The EPC dev board you linked isn't anything special, but the underlying component technology is. GaN FETs will eventually take over the throne from MOSFETs as the technology is more widely licensed and produced because it offers the promise of better component specifications: lower gate charge (faster switching, reduced dead time distortion), no reverse recovery (more ideal switching, reduced EMI), reduced switching losses, low on resistance (but not quite as good as the better MOSFETs yet), and higher power density. Right now they are being offered in mostly lower voltages (<150V) but many of the offerings would be very suitable for low and mid power class d amplifier use.
The big drawbacks include: specific package offering (reduction of parasitics are paramount to proper operation at higher switching speeds), getting heat out of the tiny packages (the PCB becomes a critical part of the thermal design), and proper gate drive. Some manufacturers offer specific gate drive ICs for GaN half bridges.
Totally agree.
https://www.powerelectronicsnews.co...apter-of-class-d-audio-amplifier-performance/
Edit: +also cost barrier at the moment.
Last edited:
Linear Tech became part of AD, not TI.
And as Dimitri pointed out, the LM4562 is a NatSemi product.
But the main problem is that the noise of a LM4562 is 2.7 nV/rt Hz.
Should have used an AD797, noise is about 0.9 nV/rt Hz, problem solved.
I add my request to Dimitri's, I would like to see some examples.
Most amps are very close to Minimum Phase so the delay is perfectly predictable, and small for a reasonably flat response.
Best wishes
David
Last edited:
Other amps not mentioned
Primare i35 I have heard it but not in controlled test, sounds very good though
Devialet never heard one but would like to
New Class D same as Devialet, but should be quite good
Lyngdorf, lot of experience with the old models. Millennium really good, the semi digital not so much
Primare i35 I have heard it but not in controlled test, sounds very good though
Devialet never heard one but would like to
New Class D same as Devialet, but should be quite good
Lyngdorf, lot of experience with the old models. Millennium really good, the semi digital not so much
Depends on the impedance of the circuit. Above about 1.5k the LM4562 will have the edge due to its lower current noise. Above about 5k and the NE5534A is going to trounce both of them due to its much lower current noise (and any JFET opamp will do too).
This is not a drop-in to a specified circuit impedance.
It was a hypothetical (and not that serious) "what they should have done".
What they should have done is keep the impedance sufficiently low to achieve really low noise.
If the impedance is above 5k then true, the AD797 isn't ideal.
But then noise will be lousy whatever the op amp.
Best wishes
David
Maybe it was NS
The datasheets from almost ebery OEM manufacturer include phase measurements.
As-series - intelligent audio amplifier power
You can search the homepages of Pascal, Hypex and Purifi for the same measurements.
Primare was actually reviewed by Sterophile Primare A35.2 power amplifier Measurements | Stereophile.com
Does New Class D actually share technology with Devialet?
I´ve never heard aboyt that, but it would be interesting to know some details.
I´ve heard some of the Lyngdorfs through the years.
I don´t remember the models.
I guess it was my comment, where I wrote, that the buffer trippled the noise on the complete Hypex nCore amplifier, that provoced a bit.
To get the proportions right, the module itself has a noise level of only 9 µV.
But @ a gain of only ~13 dB. When the buffer is connectet, the gain rises to ~30 dB with a noise level of 28µV.
We´re a few guys building our own class D amplifier, and we´ve reached ~30dB gain in the class D amp with no buffer, and a noise level @25µV.
One could just say "same same, just different".
But there is a few things that can be considered up - and downsides for either amp.
Separating the gain stages as Hypex, will put a lot of pressure on the buffer, to match the module, which will probably call for one of the super op-amps available. It will probably not be able to match the specs anyway.
But maybe one could include the buffer in a 2.nd loop with both amp and buffer. I do not know if that will work, but I imagine it could.
On the other hand you could have the full gain in the amp, with no loss of noise performance, compared to the Hypex + buffer, calling only for a unity gain buffer, if necessary at all.
We´ve obtained a Zin @ 1KOhm unballanced, and 2X1KOhm ballanced in the class D amp.
Designing a buffer with unity gain, opens up a lot of different topologies.
It can be discrete, you don´t really have any usefull benefits from feedback loops, så it could be non feedback design.
We chose a discrete buffer with NFB topology, DC coupled with servo´s, and a double regulated supply with shuntregulators in the second stage, to keep noise to a minimum.
We tried with a super op-amp (OPA1612), but the discrete one seems more transparent.
Best wishes
Soeren
A quick look at the AS1200 showed no phase, hence my question, but a check showed phase data for some of the others.
Hmm, some of the Icepower data does looks odd, further searches revealed that other people have also noticed this, is it what concerns you?
I suspect measurement error but I haven't seen an independent measurement for the Icepower units to resolve the question.
Amir at AudioScienceReview did check the Hypex and they look fairly close to minimum phase, as I said I expected.
Best wishes
David
Phase
@ Dave Zan
Sorry for the delay, I´ve been to a wonderfull summerholiday with family & Freinds, but now back and alive
To explain my considerations about Group Delay, I´ve done a bit of Cut & Paste from Wiki, they do explain it pretty well I think.
"In signal processing, group delay is the time delay of the amplitude envelopes of the various sinusoidal components of a signal through a device under test, and is a function of frequency for each component. Phase delay, in contrast, is the time delay of the phase as opposed to the time delay of the amplitude envelope.
All frequency components of a signal are delayed when passed through a device such as an amplifier, a loudspeaker, or propagating through space or a medium, such as air. This signal delay will be different for the various frequencies unless the device has the property of being linear phase. The delay variation means that signals consisting of multiple frequency components will suffer distortion because these components are not delayed by the same amount of time at the output of the device. This changes the shape of the signal in addition to any constant delay or scale change. A sufficiently large delay variation can cause problems such as poor fidelity in audio or intersymbol interference (ISI) in the demodulation of digital information from an analog carrier signal. High speed modems use adaptive equalizers to compensate for non-constant group delay. "
Does this make sense?
To know even more, you can read a bit more here: Group delay and phase delay - Wikipedia
I´ve attached two measurements of phase for class D amps.
It is pascal SPRO2 and ICEPower 1200 AS2.
As you see they both have some deviation from a straight line.
This is an often overseen parameter in both class D amps and linear ones too.
I´m sorry I don´t have phase measurements on any nCore og Purifi amps, but I have seen at least one of them at som point in time, but I don´t recall where, maybe some one can help find them.
@ Dave Zan
Sorry for the delay, I´ve been to a wonderfull summerholiday with family & Freinds, but now back and alive
To explain my considerations about Group Delay, I´ve done a bit of Cut & Paste from Wiki, they do explain it pretty well I think.
"In signal processing, group delay is the time delay of the amplitude envelopes of the various sinusoidal components of a signal through a device under test, and is a function of frequency for each component. Phase delay, in contrast, is the time delay of the phase as opposed to the time delay of the amplitude envelope.
All frequency components of a signal are delayed when passed through a device such as an amplifier, a loudspeaker, or propagating through space or a medium, such as air. This signal delay will be different for the various frequencies unless the device has the property of being linear phase. The delay variation means that signals consisting of multiple frequency components will suffer distortion because these components are not delayed by the same amount of time at the output of the device. This changes the shape of the signal in addition to any constant delay or scale change. A sufficiently large delay variation can cause problems such as poor fidelity in audio or intersymbol interference (ISI) in the demodulation of digital information from an analog carrier signal. High speed modems use adaptive equalizers to compensate for non-constant group delay. "
Does this make sense?
To know even more, you can read a bit more here: Group delay and phase delay - Wikipedia
I´ve attached two measurements of phase for class D amps.
It is pascal SPRO2 and ICEPower 1200 AS2.
As you see they both have some deviation from a straight line.
This is an often overseen parameter in both class D amps and linear ones too.
I´m sorry I don´t have phase measurements on any nCore og Purifi amps, but I have seen at least one of them at som point in time, but I don´t recall where, maybe some one can help find them.
Attachments
Last edited:
Interesting thread.
However sorry to say that this thread does not reveal technical comparison rather than impression from hearing which is not common in class D section.
Both compared have different output rating. A small rating amp will get impression as thin and flat. While big amp will always thick and full. It is always like that.
I am a diyer. I have made many amps. Or hear many chip amps. I made both UcD and double feedback with IR based topology. Both are almost equal. You can make a very good sound quality. Low gain indeed can improve sound quality.
Today most designers have push class D capability to limit. Almost no room more for improvement. Basically with post filter feedback, it can push sonic quality to limit. In this 2020 year, there are many almost perfect components for switching speed. In class D, high speed is not the key, but high speed on an off to get minimum dead time is the key. A switching speed +/- 400kHz almost meet highest quality of class D. Higher speed will be risk of distortion from dead time.
So state of the art class D will be:
1. Post filter feedback/double.
2. Low gain
3. Switching speed +/- 400kHz
However sorry to say that this thread does not reveal technical comparison rather than impression from hearing which is not common in class D section.
Both compared have different output rating. A small rating amp will get impression as thin and flat. While big amp will always thick and full. It is always like that.
I am a diyer. I have made many amps. Or hear many chip amps. I made both UcD and double feedback with IR based topology. Both are almost equal. You can make a very good sound quality. Low gain indeed can improve sound quality.
Today most designers have push class D capability to limit. Almost no room more for improvement. Basically with post filter feedback, it can push sonic quality to limit. In this 2020 year, there are many almost perfect components for switching speed. In class D, high speed is not the key, but high speed on an off to get minimum dead time is the key. A switching speed +/- 400kHz almost meet highest quality of class D. Higher speed will be risk of distortion from dead time.
So state of the art class D will be:
1. Post filter feedback/double.
2. Low gain
3. Switching speed +/- 400kHz
Last edited:
So it was a Group delay?
Very pleased to learn you are back safely.
Yes, I understand this fairly well, have you read Richard Heyser's work?
He analyses under what conditions the phase/amplitude variations lead to a true delay.
https://www.aes.org/technical/documents/openaccess/AES_TimeDelaySpectrometry.pdf
At least a certain amount of non flatness in the phase is inevitable if the amplitude response is not flat.
This is the minimum phase response. There is always some delay associated with upper frequency roll-off.
And it always must roll-off, eventually.
I don't see any point to worry about this because it is unavoidable anyway, just a consequence of bandwidth less than infinite.
The only way to reduce it is to extend the bandwidth and I don't believe there is evidence that this is audible.
And of course it happens for any class of amp.
Extra, so called Non minimum, phase is more of a problem if it occurs.
But I would not expect it in a Class D amp.
Some of the Ice power plots do show some non minimum phase but it looks like measurement error to me.
Hence my question.
The plot in your latest post looks essentially minimum phase, just from eyeball estimate.
Best wishes
David
Last edited:
I started the thread for technical interest so I am sorry too (probably more than you
)Thank you for the comments, the current state of the art is excellent so maybe there's not much more for people to add.
Best wishes
David
Personally I like the idea of going hotter out the preamp into a low gain power amp, but I don't have to be compatible with anyone but myself
Yep. Me too.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.