PicoScope 7 Software

Available on Windows, Mac and Linux

Hi, the unit of measurement of THD+N is dBc, would it be possible to have it also in % and dB?

In spectrum mode it appears the Y-axis is maxed out at -95dBV, If I measure an ultra low noise generator 95dB is not enough.

Is it possible to modify the Y axis?

Thank you

Domenico

In spectrum mode it appears the Y-axis is maxed out at -95dBV, If I measure an ultra low noise generator 95dB is not enough.

Is it possible to modify the Y axis?

Thank you

Domenico

Hi Domenico,

Sorry for the delayed response.

**To answer your 1st question:**

Units of just ‘dB’ in a measurement are only valid if the measurement is between 2 values, in which case it will have another letter, as a suffix, which denotes the reference. THD+N is the ratio of the largest Signal Peak voltage (referred to, as the Carrier in communications, which is also the Fundamental Frequency) to the Total combined harmonics and Noise voltage (THD+N). Units of dBc are Decibels where the reference is the Carrier.

**I don't know if you forgot to add a suffix, but you can’t have a THD+N measurement of just dB.**

However,**I’ll submit an enhancement request to our development team, on your behalf, for the display of THD+N as %**.

In the meantime, to do the conversion manually is fast and simple. To convert THD+N [dBc] to THD+N %, you just divide the value by 20, use the result as an exponent of 10, and then multiply that result by 100, as follows:

THD+N % = 10 ^ ( THD+N [dBc] / 20 ) * 100

So, if THD+N =**-6dBc** then as a percentage, THD+N = 10^(-6/20) * 100 = **50%**

**To answer your statement “In spectrum mode it appears the y-axis is maxed out at -95dBV”:**

If you want to measure the signals from a Generator (of signals) that has ultra-low noise then,**the smallest signals that you can display, in a PS7 Spectrum plot, will depend upon (a) the PicoScope that you are using, and (b) the Spectrum settings that you’re using.**

First of all,**if you are using a PicoScope 2204A or 2205A, then you have the wrong PicoScopes for going beyond -95dBV.** These 2 PicoScopes are our entry level PicoScopes, (they have less functionality, to reduce their cost), so they only have small Buffers for storing the data from each capture (8k samples for the PS2204A and 16k samples for the PS2205A). What this means is that when the data is converted, by an FFT, to give you a Spectrum Plot, **the PS2204A Is limited to 8192 Bins, and the PS2205A to 16384 Bins.** This will limit how much of the Y-axis can be used to display their frequency data in Spectrum Mode. (I will return to this after I have explained how the settings affect the Y-axis range).

So, in case, you're not aware, I will be using averaged Spectrum plots to get a more stable value for the signal and noise. The 1st image below shows a continuous averaging Spectrum plot to achieve this and the 2nd image shows a normal capture without averaging.

In the first image it's a lot clearer where the average of the noise is, and less of the signals are obscured by noise.

Also, in case you're not aware, you can increase the Number of Bins used for the Spectrum plot to lower the noise floor, revealing previously hidden signal content and differentiate better between signal content and noise content, by increasing the frequency resolution. The Previous 2 images used the default Number of bins (16384), while the image below uses the maximum number of Bins (1048576):

As you can see**increasing the Number of Bins makes the noise level drop** by 18dBV from -91.2dBV to 109.5dBV **and therefore reduces the lower limit of the Y-axis**. The reason for this is that the Linear Volts scaling for the Y-axis and FFT scaling (Coherent Gain) result in Amplitude (signal) accuracy instead of Power (noise) accuracy (as a Spectrum plot can't provide accuracy for both at the same time). This revealing of normally undetectable signal components for further analysis, makes Linear Spectrums useful in various fields such as, Communications, audio processing, Biomedical analysis, etc.

Note that, if we add a measurement of Signal to Noise Ratio (SNR), and then place one y-axis ruler at the signal peak and the other below the signal peak at a distance of the SNR value away from the Peak (signal), we can see where the True Noise level is, (as this is what the measurements in Spectrum Mode are based upon) which is much higher. So, this shows exactly how much of the signal content was hidden by the True noise level.

**So, to answer your 2nd question “is it possible to modify the Y axis”**

The answer is you can, i.e. you can further reduce the lower limit of the scaling for the Y-axis, in order to be able to measure everything above the artificially lowered noise floor, by increasing the Number of bins.

So, if we now go back to the 2204A and 2205A PicoScopes, I have created a plot using the Standard number of 16384 bins below:

If I now plot 524288 bins we get the following:

You can see that all that changes is the scale, the noise level has remained the same, and even down to 8192 bins, still remains the same, as below:

So, this now clarifies why having only a limited number of bins is a problem, if you want to see low-level signals. However, if you want to be able to see really low-level signals then you will need to move to lower noise, higher (than 8-bit) resolution PicoScopes, as the Y-axis is extended further because of their even lower noise and signal levels. For instance, the lower noise, smaller bandwidth, 16-bit, PicoScope 4262 can produce spectrum plots like the one below:

This was created only using the standard 16384 Number of bins.

Regards,

Gerry

Sorry for the delayed response.

Units of just ‘dB’ in a measurement are only valid if the measurement is between 2 values, in which case it will have another letter, as a suffix, which denotes the reference. THD+N is the ratio of the largest Signal Peak voltage (referred to, as the Carrier in communications, which is also the Fundamental Frequency) to the Total combined harmonics and Noise voltage (THD+N). Units of dBc are Decibels where the reference is the Carrier.

However,

In the meantime, to do the conversion manually is fast and simple. To convert THD+N [dBc] to THD+N %, you just divide the value by 20, use the result as an exponent of 10, and then multiply that result by 100, as follows:

THD+N % = 10 ^ ( THD+N [dBc] / 20 ) * 100

So, if THD+N =

If you want to measure the signals from a Generator (of signals) that has ultra-low noise then,

First of all,

So, in case, you're not aware, I will be using averaged Spectrum plots to get a more stable value for the signal and noise. The 1st image below shows a continuous averaging Spectrum plot to achieve this and the 2nd image shows a normal capture without averaging.

In the first image it's a lot clearer where the average of the noise is, and less of the signals are obscured by noise.

Also, in case you're not aware, you can increase the Number of Bins used for the Spectrum plot to lower the noise floor, revealing previously hidden signal content and differentiate better between signal content and noise content, by increasing the frequency resolution. The Previous 2 images used the default Number of bins (16384), while the image below uses the maximum number of Bins (1048576):

As you can see

Note that, if we add a measurement of Signal to Noise Ratio (SNR), and then place one y-axis ruler at the signal peak and the other below the signal peak at a distance of the SNR value away from the Peak (signal), we can see where the True Noise level is, (as this is what the measurements in Spectrum Mode are based upon) which is much higher. So, this shows exactly how much of the signal content was hidden by the True noise level.

The answer is you can, i.e. you can further reduce the lower limit of the scaling for the Y-axis, in order to be able to measure everything above the artificially lowered noise floor, by increasing the Number of bins.

So, if we now go back to the 2204A and 2205A PicoScopes, I have created a plot using the Standard number of 16384 bins below:

If I now plot 524288 bins we get the following:

You can see that all that changes is the scale, the noise level has remained the same, and even down to 8192 bins, still remains the same, as below:

So, this now clarifies why having only a limited number of bins is a problem, if you want to see low-level signals. However, if you want to be able to see really low-level signals then you will need to move to lower noise, higher (than 8-bit) resolution PicoScopes, as the Y-axis is extended further because of their even lower noise and signal levels. For instance, the lower noise, smaller bandwidth, 16-bit, PicoScope 4262 can produce spectrum plots like the one below:

This was created only using the standard 16384 Number of bins.

Regards,

Gerry

Gerry

Technical Specialist

Technical Specialist

Hi Gerry, thanks for the reply, I'll make some clarifications, my picoscope is 4424 (12bit hardware resolution) and in this case the use is for audio applications (I know that 4262 is more specific) but I have other professional audio analyzers (prismsound). however I use the oscilloscope for general control, not for measurements. if I want to display the prismsound output (only display in fft mode), I get a spectrum with the baseline (noise) off the X axis, and I can just shift the waveform to display the spectrum correctly... I would like to be able to define minimum and maximum for the X axis in FFT mode e.g. -140dB - 0dB.

P.S. I use 100kHz bandwidth FFT

Greetings.

Domenico

P.S. I use 100kHz bandwidth FFT

Greetings.

Domenico

Hi Domenico,

Could you clarify your comment of "I can just shift the waveform to display the spectrum correctly". Specifically, what do you mean by "I can just shift the waveform", how can you move it and where are you moving it from and to, or what is it moving relative to (i.e. which combination of the y-axis, x-axis, and/or the noise are you shifting it relative to)...also what do you mean by "display the spectrum correctly" (i.e. exactly how was the spectrum displayed incorrectly).

What I am trying to establish here is which features of the FFT or features of the PrismSound Audio Analyzer you may be using, to establish what can and can't be done in our Spectrum Mode plots. For instance, In a video discussing the PrismSound dScope M1 (which is available here: https://youtu.be/T93d1awPQ4o), the presenter mentions, from 6m30s into the video, that the dScope can apply Sharp Filters (presumably Digital Brick Wall Filters, where you just set everything outside the passband to zero) immediately surrounding each Harmonic. Although this sounds impressive, exactly how useful this would actually be would depend upon how it's implemented and what you would use the resulting Data for. In order to apply the Filters effectively you need to know exactly in which Bin the slope of a harmonic ends, and the next bin only containing noise begins. You also need to have selected the appropriate Windowing function that is both best suited for your measurement goal (e.g. absolute harmonic amplitude accuracy or discriminating between harmonics of the signal and relatively close by spurious peaks of distortion) and best suited for pulling harmonics out of noise, (e.g. having very low side lobes). In a niche application, such as Audio Processing, you can be more specific with your choices, so these may be chosen for you by the software, and then it's relatively easier to apply more sophisticated processing (such as Brick Wall Filters). Our PicoScopes are not Specifically Audio Processors, so, unfortunately, we can't add features for one set of users, at the expenses of crippling the flexibility of choice for all of our other users.

We could just allow users to define the limits of their own Y-axis, but then the actual lower limit of the axis plotted could easily be too small (where the Signal and noise data disappears below the bottom of the graph) or too big (where the axis has a lot of wasted space below the noise floor where nothing is plotted (this is, for instance, what happens when you choose a value for the number of Bins that is much larger than the amount of data that can be stored in the Sample buffer of a Scope, as you can see in the image 'PS2204A 524288 bins.png', from my previous reply, above).

Also, the range of the Y-axis should not have an upper limit of 0dB[x], because Spectrum Mode is still using the Input Ranges that we have for Scope Mode, which, for example, means that we have a +/-2V peak-to-peak Input Range that can be used for capturing data to convert into a Spectrum Plot. In the Spectrum Plot we can have a Y-axis scaled to dBV, where 0dBV is the reference point of 1V[rms] which is +/-1.4142V peak-to-peak. What this means is that, if you choose a +/-2V peak-to-peak Input Range, then the Spectrum Plot will not necessarily always be smaller than +/-1.4142V peak-to-peak. So by having an upper limit of 0dBV, you are artificially limiting the range of the data that can be plotted.

So, in general, it's better to let PS7 create the axis based upon the Bandwidth and Bin width that you select for your FFT plot, as PS7 is able to easily determine where the best upper and lower limits for the data are.

What it should also do, as an improvement, for the few PicoScopes that have Sample Buffer sizes less than 1M Samples (i.e. the maximum Number of Bins selectable) is take into account the Buffer Size for the Model of PicoScope being used, and set a limit for the Number of Bins Selectable based upon the Model (so this would apply to 2000 Series picoScopes lower in features than a 2206B and 3000 Series PicoScopes lower in features than a 3203D).

Having said that, it could be that the reason for your request was because we could tidy up the current implementation of our Spectrum Plot, and make it easier to intuitively interpret the relative vertical distance between plotted points on the graph, by (1) having grid lines that always match the scaling used on the Y-axis, (so that a Logarithmic Y-axis produces matching horizontal logarithmic grid lines, instead of equally spaced lines at fractional dB[x] values) and (2) always aligning the axis to the reference point (0dBu, 0dBV, 0dBm). I was just recently discussing this with another user, and I will be putting in an enhancement request to our development team for this.

Regards,

Gerry

Could you clarify your comment of "I can just shift the waveform to display the spectrum correctly". Specifically, what do you mean by "I can just shift the waveform", how can you move it and where are you moving it from and to, or what is it moving relative to (i.e. which combination of the y-axis, x-axis, and/or the noise are you shifting it relative to)...also what do you mean by "display the spectrum correctly" (i.e. exactly how was the spectrum displayed incorrectly).

What I am trying to establish here is which features of the FFT or features of the PrismSound Audio Analyzer you may be using, to establish what can and can't be done in our Spectrum Mode plots. For instance, In a video discussing the PrismSound dScope M1 (which is available here: https://youtu.be/T93d1awPQ4o), the presenter mentions, from 6m30s into the video, that the dScope can apply Sharp Filters (presumably Digital Brick Wall Filters, where you just set everything outside the passband to zero) immediately surrounding each Harmonic. Although this sounds impressive, exactly how useful this would actually be would depend upon how it's implemented and what you would use the resulting Data for. In order to apply the Filters effectively you need to know exactly in which Bin the slope of a harmonic ends, and the next bin only containing noise begins. You also need to have selected the appropriate Windowing function that is both best suited for your measurement goal (e.g. absolute harmonic amplitude accuracy or discriminating between harmonics of the signal and relatively close by spurious peaks of distortion) and best suited for pulling harmonics out of noise, (e.g. having very low side lobes). In a niche application, such as Audio Processing, you can be more specific with your choices, so these may be chosen for you by the software, and then it's relatively easier to apply more sophisticated processing (such as Brick Wall Filters). Our PicoScopes are not Specifically Audio Processors, so, unfortunately, we can't add features for one set of users, at the expenses of crippling the flexibility of choice for all of our other users.

We could just allow users to define the limits of their own Y-axis, but then the actual lower limit of the axis plotted could easily be too small (where the Signal and noise data disappears below the bottom of the graph) or too big (where the axis has a lot of wasted space below the noise floor where nothing is plotted (this is, for instance, what happens when you choose a value for the number of Bins that is much larger than the amount of data that can be stored in the Sample buffer of a Scope, as you can see in the image 'PS2204A 524288 bins.png', from my previous reply, above).

Also, the range of the Y-axis should not have an upper limit of 0dB[x], because Spectrum Mode is still using the Input Ranges that we have for Scope Mode, which, for example, means that we have a +/-2V peak-to-peak Input Range that can be used for capturing data to convert into a Spectrum Plot. In the Spectrum Plot we can have a Y-axis scaled to dBV, where 0dBV is the reference point of 1V[rms] which is +/-1.4142V peak-to-peak. What this means is that, if you choose a +/-2V peak-to-peak Input Range, then the Spectrum Plot will not necessarily always be smaller than +/-1.4142V peak-to-peak. So by having an upper limit of 0dBV, you are artificially limiting the range of the data that can be plotted.

So, in general, it's better to let PS7 create the axis based upon the Bandwidth and Bin width that you select for your FFT plot, as PS7 is able to easily determine where the best upper and lower limits for the data are.

What it should also do, as an improvement, for the few PicoScopes that have Sample Buffer sizes less than 1M Samples (i.e. the maximum Number of Bins selectable) is take into account the Buffer Size for the Model of PicoScope being used, and set a limit for the Number of Bins Selectable based upon the Model (so this would apply to 2000 Series picoScopes lower in features than a 2206B and 3000 Series PicoScopes lower in features than a 3203D).

Having said that, it could be that the reason for your request was because we could tidy up the current implementation of our Spectrum Plot, and make it easier to intuitively interpret the relative vertical distance between plotted points on the graph, by (1) having grid lines that always match the scaling used on the Y-axis, (so that a Logarithmic Y-axis produces matching horizontal logarithmic grid lines, instead of equally spaced lines at fractional dB[x] values) and (2) always aligning the axis to the reference point (0dBu, 0dBV, 0dBm). I was just recently discussing this with another user, and I will be putting in an enhancement request to our development team for this.

Regards,

Gerry

Gerry

Technical Specialist

Technical Specialist

Hi Gerry, thanks for the replay.

Attached two images, in the first the Y axis is defined by picoscope, the bottom line is not visible. In the second image I moved the track and you can see the bottom line but the Y axis is horrible.

Greetings.

Domenico

Attached two images, in the first the Y axis is defined by picoscope, the bottom line is not visible. In the second image I moved the track and you can see the bottom line but the Y axis is horrible.

Greetings.

Domenico

P.S. I would prefer an equally spaced Y-axis between a minimum and a maximum with a bit of headroom to see something slightly out of bounds.

Domenico

Domenico