[time-nuts] SMPS or conventional?

Tue Oct 20 15:14:11 UTC 2020

Hi

> On Oct 20, 2020, at 10:45 AM, Luiz Alberto Saba <las at intercat.com.br> wrote:
> 
> Hi Bob
> 
> In a previous tread, someone said that cooling PRS-10 will only make the unit consume more energy, since it has to maintain the lamp hot.
> Is this wrong?

You need to read the entire thread. Cooling a Telecom Rb to a reasonable temperature
*is* a good idea. The parts that are not attached to the oven assembly live *much* longer
at a lower temperature. Most of these small Rb’s die when this or that weird part on this or
that board dies. Caps are usually the first to go, but they can be tracked down and replaced.

If it’s a solid state part with no known replacement source …. the unit is dead. Also complicating
the repair process is the total lack of any alignment / setup information on these Rb’s. If
the part you replace requires realignment ….. you are flying blind. 

Bob

> Before I read the previous tread, I was thinking of electronic thermostat/fan but...
> 
> Em 20/10/2020 10:34, Bob kb8tq escreveu:
>> Hi
>> 
>> Not very well organized message here …. I’m still working on my first cup of
>> coffee …:)
>> 
>> First up: What do I do?
>> 
>> My “lab” is in the basement of our house. There is HVAC running pretty much
>> year round. The temperature “wanders” between about 20 and 25 C over a
>> year. It cycles about 2 or 3 C at roughly a 1 to 2 hour rate. I have three 5065
>> Rb’s running full time on a rack. I have a couple of GPSDO’s running on another
>> rack. Various OCXO’s get tested on a bench. All of it is “out in the open”.
>> 
>> I have looked into various approaches to temperature control ( thermo electric
>> heaters ….). I have a big pile of water pumps, TEC’s, piping, and radiators
>> siting in a dark corner of the room. The more I looked into the power required
>> to do that, the less I liked the idea ….
>> 
>> I do have a few cardboard boxes and a small pile of “beach towels” that I use.
>> If the measurement is particularly sensitive, I put a towel over the test set. If
>> the device is large, I put a box over it. That generally works quite well to reduce
>> fast changing temperature from drafts in the room.
>> 
>> ======
>> 
>> Are you trying to lock the PRS-10 to the UCCM? If so, don’t bother, the net
>> result is a PRS-10 that is not very stable. Just use the UCCM’s output.
>> 
>> Practically speaking, the PRS-10 is a 0-50C device. It’s no different than other
>> Rb’s in that respect. It also is no different than the rest in making those limits
>> really hard to pull out of the data sheet. They talk about 65C baseplate, but
>> that’s not ambient.
>> 
>> The UCCM has an OCXO in it that is (probably) rated too 70C. It depends a
>> bit on what version / vendor you happen to have. In still air the OCXO is ok
>> to 50, in moving air, it should be all right at 70. No idea how the rest of the parts
>> on the board will do above 50C / moving air. ( = I’d bet there will be problems
>> above 50C).
>> 
>> The PRS-10 needs a heatsink on it or it will quickly die. It will live a longer life
>> with a better / bigger heatsink. With a small one, figure 2 years. With a fan or
>> something large, figure > 5 years of life. That’s all in a indoor situation with a
>> HVAC system keeping the ambient around 20 to 25C. The small heatsink
>> gets around 50, the bigger one + fan may get down to 40C.
>> 
>> The real issue with both the PRS-10 and the UCCM  is to reduce the temperature
>> swings. The UCCM will track out long term changes. You just need to protect it
>> from quick changes. Depending on which PRS-10 you have the temperature performance
>> may be better or worse (per the specifications on the part). A wooden box with
>> heaters / fans can do this, there are other ways.
>> 
>> A lot depends on just how crazy it gets where you are in Brazil  does it get above
>> 40C? ( I looked it up, the record is 44.7C …. yikes … ). Ideally you want to keep
>> the heatsink around 40C. That’s not going to happen if it’s 40 out. Running the
>> part at 55 C all the time is not a good idea.
>> 
>> More or less, you normally need 10C of “headroom” on any sort of heated device
>> (like a heated wooden box). You also have a practical problem coming up with
>> a heatsink that gets to < 10C relative to ambient. The net is that your heatsink
>> is at “max ambient” + 20C. For a somewhat more practical heatsink, the numbers
>> are worse …. sorry about that.
>> 
>> What’s a “big” heatsink? Think of something the size of the Rb with a high speed
>> fan attached. As noted in various posts the magnetic field from the fan will not
>> make the Rb happy, some sort of ducting to keep it a bit further away might be
>> a good idea. That sort of thing probably *does* get you within 10C of ambient .
>> 
>> The bottom line is that this sort of gear normally gets used in a location that has
>> heating / cooling. Temperature may swing from 10 to 30, but that’s about it. They
>> generally have pretty big fans moving air around. A cell phone “base station” is
>> one example of this sort of setup.
>> 
>> Probably enough for now…..
>> 
>> Bob
>> 
>> 
>>> On Oct 20, 2020, at 7:36 AM, Luiz Alberto Saba <las at intercat.com.br> wrote:
>>> 
>>> Hi Bob
>>> 
>>> I will try with the toroid setup (as soon as I receive the transformer) and see what happens.
>>> I'm really lost in respect of thermal insulation (or cooling).
>>> I'll use a PRS-10 and a Samsung UCCM gpsdo.
>>> A single mass of copper (or alluminium) wil be effective for PRS-10?
>>> I saw in an old trend here people saying that forced cooling is not good for the PRS-10.
>>> Closed box or ventilation holes?
>>> It will operate at ambient temperature, in Brazil...
>>> If you or someone has some hints about what to do in this respect, I'll be very grateful.
>>> 
>>> Thanks
>>> 
>>> Luiz Alberto Saba
>>> 
>>> 
>>> Em 19/10/2020 17:35, Bob kb8tq escreveu:
>>>> Hi
>>>> 
>>>> One weird issue with big caps in a linear supply:
>>>> 
>>>> As you increase the size of the caps on the bridge rectifier, the ripple
>>>> voltage goes down. The ripple *current* goes up. In some cases (say ripple
>>>> voltage at 1%) it can go up by 10X to 100X of the minimum value.
>>>> 
>>>> Why does this matter?
>>>> 
>>>> Magnetically coupled signals and current bounce in the ground are both very
>>>> real effects. They can be tough to get rid of ….
>>>> 
>>>> There is no one solution to any of this …. :(
>>>> 
>>>> Bob
>>>> 
>>>>> On Oct 19, 2020, at 12:41 PM, Luiz Alberto Saba <las at intercat.com.br> wrote:
>>>>> 
>>>>> Hi Bob
>>>>> 
>>>>> The phase noise is not particularly important in my case.
>>>>> It will be used mainly to calibrate ocxo/rb.
>>>>> I'm pendig to a good toroid/bridge/huge caps and something like LT-1083.
>>>>> 
>>>>> Em 19/10/2020 12:38, Bob kb8tq escreveu:
>>>>>> Hi
>>>>>> 
>>>>>> What will the output of the box be used for? If low phase noise is an objective,
>>>>>> then one needs to be pretty careful about power supply noise. If low phase noise
>>>>>> is not a “need” in this case, then (good) switchers can be used for everything.
>>>>>> 
>>>>>> What’s a “good” switcher? You want one that is reliable ( = does not catch fire
>>>>>> on a regular basis). You also want to get the noise down into the 10 mv p-p range.
>>>>>> That generally means filtering the output a bit. Past that the usual requirements
>>>>>> of line / load regulation and efficiency come into play.
>>>>>> 
>>>>>> Bob
>>>>>> 
>>>>>>> On Oct 19, 2020, at 10:24 AM, Luiz Alberto Saba <las at intercat.com.br> wrote:
>>>>>>> 
>>>>>>> Hi guys
>>>>>>> 
>>>>>>> Planning to build my "reference box" with a rubidium & a gpsdo.
>>>>>>> SMPS will do a good job or the transformer+bridge+capacitor+regulator is best suited for this application?
>>>>>>> In the "test phase" I've used a switched ps for the rubidium and a conventional for the gpsdo.
>>>>>>> 
>>>>>>> Thanks
>>>>>>> 
>>>>>>> Luiz Alberto Saba
>>>>>>> 
>>>>>>> _______________________________________________
>>>>>>> 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.
> 