[time-nuts] Using a common power supply among few time standards

[Bob kb8tq]

Hi

DC/DC converters can indeed be filtered. If you do it wrong, things will get pretty crazy. In 
general, you need at least an L/C filter and may need an R/L/C design. Things like ESR
on the caps and SRF on the inductors do indeed matter. Both common mode and differential
mode filtering may be needed. With both inductors and ferrite beads, saturation performance
is important. It’s a good idea to only use parts that have detailed data. 

Three terminal regulators have a *wide* range of PSRR profiles. Some do almost nothing 
past 1 KHz, others are still doing quite well at 10 MHz. Data sheets are your friend in this 
case. For best performance, app notes are often a good resource as well. 

Yes, this *is* circuit design. There are details to it. Test / analysis / rework are all part of the 
process. That’s the way it’s done ….

Bob

> On Feb 8, 2020, at 9:59 AM, Taka Kamiya via time-nuts <time-nuts at lists.febo.com> wrote:
> 
> Right now, my problem is where to strike a reasonable yet good compromise.
> I thought of using DC/DC but these are quite noisy, and noise comes from both ends.  I had one powering distribution amplifiers.  For an experiment, I placed a ferrite bead on the output and the whole thing oscillated rail to rail.  (and killed the dist.amp)  Also, these things switch at very high speed where ability of 3 terminal regulators to reduce noise is at minimum.
> On the other hand, I was playing with red label T-bolt.  This thing has internal DC/DC converters to take 24V and turn it into +/- 12V and 5V.  Powering it with clean linear supply and a garden variety switcher, output was equally clean.  (at least far below my lab's ability to observe issues)  I saw no difference in output at all.  Tech support at SRS demanded I use linear supply for PRS-10, and declined to discuss further.  Yet, I've seen most DIY implementation uses switcher.  I *think* /tvb uses 48V common supply.  I also read a comment on this list, when one is dealing with time-nut level accuracy and stability, every milli-volt counts.
> 
> I guess I'll have to deal with this on case-by-case basis.  (pun intended)
> 
> --------------------------------------- 
> (Mr.) Taka Kamiya
> KB4EMF / ex JF2DKG
> 
> 
>    On Saturday, February 8, 2020, 9:11:13 AM EST, jimlux <jimlux at earthlink.net> wrote:  
> 
> On 2/7/20 7:01 PM, Taka Kamiya via time-nuts wrote:
>> I recall S-100 scheme had an issue of per-board regulator failing short and dumping 8V into bus killing rest of the system....
> 
> More like the 5V regulator (most likely a trusty 7805 in a TO-3 package) 
> shorting, applying 8V to all those TTL parts on the board, which die, 
> and put 8V on the address and data lines.  However, I'm not sure that 
> was a particularly common failure.  In many years of working with a lot 
> of S100 systems, I never had that problem.
> 
>> I am thinking of making use of 28V (aircraft power supply?) regulated power supply, and have local linear regulator to created needed voltages.  One concern is, should I need large amount of 5V then loss will be tremendous.
> 
> Probably a better solution is multiple bulk supplies off line AC then 
> regulate down. Or a DC/DC from 28V to 8V, then good linear regulator to 5V.
> 
> There's really no way around it, although you can do things like 
> synchronize all your PWM converters, so if you do get spurs, they're in 
> consistent places.
> 
> 
>   I just came across an article that discussed magnetic field 
> sensitivity of LPRO-101.  The author ended up relocating the transformer 
> off-board.  (in a different case)  A very timely information as this is 
> what I am actually working on.
>> Further comment will be appreciatedTakaTime-nut in training
>> ---------------------------------------
>> (Mr.) Taka Kamiya
>> KB4EMF / ex JF2DKG
>>   
>> 
>>       On Thursday, February 6, 2020, 6:51:38 PM EST, jimlux <jimlux at earthlink.net> wrote:
>>   
>>   On 2/6/20 3:14 PM, Will Kimber wrote:
>>> Hi,
>>> 
>>> Usual (best?) is supply slight over voltage and use LDO/RC filtering in
>>> each unit as Bob suggested. But that would require either supplying
>>> 24v/12v/5v/-12v (with some margin 10%?) or converting some single supply
>>> of 24v or 48v  in each unit to required voltages at each piece of
>>> equipment.  The later requires switching supplies to get all voltages.
>> 
>> 
>> Well that was the strategy used on S-100 boards. Bulk 8V unregulated
>> supply regulated down to 5V on the board.
>> 
>> 
>> 
>> _______________________________________________
>> time-nuts mailing list -- time-nuts at lists.febo.com
>> To unsubscribe, go to http://lists.febo.com/mailman/listinfo/time-nuts_lists.febo.com
>> and follow the instructions there.
>>     
>> _______________________________________________
>> time-nuts mailing list -- time-nuts at lists.febo.com
>> To unsubscribe, go to http://lists.febo.com/mailman/listinfo/time-nuts_lists.febo.com
>> and follow the instructions there.
>> 
> 
> 
> _______________________________________________
> time-nuts mailing list -- time-nuts at lists.febo.com
> To unsubscribe, go to http://lists.febo.com/mailman/listinfo/time-nuts_lists.febo.com
> and follow the instructions there.
> 
> _______________________________________________
> time-nuts mailing list -- time-nuts at lists.febo.com
> To unsubscribe, go to http://lists.febo.com/mailman/listinfo/time-nuts_lists.febo.com
> and follow the instructions there.