2022 Chapter 16
Fully Modern, Meet Old Skool
“The only thing wrong with Vidar,” Tyler said. “Is that it doesn’t have a standby button.”
And that’s it. In one sentence, all of Vidar’s problems.
Because Vidar has been a great product. It’s small. It’s inexpensive. It’s powerful. It’s super-reliable. It’s one of a handful of old-skool amps, with a big, heavy power transformer, all-linear power supply, and a ton of 15A output devices biased into Class AB. And it may be the only 3-figure amp like that made in the USA.
So it shouldn’t be a big surprise that Vidar was well-loved, sold very well, and hung around without any changes for 6 years. Yes, 6 years! Time flies.
But…Vidar didn’t have a standby button.
Of course, I thought about giving Vidar the same “debias” standby as Aegir once in a while. But it never really stuck. Adding a feature might be nice, but you might discover some gnarly gotchas that you didn’t expect with a “simple” change.
Aside: you know the old phrase, “if it ain’t broke, don’t fix it?” Yeah, that. There’s a lot of wisdom packed into those 7 words.
Worse, if you go to add a feature, you might get led down the path of redesign creep. So you get something that has a standby button and new devices and triple output topology and different biasing and a new microprocessor and and and…and then it’s a whole new product, with many, many gotchas waiting to be discovered.
So Vidar soldiered thru 6 years pretty much unchanged. We didn’t even change it over to the DC sense scheme used by Aegir (Vidar has one, but it’s more complicated). Heck, we didn’t even change the chassis to use the same tooling. It just…continued.
But time passes, and things change, and new parts come out, and you want to try new things…
…so, about 2 years ago, I started messing with Vidar.
Diversion: The Truth About Audio Power Amps (and Maybe All Audio)
Before I get into what we did on Vidar 2, maybe I should address the elephant in the room—as in, why Vidar has existed happily for 6 years with no changes.
Here’s the reason:
there’s not a whole lot of new here.
As in, there are not a whole lot of new technologies in audio power amps.
As in, a good audio power amp designed in the 1970s is probably still pretty good today.
Yeah, it might not be the be-all end-all, and it almost certainly won’t be current feedback like Vidar, and it won’t have a bunch of the power supply tricks we’re using, and it might be bigger and less efficient to put together, because those were the times of single-sided PC boards, TO-3 metal can transistors, and thru-hole parts. And some of those parts would be laughably crude by today’s standards, with maybe 5% or even 10% tolerance carbon resistors, clunky mica capacitors, and screw-terminal electrolytics.
But, you know…that ancient amp wouldn’t be
that bad.
“Whoa whoa whoa!” you cry. “I hear, like, all the time, about how there’s a breakthrough in power amps, like Gallium or virtual adaptive bias or direct digital! What you’re saying can’t be true!”
Except, well, it is. Here’s how speaker power amps break down:
- Voltage feedback, Lin/blameless topology. Class AB or Class A, BJT or MOSFEt, this is most of the speaker power amps out there. At least the ones that aren’t Class D.
- Chipamps. See above. As a chip. Typically lower power. Very limited. Can’t be customized like a discrete voltage feedback amp.
- Class G/H. See above. With rail switching to improve efficiency. Can be very good. Can be very efficient. The best schemes are closely guarded, because they are hard to get right.
- Exotic Class AB/A. This includes current feedback, differentials, circlotrons, transformer coupled and other weirdy stuff. Like everything we make—Aegir, Vidar, Ragnarok, Tyr are all current-feedback or differential current feedback.
- Tube amps. Have tubes driving output transformers. We don’t know much about these, because the key skill here is in making output transformers. Which we are not experts in, sorry.
- Class D. These are “switching” amps that transform analog inputs into pulsetrains, with very high efficiency. This includes the latest buzzword, Gallium Nitride. This is a fundamentally different kind of amp. We don’t make them. Dave has experimented quite a bit with them, including gallium nitride designs, and we haven’t been happy with the results. This doesn’t mean that someone won’t make a stellar one (and, when used inside a speaker, where the load is known and filtering can be customized to that load, they get a lot more interesting), but Class D ain’t us.
The main advances in speaker power amps in recent years have been in the Class D realm, focused on increasing measured performance and efficiency. This shouldn’t be a surprise.
But…linear speaker power amps? Even exotic ones?
Hmm…
Lin topology stuff was hammered out in the 1970s.
The first definitive treatise on CFAs was written by an Analog Devices engineer in the 1980s.
Class G/H was pioneered by Carver in the 1980s. Performance is still pretty good today.
So what does this mean? Should we give up on speaker power amp designs? Of course not. It means that we need to concentrate on building the best speaker power amps we can, and focusing on designs that are super-reliable and can stand the test of time.
Hence why Vidar lasted so long unchanged. And why our main focus of the revision was in terms of usability, efficiency, and refinement. The end result is a better amplifier—better-measuring, quieter, more convenient…and yet still an amp that you can continue to rely on, for years and years.
The Straight and Narrow… vs Off Roadin
The simplest course would be to make Vidar and Aegir consistent: add the standby switch using Aegir’s debias trick, and go to the simpler (and better) DC measurement and correction like Aegir.
Is this what I did? Of course not.
Because there were also some new parts I wanted to try out. Mainly for the output resistors. The multiple thick-film resistors we use in Vidar and Aegir were unique at the time, and also eliminated any problem we’d have with inductance (like in wirewounds), but they use a ton of parts and are generally a pain to put together. There were new metal-strip resistors with much better specs (thermal coefficient, power handling) that could be used singly. I wanted to use those.
So the most sane idea would be make Vidar and Aegir consistent, and use the new output resistors.
Did I do that? Again, no.
Because, despite the fact that there hasn’t been a ton of new things in speaker amps for a long, long time, I had BIG IDEAS! Ideas that could change the Vidar dramatically! That could push performance skyward! That would be a REALLY HUGE DEAL!
Aside: or so I thought. I think most engineers get like this, when confronted with the opportunity for a clean-sheet design. Or maybe I’m just crazy.
So what did I try? Lots of things. Some turned out to be very important.
Here’s what crashed and burned:
Output triple. Vidar, in amp parlance, is an “output double,” as in, it has a driver stage and an output stage. An “output triple” adds a predriver stage, and generally measures better. I figured we should try this out. After some weeks of fighting the compensation (super-fast current-feedback amps typically have some bizarro loop gain issues), I sat down to listen…
…and sighed. That wasn’t the way I wanted Vidar to sound. It had gone from a generally happy and warm amp to a thin and squinty kind of amp.
I went back to the compensation, tried some different devices, tried changing the bias, tried a bunch of things, fought some thermal issues (thermal stability is a whole nother thing on triples)…and in the end, decided that Vidar 2 wouldn’t be a triple.
Error correction. Another way to get great measured performance is through feedforward, or error correction. Implementing this is dead simple. Hell, if an amp has a DC servo, it can probably be converted over with a better op-amp, a couple of diodes, a couple different resistors, and losing a capacitor or two.
Unfortunately, as with the output triple, error correction sounds, ah, not great. While I would have loved to have a Vidar 2 that measures like a Magni Heretic, I didn’t like the result. So Vidar 2 isn’t an error-correcting amp, either.
Output daughtercard. I was worried for a time about the devices we use for VAS and driver stages, since they are thru-hole and, well, kinda classic. As in, they are either actual Japanese 2SC/2SA type parts, or close copy KSC/KSA type parts. So I thought, well, maybe let’s plan for the coming SMD era and do a daughtercard and a new thermal clip. But actually there aren’t a great number of SMD parts that will do the job for us at the voltages Vidar is running. So it remains thru-hole for now.
“So whadya do, nothin?” someone asks. “Hell, it doesn’t even have more power!”
And yeah, I hear you. But more power was never in the cards. That’s a different (larger) transformer at least. And the current transformer we have is straining the limit of the chassis. Remember, Vidar and Vidar 2 are small amps.
As far as what we did, here you go:
Standby (that’s better than Aegir). Vidar 2’s standby mode works like Aegir, but it’s actually a much, much better implementation. Aegir de-biases the output stage and gets power consumption down to 5-6 watts. Vidar’s standby mode powers down the entire boost supply and gets power consumption down to 1-2 watts. Fun fact: that’s why Vidar 2 takes a while to come back from standby—there’s much more oversight in this case.
Much better boost supply. Vidar’s boost supply goes to a discrete-regulated topology based on compound feedback pairs. With 10A rated parts, the boost supply is ready to provide all the current necessary to run the 12 15A-rated outputs to their limits. Fun fact: most of the reason Vidar 2 is quieter than Vidar is due to this supply.
Better parts throughout. We mentioned the metal strip resistors. There are now also exotic silicon and film capacitors used in critical stages, as well as 2SC/2SA and KSC/KSA actives throughout. As with Vidar, this is a DC-coupled amp, with no capacitors in the signal path.
New microprocessor—and new management. Vidar 2 uses a 32-bit microprocessor, rather than the 16-bit version on its predecessor. This allows us to more closely monitor operational parameters, and to protect against any fault modes.
Clean-sheet relayout. The new Vidar 2 is is a complete rip-up of the old Vidar. This allows much more optimized I/O, power supply, and microprocessor management.
Chassis consistency. Why not use the chassis from Aegir that is thermally better? Yep, done. Now both amps use the same chassis. Easier for us, keeps cost down, Vidar 2 stays cooler…win, win, win.
So, yeah. No off roadin. Vidar 2 is a better, quieter, higher-performing Vidar…with one very important front button.
What It Ain’t
“When you were playing with radical new ideas, why didn’t you try Class D or Class G or switching supplies?” someone asks.
It’s a fair ask. The easiest answer is:
That’s not what Vidar 2 is.
Vidar 2 is “The Fully Modern, Old Skool Power Amp.” As in, it has modern conveniences and protection systems, helmed by a 32-bit microprocessor that’s probably a good deal more powerful than the one in your original Macintosh…but beyond that, it’s a giant-transformer, big-heat-sink, pure-Class-AB deal all the way.
And that’s the way we want it.
Class D? Sure, tons of options out there. We still haven’t done one we like. And this includes Dave completing Gallium Nitride prototypes running at 3MHz. A lot of people forget we have deeeeeeeeeeeep development expertise here. Yes, we’ve tried Class D. We continue to try Class D. We continue to not be thrilled with Class D. This is our opinion; this is not to crap on anyone other company. Bottom line, if you want Class D, you have tons of options. Just not from us.
Class G? Oh hell yeah there are days I’d love to have a 500Wx2 Vidar, which is doable using rail-switching technology. We’ve played with it. It may happen someday. But it won’t be fast. As in, there aren’t any Class G prototypes sitting on shelves, waiting for their time in the sun. But, seeing what Carver was able to do in the paleolithic era…and what companies like Emotiva are doing today…yeah, maybe. Just not next year. Or maybe ever.
Switching supplies? I’m willing to be convinced. I’d definitely do a switching-supply+Class AB amp before I’d do Class D, based on our experience. But switching supplies have their own vagaries, not least of which being that the safest route is to buy one and drop it in, rather than develop our own, and the bought one would likely be super custom, because we are weird and like lots of strange stacked supplies—I mean hell, look at Vidar. Normal “Old Skool” amps have 2 rails, one bridge, a couple big caps, done. Vidar has 4 bridges, one for each phase of the two channels, completely separate, dual-mono-back-to-the-transformer main rails, two sets of filter caps, one for each channel, plus a regulated-back-to-ground “boost” supply that increases efficiency by allowing us to drive the output stage to the rails (note not stacked supply), plus a separate LV supply with +/-15V and +3.3V for the microprocessor and housekeeping. Want a switcher to replace the +/-50V in your old skool amp? Sure. +/-55 x 2, +/-65 x2, +/-15, +/-3.3…yeah that’s probably custom. Again, not saying no, just not next year.
So where does that leave us?
Oh yeah, back at Vidar 2…the fully modern, old skool power amp…that is also pretty darn affordable and made in California.
We hope you enjoy the heck out of it, just like you enjoyed the original Vidar!
