If you look at the dynamics of the current requirement for a high performance CPU you would see that it can't be done with a linear supply. Section 7-5 of this describes the requirements- https://www.intel.com/content/dam/w...s/4th-gen-core-i7-lga2011-datasheet-vol-1.pdf Like 135A and current transitions from small to max in very short times. Also not obvious is that the supply voltage changes depending on the mode of the processor.
A linear supply would dissipate 200W plus to support the 135W needed for this processor.
I have a vintage (1968) HP sound level meter with a low noise switching supply to power the microphone. No evidence of the supply switching in the output signal. It's not a new cutting edge technology. Just harder and with good magnetics more expensive to implement.
A linear supply would dissipate 200W plus to support the 135W needed for this processor.
I have a vintage (1968) HP sound level meter with a low noise switching supply to power the microphone. No evidence of the supply switching in the output signal. It's not a new cutting edge technology. Just harder and with good magnetics more expensive to implement.
Yeah, in a small box it may be harder to deal with the low frequency magnetic field from the transformer in a linear supply.
That's a good point a large number of famous hi-end recording mics use a switching supply right in the mic for capsule bias, that and the electrolytic coupling caps.
The sound was beautiful and it worked with no noise sources at all. Not these but another design won the triode fest contest.
I think we have a power definition vs practical issue and costs regarding smps for consumer level electronics... rather than what can be and is designed into expensive computers and ultra high volume pricing.
and a switching supply for microphone? Quit a steady low current load. a 555 timer, step-up xfmr. rect/filter could even work. And be quiet.
The Benchmark amp is closer to our DIY reality but the cost/price ($) is high for the power output compared to transformer/rect/filter PA of same pwr. We will see about long term reliability compared to a transformer etal. We bought several smps for my much higher power amp and they couldnt do it well enough.... and the cost is relatively higher. If one exists, i would want to pay for it. There is no way it will make the PA smaller with the number of OP devices and heat sink area needed etc. It was only a weight consideration we were looking for.
The idle to full maxed out running processor is very rarely encountered in consumer use..... unless you are doing some huge 3D matrix conversion in near real time display mode. I would doubt a cost cutting consumer level supply could sustain the voltage at full load. Even so, the Intel on i7 indicates that cap bank is needed to prevent sag at such max draw periods.
Just saying, it depends on the price and other factors beside Can it be done.
Sure it can be done and is done in high volume. For the DIY-er I dont think so. Why bother for the complexity involved. But if it can do all that is needed without complexity and high costs..... the weight savings alone would be great and shipping costs reduced.
THx-RNmarsh
and a switching supply for microphone? Quit a steady low current load. a 555 timer, step-up xfmr. rect/filter could even work. And be quiet.
The Benchmark amp is closer to our DIY reality but the cost/price ($) is high for the power output compared to transformer/rect/filter PA of same pwr. We will see about long term reliability compared to a transformer etal. We bought several smps for my much higher power amp and they couldnt do it well enough.... and the cost is relatively higher. If one exists, i would want to pay for it. There is no way it will make the PA smaller with the number of OP devices and heat sink area needed etc. It was only a weight consideration we were looking for.
The idle to full maxed out running processor is very rarely encountered in consumer use..... unless you are doing some huge 3D matrix conversion in near real time display mode. I would doubt a cost cutting consumer level supply could sustain the voltage at full load. Even so, the Intel on i7 indicates that cap bank is needed to prevent sag at such max draw periods.
Just saying, it depends on the price and other factors beside Can it be done.
Sure it can be done and is done in high volume. For the DIY-er I dont think so. Why bother for the complexity involved. But if it can do all that is needed without complexity and high costs..... the weight savings alone would be great and shipping costs reduced.
THx-RNmarsh
Last edited:
I think the underlying fallacy here is comparing a regulated (switch mode) supply to an unregulated linear supply. The regulated supply needs to be built to support worst case voltage and current (as in peak current). The unregulated supply does not need active components supporting the peak load and sag under load is provided for in the design. The CPU supply does need to support the max requirements of the CPU or the board will crash. Fortunately, there are thermal limits to protect against various overloads. The requirements for a power amp supply don't include mV of sag limits so its not the same challenge.
A supply for a big power amp would need to be substantially larger than the amp's rating making it a major undertaking. The benchmark uses some technology that integrates the supply requirements as part of the active system. which may get some benefits. I have seen similar efforts usually focused on high volume customers.
A supply for a big power amp would need to be substantially larger than the amp's rating making it a major undertaking. The benchmark uses some technology that integrates the supply requirements as part of the active system. which may get some benefits. I have seen similar efforts usually focused on high volume customers.
Actually the idle to full load happens in games and it happens dozens of times per day whenever I compile something large, for example. Full load without all AVX units anyway.
The cap bank is remarkably small. Most DIYers here use more capacitance in their preamps.
If you look at a teardown of a good 1200W PC power supply, there isn't anything exotic in it:
Seasonic Prime Platinum 1200W Review
It manages 12.11V @ 9A and 12.08V @ 100A on the +12V output btw...
The cost is in development, not the parts. I think high end audio manufacturers avoid SMPS at least in part because they do not know how to design good ones. It is obviously much more difficult than designing a linear supply. Then add in regulatory testing.
most of the mic switching supplies are to polarize the membrane in capacitance mics. Generally they are a single transistor oscillator and a single rectifier followed by a small capacitance (usually film or tant) The load on these is extremely small (leakage) and they are isolated from the membrane by as much a 1Gohm. Not hard to build a quiet step up supply for that.
Alan
Alan
I tried the models recommended here in DIY-land. Wasnt good enough. Also, too noisy for a amp <.001% thd. Cost is relatively and comparitively high. But, I have hopes someone will do it and I will buy it OEM and use it. I am open to recommendations.
[Something doesnt mesh on these PC supplies.... I see no pcb trace, wire gauge on cores, or I/O connector pins which would be rated anywhere near 100A continuous. ]
THx-RNMarsh
Last edited:
... Maybe because such SMPS are not directly involved with the processor. 12vdc is more like a rough prereg. On board VRM's are used to directly power the processor.
What is VRM (Voltage Regulator Module)?
YouTube
-RNM
What is VRM (Voltage Regulator Module)?
YouTube
-RNM
Last edited:
Which model and what amount of ripple at what frequency?
The Meanwell LRS-350-48 I use had under light load was 20-30 mV at 70 kHz (I'm going off foggy memories right now). It would NOT take much of a CLC to knock down the 70 kHz switching residual, although I didn't bother.
I have one class D amp with combined PSes in parallel. Linear PS helps with starting up, when loaded on big capacitances. Also, in case of peaks that SMPS can't handle it takes them. The same trick I use in a filament supply that otherwise can't start when tubes are cold.
They use many pins in parallel. The ripple is 20mVpp on the 12V though I am not sure the bandwidth. Of course they are pushing it with the wire gauge even with the 8 additional conductors over a standard 24 pin ATX connector. The tests show it meets spec.
Of course there are multi-phase buck regulators after. I mentioned this several posts ago with links to the Intersil parts.
- Status
- Not open for further replies.