[time-nuts] 10 MHz Filters

Charles Steinmetz csteinmetz at yandex.com
Sat Nov 29 01:13:54 EST 2014

John wrote:

>For the most part, you don't want transformer isolation unless you 
>plan on using balanced lines.  There are worse things than ground 
>loops out there, and lifting a coax shield away from ground is a 
>great way to find all of them.

You certainly need the shield grounded at RF, but you don't 
necessarily need it grounded all the way down to DC.  A fairly common 
solution that works well for many applications is to use isolated 
connectors (not connected directly to the chassis) with a 10nF 
capacitor from each connector shield to the chassis immediately 
adjacent to the connector.  The capacitor grounds the shield at RF 
but allows very little current to flow at the mains frequency.  NOTE: 
You want the bypass capacitor right at the connector, to make sure 
the RFI current loop is tiny and cannot radiate.

That said, I often build distribution amplifiers with 
chassis-grounded connectors and have never had ground loop 
problems.  But then, I pretty much always design transformer-isolated 
DC supplies with low-field transformers and linear regulators into 
everything I build, and do not use wall-warts, desk-warts, or 
switching supplies.

If one were building the dist amp I posted the other day and wanted 
to use chassis-grounded connectors, the per-stage transformers could 
be simplified from 1:1:1 to 1:1 -- omitting the third winding -- with 
the transistor collectors capacitor-coupled to the output 
connectors.  In this case, I would put ~100k resistors across each 
output to hold the connector ends of the capacitors at 0v even with 
an open load.

For isolation amplifiers used with mixer-type frequency comparators, 
which can be very sensitive to ground loops because of the low mixer 
output frequency, I do use RF transformers and bypass the shields to chassis.

>You definitely don't want 10.7 MHz IF transformers, unless you are 
>just trying to build a thermometer.

Hear, hear.

Best regards.


More information about the time-nuts mailing list