EEVblog Electronics Community Forum
Products => Test Equipment => Topic started by: Sphyrna on October 26, 2023, 07:44:02 am
-
Hello,
I am searching for a scope which is capable to measure in the low frequency range and is able to directly represent the measurement data in the FFT in V/sqrt(Hz). The benefit would be that I could compare the data directly with the specs given in some data sheets. I tested a LeCroy HDO6000 and a R+S RTO64 but the lowest start frequency in the FFT that can be set is 1 Hz. These low frequency measurements I am doing are actually done with a LeCroy 64MXs-B and a post data processing. Inside the FFT this device can't do a logarithmic scale and frequency's below 1 Hz. But you can extract that from the FFT data. I know there is a SRS SR780 and a HP35670A for these kind of measurements like in: https://www.renesas.com/us/en/document/apn/an1560-making-accurate-voltage-noise-and-current-noise-measurements-operational-amplifiers-down-01hz (https://www.renesas.com/us/en/document/apn/an1560-making-accurate-voltage-noise-and-current-noise-measurements-operational-amplifiers-down-01hz). But isn't there a new scope which is able to measure in the low and in the higher frequency range for measurements from 1 mHz until 20 MHz or something like that? The measurements could be done in different settings (memory limitations in combination with the sampling rate). First, the range from 1 mHz to 4 kHz would be interesting for me. Any hints?
Kind regards,
S.
-
Hello,
I am searching for a scope which is capable to measure in the low frequency range and is able to directly represent the measurement data in the FFT in V/sqrt(Hz). The benefit would be that I could compare the data directly with the specs given in some data sheets. I tested a LeCroy HDO6000 and a R+S RTO64 but the lowest start frequency in the FFT that can be set is 1 Hz. These low frequency measurements I am doing are actually done with a LeCroy 64MXs-B and a post data processing. Inside the FFT this device can't do a logarithmic scale and frequency's below 1 Hz. But you can extract that from the FFT data. I know there is a SRS SR780 and a HP35670A for these kind of measurements like in: https://www.renesas.com/us/en/document/apn/an1560-making-accurate-voltage-noise-and-current-noise-measurements-operational-amplifiers-down-01hz (https://www.renesas.com/us/en/document/apn/an1560-making-accurate-voltage-noise-and-current-noise-measurements-operational-amplifiers-down-01hz). But isn't there a new scope which is able to measure in the low and in the higher frequency range for measurements from 1 mHz until 20 MHz or something like that? The measurements could be done in different settings (memory limitations in combination with the sampling rate). First, the range from 1 mHz to 4 kHz would be interesting for me. Any hints?
Kind regards,
S.
If up to 5MHz BW is OK, I would contact Picotech pre sales support and ask them about model 4262 and present you problem.
You cannot do what you need directly in software, but they have great API and you can pull data into Matlab directly and do whatever you want there.
Best,
-
We have some picotech scopes here. But my wish would be that there is a way without post processing. Of course if it's always the same measurement I could write a script. But it would be very convenient to have a out of the box solution. For me it is not 100% clear why the modern scopes can't measure till 1 mHz. Is it because of the DC stability? Is the 1/f noise is too high?
-
with the pico
you have to slow down the acquisition speed / timebase and take looooooong measurements in the 1-5 hz range ?? below 1 hz never tried, will do when i can
i have a 3406D
-
HP89410A is one spectrum analyzer I know that spans the range from 1 mHz to 10 MHz. Scopes usually have significant 1/f noise. But for very low frequency measurements, it is often beneficial to save the data in the time domain and do off-line analysis because one see glitches and drifts more clearly.
-
We have some picotech scopes here. But my wish would be that there is a way without post processing. Of course if it's always the same measurement I could write a script. But it would be very convenient to have a out of the box solution. For me it is not 100% clear why the modern scopes can't measure till 1 mHz. Is it because of the DC stability? Is the 1/f noise is too high?
Thank you for details..
Certainly 100, 200, 500 MHz scope front ends will not be optimized for 1/f noise.
Scopes work with simple principle of getting one long buffer and then applying measurements on it, FFT and all.
On my Picoscope getting 6,1mHz bin width at 200 SPS/s and 16kbins takes 81 seconds for single sweep.
That is still order of magnitude worse than what you need..
It simply is not that practical.
-
A useful FFT from 1 mHz to 4 kHz (six orders of magnitude) would need a HUGE number of samples in it.
Maybe there's something ridiculously expensive that can do it but 1Mpts is regarded as a generous FFT in the desktop DSO world.
Why don't desktop DSOs go down below 1Hz? Presumably because you'd have to be capturing data in roll mode to make the math work and you'd have such a tiny span of frequencies that it wouldn't be worth it.
To make it work I imagine you'll need to go to a USB device that can capture a LOT of data then process it. Maybe something like the Analog Discovery or a Picoscope.
-
On my Picoscope getting 6,1mHz bin width at 200 SPS/s and 16kbins takes 81 seconds for single sweep.
That doesn't surprise me.
-
The LeCroy scope we have here is not so far away from these requirements. If you adjust it to 5 s/div and you have 10 divisions the lowest frequency in the time domain is 1/50s = 0,02 Hz. If you adjust it to 20 kS/s it has 500 001 bins which leads to a frequency resolution of 0,04 Hz. If there is a scope it can do the job without post process the data by myself and it's doing the representation in the fft directly in V/sqrt(Hz) it would already fit the needs. In my opinion these limitations are only on the software side because the hardware is capable to do these measurements. But maybe the customer base which needs these kind of low frequency measurements is so small that there is no interest in adjusting/allowing the software to do this. Yes I could use a scope or a digitizer card and do the calculations in numpy or something like that. The question is not to do the measurement. The question is to do it in a convenient way and I have some hope someone here in the forum maybe had also a similar problem and found a device which is capable to do this without that you have to do some data processing.
-
I have the same problem & wish like you!
Currently I am running a DAQ card (16bit) with a python script to plot noise densities over freq.
On the bright side - I have noticed that with the necessary averaging I get quite some promising results in the 1/f noise region (when talking about the noise floor).
I have also seen that the Red Pitaya is giving a implementation for that measurement.
So your point about the noise in the 1/f region is something I would to give energy on.
May I ask what the difference between the SR780 as a dynamic signal analyzer and an oszi approach from the signal theory point of few? Is a dynamic signal analyzer just working in the frequency domain?
-
For frequencies that low, long sweep times are quite normal. A Picoscope 6242 is on par with an e.g. HP35665 in this regard.
44 minutes sweep for a 381.5 µHz frequency step (aka bin width), measuring a 100 mHz and 200 mHz sine wave simultaneously (reply #57):
https://www.eevblog.com/forum/testgear/low-frequencies-on-spectrum-analyzers/msg1468441/#msg1468441 (https://www.eevblog.com/forum/testgear/low-frequencies-on-spectrum-analyzers/msg1468441/#msg1468441)
Notice the high dynamic range and SNR. The Pico 4262 includes a simplified frontend design which has the usual 1/f noise of a FET input stage, but not the excessive LF noise of all the wider bandwidth general purpose oscilloscopes with split path input buffer.
It is limited to 5 MHz bandwidth and I think it does not have a noise densitiy function - but that should be the least concern. Decide for a useful window function and then you know the factor to multiply the frequency step in order to get the RBW, from this you can calculate the single factor which converts the individual readings into noise desitiy, i.e. nV/sqrt(Hz).
-
Just for fun I've fired up my HP3563A and set it to look at self-noise (50 Ohm plug on the input) in 1mHz-1Hz range. Took a while ;) . Not that great at 1mHz, above the 1/f corner it measures ~20-30nV/rtHz
Cheers
Alex
-
Pico don't has a out of the box solution too. With some programming you could do it: https://www.picotech.com/support/viewtopic.php?t=41417. (https://www.picotech.com/support/viewtopic.php?t=41417.) But with some programming I would guess the most scopes could do it if you can access the data.
-
Pico don't has a out of the box solution too. With some programming you could do it: https://www.picotech.com/support/viewtopic.php?t=41417. (https://www.picotech.com/support/viewtopic.php?t=41417.) But with some programming I would guess the most scopes could do it if you can access the data.
That is very correct. But, Picoscope has direct access API (no SCPI) that is faster (way faster!) and easier to use.
For instance you have ready examples how to get data in C# or Matlab.
I used Matlab in several occasions, just copied example of getting data, and then simply run the analysis with all the built in functions.
There is Python lib too.
Picoscope API does not provide measurements though. But it is better for raw data to PC.