Boutique brands, are they really worth it?

[Matthew J Poes]

Yes, I did see that. It's an attempt to correlate AVR weight with maximum power output using ACD as the metric. Correlating AVR price vs max Po/ACD would probably have as much merit. You're including product design and marketing constraints in that comparison.

The thing is, SMPS technology has advanced, and could match or surpass that of the heavy iron/copper analog supplies, but wouldn't work well in a high end AVR because of marketing issues and perceived value:weight ratio. It's not really a question of the technology of being up to the task, it's a question of what people think they want to pay for.

And there's still a question of using ACD as a valid test condition.

Ok, add a little less than 1% additional voltage drop for a fully loaded 15A circuit with 14ga wire. Doesn't change anything in reality, plenty of voltage and current available at the outlet.

I’ve worked with SMPS supplies in amps. At this point nearly all pro amps use them and produce insane amounts of power without issue. When I designed my last amp around an SMPS the advantages were huge. They can regulate the voltage very well and have Very low ripple and noise. The only problem I ran into (which is easily fixed in the engineering stage) is that they basically shut down once you hit their limit. Where as a linear supply has this gradual reduction in voltage as current increases at the limits (which is great for low impedance loads), an SMPS goes strong until it hits its limits and then goes into protection. With my amp, long before the amp clipped, the supply would go into protection. Eventually I damaged the supply and am currently debating what I want to do with the amp.

The amp was a Class AB capable of about 300 watts rms per channel at 8 ohms. Given a sufficient supply it would double that into 4ohms. It needed +/- 74 volt rails to do that into 8ohms and really worked better on 68 volt rails into 4ohms (based on the simulations). I tuned the SMPS to sit around 72 volts and it would regulate that up to about 8 amps per rail. It was roughly a 1200 watt supply. I thought that would be sufficient but was wrong. I still really like SMPS but found that a big learning experience. I would have used a 600va to 800 va transformer and maybe 40,000uf of capacitance if I had built a linear supply. 1200 watt SMPS seemed like a similar substitute. I probably needed twice that.

 

[dc2bluelight]

That sounds like a glitch in how the supply was spec'd. Perhaps their "1200 watts" was obtained by adding 600 watts per rail? You were close to that +/- 1dB fudge factor.

I found designing analog supplies to be a real pain, though. Hitting the magic combination between desired voltage, noise and regulation over a wide current range is really quite difficult, especially if you're using off-the-shelf transformers. Thats why we saw a lot of unregulated power supplies with just rectifiers and big filter caps in power amps over the years, and even then, the power transformer was the trick. You always had over-specified transformers, except perhaps for one design I know of that was based on "dynamic headroom".

 

[Matthew J Poes]

That sounds like a glitch in how the supply was spec'd. Perhaps their "1200 watts" was obtained by adding 600 watts per rail? You were close to that +/- 1dB fudge factor.

I found designing analog supplies to be a real pain, though. Hitting the magic combination between desired voltage, noise and regulation over a wide current range is really quite difficult, especially if you're using off-the-shelf transformers. Thats why we saw a lot of unregulated power supplies with just rectifiers and big filter caps in power amps over the years, and even then, the power transformer was the trick. You always had over-specified transformers, except perhaps for one design I know of that was based on "dynamic headroom".

I’ve wondered if it was mis-specified. Anthony Holten of Aussieamps had suggested it and then had problems himself with the supply.

I think they are great supplies (SMPS) for amps but I found the one I used suggested you might have to engineer the amp a little differently.

As for traditional linear supplies, I too played around with a lot of different approaches. I built all of them with overly large toroidal transformers typically. I tried tons of capacitance vs normal amounts. I tried RC, CRC, CLC, and often even CLCRC types. The most over the top I ever did involved surplus parts from medical equipment. I had 800va toroidal transformers that I paralleled per channel for a total of 4. That went into a bank of Two 18,000uf capacitors in parallel per rail. Then a huge inductor the same size as the transformers and rated for 15 amps. That went into another bank of 18,000uf caps with small value resistors in series rates for 100 watts and in aluminum housing. It then had snubber caps, a little rc filter on the output. That then fed the output stage while a second output was made from that using a high voltage regulator to run the input and voltage stages. It was so large I had to mount it in a 4u aluminum chassis and send power to the amp with an umbillacale cord. The whole thing weighed well over 150lbs. It was that project that made me swear off linear supplies forever.

The amp was nuts in the end. It was just 150 watts per channel into 8 ohms but could double that down to 1 ohms. The rails were low, just 40 volts or so, which is what allowed that. It had way more current than was necessary.

 

[dc2bluelight]

I’ve wondered if it was mis-specified. Anthony Holten of Aussieamps had suggested it and then had problems himself with the supply.

I think they are great supplies (SMPS) for amps but I found the one I used suggested you might have to engineer the amp a little differently.

As for traditional linear supplies, I too played around with a lot of different approaches. I built all of them with overly large toroidal transformers typically. I tried tons of capacitance vs normal amounts. I tried RC, CRC, CLC, and often even CLCRC types. The most over the top I ever did involved surplus parts from medical equipment. I had 800va toroidal transformers that I paralleled per channel for a total of 4. That went into a bank of Two 18,000uf capacitors in parallel per rail. Then a huge inductor the same size as the transformers and rated for 15 amps. That went into another bank of 18,000uf caps with small value resistors in series rates for 100 watts and in aluminum housing. It then had snubber caps, a little rc filter on the output. That then fed the output stage while a second output was made from that using a high voltage regulator to run the input and voltage stages. It was so large I had to mount it in a 4u aluminum chassis and send power to the amp with an umbillacale cord. The whole thing weighed well over 150lbs. It was that project that made me swear off linear supplies forever.

The amp was nuts in the end. It was just 150 watts per channel into 8 ohms but could double that down to 1 ohms. The rails were low, just 40 volts or so, which is what allowed that. It had way more current than was necessary.

I'm not even going to think about one-upping that one!

What you were attempting, though, was the classic big, honking, stiff power supply with tons of current capability, which drove the cost and weight way up. In practice you could have built a much smaller, but more spongy power supply that would have more than satisfied musical demands with its relatively high crest factor and low statistical peak occurrence, for far less cost and weight. In fact, there's an example of such an amp from the 1980s (so, analog power supply), 100W > 8 ohms, but with 3dB of dynamic headroom that easily handled any musical peak, but with a relatively tiny transformer and only 10,000uf of filter cap per side, with no compromise in performance under realistic conditions. You could of course clip it at 100W with a sine wave, but that's not realistic audio. The old IHF Dynamic Headroom test still has merit, and last time I looked, dynamic headroom was still in the THX Ultra 2 specs. The test is of maximum undistorted power for 20ms (a 4% duty cycle) vs continuous. If you consider that as a design criteria, and that the goal is undistorted peaks, the PSU is a very different animal.

I'm fairly sure a generic SMPS would not do this at all, but perhaps a bit of careful outboard energy storage plus something custom around the protection circuit might do the trick. Rather than a maximum current shut down, you want the thing to go into some kind of current limiting while the caps recharge form a peak hit. Just thinking and typing, though, I haven't looked into this at all.

 

[Matthew J Poes]

I'm not even going to think about one-upping that one!

What you were attempting, though, was the classic big, honking, stiff power supply with tons of current capability, which drove the cost and weight way up. In practice you could have built a much smaller, but more spongy power supply that would have more than satisfied musical demands with its relatively high crest factor and low statistical peak occurrence, for far less cost and weight. In fact, there's an example of such an amp from the 1980s (so, analog power supply), 100W > 8 ohms, but with 3dB of dynamic headroom that easily handled any musical peak, but with a relatively tiny transformer and only 10,000uf of filter cap per side, with no compromise in performance under realistic conditions. You could of course clip it at 100W with a sine wave, but that's not realistic audio. The old IHF Dynamic Headroom test still has merit, and last time I looked, dynamic headroom was still in the THX Ultra 2 specs. The test is of maximum undistorted power for 20ms (a 4% duty cycle) vs continuous. If you consider that as a design criteria, and that the goal is undistorted peaks, the PSU is a very different animal.

I'm fairly sure a generic SMPS would not do this at all, but perhaps a bit of careful outboard energy storage plus something custom around the protection circuit might do the trick. Rather than a maximum current shut down, you want the thing to go into some kind of current limiting while the caps recharge form a peak hit. Just thinking and typing, though, I haven't looked into this at all.

That supply is what happens when you learn spice modeling and design to numbers. My goal was minimal ripple because I had it in my head that ripple equaled noise.

 

[dc2bluelight]

That supply is what happens when you learn spice modeling and design to numbers. My goal was minimal ripple because I had it in my head that ripple equaled noise.

Yeah, but you actually built it to find that out! That’s the really cool part. I was breadboarding some sort of audio processor when I discovered what “Power Supply Rejection Ratio” in opamp specs meant. I’d done the reverse of what you did and underdesigned the PSU (without SPICE, it was ca1980). I scoped the rails, saw a bunch of ripple and wondered why I wasn’t seeing it at the output. Hands-on learning...nothing like it.

 

[Matthew J Poes]

Yeah, but you actually built it to find that out! That’s the really cool part. I was breadboarding some sort of audio processor when I discovered what “Power Supply Rejection Ratio” in opamp specs meant. I’d done the reverse of what you did and underdesigned the PSU (without SPICE, it was ca1980). I scoped the rails, saw a bunch of ripple and wondered why I wasn’t seeing it at the output. Hands-on learning...nothing like it.

That’s very true. Those experiences also help me evaluate more objectively how and why products I buy are made the way they are made.