Hi, I'm a noob at this and would love some help.
I'm trying to figure out what power supply requirements I need for a class D amp per channel for 50W rms driving an 8 ohm load.
I am trying to follow along this article on class D power requirements:
EE Times - Understanding Class-D amplifier power supply requirements
They say:
And later:
"the power supply components must rated for the maximum output current, which is a function of the amplifier's peak output power and not its continuous output power capability."
I am confused by a few things:
Thank you very much.
Weiling
I'm trying to figure out what power supply requirements I need for a class D amp per channel for 50W rms driving an 8 ohm load.
I am trying to follow along this article on class D power requirements:
EE Times - Understanding Class-D amplifier power supply requirements
They say:
Code:
Maximum average (RMS) amplifier output power PBTL(RMS) = 50W
Amplifier load impedance RL = 8Ω
Total output resistance RT = 8.3Ω
Power supply output voltage tolerance = ±5%
Maximum amplifier duty cycle MMAX = 88%
Maximum amplifier efficiency ηMAX = 93%
The minimum power supply voltage is
VCC(MIN) = √[(2 • PBTL(RMS) • RT2) / RL] / MMAX
= √[(2 • 50W • 8.3Ω2) / 8Ω] / 88% = 33.34V
The nominal supply voltage is
VCC(NOM) = VCC(MIN) / (1 - tolerance)
= 33.34V / 0.95 = 35.1V
The power supply output current limit must be greater than or equal to
ILIMIT(MIN) ≥ (VCC(MIN) • MMAX) / RT
= (33.34V • 88%) / 8.3Ω = 3.72A
The power supply must be able to deliver at least
PPSU = PO(MAX) / ηAMP(MAX)
= 50W / 93% = 53.8WAnd later:
"the power supply components must rated for the maximum output current, which is a function of the amplifier's peak output power and not its continuous output power capability."
I am confused by a few things:
- They show VCC(MIN) of 35.1V and current of at least 3.72A. Isn't that a 130W PSU? How is that consistent with the PPSU of 53.8W?
- The second comment says I need a max. output current as a function of peak power not RMS. Does that mean 3.72A * 2 = 7.44A? Or are 35.1V and 3.72A already peak values (as shown in the chart in the article)?
- I cannot seem to find the "max duty cycle" on datasheets of most amplifiers. Is 88% a reasonable approximate value?
- I guess to put it simply, what ratings should I be looking for on a PSU? For example, can someone point me to a Meanwell PSU (Switching Power Supply - Mean Well Switching Power Supply Manufacturer) that would meet the requirements of the single channel amp in this example?
Thank you very much.
Weiling
The VCC and current limit represent peak voltage and peak current. Power is calculated from the rms values. This means 130W/2=65W. Then, you have to factor in 5% tolerance on VCC, and max duty cycle of 0.88. 35.1Vx0.95x0.88 leaves an effective supply of 29.3V. (29.3Vx3.72A)/2 is how you calculate rms power from peak voltage numbers, to get 54W.
Yes 35.1V and 3.72A are peak numbers, because the supply rail and current limit are the max voltage and current the amp can deliver.
Max duty cycle: 88% is a little lower than typical. I think they may mean max duty - min duty. Or 94%-6%. This must be an open loop amp, because closed loop can effectively go 100%. That doesn't affect efficiency, just the VCC value needed.
To order a supply, it needs DC VOLTAGE>35V and DC CURRENT > 3.72A. Which is what the article says.
Yes 35.1V and 3.72A are peak numbers, because the supply rail and current limit are the max voltage and current the amp can deliver.
Max duty cycle: 88% is a little lower than typical. I think they may mean max duty - min duty. Or 94%-6%. This must be an open loop amp, because closed loop can effectively go 100%. That doesn't affect efficiency, just the VCC value needed.
To order a supply, it needs DC VOLTAGE>35V and DC CURRENT > 3.72A. Which is what the article says.
Determining the Average Power Requirements ( from your link)
Consumers almost never need 40% power on a continuous basis. Most people operate their amplifiers well below the maximum rated power, with the outputs only approaching full power for short periods of time during peaks in the music. This difference between maximum available power and typical usage is the reason behind the power rating requirements of the various regulatory agencies.
Power supplies can be designed in different ways depending on the needs of their target market. Amplifiers for applications such as professional recording studios or laboratory applications may need to be able to support the full output power on a continuous basis. In this case the power supply would also have to be able to deliver full output power continuously. This approach is expensive and is normally only used when the application requires it.
Hi
so if you design using laboratory grade equations Pmax /(eff. class D) you will be fine for music.
you didn't say what amplifier? you get a better #s if you include that.
gathering accurate Pmax and eff. I would just use the amplifiers spec sheets to determine the max supply voltage, from that you determine the average continuous output power into a real speaker load ( Pmax ) this mostly depends on the heat sinking and ambient temp for continuous duty. (I usually use about 6 ohms per side). Also eff curves can be obtained more accurately using any curves from device OEMs for these operating conditions. ranges from ~70- 90% depending on load and power!
noted in the article> Most amplifier instantaneous current peaks come from supply decoupling at the amplifier device.
Consumers almost never need 40% power on a continuous basis. Most people operate their amplifiers well below the maximum rated power, with the outputs only approaching full power for short periods of time during peaks in the music. This difference between maximum available power and typical usage is the reason behind the power rating requirements of the various regulatory agencies.
Power supplies can be designed in different ways depending on the needs of their target market. Amplifiers for applications such as professional recording studios or laboratory applications may need to be able to support the full output power on a continuous basis. In this case the power supply would also have to be able to deliver full output power continuously. This approach is expensive and is normally only used when the application requires it.
Hi
so if you design using laboratory grade equations Pmax /(eff. class D) you will be fine for music.
you didn't say what amplifier? you get a better #s if you include that.
gathering accurate Pmax and eff. I would just use the amplifiers spec sheets to determine the max supply voltage, from that you determine the average continuous output power into a real speaker load ( Pmax ) this mostly depends on the heat sinking and ambient temp for continuous duty. (I usually use about 6 ohms per side). Also eff curves can be obtained more accurately using any curves from device OEMs for these operating conditions. ranges from ~70- 90% depending on load and power!
noted in the article> Most amplifier instantaneous current peaks come from supply decoupling at the amplifier device.
That example is sort of rubbish, STM product only? Maximum amplifier efficiency ηMAX = 93%
I think their marketing dept made them hold firm on the 93% figure (Bullcrap)
Maximum amplifier duty cycle MMAX = 88% never seen that before.
they never define this term so this is a huge fudge factor also Rds ~ 8.3/8.0 is pulled out from a hat .
I would drill down for a better Number for efficiency and just use worst case at Pmax!
I think their marketing dept made them hold firm on the 93% figure (Bullcrap)
Maximum amplifier duty cycle MMAX = 88% never seen that before.
they never define this term so this is a huge fudge factor also Rds ~ 8.3/8.0 is pulled out from a hat .
I would drill down for a better Number for efficiency and just use worst case at Pmax!
Last edited:
Also, max duty cycle numbers well below 100% are very common in open loop class D. Close loop has max duty, too, but are able to jump from, say, 95% to 100% when clipping. The error of the 5% step averages out over several switching cycles. Some architectures won't allow 100%, because the bootstrap caps need to be recharged.
No architecture likes to run just below 100% or just above zero, because the FETs don't have time to fully turn off or on, and it hurts efficiency.
No architecture likes to run just below 100% or just above zero, because the FETs don't have time to fully turn off or on, and it hurts efficiency.
In other words, that article is legit, even if they're mostly trying to sell ST amplifiers. You expect them to use a TP3116 as their example?
The maths are legit, but is it really necessary?
Even for non switching amps we oversize the capacitors and add bypass caps here and there because
1) PSU output spec depends quite a bit on the capacitance; or more precisely, output impedance and ripple under load
2) the previous 2 things mentioned above are frequency dependent
So say your PSU is beefy enough for treble, but does it have low enough impedance for bassssssssss? For example.
Overkill the psu, you would still get more performance, just diminishing returns.
Underhill the psu, reduced performance and distortion/shutdown when limit reached.
Even for non switching amps we oversize the capacitors and add bypass caps here and there because
1) PSU output spec depends quite a bit on the capacitance; or more precisely, output impedance and ripple under load
2) the previous 2 things mentioned above are frequency dependent
So say your PSU is beefy enough for treble, but does it have low enough impedance for bassssssssss? For example.
Overkill the psu, you would still get more performance, just diminishing returns.
Underhill the psu, reduced performance and distortion/shutdown when limit reached.
No not for consumers adding a good external V regulated supply like meanwells. Not those PC notebook adapter thingys which are designed for a SMPS batt. charger.
maybe if designing your own amps, but then those users realize that article makes some assumptions and leaves some other things out.
95-100 is ridiculous for any hiFi class D power conversion >50W, even the best BTL designs mounted on a real PCB without EMI filters included, show 88-90% with a single tone sine wave. The eff. of class D on loud listening levels, are between 70-85% on average which borders Class AB at its max power.
negative feedback goes open loop when anything reaches the rails, so why would this be different than an open loop design. I worked in power electronics in my early career in SMPS design so loop control theory is in my lingo,
Last edited:
Duty cycle not eff.
Infinia,
You're talking efficiency. I was talking about duty cycle. Class D IC designers don't like to partially turn on the output switches, so usually between 0 to 1% and 99 to 100% are "no go" zones for duty cycle. The duty cycle can be 0 or 100, but not 0.5 or 99.5. If there is concern about bootstrap discharge when clipping low frequency audio, then duty cycle might be limited to 97 or 98% to make sure bootstrap caps don't discharge. Nothing ruins your day like the high-side drive losing its supply voltage.
If you're talking about efficiency for power supply sizing, efficiency at max power is the only efficiency that matters. If you're curious what the average dissipation will be, then average efficiency is relevant. If you need maximum dissipation, that's at Max power. Unlike AB amps which get max dissipation near 1/2 power.
I know what I'm talking about. I designed several class D ICs at TI. I designed one amp for flat panel TV that had 94% efficiency. The TV maker did not want a hot spot behind the screen where the amp IC goes. That one had max duty cycle of 97%.
Infinia,
You're talking efficiency. I was talking about duty cycle. Class D IC designers don't like to partially turn on the output switches, so usually between 0 to 1% and 99 to 100% are "no go" zones for duty cycle. The duty cycle can be 0 or 100, but not 0.5 or 99.5. If there is concern about bootstrap discharge when clipping low frequency audio, then duty cycle might be limited to 97 or 98% to make sure bootstrap caps don't discharge. Nothing ruins your day like the high-side drive losing its supply voltage.
If you're talking about efficiency for power supply sizing, efficiency at max power is the only efficiency that matters. If you're curious what the average dissipation will be, then average efficiency is relevant. If you need maximum dissipation, that's at Max power. Unlike AB amps which get max dissipation near 1/2 power.
I know what I'm talking about. I designed several class D ICs at TI. I designed one amp for flat panel TV that had 94% efficiency. The TV maker did not want a hot spot behind the screen where the amp IC goes. That one had max duty cycle of 97%.
so what does duty cycle have to do with sizing a power supply?
to me that sounds sort of like a internal mechanism (Band-Aid) for class D topologies eg cross conduction woes. Shouldnt the DC settings be invisible to the normal class D amp user?
to me that sounds sort of like a internal mechanism (Band-Aid) for class D topologies eg cross conduction woes. Shouldnt the DC settings be invisible to the normal class D amp user?
Last edited:
If a class D amp has a min/max duty cycle of 3%/97%, for instance, then the effective VDD is multiplied by 0.94 by just that fact alone, before efficiency is even considered. It's not loss, because of the nature of class D, it's just a limit on voltage swing.
Similar to an A/B amp where you can't get within a couple of diodes of the supply, it limits the voltage swing on the output. But not similar, because it's not a loss of (2 VD* Iload) in a class D. Loss is still determined by R(on) of the MOS switches and metallization.
So, why would they do that? Give up voltage swing? Weird. Two reasons: bootstrap recharge and eliminating unnecessary power loss.
In an all-NMOS design, the bootstrap capacitor is the supply rails for the high side gate drivers. It gets recharged through the bootstrap diode only when OUT is low. If duty cycle is at or near 100%, the bootstrap cap goes a long time without getting fresh electrons. Some high-side gate-drive circuits draw enough current to discharge the bootstrap cap in a few milliseconds. Those require a max duty cycle limit to allow some bootstrap refresh every cycle. Some don't really draw any current, and can go 100% duty cycle for several seconds without significant decay. Those require no duty cycle limit. I've worked with both.
Both types do not want to send a half-assed partial turn-on pulse in the case of almost 0% or almost 100% duty cycle. This doesn't have to do with cross-conduction, but it does have to do with Qrr of the body diode of the output MOSFETs. If you can't fully turn the device on, then don't bother throwing energy at the parasitics, and unnecessarily polluting the substrate.
Similar to an A/B amp where you can't get within a couple of diodes of the supply, it limits the voltage swing on the output. But not similar, because it's not a loss of (2 VD* Iload) in a class D. Loss is still determined by R(on) of the MOS switches and metallization.
So, why would they do that? Give up voltage swing? Weird. Two reasons: bootstrap recharge and eliminating unnecessary power loss.
In an all-NMOS design, the bootstrap capacitor is the supply rails for the high side gate drivers. It gets recharged through the bootstrap diode only when OUT is low. If duty cycle is at or near 100%, the bootstrap cap goes a long time without getting fresh electrons. Some high-side gate-drive circuits draw enough current to discharge the bootstrap cap in a few milliseconds. Those require a max duty cycle limit to allow some bootstrap refresh every cycle. Some don't really draw any current, and can go 100% duty cycle for several seconds without significant decay. Those require no duty cycle limit. I've worked with both.
Both types do not want to send a half-assed partial turn-on pulse in the case of almost 0% or almost 100% duty cycle. This doesn't have to do with cross-conduction, but it does have to do with Qrr of the body diode of the output MOSFETs. If you can't fully turn the device on, then don't bother throwing energy at the parasitics, and unnecessarily polluting the substrate.
.
IDK thinking out loud ...maybe to keep power limits within the end users lowered range given poor heat sinking and/or temperature extremes. ALSO operating right at the supply rail limits, the MOSFET body diode can be overwhelmed when the amps damping disappears due to lowered feedback in large signal overload given the back EMF from speakers. motors ..to keep things in semi linear mode as it were.
RussellKinder, infinia, thank you both very much. I'm going to have to dumb down this discussion though... 🙂
infinia, I'm looking at TI digital closed loop amps like the TAS5614LA.
Going back to the first response, here's where I get confused: if I go with 40% * 54W = 22W, am I looking for reduced voltage or reduced current vs "DC VOLTAGE > 35V and DC CURRENT > 3.72A"?
I guess I want 35V and 22W? And make sure my power supply doesn't current limit so I can get bigger current draws for short bursts? Is this tolerance the same as "overloading" that I see in Meanwell specs?
So, for instance, putting aside power and voltage for a moment, here are two power bricks:
WebData
WebData
The former (GC120) is a charger which mentions constant current limiting. The latter (GC120) does not mention current limiting and has a much wider overloading range.
Is the latter then a suitable type of power supply?
infinia, I'm looking at TI digital closed loop amps like the TAS5614LA.
Going back to the first response, here's where I get confused: if I go with 40% * 54W = 22W, am I looking for reduced voltage or reduced current vs "DC VOLTAGE > 35V and DC CURRENT > 3.72A"?
I guess I want 35V and 22W? And make sure my power supply doesn't current limit so I can get bigger current draws for short bursts? Is this tolerance the same as "overloading" that I see in Meanwell specs?
So, for instance, putting aside power and voltage for a moment, here are two power bricks:
WebData
WebData
The former (GC120) is a charger which mentions constant current limiting. The latter (GC120) does not mention current limiting and has a much wider overloading range.
Is the latter then a suitable type of power supply?
1) are you planning on building a PCB or buying one?, speaker configuration / loading , packaging, cooling and heatsinking all make a difference for power expectations. determine from here http://www.ti.com/lit/ds/symlink/tas5614la.pdf look at Fig. 13 > looks like the chip wants at least 2 supply voltages. guessing 12V for digital, 12V for analog inputs, and ~35V for outputs.
I would go for a U frame style single output PS. edit > double output?
Those power bricks are pretty wonky near their limits, like I said earlier they are really designed to drive another voltage regulator for a battery charger.Tthey have decent line regulation but their outputs act unregulated with higher line ripple and noise. But, they could be useful depending on how much power you expect and the supply decoupling you need to add to support any current peaks. Some are worse than others, it's really hard to know unless you dig deeper. I recon sticking with 1st tier PC name brands will be better than the off name ones.
2) depending on what the end use is, the differential inputs and grounding can get tricky system wise. limiting the source to an external battery powered MP3 player yer good to go.
I would go for a U frame style single output PS. edit > double output?
Those power bricks are pretty wonky near their limits, like I said earlier they are really designed to drive another voltage regulator for a battery charger.Tthey have decent line regulation but their outputs act unregulated with higher line ripple and noise. But, they could be useful depending on how much power you expect and the supply decoupling you need to add to support any current peaks. Some are worse than others, it's really hard to know unless you dig deeper. I recon sticking with 1st tier PC name brands will be better than the off name ones.
2) depending on what the end use is, the differential inputs and grounding can get tricky system wise. limiting the source to an external battery powered MP3 player yer good to go.
Last edited:
so are you only figuring 1 of 4 outputs in SE mode?
GC is designed to connect direct to a lead acid batt. eg typical float voltage
IDK does the PCB yer getting have any requirements?
just look at the power eff. curves at VCC on the data sheet , using the speaker load you have, then figure PS rating =/> (Pmax x4)/ (efficiency at full load)
shop for an "off shelf SMPS", they come in big power steps . most current limit ~ 130% rated P.
if yer using 4 channels I doubt yer ever gonna hit the 100W current level. If you do you got the wrong amp, not a PS issue.
Last edited:
Infinia has the right idea trying to steer you away from using the PSU current max to set your power levels. In the long run, the PSU will probably die, and in the short run it will misbehave. At best, you will get very U graceful clipping.
35V will get you about 74 W/ch BTL into 8 ohms, if there were unlimited current. 3.72A will get you about 53W into 8 ohms, and that's just one channel. With 2 driven equally, you nly get 26.5W/ch. and your supply will be pushed to its limit.
If you only want 22W/ch, you need to reduce the PSU voltage to about 20V, but increase the current limit to about 5 A. In which case, the TAS5614 is way more amp than you need. And you have to feed it PWM input from a PWM DAC, which needs I2C control and I2S data from a USB connection. Why not use a TPA3116 instead? But still, use 20V, 5A for the supply.
35V will get you about 74 W/ch BTL into 8 ohms, if there were unlimited current. 3.72A will get you about 53W into 8 ohms, and that's just one channel. With 2 driven equally, you nly get 26.5W/ch. and your supply will be pushed to its limit.
If you only want 22W/ch, you need to reduce the PSU voltage to about 20V, but increase the current limit to about 5 A. In which case, the TAS5614 is way more amp than you need. And you have to feed it PWM input from a PWM DAC, which needs I2C control and I2S data from a USB connection. Why not use a TPA3116 instead? But still, use 20V, 5A for the supply.
Remember too, the output of the amp is rarely just one frequency at full power but a complex combination of many frequencies at much lower amplitudes all summed together to give a complex output voltage. The amp needs only be capable of creating the voltage and will not often hit the current calculated by voltage over speaker impedance since the actual current at any point in time is the sum of all the smaller voltages at different frequencies divided by the speaker impedance at that frequency. That being said, it's still a good idea to design to the theoretical maximum plus some headroom.
I should add that you can do SE x 2 into 8 ohms with 35v PSU, and get 19W/ch. You need 4.3A peak though, unless you can invert polarity of signal on one channel, so that each channel draws from the PSU 180 deg out of phased from the other. (Invert the speaker, too). That cuts the peak current requirement in half, but only for "normal" music where the bass is mostly in phase.
- Status
- Not open for further replies.