[time-nuts] Re: Dual Supplies for Low Offset Phase Noise Analyzer

[Lux, Jim]

On 7/14/22 8:01 AM, Magnus Danielson via time-nuts wrote:
> Erik,
>
> On 7/14/22 16:46, Erik Kaashoek wrote:
>> Magnus,
>> See below
>>
>> On 14-7-2022 14:50, Magnus Danielson via time-nuts wrote:
>>> I am in full agreement. While there is a certain pain in doing it, 
>>> but it removes a long set of issues relating to not having it. I've 
>>> had that standpoint from the start, but felt that Erik may benefit 
>>> from getting a lot of the other things going first and learn from 
>>> it, as he seemed so negative about it. It is only when one sees the 
>>> alternative cost that one can make the correct decisions. However, 
>>> when trying to fix many things, it can take time to spend on all the 
>>> things, so spreading things out a bit in time can serve it's 
>>> purposes. This not to be disrespectful to Erik, rather the opposite. 
>>> Then again, sometimes there is a different balance that pans out for 
>>> it's purpose.
>> No offense taken!!
> Great.
>> Unfortunately I'm the type of person that learns most by making 
>> mistakes. So I first wanted to try to get the single supply working 
>> as the PNA is not getting its own permanent supply. Performance at 
>> 10kHz offset is now at the -155dBc/Hz noise level, although still 
>> with a lot of mains harmonics but these could also come from the not 
>> so good audio cable connecting the PNA and the Audio input or a 
>> ground loop.
>> Replacing the single supply with a triple supply is still to be tested.
>
> OK, great. Then I correctly assumed your modus operandi that I was 
> trying to respect and support. Also great that you consider testing it.
>
> As for ground-loop, that is somewhat misunderstood, as most often the 
> ground-loop is not driven by magnetic fields, but different 
> ground-currents driving up potential differences, and when 
> cross-connecting causes a current to handle the difference in such 
> voltage. The two main solutions are: Break away supply currents from 
> reference voltage/grounding and make the ground-reference using bigger 
> wires (both to the common point and between the different parts). I've 
> found that considering it all as an operation of isolation helps to 
> move away from incorrect analysis. 


Well, that's one evil of a ground loop.

But magnetic coupling is also an issue. For instance, the vast majority 
of line frequency interference is coupled magnetically, so twisted pairs 
and such help a lot.  We also see magnetic coupling from switching 
transients on a power supply (where the "transmitting" loop is the 
circuit being powered, spread out across a board), and the victim loop 
is on another board, or different area of the same board.