Talk:Error vector magnitude

Prior to my edit of 2020 February 10, the talk section correctly flagged the confusion over magnitude and power. RMS Power is at best misleading and at worst meaningless and confusing. Power is found from magnitude by taking mean square value. Taking the root of the mean square value of magnitudes yields another magnitude, not a power. So I fixed the article by tying the EVM definition to that used in Keysight instruments and referencing my fixes to the manual for the Keysight instruments. Mwengler (talk) 02:02, 11 February 2020 (UTC)[reply]

Perhaps a graph would be helpful? --HelgeStenstrom 14:28, 28 March 2007 (UTC)[reply]

This discussion is confusing. The EVM is first defined as a distance between ideal and non-ideal points in a constellation diagram. This implies that it is an amplitude as opposed to a power quantity. Then it is defined as a ratio of an rms "noise" to a reference power. How can it be both? —Preceding unsigned comment added by 128.152.20.33 (talk) 22:58, 16 December 2008 (UTC)[reply]

I changed the words "power" to amplitude. Generally, when an RMS quantity is reported, this is an amplitude, i.e. a voltage, and not a power. The idea of Root-Mean-Squared is that you take a time-varying voltage, in units of volts, (whose instantaneous power into a matched load is proportional to voltage squared) then square the instantaneous voltage measurement to get a quantity (always positive) which is proportional to power--this quantity will have units of volts squared, but if we assume 50 ohm system, would have an equivalent representation in watts. Then, this squared-voltage value is averaged over a time interval, and then the average is square-rooted to get us back to units of voltage.

In an IQ modulation, both I and Q are typically in units of volts, and the instantaneous power of the signal is proportional to I^2+Q^2. For this reason, I thought the article as-written, with frequent mention of the word "power" was a bit misleading, since the error vector when expressed as a percentage, is by convention specified in units of I and Q, which would typically be volts. A ratio of powers would be EVM^2, which I have also seen in some specifications.

Sorry I left the letter P, as I couldn't change the diagram. Someone please update with what makes sense there.Com2ment (talk) 21:50, 7 February 2017 (UTC)[reply]

It might help if the RMS definition of the EVM is mentioned for more than one symbol: $EVM_{RMS}={\sqrt {\frac {\sum _{N}|{\overrightarrow {V_{error}}}|^{2}}{\sum _{N}|{\overrightarrow {V_{ref}}}|^{2}}}}={\sqrt {\frac {\sum _{N}|{\overrightarrow {V_{meas}}}-{\overrightarrow {V_{ref}}}|^{2}}{\sum _{N}|{\overrightarrow {V_{ref}}}|^{2}}}}$ or $EVM_{RMS}(\mathrm {dB} )=10\log _{10}\left({\frac {\sum _{N}|{\overrightarrow {V_{meas}}}-{\overrightarrow {V_{ref}}}|^{2}}{\sum _{N}|{\overrightarrow {V_{ref}}}|^{2}}}\right)$ or something like that --Roani52 (talk) 14:51, 17 February 2017 (UTC)[reply]

EVM in dB[edit]

I think that EVM in dB should be calculated as 20*log10(EVMrms) = 20*log10(sqrt(mean_square(EVM)) or 10*log10( mean_square(EVM ) ) and not as in the article. This is because dB relates to power (and not amplitude). — Preceding unsigned comment added by 134.191.233.199 (talk) 16:44, 21 March 2018 (UTC)[reply]