
Home  Forums  Rules  Articles  The diyAudio Store  Gallery  Blogs  Register  Donations  FAQ  Calendar  Search  Today's Posts  Mark Forums Read  Search 
Software Tools SPICE, PCB CAD, speaker design and measurement software, calculators 

Please consider donating to help us continue to serve you.
Ads on/off / Custom Title / More PMs / More album space / Advanced printing & mass image saving 

Thread Tools  Search this Thread 
1st January 2011, 07:19 PM  #1 
diyAudio Member
Join Date: Sep 2010
Location: Grove, Ok.

Noise Spectral Density and Fouier Analysis
I have been using Circuitmaker 2000 for several years to do simple sims and now I am getting into an area that I need some instruction and information.
Can anyone explain the NSD function and the associated measurements on the scale? As well as some info on the Fourier Analysis function along with its associated scale. In my amplifier design the NSD displays a nice smooth ramp from 0 to 2.000fV^2/Hz @ 16hz then tapers to a flat 2.250fV^2/Hz @ 64hz then at 131.072khz it slowly ramps up to 3.250fV^2/Hz @ 4.2meghz and of course rises sharply to peak 24.50fV^2/Hz @ 24meghz then sharply declines to 0.500fV^2/Hz @ 42meghz then to a smooth decline down to 0 @ 264.4meghz. On the Fourier Analysis tab my amp shows it starting out at 34mV then droping straight down to 0mV @ .1Hz then obviously flat at 0mV all the way across. Can someone put all that in a decent verbage that I may find revealing? If more info is needed on my circuit I can do so, just ask. I respect/appreciate everyone here. Karl 
1st January 2011, 08:22 PM  #2 
diyAudio Member
Join Date: Sep 2010
Location: Grove, Ok.

I forgot to add that the input is grounded with a +/28V supply.

1st January 2011, 09:55 PM  #3 
diyAudio Member
Join Date: Sep 2010
Location: Grove, Ok.

Restate the question please.
I'm sorry, let me restate that question.
How does one read the NSD and Fourier scales? (i.e. what does fV2^2/Hz translate into?) (and I am guessing Fourier analysis would be Xvoltage @ Xfreq?) 
2nd January 2011, 01:03 AM  #4 
diyAudio Member

In Fourier, X is frequency, Y is magnitude of the signal at that frequency. For instance, in a fourier it can be seen that a square wave can be made from many sine waves, each being an odd harmonic of the fundamental frequency.
As for noise analysis, I don't understand it either. You may know more than I do. I tried looking up noise papers by professor Leach, but it's all bizarre math stuff. If you have had some good math classes you should look up Leach's webpage. BTW, as I understand it, a proper simulator shouldn't show noise in the Fourier analysis. IE, noise is only simulated when you ask for it because most of the time it's just in the way. So the noise floor in your FFT may actually be produced from digital artefacts and maximum resolution when processing. However I have never used CircuitMaker so I can't be sure. Why not post some screenshots of what you're looking at (the graphs). And often times, I've seen people asking questions like this when they were simulating the circuit wrong; the graphs were working right but the circuit was set up wrong. So it may be helpful to post a screenshot of your circuit too.  keantoken 
2nd January 2011, 02:30 AM  #5 
diyAudio Member
Join Date: Sep 2010
Location: Grove, Ok.

Thank you kean
Screen shots of NSD and Fourier: The circuit in question uses a simple LTP with CMs, active CCS with a buffered VAS all mirrored topo with bias . No OPS at the moment. Reference input and node is a sign wave 2Vpp the output node is taken from the input of a generic 1:1 gain stage block feeding the NFB. using +/ 27.6V Rails decoupled. Last edited by kbeist; 2nd January 2011 at 02:41 AM. Reason: aditional info 
2nd January 2011, 03:01 AM  #6 
diyAudio Member

It may be more useful to set the axes on the fourier graph to logarithmic. That is how most attachments on DIYAudio are formatted. In most amps, harmonics above the 3rd aren't visible on a nonlogarithmic graph. This may be a sign of improper operation.
Same goes for the noise graph, you will see more information. Also, those spikes look like resonance peaks. It looks to me like the resolution (number of data points) taken by the graph isn't large enough to show how high they really are, as you can see by the blunted misshapen peaks. If you turn up the resolution it may be more accurate.  keantoken 
2nd January 2011, 04:35 AM  #7 
diyAudio Member
Join Date: Sep 2010
Location: Grove, Ok.

Thanks again kean
It seems as though the Fourier is inhearantly dead set on a linear X axis (nonchangeable, log selections are greyed out), y axis has various selections that seem to only change the y side bar designations, no change in the displayed data. (user input can only change freq to be tested and # of harmonics desired) I even changed the # of cycles to be ran, and nadano change. The only parameters allowable to be changed in the NSD is fundamental freq (this is of course used to see at a specific freq, so I did not want to change this), test points (I changed from 51050100200500), LinearDecadeOctave (checked with all 3). always gives the same graph shape except on linear scale (linear definitly usless at this point). I really apreciate the help. Thanks again. Karl Last edited by kbeist; 2nd January 2011 at 04:38 AM. 
2nd January 2011, 05:23 AM  #8 
diyAudio Member

If you are using a trial version that may be why.
There are a lot of folk using LTSpice, and it's what I use. It's freeeeeeeeeee. And guys here will contend that it's not cheezy. An alternative I'm looking at is superspice: AnaSoft  Analoge Simulation  SuperSpice  keantoken 
2nd January 2011, 06:19 AM  #9 
diyAudio Member
Join Date: Sep 2010
Location: Grove, Ok.

Thanks again,
Most likely that is the case then. Trial version been hacked so at least I have a basic spice program that works on the basics only. Looks as though I have advanced into the next level and I will have to spend some to make some. I like all the nice features. I think I may have that one in my favs link, if not it will be. I will be looking harder at that one now. Again I thank you for helping me. Karl 
2nd January 2011, 07:00 AM  #10 
diyAudio Member

It looks like SuperSpice is the next level as far as convenience maybe. For effective simulation, I'm not sure. It includes all the default western SPICE models, but not the japanese ones, which tend to be passably accurate. It only allows temperature steps and sweeps, but doesn't perform a realtime temperature analysis. This is a real downer. Without real time selfheating effects taken into account for semiconductors, your circuit will be very different from real life in terms of linearity and what you can get away with thermally. Another one is the lack of an effective management of the inclusion of parasitics. With most simulators you must enter in trace parasitics in a slow, tedious manner. This would be far simpler if you could simply right click on a wire and give it an LCR value, or join it with another trace to form a transmission line or inductively coupled traces. Won't necessarily be perfect, but it will be more realistic than what we have now. These are the features I'm waiting for in SPICE.
If you want to try out LTSpice, Email me and I'll help you get through the tough part.  keantoken 
Thread Tools  Search this Thread 


Similar Threads  
Thread  Thread Starter  Forum  Replies  Last Post 
density of cabinet material  andy1971  Construction Tips  10  2nd January 2010 01:50 PM 
High density Fibreboard UK Supplier?  calvert73  Subwoofers  1  17th December 2008 11:57 PM 
Spectral analysis of sunrise time variation  Circlotron  Everything Else  14  11th May 2007 04:43 PM 
noise analysis  tenderland  Tubes / Valves  5  30th March 2005 12:22 AM 
Spectral analysis  use pink or white noise?  Circlotron  Everything Else  5  20th November 2003 01:13 PM 
New To Site?  Need Help? 