IMPORTANT :

The "constant current around" misfit.

        Dimensions and effectiveness of a capacitive tuned loop.
The maximal circumference of a capacitive tuned loop is just under 1/4 wavelength. If such a loop is to be tuned to more than one frequency band, that loop should have a circumference of 1/4 lambda for the highest frequency band to be tuned to. It has maximal efficiency on that frequency band, and enables a relative thin (cheap) radiator. 

If that same circumference (in meters length) loop is also used on lower frequency bands, its relative circumference (in wavelength) is smaller. And its feed point impedance drops to a (much) lower value. Its efficiency also tends to drop (less severe but) causing the need for a thicker and more expensive radiator, in order to minimize losses on lower frequency bands.

        Radiator thickness.
Some people say, that a loop radiator should be constructed out of one piece, and very thick copper. 
"Every tenth of an Ohm series resistance should be avoided for losses reasons". 
Even tin-soldering should be avoided.
That is not always true.

A very small loop (for instance 1.3m diameter = 4m circumference = 1/20 lambda for 80m) is (very) useable for reception.

But even if optimally constructed (f.i. with an 8cm (!) thick copper radiator), it shows disappointing properties when used as a transmitting antenna.

With 100W power this very small loop generates :
-  very large currents (75Arms),
-  very high voltages (10kVp),
-  and at least 10dB losses !

These losses are not important for your own reception. 
Both received signals and noises are weaker. 
The signal to noise ratio remains unchanged. 
Your own reception quality is NOT affected.

But for other stations listening to you, these losses result in at least 10dB worse signal to noise ratio (noisy or no reception) !

        Why losses in very small loop antennas are so high?.
    In my opinion :
RF currents in a conductor only run in a very thin surface layer. This "skin" layer has a small cross-section per unit of width, resulting in a relative high value "RF resistance", which is determined by the specific material resistance, the diameter of the conductor, and the RF frequency.

Such a very small loop can only radiate energy, by running very high currents in its short antenna length. The relative high value of skin resistance will transform these large RF currents mostly into heath, instead of radiation. The loop is mainly loaded by dissipating RF energy into heath. The loading by radiation is relatively small.

        With unwanted results :
- very high Q
- unpractical small bandwidth
- very high currents 
- very high capacitor voltages
- large heath losses

A high Q is wanted for non-radiating coils.
But a transmitting loop antenna should radiate all power as good as possible. A large loop surface, thus a large circumference is therefore needed. The loop-circuit will then be loaded by radiating its energy, resulting in :
- lower circuit Q.
- wider bandwidth
- lower currents
- lower capacitor voltage
- far less heath losses

    Calculation examples using 66pacific-calculation :

            Example for a very small loop :
   Resonating at 3,65 MHz,
    circumference 1/20 lambda (1,3m diameter),
    radiator maximal thick !! 8cm !! dia. copper,

- Compared with a 1/2 lambda dipole, the maximal possible efficiency is only 11% (-9.6 dB).
- Its Q is enormous high (f.i. 5400).
- Its bandwidth is only 0.7kHz, unsuitable for SSB.
- Accurate tuning on 80m is impossible. One kHz detuned means to high SWR.
- The feed point impedance is unpractical low (f.i. 0.018 Ohm).
        - A matching transformer should have an impossible high turns ratio.
        - Matching to 50 Ohms must therefore be done by a capacitive network, a gamma match, possibly causing unbalance (TVI/BCI),
          or by a loop. Needing experimentation to get good VSWR..
- The loop current is enormous high (with 100W up to 75A).
        - It is heating the loop with 89W. Radiating only 11W.
- Resulting in enormous RF voltages at the tuningC (with only 100W up to 10kVp).
        - Even vacuum capacitors, and high voltage ceramic capacitors, can show flash-over. 

            Example for a maximal large 1/4 lambda circumference loop for 40m, 
            having a thin (1cm dia) radiator.
According to the 66pacific-calculations, the efficiency of a loop, compared with a 1/2 lambda dipole is : 
a. 1/4 lambda loop : -2dB.  (7.1 MHz)
b. 1/8 lambda loop : -7dB.  (3.65 MHz) This is 3dB better than a very thick 1/20 lambda loop !
In practice these losses are lower, as the surface RF currents in such a NOT VERY SMALL loop are not everywhere equally strong.

            Overview :
Compared with a very small loop, a 1/4 lambda circumference loop shows the following better properties :

- The Q is much lower
- Loop currents are much smaller resulting in
    - Less losses.
    - 7dB better RF radiation.
    - Much lower capacitor voltages.
    - Much larger tuned bandwidth.

- The feed point impedance is much higher (5.5 ohms) resulting in :
    - Easier matching.
    - Smaller copper diameter.
    - Much lower costs.

Very small loops are (very) good for receiving purposes,
not for transmitting.