My favorite SPICE simulator is Microcap 12. It`s freeware since a few years ago. A side from that it has an extra stability simulation which works pretty good. For anyone interested take a look at the reference manual its documentation also is pretty good. By the way great video as always :)
Fantastic video, will be following this when I finally get around to designing my own lab PSU (for fun). Where did you get the "good" margins from? I.e. 45/60 degrees and the gain?
Its found in general in technical literature; People who know way more than me about this subject came up with the values :D In theory even 1degree of phase margin would be enough to prevent the appearance of a permanent oscillation, but the exact values will have various tolerances - like depend on the circuit construction, component values, temperature, load etc. So a value of 45/60 is usually sufficient to validate the design.
@@FesZElectronics thanks, the idea of a safety gap makes sense with component tolerances and manufacturing. I have a Agilent PSU 66311b which has a fast and slow control loop mode, I'm guessing it's something to do with this trade-off you stated on margins and speed! Thanks again
It's easier if you measure the output impedance and group delay of the DUT. See the white paper at Omicron Labs "Non Invasive Measurement of Power Supply Stability"
What IC did you use ? asking since i wander if we can find 100khz in the datasheet since they mist have tested for this , also this may be lowered but the PCB design and part values but it would be interesting to see IRL value vb theoretical datasheet values. Also here is an idea: boost or buck converters on a shared output how to stabilize this .
The crossover is not usually documented, since it depends on the exact implementation and also the test load; what you might find though is records of an evaluation board being measured under specific conditions;
There is a "small" problem here. Your reasoning on the open loop is correct but you are doing your measurement in closed loop, i.e you are fighting the regulator that is trying to cancel the error and the transfer function is not the same. On top of that by changing the excitation with frequency you are adding a pole and a zero to the system. So you are not measuring the performance of the original design but a simulated combo system with virtual components. An effect of that is the spike seen in the phase response. You should bias a constant operating point with an auxiliary power source and open the loop, then perhaps add a corrector circuit based on the open loop measurements. BTW, a 30deg phase margin will give an "optimal" speed response with a single bump damped overshoot for a system with 2 dominant complex poles. Nice channel though, continue the effort, I see progress over time, and quite often learn a trick or two.🙂
Hi! Some questions... 1. What are those 2 gaps in the phase shift plot on lower frequencies? Aren't those mean also stability problem? 2. Output seems to oscillate even before applying the load step (even before ringing starts), or do I see wrong? 3. How these measurements show exactly the bandwidth of the power supply? What bandwidth of supply really mean? Thanks!
Hello. Good Questions! 1. The phase is close to 180deg; here the measurement equipment jumps between +180 and -180 since it only outputs the 180to-180 range; so +181 is expressed as -179. 2. I'm not sure what to say here, it might be a setup issue, or it might be that the supply has such a low phase margin its inherently unstable, and any small environmental perturbation causes it to oscillate; 3. By bandwidth here I am referring to reaction speed - if an external perturbation is applied, how fast can the circuit react. At any frequency below the crossover the supply can react, but not above since it has no gain above.
@@FesZElectronicsCan you provide instrcutions to diy the transformer? What should be taken care when making it? Will the ouput impedance (50 ohm？) of the connected signal generator affects the test device's loop response?
Thanks for an interesting video. Can you please mention something about the transformer itself? How can a transformer be suitable for such a wide frequency span, wont the magnetisation current run away for low frequencies? What about the flatness of the transformer characteristics? /Tomas
Yes, the type of supply should not matter. The measurement can be done as long as you have access to the feedback loop - so for example you cannot perform it on a part like 78xx where everything is integrated.
![DIY BenchSupply [2/2]](http://i.ytimg.com/vi/8E8V7Uw-3eg/mqdefault.jpg)
![DIY BenchSupply [2/2]](/img/tr.png)
Interesting.... When I finalise component choices for my solar battery charger, I'll try this test!
Very interesting and useful, thank you Fesz!! I Will also try to measure loop response when I get the necessary equipment. Keep working like this!!
Amazing as always! Thank you
Very good video. Thank you.
I like to implement power supplies with a single integrator. Stability is guaranteed with any load and start-up is clean, with no overshoot.
Good job)
My favorite SPICE simulator is Microcap 12. It`s freeware since a few years ago. A side from that it has an extra stability simulation which works pretty good. For anyone interested take a look at the reference manual its documentation also is pretty good.
By the way great video as always :)
I already dowload. Yes, it seems good. Thank you. Best regards from Portugal.
Great, thank you very much. Very useable info.
Fantastic video, will be following this when I finally get around to designing my own lab PSU (for fun).
Where did you get the "good" margins from? I.e. 45/60 degrees and the gain?
Its found in general in technical literature; People who know way more than me about this subject came up with the values :D
In theory even 1degree of phase margin would be enough to prevent the appearance of a permanent oscillation, but the exact values will have various tolerances - like depend on the circuit construction, component values, temperature, load etc. So a value of 45/60 is usually sufficient to validate the design.
@@FesZElectronics thanks, the idea of a safety gap makes sense with component tolerances and manufacturing.
I have a Agilent PSU 66311b which has a fast and slow control loop mode, I'm guessing it's something to do with this trade-off you stated on margins and speed! Thanks again
Nice Video❤
Thanks.
It's easier if you measure the output impedance and group delay of the DUT. See the white paper at Omicron Labs "Non Invasive Measurement of Power Supply Stability"
What IC did you use ? asking since i wander if we can find 100khz in the datasheet since they mist have tested for this , also this may be lowered but the PCB design and part values but it would be interesting to see IRL value vb theoretical datasheet values.
Also here is an idea: boost or buck converters on a shared output how to stabilize this .
The crossover is not usually documented, since it depends on the exact implementation and also the test load; what you might find though is records of an evaluation board being measured under specific conditions;
Great video.
Is this method related to the philosophy behind the new .fra directive in LTspice 17.1?
I think so, yes; although I did not really get a chance to check it out yet
There is a "small" problem here. Your reasoning on the open loop is correct but you are doing your measurement in closed loop, i.e you are fighting the regulator that is trying to cancel the error and the transfer function is not the same. On top of that by changing the excitation with frequency you are adding a pole and a zero to the system. So you are not measuring the performance of the original design but a simulated combo system with virtual components. An effect of that is the spike seen in the phase response. You should bias a constant operating point with an auxiliary power source and open the loop, then perhaps add a corrector circuit based on the open loop measurements. BTW, a 30deg phase margin will give an "optimal" speed response with a single bump damped overshoot for a system with 2 dominant complex poles. Nice channel though, continue the effort, I see progress over time, and quite often learn a trick or two.🙂
Hi! Some questions... 1. What are those 2 gaps in the phase shift plot on lower frequencies? Aren't those mean also stability problem? 2. Output seems to oscillate even before applying the load step (even before ringing starts), or do I see wrong? 3. How these measurements show exactly the bandwidth of the power supply? What bandwidth of supply really mean? Thanks!
Hello. Good Questions! 1. The phase is close to 180deg; here the measurement equipment jumps between +180 and -180 since it only outputs the 180to-180 range; so +181 is expressed as -179. 2. I'm not sure what to say here, it might be a setup issue, or it might be that the supply has such a low phase margin its inherently unstable, and any small environmental perturbation causes it to oscillate; 3. By bandwidth here I am referring to reaction speed - if an external perturbation is applied, how fast can the circuit react. At any frequency below the crossover the supply can react, but not above since it has no gain above.
Which transformer was used to inject noise?
I used a DIY transformer; I don't have a proper description of it.
@@FesZElectronicsCan you provide instrcutions to diy the transformer? What should be taken care when making it? Will the ouput impedance (50 ohm？) of the connected signal generator affects the test device's loop response?
I am also interested in what such a transformer looks like and how to make it
Thanks for an interesting video. Can you please mention something about the transformer itself? How can a transformer be suitable for such a wide frequency span, wont the magnetisation current run away for low frequencies? What about the flatness of the transformer characteristics? /Tomas
this is a measurement that can also be done on a digital power supply, correct?
Yes, the type of supply should not matter. The measurement can be done as long as you have access to the feedback loop - so for example you cannot perform it on a part like 78xx where everything is integrated.