IFHI pickup noise

Are you doing this testing on the AD608 evaluation board or on your own board? If you don't have an AD608 evaluation board, I think that you should get one (stock available at Mouser and Digikey) and try to mount your filter on there. It appears that the signal is leaking around the filter. If that is true you need to find the signal leakage path. There are some experiments you can do try.

1. Power VPS1 and VPS2 from separate supplies.

2. Add additional power supply decoupling capacitors to VPS1 and VPS2.

2. With the mixer circuit active, ac-couple IFHI to ground and see if the RSSI voltage is still high.

3. Completely break the link between the mixer circuit and the RSSI to see if the RSSI voltage is still high. 

I would try these tests to see if they yield any results. This part as tons of gain at both the front end and back end. On top of that. Log detectors are really good at detecting small signals. So it's not an easy part to use. 

Dear Aemon,

thank you vm for your advises! I use AD608 on my board (double sided bottom is solid GND).

1-2. Unfortunately in my board a VS1 and VS2 are tied under the IC and bypassed at both side (!). I also discovered when these bypass capacitor vales modified (say 47n->3.3n or if they are not the same and exhanged then RSSI (at 0 RF input) is changing between 0,2..0,4V.

2. If IFHI AC grounded the RSSI still ~0,6-0,7V (RFHI 50mVp)

3. IF crystal removed RSSI still high (do not change).

Meantime, I measured VS1-VS2 (note, tied together) supply using oscilloscope FFT and ca -55dBV mixer products (fdiff, fsum frequencies) can be detected on it. Also on the VMID.
It seems the supply/VMID bypassing do not work.

My board is trial "home made" (topology printed to a paper and transferred to laminated panel by iron then etched and drilled): small piece of wires used as vias.

However, the IC shall have significant PSSR, insn't? For example

the IF stage is by chain of difference amplifiers and Gilbert cell mixer uses a current source at its lower differential pair, current mirrors at the "top", etc.

Regards, Joseph

PCB: prior to crystals a tapped-C matching is used 27Ohm (pin9 load) to 1500Ohm (crystal Zin) for adding extra attenuation for IF filter. 27Ohm load + voltage ratio of tapped-C evevelops the same 900mVp than 165Ohm load.
However, using 180Ohm + series 1,3KOhm the above problem still exists.

Hi,

Meantime a made the following:

- I remove the the direct interconnect track between VSP1 and VSP2 under the AD608 and connect VSP2 separately to system 3V3: no any improvement (outband signal rejection / stopband annten.)

- Disconnect coupling capacitor from Mixer LO input and bypass it to GND: RSSI raised near to maximum (~1,8V)(!)

Gibert cell should provide 0 output product if one the input signal is 0. /*My LO is and ADF4001 + discreate transistor*/

- If RFHI is bypassed to GND, RSSI works fine: ~0,2V.

Still only ~55dB far selectivity can be reached.

J.