DIYHiFi.org

You might recall that in the Lavry DA10 thread I mention my friend in HK who bought the Lavry unit, which started the whole sordid affair and resultant threads, well he also bought some “400W” Self Oscillating amplifier modules from Belgium last year, and we finally had the chance to put them together for him….

This was my first chance to really listen to these “Self Oscillating” amplifiers in my own reference system. I once had previously listened to them in a very “Alien” system (Box speakers), after all what’s been said about there performance we had high expectations for there Audio performance…. But oh dear here we go again…

Our current reference system uses Martin Logan Aeon’s, and a very heavily modified Pink Triangle Dacapo, DC and Cardinal as the digital front end, and in fact the Dacapo serves as our product development platform for our future audio range…

So the Dacapo and CD transport are placed on the floor between the Aeon’s, upon loading the CD and pressing play I’m not even in the listing position, yet from the very first second I pressed play and still on the floor between the speakers, I commented “what the hell is that sound!!!”

The Bass was completely wrong – sound stage was L & R nothing in the middle, on single instrument sections it sounds very good, but the moment the music gets more complicated, the imaging completely breaks down. Playing a single Piano – very nice, play 2 Pianos together and they just fight each other for position, loudness, not playing in some beautiful harmony….

Now don’t laugh or comment about my musical tastes, because I listen to a whole spectrum of music, but after about 5 minutes of shaking our heads wondering what the hell we where hearing, I inserted Fat Boy Slim’s You’ve Come A Long Way. Baby – the Start of Track 1 with the very “Phasy” Bass heavy intro – what a bloody disaster, a tortured mess - (some may say Fat Boy Slim deserves nothing better…). We pulled those Amplifiers after that….

Forget the treble which is “gone”, I hate “Digital Bright”, I know good Analogue turntables and anybody who remembers those days, the differences between MM and MC treble, so trust me I’m not in need of nasty digital treble – but hey, I want some treble!!!

I find the “Loop-back Sampling clicks” on Fat Boy Slim, where the music is looped back and repeated, very revealing. These transient “clicks” really highlight the resolving power and accuracy of a system. With the self oscillating Class D, they where almost non existent… no transient information….

The Sound stage is just COMPLETYLY destroyed, BASS sounds like it’s be run though some kind of effects processor… what going on? – forget about the massive hole that’s been sucked out of the middle – strangely the imagining has moved beyond L & R to each SIDE of the speakers…. but very little between the speakers.

So what can cause such a fundamental destruction of sound stage and Bass?

“Self Oscillating” amplifiers work on the principle of variable Mark Space ratio AND oscillation Frequency – there operating frequency varies depending on just about anything – Output Power Level, Output Load Impedance - which don’t forget varies drastically across the audio BW on real speaker loads, input PSU voltage – which normally sags with higher current demands, Bass information constituting the greatest current demands etc.

If frequency changes with signal content, then so does the “Delay” though the amplifier, which results in a varying phase between amplifier channels if they are amplifying different signals which is the expection of a Stereo system.

Driving the Aeon’s at moderate levels I see about 50% changes in switching frequency – so how can it be expected that the Phase relationship will remain the same between channels????

Now normally Bass information is effectively Mono, however with modern processed music, such as on the first track of “Fat Boy Slim’s You’ve Come A Long Way. Baby” the Bass is deliberately modulated in the recording between channels, This high energy LF music content swings the Self Oscillators switching frequency all over the shop….. thereby completely destroying the recordings original phase information…

This would also explain why single instruments sound good, but completely fails with more complex music…

But with the effect being so pronounced, why has nobody commented on it before??? I suggest that all the comments about these amplifier “wide sound stage” are due to this Phase modulation….

I would be interested to hear how it would be possible to vary an amplifiers switching frequency, without affecting its “group delay” and therefore its Phase response in relation to other channels in a multi channel system?

John

_________________
Hi-Fi??? "I may not be a psychiatrist but I know insanity when I see it - what ever happened to the normal old pervert who was content to sniff bicycle seats and collect women's shoes anyway?"

Hi John!
..tough words. In fact I also got annoyed by several properties of my first self oscillating class D amp.
But I must admit, the sound was one of the few things, which did not trouble me. But I only listened to a single mono amp.
Also my entire set up is by far less high end than yours.
You made me curious... Mail comming...

_________________
Best overall performance is derived by the right trade off.

Hi Markus,

Your only down the road, your more then welcome to come and listen for yourself, also we can dig up your tweaters and binding posts, we are about 15 minutes from the Airport... Weekend coming...

John

_________________
Hi-Fi??? "I may not be a psychiatrist but I know insanity when I see it - what ever happened to the normal old pervert who was content to sniff bicycle seats and collect women's shoes anyway?"

_________________
Best overall performance is derived by the right trade off.

Hi John,

Although your listening experience exactly matches my coworker's experience (http://diyhifi.org/forums/viewtopic.php?f=5&t=665&st=0&sk=t&sd=a&start=60&sid=f5116963c10252251d5f373eb9fa780b, 11th post from the top), I think variable switching frequency has nothing to do with it. Analogspiceman posted simulations comparing linear amplifier and selfoscillating amplifier using small signal riding on a large ramp. As far as I can remember there was no phase shift difference between the two. And I would expect simulaton to show a difference.

I have given my amp to my son and he uses it as a PA amp in his band so I can not measure anything right now, but I would find it very interesting to see some measurement to back up your theory. I hope this will turn into interesting thread, since the effect is obviously real.

So do I. Any attempt to derail it will not be tolerated. Let's hope that Bruno weighs in on the subject. Likewise, any Bruno-bashing will not be tolerated, either. Anyone who wants to do that sort of thing has plenty of other forums to do that in. Keep it there. Or else.

Jocko

With any simulation, the accuracy’s in the details. I don’t know the details of the simulations performed by Analoguespiceman – however did they really simulate a self oscillator in real life conditions. For this I mean driving a complex load – such as a hybrid Cone Bass and ESL panel, which is going to have an inherently weird impedance curve – and the effects of PSU modulation (greatly pronounced at low frequency)? Without simulating these real life effects, I guess the frequency of a self oscillator is going to appear fairly well constrained.

Any phase variation would need to be doubled, as it cannot be guaranteed what the musical content will be doing between channels, they could be going in completely different directions…

I did not start this thread with a reference to a particular design, only the modulation technique. I too would like to say that this thread is not aimed in anyway to bash Bruno – he is not the inventor of self oscillating amplifiers, Bruno’s developed a cheap method of performing it – however this does not necessarily mean that they are the best at any price, and where not intended to be. Bruno’s designs perform admirably for there simplicity and low price – however as with any design topology, one should be aware of the limitations.

John

_________________
Hi-Fi??? "I may not be a psychiatrist but I know insanity when I see it - what ever happened to the normal old pervert who was content to sniff bicycle seats and collect women's shoes anyway?"

I tend to agree that it has little to do with the specific design. I have used 2 brands (and more than one model of those brands), and they both exhibit the same tendencies.

However, my observations are much didfferent than yours. Perhaps because I (along with most of my buddies) use Maggies, we hear something different than you do. Working on your assumptions, the load impedance presented by planar magnetic speakers may lessen some of the effects that you find troubling.

The main criticism that I have heard, regardless of brand or type of speaker used, is "The midrange doesn't sound right."

Back over to you, John........

Jocko

Interesting thread.
I've a few different amps here including Aksa N+,Avondale NCC200, KT88 based PP tube amp, modified UCD180 mono's,Pavels classA mosfet power follower, various chip amp based designs, P3A.
I understand you probably can't class these as high end amps but to me and others who heard them the UCD's sounded the least euphonic, only other classD amps I've heard are Belcanto and Nuforce, I personally didn't like their signature although both of those was in a different system

Forgot to say I don't have any problems regarding bass but my speakers are not too hard to drive

Leo

Hi John,

Variations in delay, or group delay, is an interesting theory. If you feel that imaging is completely ruined it makes perfect sense to look at phase information and as a result you look immediately to the variable switching frequency, but then you also concentrate most of the discussion on a lack of bass, and the precise control of it.

I think you're in error looking to phase errors, in the audio band, as a result of a variance in switching frequency, because you completely neglected the fact that the delay is an intrinsic component of the entire oscillation loop. Any change in delay therefore, for which there are many causes, some of which also have little or nothing to do with signal content, will result in a shift of oscillating frequency, and it's this shift in frequency that helps maintain the set voltage despite the shift in delay.. it's just doing the job intended of it.

So what's the shift in phase itself at audio frequencies, doing in accordance to this shift at switching frequencies? Next to nothing in the audio band, 0.1dB maybe, at the most? These changes in switching frequency occur, probably at never less than 10X the highest audio frequency of interest. You speak of observing a 50% shift in frequency during normal music reproduction, and you've made it sound like it shifts from 300khz down to 150khz. I find that unlikely! Did you mean it instead varies 50% within it's own band of variance, between the extremes? Like from say 350khz to 400khz, you observed a change of 25khz or so while listening to music?

I'd find that more likely as a variance, and more unlikely as the cause of the ruined imaging, spatial information, and especially that of low frequency accuracy.

Even at 150khz Fs it shouldn't be a huge problem, but virtually non existant any faster.

So what else is the problem. Let's assume the rest of your equipment is fair to all this, which I assume it is, being as it's all foreing to me. You may note that those who've commented poorly in this thread have made the same mistake of implementing modules in their stock format, you know, just having thrown the kit together for somebody else? A state of module for which the designer has been known to say is an "industrial" version, ready for the abuse of the masses, but having done so I thing negates one from being able to comment on the achievable quality with it, since that aspect remains very limited by the stock format.

Then there's what everyone say's about it. Consider the source, as I do here.
Alot of people build their pet projects with them and comment in accordance to the level of effort it required of them, their lack of knowledge, and the results they've achieved. For the most part, for them just having it work at all is astounding, of course they'll be giving rave reviews irrespective of the actual sound. Others, and maybe not so different, market products with them and do the same, while not wanting to ruin their profit margins by stuffing it, the module or the product, with the most expensive boutique parts known (and others do nothing but that, yet lack the understanding required to make anything of it). They'll still all do some small minor upgrades to it though!

The fact is, my amp based on the same sure does not sound or perform as you've described. My speakers are junk in terms of soundfield, my listening environment even worse, and the amps have no difficulty in making both completely vanish, leaving me with a 360 holographic soundfield, for which the sound emminates fluidly form thin air. Actually sitting in the narrow soundfield of my speakers produces some insane imaging, not only where notes can float across channels fluently, and high to low, front to back as well to a certain extent. Are my modules in stock industrial format?? Of course not! Did it take a complete reworking to get that? No of course not. It takes very little to achieve it, and a very decent power supply helps, which you also made no mention of... probably because anything production like is heavily skimped on? The total changes to the module can be done in less than an hour, I've still yet to see the very best of them, but I know the achievable sound quality is well up there compared with anything else.

So the stock UCD is said to sound thin, and "bass light". I'd say it only had pretense to bass, and lacked any level of power in its reproduction. That's one of the easiest fixes to make and goes along with lower impedance bypassing.

If you wish a list of changes you can try to remedy this let me know, but I'd like you to post your results and impressions of having done so.

Also, their latest high grade modules just released, the "HG" series, are a less industrialized /more audiophile component... I still think they'll be a little light on the bass based on the caps I see used, but will still be remarkably improved from previous versions in that aspect as well, actually having good power in the LF range, but can even likely be further enhanced, just based on what I'm able to see of them.

So I don't think the issues you experienced relate to the fact they're self oscillating amps or that self oscillating may not be 100% ideal. I think you read too much from too many people, the "hype", and as a result just expected too much of a stock module. Rather than comming up with some theory, spend just ten minutes modifying it to your liking and I've little doubt the improvements you'll see will provide you with a very different point of view.

On a different note, this is not what I expected to see of you in terms of class d, how is your own line of amps comming along? They aren't self oscillating are they? Probably don't have those nasty phase issues do they? Pretty close to completion? Told you I had to consider the source too :)

BTW: My amps are wired out of phase, with two modules in very close proximaty sharing one power supply which consists of a shielded 500VA toroid, dual hefty low noise bridge rectifiers, further snubbed, and a properly wired bank of T-network caps. Input stage reg's are still onboard. The source is a computer soundcard, 1820M audiodock with modified outputs, and speakers are cerwin vega VS-150. Nothing special.

Regards,
Chris

Let's put aside all the stuff about the op-amps and industrial grade caps used in whatever version he has. Anyone who is honest and has listened to a wide variety of self-oscillating designs, in an even wider variety of systems knows that most listeners kvetch about the midrange. Maybe John's insight into what is bugging him will yield some insight into this.

Jocko

I can't do that because it fails to even begin to address the issues properly. While I understand the desire for a nice technical thread, I dont' think false postulates are a very sound basis for one, that's best left to marketing.

You can't start honestly talking about the intricacies of high fidelity sound in relation to an industrial grade component, one that was meant to last first, be profitable, and sound good sometime later. You're expecting way too much from too little doing that, and it is unrealistic to do so, despite what any reviewer say's about just plugging them in, they do take some finess, you admitt this of course?

C'mon, that post included every single class d related myth on bad sound I've ever heard of, and relates them all directly to the fact it's a self oscillating amp, both things I find very unlikely.

The high's are bad, non existant, with sucked out midrange, so there's non of that either, or no life or realism to the little that's there since there's no high's, and bass just out of control with no power..... and all because it's self oscillating, with all the necessary evidence to support that opinion, which can all see here, did I miss it? John has more test and analysis equipment than anyone on the planet and he knows how to use it... and he did take measurements but none are on display, something is up here.

So what does it do that's any good, according to description it's not even a decent amp for subwoofer only use... give me a break. Are the outputs even in phase? For it to sound that bad, they probably aren't. Would John make that mistake, no that's only me, probably John has a clock based amp ready for production and probably he can say it's phase linear from 0Hz to Fs. I find that more likely, but not wanting to go so far as to accuse him of such a thing, rather opted to help him get at least respectable audio performance out of the stock modules.

Oh and when you do change the caps John, dont' make your first post in the first ten seconds they're burning in like has happened here, that's also taking away some weight on the old believability scale.

If you seriously want to address design shortcommings, let's then ensure all the implementation limitations, be them deliberate or not, are minimal (use your favored parts be them boutique or industrial), then we can talk intelligently should we still find some underlaying problem to exist ruining our sonic experience, and proceed with the cause/effect thing in a meaningful way.

If John bothers to do at least that, this thread will come to a quick end.

The only time any version I ever had even approached sounding that bad was while stock, poorly implemented, and poorly driven, and even still were never "that" bad. Do I need to throw out the rest of my system as a result so I can hear how truly bad they actually are, does this crazy soundfield I experience on a daily basis come from my bad speakers and computer source? Or does John need to improve his modules just a bit, before expecting greatness of them, or worse, completely dismissing them based on their very nature? I think it's obvious which should come first.

Aside from that John's insight has led him to his own, non self oscillating sollutions, where I doubt he'd use the same type of cheap electrolytic's for AC coupling, or singular high ESR bulk storage caps, before dismissing the topology as being intrinsically inadequate for even subwoofer use.

Cheers

I don't have time to find and repost the details, but sometime ago in a thread on diyAudio I reported the results of a simulation that compared a UcD style self-oscillating class d amplifier to its linear, non switching equivalent. A source consisting of a small sine wave riding on a slow dc ramp was fed to the input of both amplifiers. The output of the simulation was the difference between the outputs of the two amplifiers. The absolute value of this difference was plotted against the dc ramp level. At low to medium offsets, there was absolutely no measurable difference between the switching amplifier and its non-switching equivalent. It was only near the rails (within 70 percent or so) that the UcD style version began to exhibit a significant nonlinear response.

What this says to me is that the sound of a properly built self-oscillating amplifier should be indistinguishable from its nearest linear equivalent, at least as long as it has enough headroom not to clip (or nearly clip) very often.

One caveat, the linear equivalent of the UcD is not an ideal voltage controlled voltage source. It can have a non-flat response, phase shift, rising output impedance, etc., but what it won't do, is generate any signals (artifact) that weren't present at the input. The ideal UcD style amplifier only begins to do this as its output nears the rails. As to how noticeable or how much different or worse than normal clipping this may sound, I have no idea. -- a.s.

Hi ho!
1. Thanks John for the perfekt day! ..seems this thread moved to a battle while we where listening?
2. My 2 cents:
We listened to several amps, but only one was class D.
With all amps John's electrostatic speakers were imaging a sound stage, which I am not used to. Not so bad. ...OK incredible... About the sound stage: With these speakers the sound stage of the self osc class D was clearly on the poor side. It did NOT ADD soundstage due to phase shift errors, it reduced field due to I don't know what.
Jocko's comment on the midrange does not fit to my impression. For me it was more about the details, -the love for the small things! I complain about the treble, but here I am sometimes of different opinion than especially some sales guys. Quite often when they say listen to this clearness, then I cannot not find clearness. In this cases I often find agressivness without any contour or precision. Smooth sibilants often converted to hissing noise... OK, the UCD was not so bad that I would use the word noise, but it for my ears it was not fondly with the details.
Steping to the technics:
The self oscillating topologies does cycle up and down the switching frequency within each period of the slow audio signal.
I think that the resulting errors (if any, scratching my head) cannot be expressed in a phase shift, because phase shift is something which is defined in steady state situation and needs at least one full period of the signal in steady state to be described as an input signal input(t)=INPUT*sin(wt) resulting in an output signal output(t)=OUTPUT*sin(wt+phi).
May be step response is more helpful? But also regarding step response the UCD has theoretically the chance to perform quite well, because the output filter is inside the feedback loop, which could help very much to reduce the typical overshoot of the output filter (as long as it does not lead higher order ringing...? ...complex closed loops are to complicated for my brain simulation...)
Do we have a simulation or measurements for that?

3. ...John is living 'just down the road'... I would say instead of 'just down the road' - the road was down.. (see pic)...

Bye
Markus

P.S:
Pavel,
both Rookies are not frustrated about place 2&3 out of 5 amps in this competion. In fact there was only one amp which ended up far behind expectations. And this amp was not the UCD. It was the Mark Levinson with disappointing price/sound ratio, offering nothing but brute force and good control of the real deep bass.
I personally got many inspirations/ideas, even if I have just some partially ideas how to transform my findings into design measures...

_________________
Best overall performance is derived by the right trade off.

...btw...
John,
I know it is quite time consuming...
But I would also love to go for a comparison in measurements of all the 5 amps which we were listening to.
What's your opinion?

_________________
Best overall performance is derived by the right trade off.