Optimum Return Loss [hlp]

Hello Everyone;
What is the optimum return loss of a filter in the pass-band?

Hi

From the band pass to the next stage, it should be zero, since ideally you want the energy totally transferred within the band of interest.

cheers

Sal

Sal said:

Hi

From the band pass to the next stage, it should be zero, since ideally you want the energy totally transferred within the band of interest.

cheers

Sal

Click to expand...

No One:

Is depend of type of used filter ie. Bessel, Buttewort, Chebyschev, Cauer (elliptic).

Ideal filter without resistive loss transsmit and reflect power between ports in Different ratio depend of used filter character.

For example Chebyschev make rippel in passband depend of varying return loss in feeding port. Top of ripple in passband, say -0.044 dB give 20 dB return loss and bottom of ripple say -1 dB give 6.8 dB return loss on ideal Chebyschev filters.

Chebylschev-filters cannot have more passband ripple than 0.1 dB if you want better than 16 dB in return loss, and make slower rejection slope outside pass band compare to Chebyschev allowed 1 dB ripple in passband (and 6.8 dB return loss)

For ideal filters, all powers not transfered to output port in filters, is reflected in feeding port back to source again. If you want very sharp filter with few poles to reject power very quick outside band (chebylschev, cauer) you have always poor and very varying return loss.

make filter is always design compromisses.

Sal said:

Hi

From the band pass to the next stage, it should be zero, since ideally you want the energy totally transferred within the band of interest.

cheers

Sal

Click to expand...

Hi Sal !

You must be talking about IL (Insertion Loss) instead of RL (Return Loss).

Now for the correct answer:
The industrial and army standard is below -16dB of Return Loss of any RF/MW device.
In your case you need at least 16dB attenuation of power in the pass-band of the filer (because in this pass-band you wouldn't want the power to reflect and come back) in terms of Retun Loss (S11 and S22)

Now what really is the difference between the suggested filter which mentioned before (Cheby., Butter, Elip. etc..) is affecting the IL (Insertion Loss) of the filter (with ripples , flateness ...)

some reminder formulas:
RL=-20log|Γ|
Γ=(Z-Z0)/(Z+Z0) which is the reflection coeff.

Bye,

Itzik

itzikhaim said:

some reminder formulas:
RL=-20log|Γ|
Γ=(Z-Z0)/(Z+Z0) which is the reflection coeff.

Bye,

Itzik

Click to expand...

and I add formulas:

if z = Z/Z0 (normalized form)
Γ = (z - 1)/(z + 1)

most Smithchart using value in normalized form.

VSWR = (1 + |Γ|) / (1 - |Γ|)
|Γ| = (VSWR - 1) /( VSWR + 1); scalar version of reflection cofficient

VSWR is time to time refered in (old) litterature and impossible to make resistive and reactive value from, depend of missing phase angle information. This 'view' also using in scalar network analyzer and IMHO more or less useless.


and backward operation if you have know Γ and know Z0 from measure process.

z = (1 + Γ) / (1 - Γ)
Z = z * Z0

or using Z = Z0((1 + Γ) / (1 - Γ))

all Z0, Z,z and Γ can be complex numbers

Real part of Z is a resistive Ohm, if imaginary part have negative sign, is a capacitive reactance Xc (without '-' sign) and with know measure frequency can calculates: C = 1/(2*π*f*Xc) in Farad

If imaginary part of Z is positive, is a inductive reactance Xl and with know measure frequency calculates: L = Xl/(2*π*f) in Henry

xxargs said:

[...
most Smithchart using value in normalized form.
...

Click to expand...

And I say: all the printed Smithchart I know using normalized value form.

Of course in VNA measuring devices or any Sim. Soft. it will be not normlized.

Bye,

Itzik