Linear PSU vs Switch Mode PSU for precise\sensitive analog designs

[CVT]

I need advice from knowledgeable analog (emc) designers regarding a choice between a linear PSU or switch-mode PSU for laboratory use.

At the moment I am busy with a sensitive circuit with low level signals containing ADCs, Op Amps and voltage to frequency converters. I had a closed frame switch-mode PSU driving another piece of equipment in my office. I knew that it would be noisy but I under estimated the influence it would have on my sensitive analog circuit which it is not even connected to it. I know this is because of radiated emissions and susceptiblity.

In future I will be doing even more designs with such sensitive circuits. I want to buy a general dual output laboratory bench PSU that will have very low radiated and conducted emmisions.

My question is:
Will a good quality linear PSU always be much better than a good quality switch-mode PSU for such sensitive work?
Maybe there are good modern implementation of Switch-Mode PSUs that are comparable to good Linear PSU?
Please elaborate on your views or point me to good switch-mode PSU i.e. noise comparable to linear PSUs.

I am looking at:
Linear PSU (2 x PAN60-6A) at a cost of $4800 (https://www.kikusui.co.jp/en/product/detail.php?IdFamily=0003)
Switch MOde PSU: (1 x XPF 60-20DMHV) at a cost of $3000 (https://www.signaltestinc.com/v/vspfiles/assets/datasheet/Sorensen_XPF_Datasheet.pdf)
I am not concerned about spending the $4800 if I have to.

 

[sky_123]

I think it would be a waste of money better spent on other test gear. If you're concerned with low level signals, and driving ADCs etc., I suppose you don't need a 60V 6A PSU. A linear bench supply for (say) 4A 30V can be obtained from Farnell for a few hundred dollars. If you're really concerned about noise, use a battery (primary battery or charged up battery). If the circuit is so sensitive (e.g. for some scientific experiment) then I suppose you need to take additional precautions anyway (filtering, screening and so on) - and for that I'd rather a dedicated power supply anyway (even if it needs to be custom designed), for the task at hand, rather than a bench supply.

 

[CVT]

Hello sky_123

Thanks for the quick response. You are right about not needing a bench PSU with such high current and voltage outputs when doing low level signal development. My idea is/was to have a single PSU that can cover all/most of my use cases. There will be cases where I will need high power and not necessarily low noise. I suppose it would be better to have a separate supply for those cases.

Your also right about screening for RF interference etc.. I was caught a bit of guard regarding this issue in my current design. I am just a bit unclear about what you mean with a dedicated power supply. Does a dedicated supply mean a DC to DC using regulators on the PCB with the sensitive electronics? How far is the linear PSU (AC to DC) with conducted and radiated emmisions 50 Hz emissions removed from your sensitive electronics?

 

[sky_123]

It is possible to get DC-DC converters with low noise, in the hundreds of uV (Linear Tech have some nice ICs) and can be followed by an linear regulator. This doc has lots of tips to minimise switching noise.
I needed a negative supply for a project, and I'm using a charge pump method in an effort to keep harmonics noise power low, and low freq
ripple of a few hundreds of uV should be easily possible (I appreciate that may be too high for some scientific applications).
Personally I prefer to use dedicated and smaller supplies (switched, linear or combinations) where needed, and this allows me to power the entire
project from a single AC-DC converter power brick, I tend to have lots of 12V and 48V power bricks! My adjustable bench PSU sees little
use as a result. By dedicated I just mean a supply designed for the specific requirements in the cicuit, i.e. expected input voltage and
expected current output requirement.

 

[picgak]

Hi,
if your requirement involves very highly sensitive devices and need to be tested in open without shielding it is advisable to use a series pass regulated power supply with appropriate transformer.
regards ani

 

[jiripolivka]

All switched-mode power supplies generate high levels of noise, recently the spectrum exceeds 100 MHz. Such devils can and do affect many nice circuits quite badly.

If you mean to run a good laborAtory and make good analog circuits, forget the devils! A linear power supply is always the best, and predictable. Remember, even a battery generates noise!

 

[goldsmith]

All switched-mode power supplies generate high levels of noise

Hi Jiripolivka
I'm disagree about your statement . a bad SMPS ( a bad design ) will have noise ! with a simple differential filter or best selection of filter elements , it is impossible to have noises !

recently the spectrum exceeds 100 MHz. Such devils can and do affect many nice circuits quite badly.

most of my designs are using SMPS ! ( some times linears are good too )
( the secret is that i always using my own SMPS circuits ! i prefer to design each thing myself because i can consider all of the parameters .

If you mean to run a good laborAtory and make good analog circuits, forget the devils! A linear power supply is always the best, and predictable. Remember, even a battery generates noise!

I'm sorry but i'm disagree again . i have an SMPS ( variable one ) on m y desk . i have designed it myself . it is pretty nice ! ( i'm an analog designer and all of my circuits are analog or most of the times in HF band )
Again , if an SMPS correctly has been designed you won't have any problem with that ! ( i'n my opinion they are not devils ! they are angels ! ) ;-)
Best Regards
Goldsmith

 

[jiripolivka]

Hi Jiripolivka
I'm disagree about your statement . a bad SMPS ( a bad design ) will have noise ! with a simple differential filter or best selection of filter elements , it is impossible to have noises !
most of my designs are using SMPS ! ( some times linears are good too )
( the secret is that i always using my own SMPS circuits ! i prefer to design each thing myself because i can consider all of the parameters .

I'm sorry but i'm disagree again . i have an SMPS ( variable one ) on m y desk . i have designed it myself . it is pretty nice ! ( i'm an analog designer and all of my circuits are analog or most of the times in HF band )
Again , if an SMPS correctly has been designed you won't have any problem with that ! ( i'n my opinion they are not devils ! they are angels ! ) ;-)
Best Regards
Goldsmith

I accept your disagreement but I insist that power supplies should generate only the desired DC power. Anything else is simply wrong. Sensitive circuits are very vulnerable to stray signals. If you allow your power supplies to generate unwanted signals, you will spend your time searching for the causes of interference your equipment is generating.

In radio astronomy, the main problem of interference are computers and other digital stuff. Today the processors run at 1...3...5 GHz, and high harmonics saturate the receivers. If you handle sensitive communication systems, the worst cases of phase noise come from the switching power supplies.

Keep living with your "angels" till they sting your sensitive tongue...

 

[goldsmith]

In radio astronomy, the main problem of interference are computers and other digital stuff. Today the processors run at 1...3...5 GHz, and high harmonics saturate the receivers. If you handle sensitive communication systems, the worst cases of phase noise come from the switching power supplies.

Hi again
Do you know about EMI filters ? what is the duty of such heavy filter ? simply to don't allow any unwanted harmonic to go through power line ! just 50 HZ ( here frequency of line is 50 HZ but i think in your country it is 60 HZ ? ) just 50 HZ can go through rectifier , and any other components can't enter or come out ! so it means there won't be any problem . however if you have a rubbish signal as input with heavy noises ! that EMI filter will optimize all things as we want ! ( i'm using EMI filter in all of my transmitters and amplifiers . )
ANother thing : consider please that we want a variable PSU with current of 0-5 Amperes or even 0-500 A ! 0-30 volts or 0-30kv ! ( but for educational applications usually a simultaneous 0-50 0-5 A would be ok . ) so can you guess size and price of an iron core transformer for that ? and another issue is can you imagine if we set out put voltage , as 0.5 volts and out put current as 5 A , how much will be the efficiency ? what about power loss ? So i think now you are agree that linear PSUs are not good . ( i.e i have many linear PSU in my laboratory but most of the times i prefer to use an SMPS ! ( another parameter is lower weight )

Keep living with your "angels" till they sting your sensitive tongue...

I think you know English is not my native language hence i can't understand some of the local sentences . can you please tell me what do you mean by that ?
Best Regards
Goldsmith