Hi all.
Im looking for a neutral and non coloring Class D amp for experimental music production.
There are so many options that i just lost track.
My first online contact with Class D/T was a thread about the SMSL SA-50
at the german analog-forum.de wehere everyone is/was raving about it.
But im not sure because of bad THD and/or power supplies and stuff.
Is a PWM direct digital amp the way to go when working with dsp only?
I like the idea of a pure signal path without any a/d or d/a conversion applied.
The amp im looking for :
- neutral and transparent without audible distortion within the 20hz-20khz range.
- fast and accurate transient response
- good low-end separation
- natural presentation of depth and width
- ground tone based representation
I am looking forward to your constructive ideas.
Rico
Im looking for a neutral and non coloring Class D amp for experimental music production.
There are so many options that i just lost track.
My first online contact with Class D/T was a thread about the SMSL SA-50
at the german analog-forum.de wehere everyone is/was raving about it.
But im not sure because of bad THD and/or power supplies and stuff.
Is a PWM direct digital amp the way to go when working with dsp only?
I like the idea of a pure signal path without any a/d or d/a conversion applied.
The amp im looking for :
- neutral and transparent without audible distortion within the 20hz-20khz range.
- fast and accurate transient response
- good low-end separation
- natural presentation of depth and width
- ground tone based representation
I am looking forward to your constructive ideas.
Rico
Last edited:
You could try the Flying Mole successor; Hypex, unless you want to make one and then I would utilize the IR20** series of chips.
Nerdbeere, you may look at FX-AUDIO D2160. It is a decent build amp from TI's TAS5548 + TAS5614A full digital chipset. It goes for 270 EURO.
Half priced, lesser quality alternative is SMSL AD18.
Both can be run from USB, SPDIF and BT aptx without DAC component, but with DSP based direct PWM conversion.
Half priced, lesser quality alternative is SMSL AD18.
Both can be run from USB, SPDIF and BT aptx without DAC component, but with DSP based direct PWM conversion.
I bet 99.99999% of direct digital "solutions" either have
1. ASRC ( approximating an AD-DA conversion to make use of a PWM side sample rate/clock )
2. jitter ( so, synchronous / recovered clock supposed to be even worse)
3. non-eloquent volume attenuator , maybe with the exception of DDFA where point 1 applies
+1 , open loop (uhh)
but then digital amplification is more or less a canard anyway.
1. ASRC ( approximating an AD-DA conversion to make use of a PWM side sample rate/clock )
2. jitter ( so, synchronous / recovered clock supposed to be even worse)
3. non-eloquent volume attenuator , maybe with the exception of DDFA where point 1 applies
+1 , open loop (uhh)
but then digital amplification is more or less a canard anyway.
zx321, would you please explain what is a non-eloquent volume attenuator? I don't find
any audio related hit by Googling.
any audio related hit by Googling.
Last edited:
TI used to have a reference design , a variable voltage power supply for open loop digital input class-d , usually omitted by designers, except Lyngdorf / TACT .
You didn´t search deep enough.
It´s in the same Google area where they store hard data about soundstage, granularity, micro-quantum, micro diode, sound blackness, rhythm, and a thousand other Audiophile keywords.
Thanks the clarification! TAS5508/18 indeed had this option bulit in.
BTW with the above mentioned amps you can still have this feature if you wish so by swaping external fixed PSU for a variable bench top lab unit.
Principle works with closed loop PA stages as well, like TAS5614A, down to 12V PVDD. So that is not a DSP based miracle.
BTW with the above mentioned amps you can still have this feature if you wish so by swaping external fixed PSU for a variable bench top lab unit.
Principle works with closed loop PA stages as well, like TAS5614A, down to 12V PVDD. So that is not a DSP based miracle.
Last edited:
You didn´t search deep enough.
It´s in the same Google area where they store hard data about soundstage, granularity, micro-quantum, micro diode, sound blackness, rhythm, and a thousand other Audiophile keywords.
Instinct saved me from searching deep, but you can never tell in advance
I have a low cost yet-to-qualify candidate suggestion, calling all to help investigate & share your thoughts:
How about this: Alientek D8 ~ a STA328 PWM direct digital open-loop, built-in XMOS USB/Optical asynchronous input, using WM8805 for S/PDIF recovered clock having low intrinsic jitter of 50 ps RMS.
I think it is for music reproduction, not sure if works for music production 😉
How about this: Alientek D8 ~ a STA328 PWM direct digital open-loop, built-in XMOS USB/Optical asynchronous input, using WM8805 for S/PDIF recovered clock having low intrinsic jitter of 50 ps RMS.
I think it is for music reproduction, not sure if works for music production 😉
Last edited:
The DDFA way of volume adjusting is rather eloquent , because adjustment is done at higher sample rate than the incoming eg. at least 88.2 or higher,
and when set to low level, it won't result in true low level (eg. word length reduced) signal but the shaped noise outside of audible frequency range makes sure audible frequency range low level informetion is preserved, so it's like a pseudo low level signal ( I guess) .
and when set to low level, it won't result in true low level (eg. word length reduced) signal but the shaped noise outside of audible frequency range makes sure audible frequency range low level informetion is preserved, so it's like a pseudo low level signal ( I guess) .
I think you mixed up volume control and PWM generation/noise shaping functions here. All DAC and FDA implementation I know of does them separately. Volume control is simply done buy multiplying the incoming stream by a fixed gain, like what foobar2000 volume control does.
What different in DDFA compared to other FDA is that it's noise shaping heavy relies on analog feedback from the output. So it is not really full digital. This feedback can push measurable THD+N down but may also introduce the usual design problems with analog feedback.
I think TI's FDA with closed loop power stage is a healthy compromise between open loop FDA and DDFA. Noise shaping is still full digital, but PSSR and distortion of the power stage is improved by local feedback inside the
ic.
D8 looks to be a decent design as well, eg. I like the big output inductors compared to the tiny ones in SMSL AD18.
However I never used D8 or any ST based FDA chipsets, so I cannot comment on their sound and etc.
But in general I would stay away from FDAs with open loop power stage. Those have 0 PSSR so power supply noise will find its way to the speaker.
And with the supplied low cost notebook type power brick you may find >> 100 mVrms audio band noise on the power rail if you push the amp harder.
For a desktop amp they can be ok, when the output is low, but not for bigger volume.
Using D8 with meanwell LRS150-36, 150W rated 250mVrms.
At low input, it is virtually total silent. Cant say i spot any issue with loud volume input, i guess noise masked.
At low input, it is virtually total silent. Cant say i spot any issue with loud volume input, i guess noise masked.
Last edited:
I can't argue your experience. My approach was a bit more pragmatic. I measured a TI based FDA running from 24Vdc power brick, with 0 dBFS real music into 8 ohm and found that the power supply noise can go up to 400mVrms, depending on the dynamics of the music.
It was not the switching/ripple noise of the supply but mostly audio band noise due to the limited load regulation of PSU. I tried different PSUs in the 60..200W range, all performed similar.
After this I decided that for a general purpose amp I wouldn't use open loop power stage with 0 dB PSRR. TI has many closed loop ones since years.
I must admit that occasionally I use the open loop SMSL AD18 as a desktop amp. I find it ok, but I use it only in the -20..-13 dBFS range at the desk.
It was not the switching/ripple noise of the supply but mostly audio band noise due to the limited load regulation of PSU. I tried different PSUs in the 60..200W range, all performed similar.
After this I decided that for a general purpose amp I wouldn't use open loop power stage with 0 dB PSRR. TI has many closed loop ones since years.
I must admit that occasionally I use the open loop SMSL AD18 as a desktop amp. I find it ok, but I use it only in the -20..-13 dBFS range at the desk.
No mix up , it was already there in SM5803, SM5813 , PMD100, new DAC chips, and foobar has it too.
Foobar volume control does not dither, or attenuate correctly!
See, the noise floor is "cranked", wouldn't happen with higher sample rate. It's an intricate detail , shouldn't go unnoticed.
I think the information you refer to is a bit outdated. Modern systems are not limited to 16 bit processing or output devices where dithering was considered important. For instance the TAS5548 series which I'm familiar with has 32/60 bit internal processing. The processing error is so low that it matters little whether they use or not dithering for the volume control function or at what upsampling stage they apply VC. Actually it is placed at 96Khz by default, so it should be just as 'eloquent' as DDFA.
Please note that in our discussion the noise shaping principle has two applications. One is dithering which is useful in bitdepth limited systems when truncating intermediate results. Increased bit depth makes this application unnecessary.
But NS principle used also when converting input PCM stream to PWM. This is the tricky part of these amps, it always needs to provide enough audio dynamic range regardless how volume control is implemented.
Please note that in our discussion the noise shaping principle has two applications. One is dithering which is useful in bitdepth limited systems when truncating intermediate results. Increased bit depth makes this application unnecessary.
But NS principle used also when converting input PCM stream to PWM. This is the tricky part of these amps, it always needs to provide enough audio dynamic range regardless how volume control is implemented.
Last edited:
I think it's perfectly relevant, especially the 16 /44.1 NS plot , given 16/44 is about the same data rate as the DDFA 1bit / 744 khz .
Is there a chip by TI or others, operating at such frequency? That's just enough for eloquent in my book.
Is there a chip by TI or others, operating at such frequency? That's just enough for eloquent in my book.
How does audio band noise sounds like?
There is no lack of closed-loops, but the idea of feeding back output to input, put me off 😀
Are you using high sensitive spkr? Sorry, anyway i am clueless.
Uuhhhhhh where is this thread going? TS asked for simplification......... 😛
TS; get yourself a good tpa3116 or tda7498 (E) board and a 19V laptop power supply and be amazed with your first class D experience.
To me all my class D amps sound less colored / more neutral then all a/b amps I compared them to. My best sounding AB amp was a yamaha ax-700 natural sound, which gets blown away by these little 50 times cheaper and lighter chinese amps.
TPA3116 2.1 50Wx2+100W Amplifier Board Digital Power 12-24V Super Bass Speaker | eBay is my favorite board; combine it with 1 or 2 sealed (2nd hand car-) subs which you put in a corner of the room.
There is a variable crossover for the sub and a switchable ~150hz high pass filter for the tops.
TS; get yourself a good tpa3116 or tda7498 (E) board and a 19V laptop power supply and be amazed with your first class D experience.
To me all my class D amps sound less colored / more neutral then all a/b amps I compared them to. My best sounding AB amp was a yamaha ax-700 natural sound, which gets blown away by these little 50 times cheaper and lighter chinese amps.
TPA3116 2.1 50Wx2+100W Amplifier Board Digital Power 12-24V Super Bass Speaker | eBay is my favorite board; combine it with 1 or 2 sealed (2nd hand car-) subs which you put in a corner of the room.
There is a variable crossover for the sub and a switchable ~150hz high pass filter for the tops.
- Status
- Not open for further replies.