For reasons that may or may not be related to Stockholm syndrome (10 years in the lab...), my favourite oscilloscopes ever are the Tektronix TDS 700 series.
My personal scope is a TDS744A from 1995, recently upgraded to 784A specs (1GHz bandwidth and 4GS/s) and calibrated (2017). A search on Tektronix warranty website shows it was originally the property of National Semiconductor, where it's service contract expired in december 2009. I wonder what strange and marvelous things it had to measure before coming to spend it's retirement years in my lab...
Most TDS 500, 600, 700, 2000 and 3000 series scope can do a reasonnably good FFT. They all share the same limitations and controls so the following discussion might be useful for owners of any of those scopes. Tons of them are also still in use in Uni labs everywhere.
I was recently re-watching EEVblog episode 845 where Dave compares the FFT of several different scope using a rather difficult test where the sidebands are very close to the carrier. I was wondering if the TDS could do it, or if it would just show "dick and balls"...
TDS BasicsThe TDS series graticule is a classic 10 by 8 divisions where each division is subdivided in 50 points horizontally and vertically.
Thus the graticule can show a maximum of 500 samples at a time, with 400 possible levels for each point. When making measurements, the horizontal and vertical cursors can be positionned anywhere on this 500x400 matrix.
The TDS can acquire much more than 500 points per record (1 acquisition or measurement after trigger) in my case up to 500 000.
It will still show only 500 points on the graticule, evenly spaced and selected from the whole record. A Zoom function then let's you drill down on sections of interest. (slowly)
TDS FFT FFT points : The number of FFT points is based on the number of points from the source waveform, up to 10 000, always starting at the beginning of the record. The more FFT points, the more accurate the FFT, but also the longer it will take to update. The default is 500 points, which is what fits on the screen directly.
Span : The frequency span of the FFT is always from 0 to half the sample rate. You control the sample rate with the horizontal SCALE knob when the source waveform is active. Sample rate is indicated in the upper left corner of the display. Turning clockwise increases sample rate.
Center frequency : The center frequency is always the sample rate / 4.
Say I want to look for distortion on a 1kHz sine wave... Hint #1 : Don't try to look at the signal and it's FFT at the same time. If you set the horizontal time base to 200us so you can see 2 cycles of the waveform, the resulting FFT span will be much too large at 0-125kHz, and all the interesting bits will be completely crushed at the left side of the plot.
What I do is I set the sample rate to 10.0kS/s (Timebase at 5ms/div). This places the fundamental about 1/4 of the way in and let me see up to the 4th harmonic. If I want to see further, I use 25kS/s, giving me a span from 0 to 12.5kHz.
The channel 1 waveform is no longer useful and can be turned off to reduce visual clutter. It will need to be selected again to change acquisition parameters, but I just turn it off again when I'm done so I can look at my FFT in peace.
Cleaning up the FFT Now I can clearly see the fundamental, but everything else is a huge noise band. This can be cleaned up a lot, but performance will suffer. Such is life.
The first step actually has a very low performance cost. Use high-res acquisition mode if available, and your sample rate is below 250MS/s.
(Activate channel 1, Acquire menu, High-Res, Clear menu, waveform off)
To improve frequency resolution, we need more FFT points.
(Activate channel 1, Horizontal menu, record length, select from 1000pts/10div to 15kpts/15div)
1000pts is still very fast when looking for stuff.
5000pts is a good compromise of resolution vs speed.
In the case of a difficult test like 5kHz sidebands on a 1MHz carrier in episode 845, you will need the full 10k FFT points, achievable with the 15kpts/15div setting (there is no 10k setting in the 700A
).
To further reduce the noise, turn averaging mode on in the FFT menu.
(Math Ref, Math 1, Average : 16)
At 5000pts, high res and 16 average, we can now see that the 1kHz sine wave from the HP 3312 is very clean, with all harmonics at least 60dB below the fundamental.
But that's a bit BORING.
Let's look at the square wave instead.
For square waves, you want to look much further in the harmonic series so you need a larger span. Let's start at 100kS/s.
You can quickly conrfirm that only the odd harmonics are present using the vertical cursors. You can also clearly see the harmonic series folding back on itself at the top of the span (the Nyquist frequency). This is called aliasing. A trick to see if a tone is real or an alias is to increase the sample rate (and thus the span). If the tone disappear then it was an alias (it will be shown at the correct, higher frequency).
Now for the final test.I do not have a nice arbitrary generator like Dave uses in episode 845, so I am faking it with the 3312 and 2 counters. I set the main generator to a 1MHz sine, the modulation generator to a 5kHz sine, the FM delta-F control almost all the way down (no way to measure that, but Dave's test was 500Hz for a 1MHz carrier for a delta of 0.05%, which is very low and the signals showed very few sidebands and in fact looked almost like AM on the spectrum analyzer)
Set Sample rate to 5MS/s, record length to 15kpts/15div for a full 10kpts FFT. High res mode, 16 FFT average.
Well, we clearly see the carrier...
This is where the Zoom is useful.
Hint #2 : Manipulating a zoomed, averaged 10kpts FFT is VERY slow. Here are some tips to minimize aggravation :
1. Center the zone of interest using the Horizontal position control while at 0.1x horizontal zoom factor.
2. Bring both vertical cursors to the center BEFORE zooming in further.
3. The paired cursor are just aggravating, don't use them.
4. The vertical and horizontal cursors remain useable.
5. Turn the Horizontal Scale control until you can discriminate the sidebands from the carrier.
So if you wanted to know the frequency delta and amplitude of those sidebands, the TDS would show them to you, if you asked VERY nicely...
Home
Help
Search
Login
Register
