DIY 30MHz LPF having problem on the matellic housing

[trxguy]

Hello all,

I designed a 30MHz LPF for SWL, the result is quite good, cut-off point at 33MHz, max attenuation more than 50dB, the curve rising up at 200MHz. I soldered all the component on a tiny single layer pcb (20mmx26mm). Total 4 smd capacitors and 3 smd coils.
But I pull it into a small matellic housing, the result is getting bad, attenuation decreased to 20dB, the curve is rising up on 60MHz.
I have tried to isolate the matellic housing from the gnd (outer BNC connection), it returned to normal (att. >50dB).
It would be affected by parasitic capacitance....
My question is
(1) If I isolate the housing from the gnd, does it be affacted by EMI easily? Does it be an odd design?
(2) How much parastic capacitance should be taken in account?

Is there anybody giving me some idea?

mkchan

 

[vfone]

If you isolate the metal case from the ground (floating ground), will be the same result as you don’t have any shield. Actually maybe worst results, because the case will behave like an antenna and also detune your circuits.

So, at least for these relative low frequencies I would recommend to don’t put the LPF inside of any metal shield. I’ve seen very sensitive professional radios up to 50MHz that don’t use any shielding on the front end.

More important for these stages to be immune to the external interferer, is the impedance matching between stages.

 

[biff44]

Agreed, how you package it is very important.

if you want to use a metal housing, you have to be extremely carefull about how you mount the board and connectors. the slightest resistance or inductance between those components and "gound", and you will have poor filter rejection due to ground loops/re-radiation. Something like 0.1 ohms between the board ground and the connector shield will doom you to probably 20 dB rejection.

1st, you need a very short run between the connector center pins and the board. Otherwise you will create an antenna inside of the box.

2nd, you need an extremely good, short connection between the connector metal shell and the board at the filter inpur. Solder the board ground plane (I assume/hope you have a board ground plane) right to the metal box. If no can do, use two or three screws/nuts/star washers right where the board meets the housing floor closest to the input connector.

Repeat above at the output connector.

 

[trxguy]

Thanks for reply, I have uploaded my LPF design and the result, please help me check there is any problem or not.
Thanks,
mkchan

 

[kspalla]

Instead taking two line transmission line to connect BNC to board use a coax cable.

 

[trxguy]

Hello,
Because the housing is very small, and the connection between the BNCs and the PCB is lesser than 1 cm, I cannot use coaxial.
I will try to use a segment of coaxial shielding as the GND connection on both end, reduce the inductance of the GND connection. please wait for my result.
Thanks of all
mkchan

 

[biff44]

Looking at your sketch, I would lose the ground wires (they are just a joke at 200 mhz anyway, so no great loss). I would make sure the two connectors are well attached to the metal housing (no paint in the way, star washer to bight into the metal surfaces). I would attach the assembled LPF board into the housing with either:

1) conductive epoxy between the backside board ground plane and the bare metal housing floor,
or
2) 3 screws on the left side, and 3 screws on the right side, holding the board's ground plane tightly to the bare metal housing floor.

 

[trxguy]

Thanks for giving me many ideas. Due to too small of the metallic case, I cannot add screws on the both side of the pcb.
Finally, I made the changing,
(1) isolated the connection between the BNCs and the housing (replaced the BNCs with an isolating type BNCs).
(2) connect the PCB GND through two 0.1uF capacitors to the housing.
Now, there is no any DC connection between the circuit GND and the metallic housing. The result is that, please see the attachments.
one is 100kHz to 100MHz, another is 100kHz to 500MHz, both are -10dBm input and 10dB/dev.
The result seem to be fine? give me some comments,
thanks,
trx