The output inductors he’s taking about, is a ferrite on the speaker wire outputs, the speaker wire is coiled around it. Not a coil on the power supply.
@donnylovely Yes- that was very clear in his prior post. FWIW, the choke used in a class D is not in the power supply; I don’t think anyone here had thoughts otherwise! A choke used in series with the speaker output (as mentioned prior) wouldn’t be part of the power supply either.
However if the amplifier suffers parasitics on either account, those can cause noise to be radiated via the power supply.
|
@atmasphere do you care to comment? I think the removal of an output filter for Class-D will be very measurable. I guess when you don't have to worry about compliance testing, anything is possible. @atmasphere what would happen if you hooked these up to a big electrostatic speaker or something like a Magnepan? Aren't these switching at frequencies right in the middle of the AM band and then hooked up to all that wire in the speakers? Aren't those speakers big capacitors? What is going to happen without those filters?
The extra output coils mentioned are only part of the filter. Certainly the result will be measurable, and likely audible as well. As best I can make out, the output filter choke remains. The amp would not function without it!
A zero feedback class D amplifier will exhibit an electrical resonance when used on an ESL; this being the result of the choke inductance and the capacitive load of the speaker. I've seen this resonance cause some class D amplifiers to fail.
I'm sure those series inductors are there for a reason. For example, most solid state amplifiers have a resistor/inductor network across their speaker terminals to increase stability at ultrasonic frequencies. In a class D, the concern is always radiated noise. To this end, with any changes made to the amplifier, its a very good idea to test the radiated noise to the AC line, since that noise can leak into other parts of the system (such as digital devices) and really mess with them. That is why its important to not just meet EU Directives, but to actually have noise figures well below that, since a lot of consumer digital devices aren't that good at dealing with switching noise in the AC power. Radiated noise which might emit from the speaker cables (behaving as an antenna) can also get into other equipment, increasing noise in the system.
Because most class D amps (if designed competently) are meant to meet EU Directives and similar emissions ratings of other countries, its often a Bad Idea to mess with the components in the output filter since they are often chosen for their noise characteristics; for example many capacitors have leads which have inductance; if the inductance is increased you can wind up with parasitics, which in turn if they don't mess with the amp directly can mess with other parts of the system.
Put another way, reducing noise in a class D can have a very direct improvement on how it sounds!
|
As the GAN1 is a PWM amplifier controlled by digital timing signals mainly, and not attempting track an analoig signal with high precision as would be required in a Class D amplifier, would the GAN1 be less sensitive to some of the issues you have indicated?
The digital input likely reduces concerns about the encoding scheme- I would expect it to be quite accurate. But otherwise no. The major distortion source otherwise will be caused by the deadtime required in the output section. Despite the ridiculously high speed of the output devices themselves, quite a bit more deadtime is required because the GaNFETs won't turn off when you tell them to; they need a little 'kick' which is supplied by the output filter choke. When the magnetic field in it collapses, the flyback voltage produced provides the kick. You need time to allow that to happen; hence the deadtime. If it were up the output devices otherwise it would likely be between 1/10th to 1/100th of what it is now, and distortion would be that much lower too.
|
@donnylovely There's no hate involved. If that were so my posts would be a lot different 😉 The trick on the internet is to not take things personal. That's when you get into trouble. I don't claim to know everything either- but I've learned that if I have doubts to just not say anything.
We found that you have to have really expensive test equipment in order to see the waveforms that make noise in class D GaNFET amps! This is because they can switch at such incredibly high frequencies. Our equipment was right on the edge of that kind of resolution; we had to buy some really expensive probes and the like to be able to sort things out. Otherwise I don't think we'd have been able to meet EU Directives, which one must do in order to export to Europe. Most countries use similar rules (like the UK, China or Japan) so if you can use the CE mark then you're in good shape.
So I've trying to be helpful. If Ric did the measurements for noise (I'm not talking distortion, just RFI and noise radiated on the AC line) then its a guarantee that his mods would be more effective too. They go hand in hand.
|
You have already stated that you thought op amps have no sound (are neutral). This has been refuted by hundreds of posts all over the net. Let us hear your story....if you think I am making up a story about you.
Opamps: the more important aspects are the open loop gain and Gain Bandwidth Product. These days most opamps have enough of either that as long as you don't ask more than about 20dB of gain then they will be neutral. Ask more and their 'sound' comes out. Older opamps from the 60s and 70s in particular weren't so good- so if repairing older guitar effects pedals you can mess up their 'sound' using newer opamps.
How GBP works is if you don't have enough, feedback falls off on a 6dB slope. When the feedback falls off distortion goes up. So when designing opamp circuits its a good idea to know these specs and their implications; otherwise expect colorations. Obviously 'hundreds' of posters are either using older opamps or don't know what they are doing.
none of my customers of my modded class D amps for 20 years have ever had a problem....
That you know of...
A lot of people new to class D are concerned about its noise interfering with other equipment. When you don't test, you don't know if your mod did damage on its way to trying to get the amp to sound better. I'm not saying you can't make an amp sound better but I am saying that if the amp injects more noise on the AC line or simply radiates more thru the air, it can interfere with other parts of the system, the TV in the next room, that sort of thing. I don't need a faraday cage to live in because we tested our amps...
Having built zero feedback class D amps (that is where we started) the thing we noticed is that as you have said, everything makes a difference since there is no correction. Noise is really hard to control; tiny changes you could hardly see in the board can have audible effects on the noise floor of the amp (as heard through the speakers). So I know Peachtree spent quite a lot of non-trivial time getting it right. They had to have been really careful about specs to make sure the parts they were using weren't shooting them in the back. If I were modding one of those amps, I'd want to know everything about it and have talked to the manufacturer (with whom I'd want to be on very good terms), before digging into their work. This might be because I respect how much work goes into the design, the designer that has done the work, and brought home the bacon.
|
However, that is where you stop. Caddock resistors do not measure better than other low inductance resistors....in fact, they do not sound nearly as good as parallel Nude Vishays....
Caddock has several model lines. The most commonly available ones are as you say. But they make others which have to be custom ordered and they are better than the Vishays. Vishay owns Caddock FWIW, and has seen fit to not mess with them when they bought Caddock out.
Ralph.....you sure have a fear of radiation.....maybe make a faraday cage to live in. I have been modding class D amps for 20 years and no one has died of radiation poisoning.....or had birdies or whistles or whatever your fear is.
We have to meet FCC and EU Directives (for the CE mark) since we’re an actual company. The CE mark is required for export to Europe. Some of the equipment that you worked on may have borne the CE or FCC mark; which means that it may have done so illegally after modification. Quite often that escapes notice but it can be a pain for the owner if they don’t realize why their tuner has so much interference. If you’re not doing testing you are doing your customers no favors; makes no difference how long you might have been doing the work other than it might mean a greater legacy of interference.
If you were to measure and certify the noise generated, you might on occasion find a unit that does not meet criteria right out of the box and you might be able to come up with a fix. But without radiation measurement you’ll have no idea.
No one....including me.....listens to everything....there is just not time. That is why the first thing to do when tweaking is to eliminate everything that is not needed......like LEDs, output connectors, extra connections, extra fuses, steel bolts and plates holding transformers down, etc......all these things mess up the sound.
This is speculation on your part.
I’ll give you a little tip: most toroid transformer manufacturers will include a steel bolt for mounting the part. But if you use it and spend some time with the installation, you’ll find that the bolt gets hotter than the transformer since its a magnetic short. We found this out back in the 1980s and so use non-magnetic stainless bolts. This causes the transformer to draw less current and run cooler.
My point here is you have jumped to a conclusion based on a made up story about what we do or don’t do. If you want street cred, refrain from talking about things of which you have no idea.
|
So, you are clearly stating that changing parts does not matter because you have no test results that show that they do? As long as the amp measures the way you want then it does not matter what part you use? You earlier stated that a LOT of what I said could be measured. Now you are stating that you just measure the amp and as long as it measures OK..it is. You just keep measuring it so it has no rf? You are clearly not measuring A LOT of what I said. You are measuring NONE of what I stated. You are just measuring the entire amp and that is it. You have no idea of what parts SOUND LIKE.
You have never tested parts for distortion because it is impossible and you and everyone knows that...
A lot of what you are saying here is false, for example the last two statements. We use Caddock resistors in our tube amps because they sound better, which you can see in their specs. Our tube amps are zero feedback and so have no correction; as a result we have to give them the best opportunity. Its funny how parts that measure well can sound good too.
The opening statement is false. The second statement is misleading. And yes, you test the finished product. That is how everyone including yourself does it, no matter how that testing is done; even if its just listening, which in the case of a class D is asking for trouble. If you mess up the radiation, you’re doing no-one any favors.
So I’m sure you do the testing and present your customers with a certificate showing that it still meets FCC part 15, right?
|
Ralph, You have stated a couple or more times now on Audiogon that you can measure parts....here is your latest: "Actually, a lot of the things in the list above do have measurable effects. The trick is knowing what to measure." You have never given us an example of measuring something and changing to another part and finding a measurement difference and then that difference results in a change in sound. As far as I know.....there is no way to measure resistors, capacitors, diodes, solder, wire, jacks, damping, directionality of wire, etc. If you use a large inductive part in a sensitive place...like a feedback loop.....then of course, it would change the distortion measurement.....Who the heck would use a large inductive part there? No one.
Seriously? If you want to know the effect of a part in an amp, measure the amp not the part. In a class D tiny things can have a big effect, like how a lead is terminated inside a capacitor. If the amp is noisier it can radiate noise through the air and thru the AC wiring (as well as audio grounds). When this happens, it can interfere with other equipment producing things like hum or buzz, or increased errors in digital gear.
If the circuit is also zero feedback, just about anything can affect how it sounds. Combined with degraded noise floor its hard to imagine how that will turn out well. That is why I recommend that any modified class D amp be tested to meet or exceed EU directives for radiation. Its common sense.
In case its not really obvious, getting class D amps to behave and not make noise is most of the design work. Even if your layout is good, changing a single component such as a bypass capacitor can blow that out of the water.
Sound by itself isn’t enough- if the amp is noisy its a liability. I mentioned earlier that GaNFETs can switch really really fast. The deadtime in any GaNFET amp isn’t there for the output devices - it actually has a different purpose in GaNFET amps (extra credit: what is that purpose? If you can’t answer that question you are not qualified to work on the amp). If you make a change and the amp starts switching at 40MHz for no good reason, its a simple fact that the output devices will run hotter. Heat and cooling cycles impose limits on all electronics- GaNFETs are no exception. If they fail early on because they are running hotter, that’s not a good thing, right?
The bottom line for anyone looking inside a class D amp is if you plan to modify it for any reason, there is no integrity in doing that if you can’t be sure the noise isn’t affected.
|
YOU CANNOT measure the sound of solder, wire, jacks....removing jacks, damping, wire directionality, capacitors, resistor brands, inductor brands, removing noise creating things like fuses and LEDs, adding exotic noise filters, etc, etc. into infinity.
Actually, a lot of the things in the list above do have measurable effects. The trick is knowing what to measure. In almost any class D circuit, the specs of the resistors, caps and inductors have to be examined to make sure they will work in the circuit. Different brands often have different specs and different ways of presenting them, as well as different models that are not equivalents from line to line. Solder makes a difference too, since poor solderjoints can result in unwanted effects. I can go on but you get the point.
If the parts/materials do not perform correctly, in a class D amp its a very good bet that noise will be the result. In a GaNFET amp, where the output devices can switch at 60MHz and higher, it might be hard to detect parasitic oscillations (you need a really fast oscilloscope and fancy probes that can cost more than the amp you're working on; just getting the probe near a parasitic can shut it down...)- which can result in the GaNFETs heating up more (not to mention interference with other services like FM broadcast).
So you can take the quote above with a enormous grain of salt.
|
Sometimes you just can't explain why something works when the scientific mind says it shouldn't. I see Picasso and am dumfounded others see his stuff and marvel at it. Audio is like this.
@dolfan That was true in 1985 but measurement tech has advanced quite a lot since then as well as our understanding of how the ear/brain perceives sound. So we can now make measurements that relate directly to what we hear and if enough measurements are done we can forecast how the amp will sound. That certainly was not true 30-35 years ago, so a lot got ascribed to 'magic'.
|
Why continue raining on other's parade when you don't agree with their enthusiasm, or beliefs. It's rather confounding.
@kitsap2 FWIW, the manufacturer of the amp did get on this thread and confirmed something that @kuribo caught quite a bit of heat for here: the issue that a zero feedback class D amp will have frequency response that is load dependent.
That issue does appear to be minor, but kuribo was not wrong.
WRT the issue of mods. I'm just using these two bits as examples: The output filter is a critical bit of engineering in any class D amplifier. It can have an enormous effect on the spurious noise generated by the amp- noise that can radiate into other components and can interfere with things like WiFi operation and digital circuits in general. Bypass caps in the circuit have a similar effect and are often chosen with care for their characteristics.
In a zero feedback amp these things can be quite critical because there is no feedback to offer any kind of correction.
When these things get changed out or removed, its guaranteed that the amp will need testing to know if it still meets important critical specs such as noise radiation thru the air or on the AC line. If this testing is not performed, if the mods are not done properly, don't be surprised if certain FM stations don't show up on your FM tuner because a buzzy sound has replaced them. So make sure that adequate testing has been performed when you have mods done. This testing should be done at the same level as if the amp were to be exported to the European Union. If it can surpass those marks it should be OK.
|
Is this what you are talking about?
An 16 bit word length doesn't work as well as 24 bits, 32 or 64.
|
So Ralph, are you saying that the data output (to be sent the the input of the Gan 1) sampled at 96kHz would not necessarily be a restrictive limiting factor to the ultimate performance of the Gan -1?
I am. The limitation is in the word length, not the scan frequency.
Gentlemen. I see that @kuribo has entered this thread. I've crossed swords with him in the past, but on this point of load dependent frequency response he is correct. @ricevs , if you want to find out what the effect of this might be, try a load that varies between 3 and 16 Ohms over the audio band (which is typical of many speakers) and find out what actually happens. Its not sufficient to look at the difference in FR with a simple resistive load.
Now I made a career of building zero feedback OTL tube amplifiers so I have a fair amount of hands-on experience with this. The output impedance of the amplifier plays an enormous role here. So I would expect that the FR of a zero feedback class D amp to be less 'reactive' (if you see what I did there) than that of a tube amp. The variable here though is the output choke, which is a critical part of any class D design, especially a GaNFET, which will require the choke to operate at a lower frequency rather than a higher one, on account of its inductive kickback being used to initiate the off state of the output device.
Its complicated.
Now how much this will affect the audio FR with a variable load is a good question. The On resistance of GaNFETs is stupidly low, so if it were up to them this would be a non-issue. IOW the output filter choke is really the variable. Based on my own experience (since a number of our prototypes were zero feedback) my guess is its a minor issue.
|
Sounds like this would be a limiting factor going into the GAN 1 which can accept up to 192khz rates. Am I missing something?
Yes. Digital audio uses digital 'words' whose resolving power is determined by the number of bits in the word. IOW there is a chunk of data representing a voltage at a certain time. A string of these words makes up the waveform. In a class D amplifier, the on and off states have no such meaning. So a conversion from digital to a pulse stream that has wider and narrower pulses has to occur. FWIW Almost any GaN amp is really switching at 500KHz and beyond.
|
@atulmajithia Its not a simple as you are asking- your question suggests to me that the amps you've heard that use feedback didn't use enough (which describes most amps made). Imagine a bell curve with the bell being the amount of harm caused by feedback, with zero feedback on the left and really a lot of feedback (+35dB) on the right. The reason it happens this way has to do with non-linearities present where the feedback is returned to the input of the amplifier, causing the feedback signal to be distorted before it can do its job, so distortion is created while innate distortions are suppressed (for the most part). If you can add enough feedback though (+35dB or more at all audio frequencies) then you overcome this problem.
When you get over that hump, then the feedback isn't doing the damage it does when its somewhere in the middle. Strictly in our own comparisons to our zero feedback triode class A OTLs, our class D sounds remarkably similar.
|
