Yes, you got it right. This is, as above, the hard part of designing a scope. I work directly with R&D people that do this - design PC based scopes - in my current job, and have since 2003. So I'm pretty close to understanding what goes on inside.
The point is the A/D chip has one max input level - say 2V, in many cases it must be balanced. That's not very friendly, so the instrumentation "front end" has to add quite a bit of "other" capabilities - AC,DC coupling, 50ohm/1Mohm inputs, filtering, triggering, and gain/attenuation. In most cases, the bandwidth of the scope will be determined by the front end choices.
That why it's so hard to build an equivalent type instrument DIY. And yes, sound cards are also poor in the front end capabilities, but the 24 bit range gives them lot of dynamic room.
Yep, did you miss the mention of a sound card for that in the post above?
Turns out when you measure effective bits, which includes noise, THD, etc, it measures great at low frequency, and poor at high frequency. The quoted number is usually in the BW of the scope, or broken out for different frequencies.
So a 500MS/s scope has really GOOD effective bits in the 10MHz range and below, and will be significantly worse at say 250MHz.
A perfect 8 bit scope would have 8 effective bits, but it's not that uncommon to see 7 or 7.2 bits for low frequencies like 10MHz - hence the "ballpark" number of 45dB above. Also, you can use averaging to increase the effective bits, and FFT techniques to extend the spur free dynamic range beyond the 45dB number.
And yes, scope front end design separates the men from the boys. When not done correctly, it can add a lot of unwanted "things" to the signal. The best companies know how to do it right.
Actually, in my experience, no, and no. Generally all commercial scopes I've worked with have both gain and attenuation. And the harder thing to do is gain.
Unless you are talking about whopper voltages, like way over 50V, and then there are other problems where bandwidth usually suffers as compared to 50 ohm bandwidth at lower voltages.
Can I ask a couple of questions for my general understanding of scopes and spectrum analyzers –
Does this all apply to spectrum analyzers also ? I am assuming spectrum analyzers have a similar input gain/att setting as well ?
And is it possible to export data from a digital storage oscilloscope out to a PC where a third party program can be used to do spectrum analysis on it ?
Thanks.
Does this all apply to spectrum analyzers also ? I am assuming spectrum analyzers have a similar input gain/att setting as well ?
And is it possible to export data from a digital storage oscilloscope out to a PC where a third party program can be used to do spectrum analysis on it ?
Thanks.
It's all about the ADC real resolution (not equal with the number of bits). Faster ADC's like the ones in oscilloscopes will have less bits than a dedicated audio ADC. Digital scopes have usually 8-12 bit ADC. Noise floor is usually around -50...-90dB depending of the modes (slow average gets better SNR).
A high quality scope costs 10-50 times the amount that you pay for a professional sound card. So it's all about what you really need to measure.
A high quality scope costs 10-50 times the amount that you pay for a professional sound card. So it's all about what you really need to measure.
Last edited:
My Tektronix TDS1002B (a low-end digital 'scope) can export a text file of the sample values to a USB Flash drive. I have never done so, but I'm sure the file can be read by Excel, Matlab, etc - perhaps converted to a *.wav file for analysis by LTSpice, or an audio-processing program.
Dale
That low-end costs some $1000, has 8 bit resolution (except on 2mV/div). It's great for some things, but lousy for high-end audio measurements
Last edited:
That low-end costs some $1000, has 8 bit resolution (except on 2mV/div). It's great for some things, but lousy for high-end audio measurements
may want to read
http://www.diyaudio.com/forums/equi...ution-diy-spectrum-analyzers.html#post2934221
again, especially the 2nd part of the post looking to do measurements ABOVE 2MHz to understand what we are discussing here.
Yes, in general, spectrum analyzers have a front end with attenuation and gain. The way the measurement is done vs. scopes is totally different in "old style" SA's, while some of the newer high end units have hybrid modes, where real time scope type measurements can be done over a limited band of frequency, that the user can select. This is needed to analyze digital communication type wireless signals.
In general, if the scope has a disk drive, usb port, GP-IB, LAN, RS-232, etc. you can export the data. How it's done is a little different for all.
Most people use Matlab, Labview, or something like that for extensive post processing.
I think you need to read that again, there was no 2MHz mention:
4MSPS at 24 bit means less than 350kHz bandwith...
The oscilloscope is not closer of the original post requirements either.
Last edited:
Just a few additional comments:
All of our "real" Agilent RF spectrum analyzers at work (e.g. 4407B models) can have noise floors down to well below -110 dBm power, from 10 MHz to 6 GHz. And those are just the "medium quality" $50,000 ones, with RBW (resolution bandwidth) that only goes down to 1 kHz. The noise floor depends mostly on the RBW, and the frequency span being analyzed. Even for those models, I believe, there are options available to get RBW of down to 1 Hz, which should take the noise floor down to better than -140 dBm.
Also, the ones we use all have GP-IB (HP-IB) ports and usually the most-convenient way to get the data from those analyzers to a PC is, for me, with a GP-IB/USB adapter with either the Agilent or National Instruments HP-IB/GP-IB drivers on the PC. That puts a special spectrum analyzer menu bar right in Microsoft Excel. One click puts all of the frequency points in a column and each trace's data into a column. All relevant analyzer settings are also placed into the spreadsheet at the same time, along with a color plot of all of the data, exactly like the analyzer's screen. Another button click on the spectrum analyzer menu bar in Excel adds a color-correct screen capture into the spreadsheet (showing everything on the analyzer's screen, not just the traces). With just those adapters and some included driver software, we can also write Visual Basic macros in Excel that can completely control the analyzer, perform similar data captures, control other instruments, etc. Yes, Labview or Matlab would be more powerful, but the simple adapter and driver software are probably the most convenient, and can be used just as easily in the field, with a laptop.
In an Agilent "RF/Microwave Measurement Fundamentals" course that I was able to take, one part of what we learned was how to make better Spec-An measurements by understanding how the analyzers work internally (We also covered Oscilloscopes and RF Power Meters, similarly, and some other interesting and useful stuff.). We spent a couple of days doing deep dives on each block in the block diagram of a traditional analog spectrum analyzer. Later the Agilent instructor mentioned that it turns out that everything we had learned was also applicable to analyzers that are all-digital internally, since the math all turns out to result in the same effects, whether it's implemented in analog or digital.
Spent 25 years at HP/Agilent, so this rings a bell
The best way to understand how spectrum analyzers work is to think about how FM radios work - see superheterodyne receiver
Superheterodyne receiver - Wikipedia, the free encyclopedia
In a very general simplistic view, with spectrum analyzers, the local oscillator is swept, and the IF bandwidth determines the resolution bandwidth. There is a log amp detector after the IF, this drives the Y or amplitude axis, and the sweep drives the X or frequency axis. This gives the log amplitude versus frequency display. In general, the user sets a center frequency, and sweep bandwith when looking at modulated carriers.
On communication analyzers, or vector signal analyzers, you use the same approach. But instead of sweeping the local oscillator, it usually stays fixed, on some center frequency. The IF can be fairly wide, say 50MHz, and the output from the IF is digitized by a 12-16 bit or greater ADC. So this way you can capture filtered time domain signals at say 2.4GHz in a 25MHz band using a ballpark 200MS/s digitizer. Post processing is then done using specialized software to de-modulate and analyze the waveform. See Agilent 89600 for more on that.
As mentioned, the superheterodyne frequency translation mixing process can also be implemented in software. In most cases, this is done after the ADC chip, in a really fast FPGA, say like a Xylinx Virtex series.
I thought about how one might modify an old analog FM radio to play around with this. You could disable the AGC, and feed the 10.7 MHz IF output to a scope, or digitizer, to play around. The downside is you are limited to 88-108MHz, unless you figure out how to tune it over a wider range (use an external LO and bypass the tuned front end). But still may be fun as a learning exercise.
The best way to understand how spectrum analyzers work is to think about how FM radios work - see superheterodyne receiver
Superheterodyne receiver - Wikipedia, the free encyclopedia
In a very general simplistic view, with spectrum analyzers, the local oscillator is swept, and the IF bandwidth determines the resolution bandwidth. There is a log amp detector after the IF, this drives the Y or amplitude axis, and the sweep drives the X or frequency axis. This gives the log amplitude versus frequency display. In general, the user sets a center frequency, and sweep bandwith when looking at modulated carriers.
On communication analyzers, or vector signal analyzers, you use the same approach. But instead of sweeping the local oscillator, it usually stays fixed, on some center frequency. The IF can be fairly wide, say 50MHz, and the output from the IF is digitized by a 12-16 bit or greater ADC. So this way you can capture filtered time domain signals at say 2.4GHz in a 25MHz band using a ballpark 200MS/s digitizer. Post processing is then done using specialized software to de-modulate and analyze the waveform. See Agilent 89600 for more on that.
As mentioned, the superheterodyne frequency translation mixing process can also be implemented in software. In most cases, this is done after the ADC chip, in a really fast FPGA, say like a Xylinx Virtex series.
I thought about how one might modify an old analog FM radio to play around with this. You could disable the AGC, and feed the 10.7 MHz IF output to a scope, or digitizer, to play around. The downside is you are limited to 88-108MHz, unless you figure out how to tune it over a wider range (use an external LO and bypass the tuned front end). But still may be fun as a learning exercise.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.