Need some guidance about simply measuring signal levels in decibels

[wobbly]

I've been watching a few videos about how to use the decibel system for evaluating signal levels.
Notably EEVBlog #49 and several from Alan Wolke's great channel (https://www.youtube.com/user/w2aew/videos).

These tutorials made me want to try this for real with some actual signals.

I have access to some test gear:

• DSO with built-in signal generator - Keysight EDUX1052G
• Attenuator box - Marconi TF2163S (July 1969 - Apollo 11 vintage!  In beautiful condition too)
• Spectrum Analyser - TinySA (tinysa.org)

Here's what I'd like to do (in fact I've already tried but I didn't understand my results)...

• Generate a sinusoidal signal: 10 MHz, 1 mV p-p, no DC offset, output load set to "50 Ohm"
• Pass that through the attenuator box set to 30 dB or similar
• Measure the resulting signal using the spectrum analyser, set to have no internal input attenuation, display units set to dBmV

My thinking was thus:  If I set the spectrum analyser to use dBmV (1 mV reference) units, and give it a signal from the sig-gen of 1 mV, then I should see a spike on the display which matches the attenuation on the Marconi.

But that's not what I see.  I have the Marconi attenuator set to 30 dB, but the spectrum analyser tells me that the peak is at -40 dBmV.
So to my mind there is a discrepancy of 10 dB between the sig-gen and the spectrum analyser.  I must be missing something of course.

When I change the attenuation on the Marconi, the spectrum analyser's measurement changes linearly.
In other words the measured signal is always 10 less than the attenuation...

• 10 dB attenuation measures as -20 dBmV
• 20 dB attenuation measures as -30 dBmV
• 30 dB attenuation measures as -40 dBmV
• 47 dB attenuation measures as -57 dBmV ... etc

So where is this 10 dB difference coming from?  I'm using relatively short coax cables and proper coax connectors (N-Type, BNC, SMA with 50 Ohm adapters throughout).  None of this is hacked together.

What am I doing wrong or misunderstanding?

Thank you for reading, and your time.   

[Benta]

A 0 dBmV level is a voltage of 1 mV_RMS. But you're operating with 1 mV_PP, which is something else. Get that right first.

[fourfathom]

Set your sig gen to 1mV RMS. This will be 2.828 mV p-p. This is almost a 10dB difference (3.16 is roughly 10dB.)

 

[wobbly]

Okay, that's better.  I'm getting a difference of 1.3 dB now.  I reckon that's reasonable since the sig-gen is slammed against its lower limit.

Am I right in thinking that I have to multiply the 1 Vpp by 2*root(2) because it's a voltage measurement, rather than a power measurement?

Thank you both for your help.

[RoGeorge]

multiply the 1 Vpp by 2*root(2) because it's a voltage measurement, rather than a power measurement?

That is because of Vpp to V_RMS conversion.  Here is how pp and RMS values are related in a sinusoidal waveform no matter what that sinusoidal waveform represents.  Can be for example the force in a spring oscilator, and the RMS and peak-peak values will have the same conversion ratio if it a sinusoid.

Other waveforms have other conversion ratios, and that's the catch with the so called True RMS multimeters.  The cheap ones always asume a perfect sinus waveform and just multiply its Vpp to get the equivalent V_RMS, while the TrueRMS instruments properly measure the true RMS.

Random search result from  https://testguy.net/content/%20270-Peak-vs-Average-vs-RMS-Voltage:


V_PK is often called the amplitude of an oscillation, in formulas wrote usually as A (and that A is not about Amps), e.g. v=A*sin(2*pi*f*t)

[TimFox]

For a sinusoidal waveform, V_RMS = (V_pk-pk)/(2x2^1/2) = (V_pk-pk)/(8^1/2) = (V_pk-pk)/(2.828).
You divide by 2.828 to decrease the pk-pk value to RMS, not multiply.
As discussed above, a "True RMS" meter measures the actual RMS value of the waveform, which is different for sine waves, square waves, DC, random noise, etc.
Other meters measure the "average value", which is actually the mean absolute value, and apply a conversion factor to RMS that is only valid for sine waves.

[wobbly]

So I should have been using 1 mV RMS instead of 1 mV pk-pk.

1 mV RMS
= 2.828 mV pk-pk
= +1.414 mV to -1.414 mV amplitude

Is "peak-to-peak" actually a useful term in this context?  Is it better to use amplitude instead (as shown in the picture RoGeorge posted as "a" and "-a")?
I think I'm getting confused by the numerous ways one can measure a sine wave.  My sig-gen gives "Amplitude" in units of "mVpp", so I have to remember to halve that.

Ok, I think I get it now.