ProSoundWeb Community

Sound Reinforcement - Forums for Live Sound Professionals - Your Displayed Name Must Be Your Real Full Name To Post In The Live Sound Forums => AC Power and Grounding => Topic started by: Stephen Swaffer on July 20, 2014, 03:04:18 PM

Title: Switch Mode Power Supplies
Post by: Stephen Swaffer on July 20, 2014, 03:04:18 PM
This question is perhaps better answered by giving refernces, since it is probably a rather complex answer.

If a switch mode power supply uses a bridge rectifier and capacitor to create a DC voltage that is then mode into square waves and the switching action done, why is it any harder on a power source than a linear power supply that uses a bridge rectifier and capacitor to create a DC voltage?  I can see that it might tend to create more noise on the supply since you don't have an inductor (tansformer) to help attenuate high freqs.

If the switch mode power supply worked as many triac dimmers do and only passed part of the AC waveform on, the problem is obvious.
Title: Re: Switch Mode Power Supplies
Post by: Mike Sokol on July 20, 2014, 05:22:22 PM
Your inverter apparently is seeing the low impedance load and thinks it's a short circuit.

I saw this sort of thing a LONG time ago (25 years) with a big computer hard drive starting from a UPS. It would only start successfully maybe 10% of the time, and the other 90% of the time would send the UPS into shutdown mode. I speculated that the tripping was due to the random starting point in the phase of the 60-Hz cycle creating an inrush current causing the trip. So I built a zero-crossover starting relay (basic Solid State module) that waited for 0-degrees phase to energize the hard drive motor. Worked like a charm, and it never had a starting trip after that. Don't know if that will help your inverter tripping problem, but perhaps that helps explain what's happening.   
Title: Re: Switch Mode Power Supplies
Post by: John Roberts {JR} on July 20, 2014, 08:50:01 PM
This question is perhaps better answered by giving refernces, since it is probably a rather complex answer.
references... I don't need no stinkin references.  8)
Quote
If a switch mode power supply uses a bridge rectifier and capacitor to create a DC voltage that is then mode into square waves and the switching action done, why is it any harder on a power source than a linear power supply that uses a bridge rectifier and capacitor to create a DC voltage? 
The difference is that conventional PS run through a transformer primary, magnetic circuit, and then secondary, so the series impedance is higher and peak current is lower, than hanging a diode and cap directly on the primary.
Quote
I can see that it might tend to create more noise on the supply since you don't have an inductor (tansformer) to help attenuate high freqs.
Think about rate of change driving a cap directly with no series impedance.
Quote
If the switch mode power supply worked as many triac dimmers do and only passed part of the AC waveform on, the problem is obvious.
I see little obvious about this.

JR
Title: Re: Switch Mode Power Supplies
Post by: Steve M Smith on July 21, 2014, 01:50:02 AM
I built a zero-crossover starting relay (basic Solid State module) that waited for 0-degrees phase to energize the hard drive motor. Worked like a charm

A bit of inductance in series with the incoming supply might sort it out too.


Steve.
Title: Re: Switch Mode Power Supplies
Post by: Guy Holt, Gaffer on August 04, 2014, 01:44:54 PM
If a switch mode power supply uses a bridge rectifier and capacitor to create a DC voltage that is then mode into square waves and the switching action done, why is it any harder on a power source than a linear power supply … If the switch mode power supply worked as many triac dimmers do and only passed part of the AC waveform on, the problem is obvious.

What makes SMPSs tougher on power supplies is that, like triac dimmers, they do in fact pass only part of the AC waveform,  but unlike triac dimmers, they also draw current in harmonically rich pulses that are many times larger in amplitude than what a comparable size (wattage) linear load on a triac dimmer would draw. To understand why this is the case, let’s look at how SMPSs operate in more detail.

(http://www.screenlightandgrip.com/images/generators/Pro_Sound_PP_Image85_IA.jpg)

The SMPSs we typically encounter in motion picture lighting convert AC power to DC, and then use some type of switch to convert the DC back to AC of a higher frequency (20-50k Hz in the case of flicker-free fluorescents) or different waveform (the square wave of flicker-free HMI ballasts.) Regardless of the end result, all SMPSs use a diode-capacitor front end to convert the supply AC to DC.

(http://www.screenlightandgrip.com/images/generators/Pro_Sound_PP_SMPS_Kino_Ballast_Sch.jpg)

As can be seen in these power point slides from a workshop I developed for the members of New England Studio Mechanics Union (IATSE Local 481) on power quality, the diode-capacitor section converts the AC power to DC power by first feeding the AC input current through a bridge rectifier, which inverts the negative half of the AC sine wave and makes it positive. The rectified current then passes into a conditioning capacitor that removes the 60 Hz rise and fall and flattens out the voltage - making it essentially DC. The DC is then fed from the conditioning capacitor to the Switch-mode converter, which in the case of fluorescent ballast, is a high frequency inverter that utilizes a pair of MOSFETs to generate the high frequency (20-50kHZ) AC power.

(http://www.screenlightandgrip.com/images/generators/Pro_Sound_PP_SMPS_Diode-Cap.jpg)

To obtain a low ripple on the DC output of its’ front-end diode-capacitor section, the smoothing capacitor of a SMPS is typically quite large. Since the smoothing capacitor of a SMPS can only charge when the input voltage is greater than its stored voltage, as illustrated above the instantaneous line voltage is below the voltage on the capacitor most of the time.  Consequently, as illustrated below, the time the capacitor has to charge is only a very brief period of the overall cycle time while voltage ascends to its’ peak.

(http://www.screenlightandgrip.com/images/generators/Pro_Sound_PP_SMPS_Pulse_I.jpg)

That is because after peaking, the half cycle from the bridge drops below the capacitor voltage, which back biases the bridge, inhibiting further current flow into the capacitor. Thus the rectifiers conduct current only for a small portion of each line half-cycle and only during the ascending portion of the supply voltage waveform - which pulls the current out of phase with the voltage.  And since, during this very brief charging period, the capacitor must also charge fully as can be seen in this illustration above, large pulses of current are drawn for short durations. The end result is that the current drawn from the mains is a series of narrow pulses that peak before the voltage peaks and whose amplitude is 5-10 times higher than the resulting DC value. As such, SMPSs are a very inefficient power supply.

(http://www.screenlightandgrip.com/images/generators/Pro_Sound_PP_SMPS_Kino_vs_300.jpg)

The inefficiency of SMPSs is apparent if we contrast the 300W incandescent Fresnel light in the slide above to a 4’-4 Bank Kino Flo. The 4’-4 Bank Kino Flo uses four 75W T-12 tubes for a total wattage of 300W (same as the 300W Fresnel. As the slide below indicates the current drawn by the 300 Fresnel on a power meter is only 2.44 amps and it’s nice sinusoidal waveform has a crest factor of 1.3.  If we look at the current drawn by the 4’-4 Bank Kino we see it draws 4.12 amps and draws a distorted waveform with a much higher crest factor of 2.5. Because it draws current in short bursts to refresh its’ smoothing capacitor, it must draw more current (4.12A) for the same wattage (300) as the 300 Fresnel (2.44A.)

(http://www.screenlightandgrip.com/images/generators/Pro_Sound_PP_SMPS_Kino_vs_300_WF.jpg)

One measure of the inefficiency of SMPSs is “power factor.” To explain power factor, let’s take a second look at our 300W incandescent and 300W Fluorescent Lights. The significant difference in the amplitude of the current drawn by them can be attributed to the 4’-4 Bank’s relatively ineffective means of using power. Because it must charge its smoothing capacitor in short bursts, it draws significantly more current for roughly the same wattage as the 300 Fresnel. Because this difference in the amount of current drawn does not contribute to an increase in work (wattage) it is called reactive power. There is then two components to the energy expended by the 4’-4 Bank: there is what we call “real power” or that which generates work (measured in kilowatts or kW) and the reactive power or that which does not contribute to the work (measured in kilovoltamperes or kVA.) The sum of the real power and reactive power is called the “apparent power” (also measured in kVA.)

(http://www.screenlightandgrip.com/images/generators/Pro_Sound_PP_SMPS_Beer_Analogy.jpg)

Power factor is the ratio between real power (in kilowatts) and apparent power (in kilovoltampere). The favorite analogy film electricians like to use to explain these terms is that if apparent power is a glass of beer, reactive power is the foam that prevents you from filling it up all the way, so that you are left with less beer or real power. In other words, the thirst-quenching portion of your beer is represented by kW in this illustration. The foam is represented by kVAR. The total contents of your mug, kVA, is this summation of kW (the beer) and kVAR (the foam). In our beer mug analogy, power factor (pf) is then the ratio of Beer (real power) to the entire volume of the mug (beer plus foam or apparent power.) In the case of the 4’-4 Bank Kino, its’ apparent power of 490W (kVA = 4.12A x 119V= 490.28W) is the actual power expended by the fixture, the load that the generator and cable must support.

(http://www.screenlightandgrip.com/images/generators/Pro_Sound_PP_SMPS_PF.jpg)

To determine that this is in fact  the case with our respective 300W loads, let’s now take power factor readings for each. In the case of our two 300W light sources, the 300 Fresnel would considered to be more “efficient” because its kW (.29) and kVA (.29) are equal and so the power factor is 1.0 (the more efficient a source the closer the apparent power will be to the real power and the closer the power factor will be to  1.0.) Where the 4’-4 Bank Kino takes 4.12 Amps at 119 Volts it has an apparent power of roughly 480W, to generate 310 Watts of light (KW), our 4’-4 Bank Kino ballast wastes roughly 35% of the power that it uses in charging its’ smoothing capacitor and has a pf of .65.

(http://www.screenlightandgrip.com/images/generators/Pro_Sound_PP_SMPS_Harmonic.jpg)

How do we account for this wasted power? Where does it go? To answer that question let’s look at the harmonic currents generated by each load. When compared to a linear load like a 300W Fresnel, that generates virtually on harmonic distortion of the current waveform, a non-linear load like the 4’-4 bank Kino generates considerable harmonic components that distort the waveform. The harmonic components of the current drawn by the 4’- 4 Bank Kino produce no useful work and therefore are reactive in nature, but none-the-less are part of the load that the generator and cable must support making them tougher on power supplies than the 300W Fresnel light. These power quality issues have been vexing film electricians for years, to learn more about how we have learned to remediate the adverse effects of harmonics, read a white I have written on the use of portable generators in motion picture production available at http://www.screenlightandgrip.com/html/emailnewsletter_generators.html (http://www.screenlightandgrip.com/html/emailnewsletter_generators.html).

Guy Holt, Gaffer
ScreenLight & Grip
[email protected]
Title: Re: Switch Mode Power Supplies
Post by: Guy Holt, Gaffer on August 04, 2014, 01:49:24 PM
A bit of inductance in series with the incoming supply might sort it out too.
A bit of inductance in series with the incoming supply might sort it out too.[/quote]

Since, the smoothing capacitors of SMPSs draw current only at the peak of the voltage waveform, distorting it from a sine wave, simply adding linear components, such as inductors, will not counteract their capacitive reactance. To bring current back in phase with voltage, requires that the ballast’s smoothing capacitors draw current throughout the AC cycle rather than just a brief portion of it. A Power Factor Correction (PFC) circuit accomplishes this task by boosting the full-wave rectified supply when it is lower than the voltage on the storage capacitor so that the capacitor charges throughout the AC cycle.

PFC circuits are quite complicated. But, as can be seen in the illustration below, in the case of Kino ballasts consist of an Inductor, MOSFET's, and a Control Chip inserted between the bridge rectifier and smoothing capacitor (the low value capacitor at the front end of the circuit is there only to stabilize the power to the PFC circuit). The basic concept is fairly simple. The control circuit switches the MOSFET to draw current through the inductor to fill in the gaps in the current waveform that would otherwise create the harmonics.

(http://www.screenlightandgrip.com/images/generators/Pro_Sound_PP_SMPS_HMI_PFC_Circuit.jpg)

To accomplish this the PFC control chip constantly adjusts the on time of the MOSFET's as the input voltage changes, such that the inductor stores the same energy regardless of the instantaneous input voltage. The net result is that the current into the large smoothing capacitor approximates a sine wave rather than abrupt pulses. To accomplish this task, the PFC chip must track the input waveform in real time, making adjustments for both the input voltage and load current. As illustrated below the end result is that the AC current waveform will be Sinusoidal, with low distortion and perfectly in-phase with the applied voltage. Now that the capacitor(s) charge throughout the AC cycle rather than just a brief portion of it, the peak current is reduced and harmonic currents are not generated.
 
(http://www.screenlightandgrip.com/images/generators/Pro_Sound_PP_SMPS_PFC_Results)

The waveforms shown above for VC, IL, VE and IE correspond to the points marked on the overview circuit.
Note that the inductor current is pulse-width-modulated to simulate a sine-wave input current.

To see how effective power factor correction circuitry can be, let’s compare a power factor corrected Kino, the new 4’ Tegra 400, with a non-power factor corrected equivalent, the classic 4’ 4 Bank Kino.

(http://www.screenlightandgrip.com/images/generators/Pro_Sound_PP_SMPS_Kino_Options.jpg)

As can be seen in the comparisons of the current and voltage waveforms of the ballasts above, without power factor correction (bottom) and with power factor correction (top), PFC circuits can substantially increase power factor (to as much as .98), making ballasts with it near linear loads. As a result, the ballast uses power more efficiently with minimized return current and line noise and also reduces heat, thereby increasing their reliability.

Guy Holt, Gaffer
ScreenLight & Grip
[email protected]
Title: Re: Switch Mode Power Supplies
Post by: John Roberts {JR} on August 04, 2014, 03:44:04 PM
Nice pictures but an overly complex explanation for a simple question. The OP was asking why switch mode supply hanging directly on the mains was harder than conventional power supply going through a transformer.

The conventional non-PFC supply will have the same cap charging issues as you describe, but with the impedance of a transformer primary and secondary in series to limit the peak current.

Some high power amplifiers use PFC and many consumer products are mandated to be PFC to reduce wire losses in mains distribution infrastructure. Audio power amps have pretty much dodged that regulatory bullet so far.

JR

[edit- for an amusing related tidbit about lighting... the CFL lamps typically hang a diode and cap on the mains so in practice would load the mains with larger peak current than incandescent lamps, but the regulators allowed this because they use so much less total current.  /edit]
Title: Re: Switch Mode Power Supplies
Post by: Stephen Swaffer on August 04, 2014, 05:57:26 PM
I need to read and study a bit, but part of what makes this odd to me is that in the industrial world where I "grew up" as an electrician, our power factor issues were caused by too many inductive loads, thus we would add capacitance across the incoming lines, switched in and out as needed, to correct the power factor.

It just seems that in if you have a bunch of inductive loads-ac compressors, etc the capacitance would not be all bad.  I do understand that a pure capacitive load would act differently than one that is supplying a dc load.

Another interesting tidbit about CFLs is that their actual energy usage was higher than nameplate watts.  Marketing hype ? Government sanctioned to justify the regs?
Title: Re: Switch Mode Power Supplies
Post by: Frank Koenig on August 04, 2014, 07:17:18 PM
I need to read and study a bit, but part of what makes this odd to me is that in the industrial world where I "grew up" as an electrician, our power factor issues were caused by too many inductive loads, thus we would add capacitance across the incoming lines, switched in and out as needed, to correct the power factor.

It just seems that in if you have a bunch of inductive loads-ac compressors, etc the capacitance would not be all bad.  I do understand that a pure capacitive load would act differently than one that is supplying a dc load.

Another interesting tidbit about CFLs is that their actual energy usage was higher than nameplate watts.  Marketing hype ? Government sanctioned to justify the regs?

The confusion here results from the distinction between linear and nonlinear loads.

Power factor is  defined as the ratio of average (true) power to apparent power. In the case of linear loads (that can be modeled as an ideal resistor in combination with an ideal inductor or capacitor and in which load current is sinusoidal) the power factor is simply the cosine of the phase difference between the voltage and the current. In the linear case the power factor can be "corrected" to unity by adding an external capacitor or inductor that resonates with the inductance or capacitance of the load, respectively, to make the load appear as a pure resistance, thereby bringing the current in phase with the voltage.

In the nonlinear case, such as the capacitor-input-filter rectifier discussed here, the math is more complicated but the above definition of power factor persists.

The topology of the correction circuit in Guy's post can be recognized as a (voltage) boost convertor.

A good, if somewhat dated, reference for all this stuff is "Power Electronics; Convertors, Applications, and Designs" by Mohan, Undeland, and Robbins (Wiley 1989, 1995). There may be a newer edition.

--Frank
Title: Re: Switch Mode Power Supplies
Post by: Jerome Malsack on August 05, 2014, 07:50:14 PM
Ok and from another question the capacitor holds and releases a charge to keep things smooth.  That charge goes both directions on the circuit and back to the diodes and creates a reverse bias on the diodes pushing noise back at the mains.   ??   Would it not also push some DC onto the AC mains ??
Title: Re: Switch Mode Power Supplies
Post by: John Roberts {JR} on August 05, 2014, 09:11:14 PM
Ok and from another question the capacitor holds and releases a charge to keep things smooth.  That charge goes both directions on the circuit and back to the diodes and creates a reverse bias on the diodes pushing noise back at the mains.   ??   Would it not also push some DC onto the AC mains ??

No, by definition diodes only conduct current in one direction. A diode with a reverse bias is just a high impedance, like an open circuit, within normal operating voltages. Caps in power supplies after rectifier diodes are often called "reservoir" caps since they store up and release charge to the power supply between charging pulses.

Note: an asymmetrical load that pulls more current in one direction than the other does in fact presents an effective DC term back to the mains.  I recall old light bulb life extenders that were simple diodes in the base of the lamp socket so the lamp only drew current for one half of the wave form.  One lamp no problem, lots of lamps a possible DC problem, unless you carefully alternated the direction of each of them, so 1/2 blocked one polarity and half the other.

JR
Title: Re: Switch Mode Power Supplies
Post by: Stephen Swaffer on August 05, 2014, 10:45:22 PM
IIRC, it was fairly common for early tube equipment to use half wave rectifiers in their power supplies due to the cost of tubes.  Full wave required 2 rectifier tubes plus a costly center tapped transformer.  Full wave bridge rectifiers did not become common until solid state diodes become common.

Probably the strangest tech support call I ever placed was in the early 2000's.  We had a magnetic clamp on a surface grinder that had quit working-and it had a vacuum tube rectifier.  I was fairly certain the rectifier tube was bad-but figured I would never get any help with it.  The first tech I spoke with knew the problem and the part number and I had a replacement on the way in under 5 minutes on the phone.  In a day when 5 years old is obsolete, the 40 yr old tube stuff was a simple fix, go figure.

Would active PFC be more economical to properly implement in one large supply vs a number of small ones?
Title: Re: Switch Mode Power Supplies
Post by: Steve M Smith on August 06, 2014, 03:48:40 AM
I recall old light bulb life extenders that were simple diodes in the base of the lamp socket so the lamp only drew current for one half of the wave form.

Many US voltage photographic enlargers are wired like this.  I had one which I ran via a 240 - 120v transformer.  They are fitted with a bulb rated at 82 volts.

I have no idea why they didn't just fit a 120 volt bulb and do away with the diode.


Steve.
Title: Re: Switch Mode Power Supplies
Post by: John Roberts {JR} on August 06, 2014, 10:42:19 AM


Would active PFC be more economical to properly implement in one large supply vs a number of small ones?

Interesting question... Probably not, but PFC is not cheap so not widely used for audio amps. The Euro zone government was talking about making it mandatory so most major amp makers developed the technology, and even one or two amp models were introduced but AFAIK this is still not mandatory. 

I guess we could make an AC to AC PFC correction unit, that works kind of after the fact so it puts out a low impedance AC waveform, then spreads out that current draw proportionately over the input sine wave.  The output wave form doesn't even need to be a perfect sine wave, like a cheap inverter, but still not trivial or cheap. 

JR
Title: Re: Switch Mode Power Supplies
Post by: Mike Karseboom on August 07, 2014, 11:34:29 AM
This discussion is interesting to me because I have 8 QSC PLX (1) series amps which are, I believe, switching amps with no PFC.  I have run these off a variety of power sources including generators.  I have tried to reconcile some of the general statements I read/hear about these type amps with my personal experience.  The clarification of what happens to the current wave form seems to help explain some of the general statements.


Statement:  You need a "stiff" power supply to get the most out of the PLX amp. 


Well that is no doubt true for all amps but I have definitely noticed a subjective improvement in sound quality when I can dedicate a 20A circuit to a maximum of 2 amps.  Subjectively the sound is just more "present", "spacious", and "punchy" at the same time.  I also have some older RMX 2450's and some even older Crown Powerbase 2 amps and have not detected the same "improvement" when the power supply was very robust.


Statement:   PLX amps suck on subs,  old iron amps work better


I suppose this relates back to the all too typical situation where the power supply is less than optimum. 


Statement:  PLX amps "sip" power compared to old iron amps


The arguments put forth with the kino light vs. fresnel light seem to indicate that switch mode amps are less efficient and draw more peak amperage and waste a lot of power.  So how is it that switch mode amps get the reputation for not needing as much power?



Title: Re: Switch Mode Power Supplies
Post by: Steve M Smith on August 07, 2014, 11:47:26 AM
With enough capacitance in the output, it shouldn't really matter if a power supply is linear or switch mode.  I suspect the perceived differences are due to manufacturers using the smallest capacitors they can get away with when designing switch mode supplies.


Steve.
Title: Re: Switch Mode Power Supplies
Post by: John Roberts {JR} on August 07, 2014, 01:03:12 PM

Statement:  PLX amps "sip" power compared to old iron amps
Kind of depend on how old that "old iron" is... CS1200 old, or RMX2450 old? The modern amps like 2450 are more efficient so deliver more power to speakers for the same line cord draw.
Quote

The arguments put forth with the kino light vs. fresnel light seem to indicate that switch mode amps are less efficient and draw more peak amperage and waste a lot of power.  So how is it that switch mode amps get the reputation for not needing as much power?
There seems to be some confusion regarding "peak current" and "efficiency"... Efficiency is about the average power output vs average power consumed. Power supply topologies like PFC that smooth out the current draw over the entire waveform cause less peak IxR voltage drop, in the mains wiring, so actually allow the amp to pull more voltage and power from a given mains distribution. This "lost" power difference is not literally wasted or dissipated in the mains wiring but is more of a missed opportunity. The power that an amplifier can put out is limited by the internal DC power supply rail voltages. The voltage of these internal rails is a simple function of the Peak mains voltage. With high peak current draw concentrated at the top of the mains power sine wave, IxR voltage drop at the line cord limits how much voltage the internal PS rails get charged up to.   

I can imagine a rack full of power amps putting a hurt on the top of sine waves, with or without switching supplies, while the raw diode cap front end of many switchers will try to draw the same current over a narrower window, so higher peak current for the same average current compared to big transformer amps.

JR

PS: One of my old amp ideas, that I never developed due to cost/complexity was to perform a variation on the classG/H multi-rail power amps where instead of using separate transformer windings for each rail voltage, each charging their respective PS caps only at the top of the mains sine wave, use a high power switch to connect and disconnect from the same transformer winding, before and/or after the peak voltage to charge the lower voltage amp rails. The high voltage rail still draws current from the top of the sine wave waveform, but the other rails draw their power away from the sine wave peak. This approach would be like a poor man's PFC and spread the current draw away from the top of the sine wave so effectively reduce the IxR voltage drop that limits max power.  I did not pursue this because in my estimation it would not be clean enough to qualify as true PFC to meet european harmonics regulations, so relatively expensive for the partial benefit it could deliver.  Note: This switch disconnect could even vary the actual voltage of the lower rails to optimize amp efficiency for real world audio signals.

PPS: I have other unexplored amp ideas that are even more bizarre but academic at this point with class D now cost effective. All we need now is cost effective PFC.