Suggestion for cheap network analyzer / impedance analyzer

That's how it's done right. But in most scenarios one also needs a reference input for A/R measurements.

--
Regards, Joerg 

http://www.analogconsultants.com/

With a good mixer it could go down to DC but for LF stuff the sound card of a PC is also very good. Up to around 20kHz. This may even be one of the best options for a low cost solution, a sound card input with (possibly switchable) pre-mixers. An advantage would be that you already have a reference channel because they are stereo. Best to get one with separate RCA connectors and not those dreaded 3.5mm phono plugs. Better channel separation, no erratic ground contact. I prefer external USB sound modules for such jobs.

Don't forget your wife. They don't take it lightly when not specifically mentioned :-)

--
Regards, Joerg 

http://www.analogconsultants.com/

ode 100 and it's decent

or my home lab)

pectrum analysis mode, peak hold and using a HP function generator as a swe ep generator, but since it captures the entire spectrum, a high level signa l at low frequency makes high frequency low signals disappear in noise - e ffective about 70 dB dynamic range

No one's mentioned the Rigol DSA815-TG Not quite your frequency range ~(10 kHz - 1 GHz)

George H.

Den mandag den 18. august 2014 00.55.50 UTC+2 skrev Klaus Kragelund:

at that speed +14 bits gets awfully expensive

I know a guy that just bought on of these:

haven't heard if it is any good, look like the software is java and connects as a comport

-Lasse

Right, that's my real boss. Any spare time goes into her projects :-)

Cheers

Klaus

I've got a couple of very good National Instruments USB modules, 24 bit ADC up to 25kHz, but they are no good for higher frequency and mixing different front ends for various frequencies just adds work

Cheers

Klaus

bode 100 and it's decent

for my home lab)

spectrum analysis mode, peak hold and using a HP function generator as a s weep generator, but since it captures the entire spectrum, a high level sig nal at low frequency makes high frequency low signals disappear in noise - effective about 70 dB dynamic range

I really need something that can do the gain/phase measurement of filters/p ower supplies and from what I can read, the Rigol does not seem to do that

Regards

Klaus

I e

Great

Sounds interesting

I

I have written a good portion of code. I like working on the HW stuff, but often the good system design only emerges when you can gap the HW/SW discip lines. Mostly C, a great deal of visual programming in Labwindows, which al lows the use of standard functions in libraries (so when you need a FFT fun ction or whatever, you don't have to start from scratch)

t

Yes, my email takes it all. The email is: snipped-for-privacy@hotmail.com

For a project like this splitting the HW and SW up wisely makes good sense, so if the front-end needs to be changed or another type of front-end is us ed, it's only writing a new SW driver and tweeking the configuration settin gs

Cheers

Klaus

That's why, when faced with needing a small spectrum analyzer for the road, I went the decadent route and just bought the Signalhound plus matching track gen. It felt like cheating because real men build this kind of stuff but I didn't have the time.

--
Regards, Joerg 

http://www.analogconsultants.com/

. We need another that can be used for more mobile uses. I have tried the b ode 100 and it's decent

one for my home lab) I have tried with the Picoscope 3424, 12bit USB scope, hooking it up in spectrum analysis mode, peak hold and using a HP function generator as a sweep generator, but since it captures the entire spectrum, a high level signal at low frequency makes high frequency low signals disa ppear in noise - effective about 70 dB dynamic range Could even be a kit, but I would like specs: 100Hz to 10 MHz Minimum 80 dB dynamic range or bett er Network and impedance analysis

eries where I have both the track gen and analyzer. Supposedly one can do scalar network analysis with it but I've never done t hat:

at I have does cover down to 100Hz and lower:

h the usual 8-bitters there is no chance to achieve what Klaus wants to do. Gussied up oscilloscopes are generally not the answer in this scenario. Ma ybe SDRs are.

running average filter (if the frequency is low enough to do that), the re solution is boosted to 14-16 bit.

ou have ideal signals all the time with no pulsating stuff in there.

The process doesn't work too well if your signals really are ideal and nois e-free.

To get real resolution enhancement you need enough noise on the signal to " dither" it across at least two adjacent A/D thresholds. Depending on the st yle of A/D - successive approximation A/D converters don't have their thres holds all that nicely spaced - you may need more than the bare minimum of d ither.

se are normally only found in gear such as SDR. They can do this because they use a down-converter. For example, the last stage of my Signalhound has a b andwidth of only 250kHz for this reason, so they can use a hi-res ADC. In medical we call this instantaneous dynamic range. If you want 80dB for t hat you realistically need a 16-bit converter with 14 bits or so of ENOB.

mV/div), function generator with similar output range levels (max 10V down to 10mV into 50 ohm) and syncronized averaging with swept single frequency sine waveform, yields dynamic range over the entire frequency range of exce ss of 100dB (only problem it is slow).

What *do* you mean by "pulsating"? If you have low frequency components in your signal that nearly coincide with with one of the scanning frequencies of your system you can get very obvious beat frequencies low enough to matc h the nominal 1 Hz of your pulse rate.

--
Bill Sloman, Sydney

sent you a block diagram, alas, it is of the lower frequency version.

[...]

What I mean is irregular signal amplitude and phase like you have it in modern digital communications, also pulse-echo with agile PRF, and so on. There you need a "real" analyzer.

--
Regards, Joerg 

http://www.analogconsultants.com/

(Not a criticism of you, Klaus) ...

I think we're jumping the gun here. What we need to do is first decide how the measurements are to be taken in the analogue world, and only then speculate on what might be the computing approach to present the information.

(Speaking as a 40-year real-time embedded softy, the computing stuff would be a relative doddle.)

I think you are right, when the signal is down-mixed correctly, the sampling stage is simple. What is not simple is the SW on the PC, takes quite a bit of coding for something that is really usefull

For example for network analysis, the SW must continously monitor the output of the DUT and regulate the generator amplitude to provide sufficient signal but at the same time not entering the non-linear region of the DUT

Regards

Klaus

Then, clearly I'd have a role on such a project. Everything that you wish to do is striaghtforward, and merely needs the cranking of the handle to turn it out. Although even things that are a doddle still take some time to implement.

Multi-tasking to achieve the above is feasible even in single-processor PCs.

what network analyser does that? I've not seen any

-Lasse

I haven't seen any either. This is one of the problems with Bode analysis. If the amplitude is too low you see erratic signals a lot. If it is too high the regulator loop can go haywire and you might experience a loud bang, followed by smoke and a standing ovation from the colleagues.

What you can sometimes do is add some weight to the amplitude so it slopes with frequency. But I've never seen active adaptation features.

--
Regards, Joerg 

http://www.analogconsultants.com/

yeh, I'd think something active would just add confusion

-Lasse