conversion to baseband from
the 300 MHz or whatever the first IF is....

At best it might get 40-50 dB of suppression of the other sideband, but it
would remove most of the undesired spurs and something like an AD8307 log
detector could be used to convert to a DC level.

After that, feed it into an A/D converter and into a PC using the printer
port.

Given a dual channel A/D converter, I'd also sample the tuning voltage for
the VCO and feed that into the PC as well.

All of this would move the programming from a possibly complex DSP problem
to a relatively simple Parallel port input and then number crunching to a
display.

There would be spur responses 40-50 dB down, but generally speaking, it
should give useful information.

===========

Granted, one problem is how fast the data transfer is through the parallel
port, that would limit how many samples/second the system could process.

There is the second issue, about maximum sweep speed Vs the IF bandwidth,
and the maximum useful bandwidth would be a function of the Phase shift
detector system. Still, I think this approach has some possibilities.

Almost forgot... WIN XP and WIN2000 make getting to the parallel port for
something like this MUCH more difficult then it was under WIN 98.

You could use RS232 and perhaps two COM ports to allow getting the AD8307
and the VCO sweep voltage in parallel, so to speak. Once again, it would be
a limit on how many samples per second the systme could provide.

If you want to be more complex, you could feed OUT commands to a
synthesizer to tune the VCO and this would allow really slow sweeps and very
precise accuracy, but it complicates the design.

I have to admit, I've been looking at the synthesizer design John Miles,
KE5FX, did which tunes from 1 GHz to 2 GHz with very fine tuning steps and
all osrts of ideas for various test gear built around it keep dancing in my
mind. I suppose I should see if I can build one and get some idea as to
the cost and so on. But it'd make a wonderful starting point for a
digitally tuned spectrum analyzer !!!

Admitted, the close in phase noise could limit dynamic range 'way below
the range pssible with an AD8307 detector, but even so the frequency
accuracy of an analyzer would be fantastic.

=========

Overall, Avery is probably correct that the result may be a lot of work an
less than perfect performance, but as a homebrewer it does seem like a fun
way to get something fairly useful and learn a lot as the project goes on.




Jim Pennell
N6BIU

hans is right about the PC being a kludge when compared to an
oscilloscope. I live in India and second-hand oscilloscopes are a
rarity. The new ones cost an engineer's whole year's salary. I have
purchased a tektronix 454 last month on ebay for $300, it is still on
its way to India. So, while i can imagine that in a number of more
developed countries oscilloscopes are not really a problem, they do
remain a problem in many parts of the world.

That puts another angle on it. In that case I'd say you could produce a nice
instrument by using a complete analogue analyser such as my design
http://www.hanssummers.com/electroni...yser/index.htm
or similar, and feeding the sweep (X-axis) and logarithmic output (Y-axis)
into the PC via ADC's on the parallel port. 8-bit ADC's would probably be
marginally good enough, 10 bit for sure.


More often than not, an amateur already has a PC. Adding a simple
hardware to be able to do quick and dirty spectrum analysis might be
an interesting option. As Wes writes in SSD and EMRFD, the purpose of
test instruments is to help with the projects on hand, rather than be
projects themselves (then he went on to homebrewing a spectrum
analyser, hehe).

It became a project in its own right for me. So much that it spawned a
second project
http://www.hanssummers.com/electroni...ser2/index.htm.
But not wasted time - I learnt so much making it and got my first experience
at VHF work etc, so it was time well spent.

Hans G0UPL
http://www.hanssummers.com

i just ran a check with the sound card:

a) the sound card contributes noise. i shoved in a 50 ohms resistor
into the mic input, cranked up the mic level, save a 'silent' wave
file and saw it through a hex editor. There is a random noise of about
2-3 bits. That means, the effective range is not 16 bits but 12 bits.
12 bits can represent 4096-1 as the highest number. That means, the
dynamic range is 4096 squared: about 16 million or 72db. A little low
really for interesting work. there isnt enough 'roof' to resolve a
spur about 60db down without touching the grass

b)the sampling is at 44khz. some of the posts got confused between
20khz b/w and 20khz sampling rate.

hans is right about the PC being a kludge when compared to an
oscilloscope. I live in India and second-hand oscilloscopes are a
rarity. The new ones cost an engineer's whole year's salary. I have
purchased a tektronix 454 last month on ebay for $300, it is still on
its way to India. So, while i can imagine that in a number of more
developed countries oscilloscopes are not really a problem, they do
remain a problem in many parts of the world.

More often than not, an amateur already has a PC. Adding a simple
hardware to be able to do quick and dirty spectrum analysis might be
an interesting option. As Wes writes in SSD and EMRFD, the purpose of
test instruments is to help with the projects on hand, rather than be
projects themselves (then he went on to homebrewing a spectrum
analyser, hehe).

- farhan