Topping PA5 Review (Amplifier)

[enioentity]

That will be interesting, excellent measuring class AB vs class D. Please write your observations.

Yep, definitely do please..

 

[mocenigo]

In addition to everything you said, let's not forget that even double blind comparisons of electronic equipment have many issues:
-- They have nothing to do with fidelity (except, possibly, in very rare instances)
-- They are corrupted by many factors:
- room acoustics
- chosen recordings
- personal taste
- other equipment in the reproduction chain
- etc.?

The definition of "better" gets completely blurred by all of these factors, except for the owner of the listening room and the rest of the reproduction equipment.

Double blind tests are more about distinguishability. In an ABX you hear first "A", then "B" and then "X" and you have to establish whether X=A or X=B. It is not about determining the better amplifier or DAC.

 

[antcollinet]

Per Amir benchtest "Sweeping the frequency vs power and distortion we get:"
View attachment 183148

1khz granularity would be nice in this plot but we only get 1khz 5khz 10khz and 15khz. plots. With 1khz plot granularity you would see all frequencies above 2khz are rising in distortion with power.
"normal distortion rise" ? Its not just frequency but power levels as well. Normal Class-D maybe.
This is with a RESISTIVE Load not REACTIVE via speakers or emulation. I am resisting debating this as a factor.
The sinad test is based on a 40 ohm resistance at 1khz.
With this above plot Noise even 1khz @ 5w = -95db @4ohm not 40. 15khz is at -75db
However The TPA32xx(3245 or 3251) is known for issues with reactive loads as it requires LC filtering on its output.
Phase however isn't even tested at any frequencies on ASR.

Here is a paper regarding how Harman tries to deal with reactive(speaker) loads with Class-D using DLA technology.

https://media.flixcar.com/f360cdn/Harman_Kardon-75965644-Digital_Loop_Amplifier_White_Paper.pdf

In short via Harman:
"The key breakthrough is an ultra-fast, low-latency analog-to-digital conversion (ADC) of the analog output signal from the speaker terminals back into the digital control loop. The DLA can achieve speeds of 20 nanoseconds." This done in conjunction on the output side with speakers attached so that reactive load is considered.
Yes companies have marketing hype, but at the same time they do have to innovate sometimes to achieve design needs.

Topping isn't an IC chip designer. They just implement off the shelve parts Like Dacs, op amps etc. There is little talk about how they achieve the PA5.

Understandable for propriety sake but at the same time, how does it really do with "Real World" Speaker loads?

Ok

Noise distortion is not a thing. Noise and distortion are two different phenomena. The plot above is a plot of the sum of noise and distortion.

 

[NTK]

Yes my posts are a bit Didactic,

You are talking about 2nd Harmonic Distortion. Harmonic blends nicely. What about Noise Distortion?
Try playing a 10khz Fundamental signal next to a 10.001 khz signal, maybe a 10.002khz and then a 10.003Khz and slowly sweep up and see what level will you begin to notice?

Beats me...

@tonycollinet has already asked you what "noise distortion" is, so we're waiting for your explanation.

Do you also have an explanation on where those 10.001, etc. kHz signals came from? Anyway, even if we assume that they are there, the effect is well studied under the topic of auditory masking. The differences in frequencies will create beats (beat frequency is the difference between the 2 frequencies), which manifest as variations in loudness. You can go look up how strong a 10.001 kHz needs to be next to a 10 kHz for its presence to be audible.

 

[IPunchCholla]

I am genuinely excited to hear if everyone likes their PA5, I hope they do. Please leave you feedback.

I’m really liking mine. It doesn’t seem to do anything except amplify! Music sounds the same, but at high volumes there isn’t any distortion in the base and everything in the upper frequencies seems cleaner too. I was coming off an Dayton ada150.

 

[IPunchCholla]

Per Amir benchtest "Sweeping the frequency vs power and distortion we get:"
View attachment 183148

1khz granularity would be nice in this plot but we only get 1khz 5khz 10khz and 15khz. plots. With 1khz plot granularity you would see all frequencies above 2khz are rising in distortion with power.
"normal distortion rise" ? Its not just frequency but power levels as well. Normal Class-D maybe.
This is with a RESISTIVE Load not REACTIVE via speakers or emulation. I am resisting debating this as a factor.
The sinad test is based on a 40 ohm resistance at 1khz.
With this above plot Noise even 1khz @ 5w = -95db @4ohm not 40. 15khz is at -75db
However The TPA32xx(3245 or 3251) is known for issues with reactive loads as it requires LC filtering on its output.
Phase however isn't even tested at any frequencies on ASR.

Here is a paper regarding how Harman tries to deal with reactive(speaker) loads with Class-D using DLA technology.

https://media.flixcar.com/f360cdn/Harman_Kardon-75965644-Digital_Loop_Amplifier_White_Paper.pdf

In short via Harman:
"The key breakthrough is an ultra-fast, low-latency analog-to-digital conversion (ADC) of the analog output signal from the speaker terminals back into the digital control loop. The DLA can achieve speeds of 20 nanoseconds." This done in conjunction on the output side with speakers attached so that reactive load is considered.
Yes companies have marketing hype, but at the same time they do have to innovate sometimes to achieve design needs.

Topping isn't an IC chip designer. They just implement off the shelve parts Like Dacs, op amps etc. There is little talk about how they achieve the PA5.

Understandable for propriety sake but at the same time, how does it really do with "Real World" Speaker loads?

I’m not sure if I live in the real world, but my PA5 seems to do real well. At low levels, I can’t honestly tell the difference to my Dayton ADA150. But turning it up, both the lower end and the higher frequencies sound a lot cleaner. Which is all I was wanting. An amp that was transparent. I don’t care if topping is just plopping pre-existing components down or gaming SINAD. They seem to have made a transparent (for personal listening) amplifier for sub $400. Same for their DACs and pre-amps. My entire stack cost less than $600. And it doesn’t seem to anything except it’s job. At least to my ears, which are normal human 50 year old ears.

I have no doubt Benchmark, Hartman, etc. could do the same. But they haven’t.

 

[IPunchCholla]

I just did this yesterday. I’m not electrically very competent, but it was simple enough. I cut an RCA cord in half. Tip wired to tip. Shield to shield. And then bridged ring to shield at the TRS end. Worked great.

But also RCA to TS also work just as well and should do the exact same thing; flip the shield signal to eliminate any noise picked up on the 6 inch journey from preamp to amp.

I’ve used both on my amp. And they work well.

I wish I had come across the solution before dropping $50 at proaudio LA. Though, I am sure, their cables would be much better.

 

[mocenigo]

However The TPA32xx(3245 or 3251) is known for issues with reactive loads as it requires LC filtering on its output.
Phase however isn't even tested at any frequencies on ASR.

Here is a paper regarding how Harman tries to deal with reactive(speaker) loads with Class-D using DLA technology.

https://media.flixcar.com/f360cdn/Harman_Kardon-75965644-Digital_Loop_Amplifier_White_Paper.pdf

In short via Harman:
"The key breakthrough is an ultra-fast, low-latency analog-to-digital conversion (ADC) of the analog output signal from the speaker terminals back into the digital control loop. The DLA can achieve speeds of 20 nanoseconds." This done in conjunction on the output side with speakers attached so that reactive load is considered.
Yes companies have marketing hype, but at the same time they do have to innovate sometimes to achieve design needs.

That paper deals with digital control loops. So this is for actual "digital amplifiers" which get a digital signal as input and output an analogue power signal. Since there is no DAC, the digital input signal is processed in the digital domain to directly drive the output stages. If you want feedback then you need an ADC to feed back "numbers" from which to compute the error to correct.

This does not apply to the TPA32xx base amplifiers we are talking about here. The input to the chip is analogue.

 

[steve-z]

You have my sympathy - if your purely subjective evaluations are correct, and confirmed by rigorous double blind testing, you've owned some really crappy - or broken - amplifiers.

The only crap one was the Yamaha, all were bought new and none were broken.

 

[antcollinet]

@tonycollinet has already asked you what "noise distortion" is, so we're waiting for your explanation.

Do you also have an explanation on where those 10.001, etc. kHz signals came from? Anyway, even if we assume that they are there, the effect is well studied under the topic of auditory masking. The differences in frequencies will create beats (beat frequency is the difference between the 2 frequencies), which manifest as variations in loudness. You can go look up how strong a 10.001 kHz needs to be next to a 10 kHz for its presence to be audible.

Furthermore, the "beating" effect is not a characteristic of the amplifer (or any other electronics). It is simply the effect of two frequencies drifting in and out of phase, and will happen anywhere that two similar frequencies are present - including in the air.

 

[jokan]

I seriously can't believe that this amplifier needs or deserves 170 pages worth of posts. Probably 20-30 pages are useful.

Thankfully there is an unwatch thread button!

 

[AudioArchitech]

And yet you added another useless post

 

[Rottmannash]

Yes my posts are a bit Didactic,

You are talking about 2nd Harmonic Distortion. Harmonic blends nicely. What about Noise Distortion?
Try playing a 10khz Fundamental signal next to a 10.001 khz signal, maybe a 10.002khz and then a 10.003Khz and slowly sweep up and see what level will you begin to notice?

Beats me...

Do you believe the Purifi modules also exhibit this undesirable behavior as well? You appear to be lumping all Class D offerings in the same flawed basket.

 

[BoredErica]

I seriously can't believe that this amplifier needs or deserves 170 pages worth of posts. Probably 20-30 pages are useful.

Thankfully there is an unwatch thread button!

The signal to noise ratio of pa5 is far higher than this thread.

 

[Abe_W]

Oh no...Oh no woe is me....I spent so much cash on my ATI 6005.

Even if I gave my ATI amp to a ASR scientolog...i mean scientist for free, he wouldn't accept it because the "holy grail" Topping is so much better.

But, since I've seen the Sinad, let me hop on this grand topping bandwagon. Oh wow. Topping Topping Topping. Oh wow wowowow

 

[pkane]

Oh no...Oh no woe is me....I spent so much cash on my ATI 6005.

Even if I gave my ATI amp to a ASR scientolog...i mean scientist for free, he wouldn't accept it because the "holy grail" Topping is so much better.

But, since I've seen the Sinad, let me hop on this grand topping bandwagon. Oh wow. Topping Topping Topping. Oh wow wowowow

Very funny, just not in the way you meant it. Says more about you than about anything else.

 

[antcollinet]

Yet another sleeper account comes to knock over all the pieces, crap on the board, and strut around like he's won.

 

[mocenigo]

I disagree that 1khz sinad would determine if two things sound the same.

To be fair, he said "likely"

 

[Andrej]

Double blind tests are more about distinguishability. In an ABX you hear first "A", then "B" and then "X" and you have to establish whether X=A or X=B. It is not about determining the better amplifier or DAC.

Not true. It is whatever you choose to evaluate. Comparing electronic (audio) equipment is so hard that two approaches are common: can you tell the difference (is A and B the same or different) or the example you provide which includes identification (harder to do). But defining preference (an individual's metric for "better") is easily conducted within the same setup. It is just much harder to get statistically significant results, requiring many, many experiments making it less desirable as an approach. If you fail at the two easier ones, it is a moot point anyway.

 

[J-Sine]

Ok

Noise distortion is not a thing. Noise and distortion are two different phenomena. The plot above is a plot of the sum of noise and distortion.

I appreciate your respectful tenacity in response while remaining neutral and objective to further pose me to clarify.

Noise isn't THDistortion. Albeit it can create a "Distortion" of Fidelity potentially. In practicality we lump these Noise and THD together as undesirable,(metaphorically this chart and I did the same) THD is separate and relates to the fundamental. Humans tend to prefer EVEN harmonics over ODD. Tubes are known for EVEN, Class-D are known for ODD (3rd 5th etc) harmonics. Noise is generally more offensive.

So from this chart we can begin to deduce:

For some reason Sweeping Frequency with Power generates NOISE with the PA5. The NOISE only plot is missing.
This begs further scrutiny as to what Kind of NOISE and cause?
Is it square, saw, triangular or an amalgamation of these at said frequencies/power? Since we can extrapolate it at least begins at 2khz to 5Khz. It seems to be pervasive depending on power levels. In fact it seems frequency and power is correlative. Making it Dynamic Noise. Being Dynamic it will likely alter Percussive sounds more so. Coincidentally snares and cymbals use a lot of the 2khz to 15khz bandwidth.

This PA5 is a benchmark case of Low THD with much higher NOISE. Exposing the Anomalies of this Class-D design but allowing one to hear DYNAMIC NOISE in a isolated way.
The Noise however is likely Power Switching related.
A square wave plot @frequencies above 2khz in 1 Khz steps at various power levels would bring more to light in this situation. Phase plots are needed as well as we can not assume linear phase response in any device.

@tonycollinet has already asked you what "noise distortion" is, so we're waiting for your explanation.

Do you also have an explanation on where those 10.001, etc. kHz signals came from? Anyway, even if we assume that they are there, the effect is well studied under the topic of auditory masking. The differences in frequencies will create beats (beat frequency is the difference between the 2 frequencies), which manifest as variations in loudness. You can go look up how strong a 10.001 kHz needs to be next to a 10 kHz for its presence to be audible.

Forgive the sarcastic humor, I meant no offense. A lot of snide Dunning Kruger exists in forums. All forums seem to breed a Cultism. Only an "A" difference between SNIDE & SINAD.

Furthermore, the "beating" effect is not a characteristic of the amplifer (or any other electronics). It is simply the effect of two frequencies drifting in and out of phase, and will happen anywhere that two similar frequencies are present - including in the air.

Yet NOISE has the potential to cause beatings. Just as some use it on Forums. Level of NOISE does matter. Granted masking can make it unheard.

That paper deals with digital control loops. So this is for actual "digital amplifiers" which get a digital signal as input and output an analogue power signal. Since there is no DAC, the digital input signal is processed in the digital domain to directly drive the output stages. If you want feedback then you need an ADC to feed back "numbers" from which to compute the error to correct.

This does not apply to the TPA32xx base amplifiers we are talking about here. The input to the chip is analogue.

Just some technical insight of an approach to deal with Reactive Loads with Class-D for others to understand it is an actual issue. Since the tests are RESISTIVE on ASR we are limited and use it for ease of reference. Amir can not test every speaker with every amp.

Do you believe the Purifi modules also exhibit this undesirable behavior as well? You appear to be lumping all Class D offerings in the same flawed basket.

Actually I have stated Purifi has made efforts to address Switching NOISE/Distortion/Phase/Reactance to Load in Class-D. I believe the Harman Kardon Citation Class-D has as well. Yet Class-D including the TPA32xx have similar problems as it is an older topology approach. The recently tested Vera exhibits Dynamic Frequency Noise similar to the PA5 yet tests at over a 102db Sinad.

If comparing a TRUE transparent AMP to the TOPPING PA5. It is likely some will prefer the PA5 as it will seem to be "bright clear and alive with sizzle" to high frequency peaks while ignoring sound stage localization, Dynamic Noise, Phase and other anomalies. Just like there are those who like "tube" amps. It is just a new flavor of the same argument and is destined to create a new BIAS/reference for many not willing to put any more effort to understand what they are hearing. Manufacturers will learn the SINAD metric sells gear and will "Game" the system.

The PA5 isn't a Purifi or AHB2 in design philosophy approach. The SINAD metric and substantiating tests by ASR does not show "transparency" and tries to (RATE) by SINAD. Many equate them to being the same. Some will use the TOPPING PA5 as an "End Game". Further Bench tests and understanding of NOISE, Load reactance stability, phase linearity and current transient ability is needed with Class-D.
BENCHMARK, HK and PURIFI seem to consider and implement these important design characteristics in philosophical methodology in concert with SINAD.

However the KISS SINAD metric does seem to mostly work with modern DACS now. DAC I/O is resistive and uses proper built-in filtering without phase, frequency artifact issues, essentially being direct from chip output. AKM and Sabre have made it easier to implement proper "Transparent Design" High Fidelity. NOTE: This applies to their upper tier recent DAC offerings in the last 2 to 4 years.

Class-D however needs more Transparency and Bench Tests then what is presented on ASR to be objective to ascertain Class-D "True Fidelity" I:E Transparency.

"Whats good for the DAC isn't good for the Quack"

Enough of my Dynamic Noise for now...

J-Sine