We knew that from the performance and from the PFFB traces visible on the bottom of the PCB.
From Post 1078.
From Post 1078.
What kind of music files do you use to keep phase shifts unchanged up to 300kHz? What kind of speakers, microphone? Are you saying that the amp must keep the phase up to 300kHz, however the signal source may be 10 - 100 x degraded in this regard?This is why I think it is important for an amp to be able to reproduce sound without imparting too much phase shifts below 300kHz. When I design a discrete Class A or Class AB amp, I look at the phase shift and optimize the design in LTSpice to minimize phase shifts below 300kHz. In listening tests, this seems to improve the amplifier’s ability to provide better imaging and soundstage (all subjective measures).
Please read the tech note on PFFB carefully, Ti states the THD improvement is max 5 db with pffb. So it is not only PFFB!We knew that from the performance and from the PFFB traces visible on the bottom of the PCB.
From Post 1078.
And you seem to believe that the captured music has the same "phase accuracy" in microphone paths. Do you have an idea how their response looks like? And have you ever analyzed the "768kHz data files" for their spectral content? So where does it end?Digital audio files do not need to be at 600kHz data rate to have 300kHz phase accuracy. Although there is 768khz rate dats files. Also much faster for DSD and other formats. Typical clocks that control ADC’s and DAC’s run at 12MHz and have typically 50ppm stability and jitter measured in the picoseconds range. More than enough to preserve phase accuracy to 3usec (300kHz).
Post Filter Feedback is a design technique, not a specific circuit. The passive TI example is just one simple variation with a lot of specific conditions, like not loosing too much circuit gain and simplicity of the compensation (2-pole).Please read the tech note on PFFB carefully, Ti states the THD improvement is max 5 db with pffb. So it is not only PFFB!
Only depends on the maximum level you can get before you run into current clamping or thermal issues. This is true for all amplifiers.Any idea if this amp would work properly with speakers that has impedance down to 2.9 ohm? (focal aria 926)
Topping specification mentions load only to 4 ohm
Would you mind explaining why maxing the volume on the dac and turning down the amp makes a difference, and why this is more important for unbalanced interconnects? I always found this confusing. Do I turn the volume down on the dac, the amp, the preamp (if there even is one in the system), or digitally.Max the volume on the DAC. This gives the best signal to noise/hum/buzz ratio, notably when you run unbalanced interconnects.
Fair enough.Because...
Topping PA5 Review (Amplifier)
Topping PA5 manual available here: https://www.tpdz.net/newsinfo/575024.htmlaudiosciencereview.com
"Only have rca output on DAC. Is it possible to DIY a rca to TRS balanced cable? Will this work?
It should but you won't get maximum power unless the DAC can produce 2.5 volts (some can, most cannot)."
With unbalanced connections any mains leakage current that flows through the RCA cable shield causes a voltage drop which is part of the signal as seen by the receiving end.Would you mind explaining why maxing the volume on the dac and turning down the amp makes a difference, and why this is more important for unbalanced interconnects? I always found this confusing. Do I turn the volume down on the dac, the amp, the preamp (if there even is one in the system), or digitally.
Seems like the volume knob is an extra bonus for me then. I think I can go without a preamp and just use the volume knob on the amp!With unbalanced connections any mains leakage current that flows through the RCA cable shield causes a voltage drop which is part of the signal as seen by the receiving end.
Thus, the higher the signal level on the line the larger the ratio of signal to noise (hum/buzz).
With balanced, the problem is less severe but still there.
General rule for signal transmission with cables is: always use the highest possible signal levels. Of course not so high as to overload/clip the input of the amp before it has a chance to attenuate it.
TI note only has a passive PFFB implementation. That's why I've said before that this level of performance can be only done with an active PFFB.Please read the tech note on PFFB carefully, Ti states the THD improvement is max 5 db with pffb. So it is not only PFFB!
Yet, there is virtually no spectral content above 15k from an orchestra at a normal listening position.To be clear, I am talking about amplifiers needing to have phase accuracy in one channel. I am just saying a recording should have a time base accuracy equivalent to a bandwidth of 300kHz (or 0.3MHz). Not a hard thing to ask for given that DAC clocks typically have a time base originating from a 12MHz Xtal oscillator. An amp should be able to reproduce a time series signal with phase accuracy relative to itself of at least 2usec. This can be confirmed with an auto correlation function.
It is a well known fact that people totally deaf in one ear can spatially locate sound sources in 3D space with their remaining ear. There have also been experiments where the pinna (external ear lobes) was taped over with a smooth funnel and the subject lost their ability to locate the directionally with one ear. (Perhaps try this experiment yourself). There are a lot of scientific papers on this. Just search “Human Pinna Sound Localization”. It’s no accident that the pinna is asymmetric so that directionality of the sound wave can be resolved through the phase differences imparted by the pinna, and the the brain applies real-time “bio-DSP” aka psychoacoustic processing, to determine the spatial location.
A 1mm path difference in the pinna can impart about a 3usec shift to a wave traveling at 342m/sec (speed of sound).
If you look at the paper here:
According to section V in the paper, the time domain of the pinna is 2usec to 300usec (0.3ms).
Look at Fig. 4 which shows the transformation of a sound pulse from a source from the side by a cast mold of a human pinna fitted with a microphone. We can see that it imparts a transformation of the pulse in the time domain.
View attachment 171186
More recently, there have been papers discussing the microsecond resolution of the human hearing:
Krumbholz K, Patterson RD, Nobbe A, Fastl H. Microsecond temporal resolution in monaural hearing without spectral cues? J Acoust Soc Am. 2003 May;113(5):2790-800. doi: 10.1121/1.1547438. PMID: 12765396.
Here is a very good review paper on sound localization by the human ear. Tons of references in there for those interested.
You can look for many more scientific publications discussing this.
This is why I think it is important for an amp to be able to reproduce sound without imparting too much phase shifts below 300kHz. When I design a discrete Class A or Class AB amp, I look at the phase shift and optimize the design in LTSpice to minimize phase shifts below 300kHz. In listening tests, this seems to improve the amplifier’s ability to provide better imaging and soundstage (all subjective measures).
From the TI thread... "We are developing a hiend amp based on tpa3255 (under 400$ including linear PSU 250W" ..TI note only has a passive PFFB implementation. That's why I've said before that this level of performance can be only done with an active PFFB.
Looks like they've figured out how to implement active PFFB properly.
Take a look at this TI Support thread
I already read this thread before and I thought back then that this is cdsgames which is the designer of Allo
Pretty sure the PA5 is a composite with some opamp being the master and the class-D chip being the power slave. Which makes a bit easier to squeeze out the last bit of loop gain even at higher frequencies, ~5kHz.
I guess it is related to the bandwidth limiting in the AP-filter.What about the 15kHz (45kHz BW) THD dip? An aberration in the AP?
AP abberation? No ;-)What about the 15kHz (45kHz BW) THD dip? An aberration in the AP?
View attachment 171266
That suggests an LF and HF separate FB network, perhaps trimmed for minimum THD at specific frequencies. There were a bunch of designs that utilized similar topologies in the 1980s, including some non-NFB, direct distortion cancelling offerings.