Electric Automation Forum
Forum » General Discussion » Digital control solutions vs. analog control solutions?
Topics: Digital control solutions vs. analog control solutions? on General Discussion
#1
Start by
Marty
09-09-2013 04:39 AM

Digital control solutions vs. analog control solutions?

Some of the digital control sites insist that good digital control can be achieved by less parameters being sensed (gain only). Sensing typically causes additional losses/costs. For me, there are four rules of "closed loop" compensation:
1. Present the highest gain at DC to present the highest output regulation accuracy.
2. Phase margin: never go above 345 deg for transient stability.
3: "Excess phase" - one decade above the output filter pole there can be a -90 to -180 degree phase lag (with a closed loop starting point of -270 that only leaves -45 deg). Phase boosting must be provided.
4. Gain Margin: Whenever the gain is greater than 1, make sure that your phase is away from -360. Sometimes ESR and parasitic elements caused zeroes - pay attention to those.

I disagree with the digiofiles about ignoring phase by only measuring the slope of the gain curve. It is true that there are dependencies between the two, but it could also cause instabilities, caused both by computing latencies and resolution, that could cause problems. Phase ultimately causes instability, regardless of gain.

Both systems should be measured by a network analyzer, such as Ray's. Both are "black box" controllers and should perform similarly under analysis. There are, of course, non-linear digital transient algorithms, which can be examined separately.

Thoughts?
09-09-2013 04:39 AM
Top #2
Ray
09-09-2013 04:39 AM
On the gain margin, you cannot always satisfy 4 above. Many power systems have conditional stability where the gain is high with no phase margin. It is not a stability issue as long as the phase margin is OK at 0DB.

Agreed, Marty, you cannot just look at the slope. A converter can have a constant -1 slope, but within that is a RHP zero and a pole to keep the slope the same, while the phase drops off rapidly.

Even when you do a great job with a digital controller, the analog power stage remains unpredictable in some regions, and it should always be confirmed with the analog measurement for rugged stability. Our AP300 analyzer is embedded in the design process of many of the digital companies for this purpose.
09-09-2013 04:40 AM
Top #3
Amita
09-09-2013 04:40 AM
As Ray said "Many power systems have conditional stability where the gain is high with no phase margin. It is not a stability issue as long as the phase margin is OK at 0DB" This could be a Inverter or a BLDC motor drive controller , which are inherently stable? A Single pole system....? Where as with SMPS, a 2 pole system where stabilization is not as simple as Inverter with simple gain adjustment without phase adjustment? Is it something like this?
09-09-2013 04:41 AM
Top #4
Ray
09-09-2013 04:41 AM
The simplest example is of the slow control loop of the PFC circuit. The transfer function from control to output is a 1/s characteristic when the load is a switching power supply. This must then be cascaded with an integrator in the feedback, creating a -40 dB/decade slope and 180 degrees phase delay. As long as the zero kicks in before crossover, it is fine.

Another example is a high-Q buck converter with a sharp resonant drop in phase. When compensating this with a pair of real zeros, it is impractical to stop the phase temporarily dropping below -180 degrees.

More complex examples can involves input filter interactions.

Conditional stability is much more common than you would think.
09-09-2013 04:41 AM
Top #5
Amita
09-09-2013 04:41 AM
Thanks sir . If this is the case with digital control system, why all the control loop is implement using digital controller with DSP/DSC in the system like UPS, Solar inverter etc where it is a combination of DC-DC converter and inverter in most of the cases....?? Any comments......??
09-09-2013 04:42 AM
Top #6
Hoon
09-09-2013 04:42 AM
Hi Ray, Could you provide more details or data to support that the system is still stable while the conditional phase margin is zero and Gain is still high ?
Thanks.
09-09-2013 04:43 AM
Top #7
Merren
09-09-2013 04:43 AM
How about "Constant On Time" control scheme.. for simple digital control with minimum sensing parameters... with excellent stability....
I refer to this as variable frequency controller ....
09-09-2013 04:48 AM
Top #8
Mohammad
09-09-2013 04:48 AM
"The Bode criteria only says that the phase must be above -180 degrees when it crosses over 0 dB. There is nothing that says it can't do that before 0 dB."

"Gain Margin: Whenever the gain is greater than 1, make sure that your phase is away from -360. Sometimes ESR and parasitic elements caused zeroes - pay attention to those."

gain greater than 1 means greater than 0db then the phase should be above -180 degrees
what does than "away from -360" means in the above statement.
09-09-2013 04:49 AM
Top #9
Marty
09-09-2013 04:49 AM
Mohammad, I refer to a closed loop system where the error amp immediately provides a -180 degrees of the closed loop phase contribution. Add to that, there is typically a feedback capacitor which contributes another -90 degrees of phase. That leaves only 90 degrees of difference between -270 and -360. So there is a difference between how it is defined. .
There are two critical points, in my opinion, 10X the output filter pole frequency (either -90 or -180 deg) and the gain crossover point. Gain should be killed higher than the crossover point due to the parasitic influences (PCB or load characteristics).
So it is a matter of definition of the system.
09-09-2013 04:49 AM
Top #10
Ron
09-09-2013 04:49 AM
With any control, but especially digital, you need to consider the contribution to phase of the transport delay around the loop. In many (most) digital system the delay created by the signal processing can be significant. You can have a stable system but low effective bandwidth due to this delay. There is no way around it unless you re-architect the signal path.
Reply to Thread