new GAN amplifier

I am going to go out on a limb and say you wouldn't know without being told if the amplifier had FETs, BJTs, etc.   Topology is going to make far more difference to the sound.

How can you know "what to listen for" wrt transistor type when you don't even know the underlying amplifier topology and distortion mechanisms, intentional voicing, damping versus frequency, bandwidth, etc.   You simply can't.  You can't take something you heard on a limited set of products and apply that to all products.

This, what you wrote below, I cannot just accept on the face as factual. The drive requirements, biasing, temperature compensation, bandwidth vs. drive, etc. are so significantly different between a BJT pair and MOSFET pair, that unless you essentially modified the whole amplifier, you could not do this and at which point you essentially have two different amplifiers and at that point, any comparison has lost all meaning.

georgehifi6,173 posts11-16-2019 2:38pm

How can you know "what to listen for"

There’s a post a member put up a couple of weeks ago (go find it), who does mods for his customers, where he takes a Mosfet amp removes the complimentary N/P channel Mosfet output stage and replaces it with complimentary NPN/PNP Bi-Polar (BJT), and he and his customers say it all in their description of the sound change.

No that is not an opinion that is a experience. The drive requirements for a set of bipolar transistors and a set of mosfets is much much different. Add in different biasing requirements, different thermal compensation, different bandwidths for a given drive circuit, etc. If you change the two you don’t have the same amplifier anymore. Do you design amplifiers or are you just repeating others?

You can't just swap out the output FETs for the BJTs. There are a lot of other things that need to be addressed. It is not the same amplifier any more.


Again, do you know how to design amplifiers? I get the impression no or your replies would be more technical. Hence what you are posting is mainly heresay.

This is what you said "sunshine" and to my point, you are essentially making a different amp.  The one who doesn't have any decent tech knowledge is you. I have designed many amplifiers, discreet and on integrated circuits, hence why what you post sounds so ridiculous to me. You can't simply change an output stage on a power stage between MOSFET and BJT. It don't work that way sunshine.   Now if you had any tech knowledge you would know:

• Typical output stages have 0 voltage gain so quoting in this case that they have higher voltage gain is silly.
• In an amplifier, output current is mainly a function of thermal limitations. I can cheaply buy a 100A MOSFET, more than enough for almost any power level so to say a BJT can handle higher current is just wrong. That has not been true for decades.
• Output impedance is a function of device and circuit, not just the device so making a claim a BJT has lower output impedance is specious. You can try to make the claim there is less difference in base/emitter voltage compared to gate/drain voltage over loading, but higher biasing of the output stage with modern FETs can negate that advantage. As well a BJT requires a very high current to drive it and that current gain is variable w.r.t. output current leading to distortion inducing mechanisms elsewhere in the amplifier that a FET amp may not experience ..... And also why you cannot simply swap the output stage! ...and have it work properly.

Actual "tech", not a cut and paste tech ....

There’s a post a member put up a couple of weeks ago (go find it), who does mods for his customers, where he takes a Mosfet amp removes the complimentary N/P channel Mosfet output stage and replaces it with complimentary NPN/PNP Bi-Polar (BJT), and he and his customers say it all in their description of the sound change.

And it’s a different amplifier when done, and not indicative of a comparison of two sets of output devices .... sunshine. Yes, lets get back to the OPs topic which you side-tracked ....

georgehifi6,190 posts11-15-2019 2:13pmMany (me included) just don’t like Mosfet sound.
Listen to some older Threshold Pass amps that used bi-polars, you’ll change your mind.

Cheers George

Output stage, the stage before the output stage, hopefully the feedback network, the bias circuitry for the output stage ..... It's a different amplifier.

Can you even buy TO3 transistors any more? (TO3, as opposed to TO3P which is a different beast). I think with the exception of MIL they have all been long obsoleted.

On the TO3, I mean that quite seriously. I can’t remember the last time I saw anyone use a TO3 in a production design. Your idea of plentiful and cheap is likely different from mine. I don’t buy components off EBay unless I am finding something obsolete. You can find the odd TO-3 in stock at stocking distributors, but they are a rarity these days and legacy devices.


If you change the output stage, bias circuit, and the drive circuitry before the output stage, which means 2 stages of what is likely a 4-6 stage amplifier (and likely feedback to account for different bandwidths), then it is not the same amplifier any more. You can’t change 33-50% of an amplifier and say you are just comparing the output devices.  Yes end of story. Your claim that this allows you to simply compare output devices is not true.

Pot, meet kettle ....

georgehifi6,196 posts11-18-2019 4:17pm

Now lets get back onto the OP’s topic.

Not sure what all this has to do with a new GAN amplifier, but the Wilson Alexia, at least the current model does not seem that hard to drive. Even the older ones only drop to 2ohms.


Can you put some more details into your hypothesis that being able to double power into 2 ohms is a guarantee of "quality". I can see reasons why it could be, but not a guarantee, just as the way I can see why power limitations at 2 ohm are not a guarantee of a lack of quality.

Back to Class-D, there is of course no inherent reason why a Class-D can't work at 2 ohms, or less. If you want best operation with low impedance loads, you would want to optimize for that, but then that is true for a linear amplifier as well.



https://www.stereophile.com/content/wilson-audio-specialties-alexia-series-2-loudspeaker-measurement...

https://www.stereophile.com/content/wilson-audio-specialties-alexia-loudspeaker-measurements

georghifi,

You appear to be misinterpreting what EPDR means. An EPDR of 0.9 ohms does not mean the amplifier sees an 0.9ohm load, it means that the power dissipation in the amplifier is equivalent to what it would see with a 0.9 ohm load. It sounds the same, but it is much different. When the load and current are in phase, the output transistor has the lowest drop across it (rail to load) when the current it at a maximum, hence power in the device at this voltage is not high. When the current is out of sync, you can have a maximum current when the voltage drop across the transistor is higher, hence the peak power across the transistor is much worse. Again, this is not "load" impedance, this is a mathematical formula to generate a convenient number to represent dissipation in a linear amplifier.


Here is the kicker. This problem is much worse for BJT transistors because BJT transistors experience secondary breakdown at high voltage and current, especially at high temperature. So in addition to taking into account the additional heat dissipation, you need to take into account secondary breakdown mechanism, which may mean doubling up devices, where one would normally be okay. MOSFETs don't experience secondary breakdown and don't mind that peak power, as long as average power is taken into account.

Here is the 2nd kicker, this problem does not exist at all for Class-D amplifiers.

Given that you have been and are repeatedly misusing the term EPDR, I am not sure how you can make a definitive statement about the OTLs.

georgehifi,

Considering most tube amplifiers with transformers, which many audiophiles love, and use regularly even with hard to drive speakers, have a damping factor <=20, your overall argument at least w.r.t. listening enjoyment is questionable.

If you look at the amplitude response curve and impedance curve, you will notice throughout the base, they are somewhat inverted. So the amplifier will not act as a "tone control", it will act as an equalizer to smooth out the response. That may not be a bad thing euphonically.  Are you 100% certain that low damping factor is always best?

Wrong on both counts.

A lower damping factor will, and especially in the case of the Alexia where one just has to look at the frequency response/impedance curve to see it, often correct peaks (valleys) in the frequency response. It is like a built in equalizer :-)

And .... most people don't know this, so I will not fault you, but high damping factor often results in more distortion in dynamic drivers. So no, high damping factor is not always better. Different yes, but better ... not necessarily so.

I will ask you this once. Stop insinuating I have any sort of business relationship with atmasphere. It is inflamatory and wrong. Stick to the facts, period.

I see you are parroting what someone told you about using the AP to measure a Class D amp. Funny I have tested Class D at 100’s of watts on our AP. Can’t imagine how I did that ....

I am going to guess I have far more experience with AP equipment and getting the most out of it than the average Stereophile writer who is a casual user. Hint that filter is optional dependent on external circuitry and amplifier performance.

Thanks Pirad. You just confirmed what I already expected that he doesn't know what he is talking about and just parrots what he reads. It would never occur to him that external custom filters could exist Or that the amplifier would filter out the high frequencies themselves rendering the filter use unnecessary.


John Atkinson is just using the tools available to him. He doesn't have an engineering or signal processing background (I expect you don't either). Many of us design custom interfaces to do all sorts of things with our AP units (or build whole devices to achieve what AP products can"t).

You make me laugh Georgehifi. Anyone who knows my posts knows how ridiculous your statement is. Actually you are just like uber and Frank and geoffy. You take things you read, without fully understanding them, then try to apply them to all situations.

You have misapplied EPDR repeatedly in Class-D threads. You wrote those threads, not me. EPDR does not apply to Class-D. People that actually design amplifiers, test them, etc. know this. That is a big statement about your knowledge.

You also believe that AP equipment cannot test Class-D beyond a few watts. That is also not true. That is another statement about your knowledge.

I’m not the one idolizing a reviewer of an audio magazine reviewer that would be you, and because you just basically parrot what your read you have no clue about what’s possible with this equipment so I don’t know why you were claiming limitations you don't have the experience to claim?? You take one thing read about one accessory and assume that must apply to any time you use AP equipment. That is not the case. It is no different than with epdr which you continuously referenced even though it had absolutely nothing to do with class D amplifiers.

georgehifi,
EPDR does not define "difficulty" to drive the load per-se, even though it was defined that way in consumer audio media. It is illustrative of the thermal stress that is put on the amplifier output devices instantaneously. The peak current at the worst case EPDR is not the peak current of the amplifier.

Good current delivery with complimentary P-Channel is certainly possible, but better performance these days can be done with N-channel so that is normally the preferred architecture.

Note that MOSFETs don't have secondary break-down effects, hence EPDR is not nearly as critical for FET output stages. They don't mind high instantaneous power.

You don't even understand what EPDR is and you just proved it yet again. EPDR is a number exclusively for linear amplifiers to represent equivalent PEAK power dissipation in the output devices. You have brought it up repeatedly in Class D threads. EPDR has no meaning in a Class D amplifier and that you keep bringing it up shows you don't understand EPDR or you don't understand Class D.  I will let you tell us which that is.


I read an article not long ago that said JA has tested over 300 amplifiers. We have ran that many tests in design variations across synthetic loads in a week on automated stations running 24/7. Not faulting JAs work, but putting his expertise in perspective. You don't seem to appreciate just how sophisticated companies and design groups can be set test and measurement. What Stereophile does only scratches the surface.  You idolize that because you don't have experience to put it into perspective.

Bingo arty_vandelay. However, it is specific to linear amplifiers where that instantaneous voltage represents a large voltage drop across an output device operating in the linear region at a given current and hence high instantaneous power. As the output devices in Class-D are not operated in the linear region, it has little meaning.


What EPDR hits really hard is BJT output stages as BJT devices have secondary breakdown mechanisms impacted by instantaneous power (or accurately power and voltage), even if they can handle the current delivery.

arty_vandelay23 posts11-24-2019 8:43pmEPDR is obviously a means of expressing power dissipated in the amplifier based on instantaneous values of voltage and current at a specified frequency. In other words, a reverse engineered value.

Admittedly most MOSFETs designed today are designed for pulse operation and the long pulse/DC rating of those MOSFETs is less than the rated current. That is a thermal derating though due to rapid heating of the usually quite small die. You can see that on the data sheet if it has a thermal response graph. MOSFETs designed for operation in the linear region (market that seems to be getting smaller) have larger die. This is different from the secondary breakdown of BJTs cause by current concentrating in a small area causing thermal runaway.

When did you last look at a MOSFET SoA plot? Its true that going back 30 years or more they generally didn't have secondary breakdown issues but these days its rare to find an SoA plot which is purely thermally limited.

I am not sure what the goal is with these quotes. Pretty much none of these people are experts in signal processing, and to long term advance the start of the art of class-D, it will require a different knowledge set than linear amplifiers. Thorsten Loesch  particularly shows his lack of knowledge. I have no idea why he even made a comment.

"Perhaps more crucially, so called Class D Amplifiers, which have in recent times sprouted up like mushrooms after a warm rain, continue to use the straight two or three level modulation scheme described above. And thus they still require the use of heavy handed noise shaping to attain anything like acceptable 16 Bit Audio performance.The clock frequencies for these amplifiers are usually at 300 KHz to 1MHz in the best cases. That is 3,000 to 10,000 times lower than what is required to attain 16 Bit / 44.1 KHz performance without noise shaping and other forms of signal manipulation!?"

How do you take someone seriously who has such flawed views of what is possible or the underlying technology?

Fortunately, we can count of people such as Nelson Pass, and Jeff Rowland, even John Curl to offer a more balanced (and probably more informed) view of where Class-D is and can go.

GaN FETs suitable for a Class-D output stage and superior to silicon devices are relatively inexpensive now. Even in a $2,500 unit, they use of GaN shouldn't command that much of a premium.

Given the target market, you can't fault someone for making it look like a tube amplifier. Can't comment on the 2 ohms, or the sound, but fairly sure that clear acrylic cover would not cut it as a required fire barrier required for AC products in North America.