Amplifier -> Filter, won't the off band mismatch cause a problem?

[daqq]

Hi guys,

I'm new to RF stuff, so I've been wondering - when an amplifier/LNA is matched to 50 Ohm (such as the ADL5536 http://www.analog.com/media/en/technical-documentation/data-sheets/ADL5536.pdf ) and the output is connected to a filter  ( such as the BPF-C45+ https://www.minicircuits.com/pdfs/BPF-C45+.pdf ), won't there be a problem when operating amplifying frequencies beyond the passband? My understanding is that the input impedance of the filter when the input is beyond the passband effectively changes and starts reflecting.

Won't the amplifier be unhappy with this? Assuming small signals, not a power amplifier, just something at ~10dBm.

Thanks,

David

[T3sl4co1l]

You're going to love this answer...

...It depends.

If it's sensitive to load (i.e., the amplifier's Zo isn't 50 ohms resistive), then the filter response will be wrong, if nothing else.  If the amplifier is really bad, it may even become unstable and oscillate at some unlucky frequency.

If you absolutely need a guarantee, and can't obtain any detailed information about the amplifier, then use an attenuator, or use a constant-resistance or diplexing filter (terminate the stopband port).

Tim

[daqq]

...It depends.

One of those, eh? 

The specific amplifier and filter are both in the link. Seems the ADL5536 does not have a simple resistive Z on the output side.

Is there any book/app note you could recommend? Or what keywords search for?

The attenuator seems the simplest solution. Though what kind of attenuation would be needed if it's 50 ohms in and 50 ohms out?

If I understand it correctly: At the proper frequencies (impedance of the filter ~= output impedance of the amplifier), the attenuator would just skim a little of the power going into the filter. At the improper frequencies, the power would mostly go into the filter and not mess with the amplifier by making the mismatch smaller?

[1design]

More or less correct. As an example:

If you use a 3dB attenuator:
-3dB will be lost in the forward direction to the filter(insertion loss in the forward direction)
-Additional 3dB will be lost on the signals reflecting from the filter(insertion loss in the reflected direction)

So you should improve the S11 (reflection coefficient as seen by the amplifier) of your filter by roughly 6dB in this case, in practice a bit less due to the matter mentioned below.

This works while the summed back and forth attenuation is smaller than the reflection coefficient of the attenuator, from than on that one is the driving mismatch in the circuit.

I hope the explanation is not to messy and will help you get a feeling for what is going on.

BR

[T3sl4co1l]

Hmm, S22 is pretty low ( < -9.58 dB over the rated frequency range), so it's not a terrible match to 50 ohms, i.e., it contains internal resistance on the order of 40 ohms.  And, curiously, it looks capacitive(?) at very low frequencies, possibly coupling capacitor action..?  Or maybe a supply inductor, or gyrator (the latter could be internal).

Probably fine to filter after it.  Contact an FAE if you need more clarification.  They'll probably be familiar with examples.

Tim

[daqq]

Thanks guys! I'm new to all of this, so it does get confusing.

[AF6LJ]

A little bit of attenuation fixes many evils...

[daqq]

A little bit of attenuation fixes many evils...

I don't suppose there's a simple answer on How much? 0.3 dB?

[AF6LJ]

A little bit of attenuation fixes many evils...

I don't suppose there's a simple answer on How much? 0.3 dB?

There is actually a simple answer, as much as it takes.

I've used attenuators in all kinds of places to get amplifiers to work well with filters, mixers and power amplifiers to work well with narrow band isolators.
In the case of the Isolator to power amplifier problem...
The isolator had reticence out at the band edges, the transmission line between the isolator was made longer by three inches. There was still plenty of power, and the instability the amplifier had went away.

It never hurts to have more signal than you need, provided you meet your current spec...
You can always pad it down...

[G0HZU]

You could download the s parameter data for both the filter and the amplifier and see how the filter response changes on an RF simulator when you stack both of them together. It only takes a few minutes to do this. You can look at stability and return loss as well.

Looking at the official 5536 s parameter data file I would guess that this part uses internal (resistive) negative feedback to control the gain and the I/O matching. Their s parameter model data suggests that the feedback is band limited (using a relatively small blocking cap?) so the gain will begin to rise quite sharply at lower frequencies (eg below about 20MHz) and the I/O matching will degrade sharply here and the part won't be unconditionally stable below about 25MHz.

Because you want to use this part with a BPF in the low VHF region I'd question if this amplifier is the ideal choice here. Do you need the low noise figure and high IP3 it offers? I think that you will (easily) be able to get the 5536 amplifier to be stable with this BPF on a simulator and in practice,  but why not go for a classic gain block from Minicircuits if you can afford to trade a bit of NF and IP3?

[hendorog]

Could you put the filter in front of the LNA instead of behind it?
The 2nd harmonic from the LNA is about -60dBc.

[bson]

Build it both ways and see what the trade-offs are!