Topping G5 Review (Portable DAC & HP Amp)

[CedarX]

Yes Hiby Player and with 10 PEQ's but for headphones only (not all filters).
@CedarX and how that have anything to do with it? It has 2x M2 CPU's and two DSP's. One of each to be used for it on OEM side and on 32 bit FP (that's why it's transparent). You acces it trough USB debugging bridge and transfer it to it's memory and then it use it in the path. Nothing what so ever to do with UAC.

I still find strange that Topping would select a quite powerful XU316 USB bridge, channel everything—the data, but also PEQ -settings commands—onto the QCC5125 for PEQ processing… and not even implement PEQ for the BT input of that very same QCC5125. And why did they choose the low-end of the QCC chips—single DSP core, not dual?

But OK, you evidently have access to the architecture/design details on how Topping implemented their PEQ on the DA50-III that I don’t have. I’ll take your word!

 

[ZolaIII]

I still find strange that Topping would select a quite powerful XU316 USB bridge, channel everything—the data, but also PEQ -settings commands—onto the QCC5125 for PEQ processing… and not even implement PEQ for the BT input of that very same QCC5125. And why did they choose the low-end of the QCC chips—single DSP core, not dual?

But OK, you evidently have access to the architecture/design details on how Topping implemented their PEQ on the DA50-III that I don’t have. I’ll take your word!

Because other pair is locked to vendor. I don't have access to what ever Topping is doing but I whose QC developer some time ago.

 

[gvl]

So sounds like enabling PEQ in G5 via new firmware is a possibility? They could offer it as a “premium” upgrade, many people would buy it I’d think. I know this is not how they operate, but perhaps they can consider it.

 

[CedarX]

Because other pair is locked to vendor. I don't have access to what ever Topping is doing but I whose QC developer some time ago.

Just curious: the JDSLabs Element-IV also has PEQ, implemented in the XU316 (it has no BT chip). Why would Topping go through this complex (?) memory transfer from the XU316 to the QCC5125 to implement PEQ… and not simply use the XMOS PEQ libraries like JDSLabs did?

FiiO—I know, not really a “model” for SW development—in their QC15 and BTR17 implements PEQ on the XU316 for USB input, and in the QCC5125 or QCC5181 for BT input. Keeping these two separate PEQ paths in sync, whether you choose to enter your settings through USB or BT, is a nightmare and appears buggy as h@#ll… but, for whatever (good?) reason, they don’t channel everything through the QCC chip.

 

[ZolaIII]

Just curious: the JDSLabs Element-IV also has PEQ, implemented in the XU316 (it has no BT chip). Why would Topping go through this complex (?) memory transfer from the XU316 to the QCC5125 to implement PEQ… and not simply use the XMOS PEQ libraries like JDSLabs did?

FiiO—I know, not really a “model” for SW development—in their QC15 and BTR17 implements PEQ on the XU316 for USB input, and in the QCC5125 or QCC5181 for BT input. Keeping these two separate PEQ paths in sync, whether you choose to enter your settings through USB or BT, is a nightmare and appears buggy as h@#ll… but, for whatever (good?) reason, they don’t channel everything through the QCC chip.

I didn't know, didn't find any software EQ for XU316, I did look at dataset. It's tile based synchronous micro architecture (5 to 8 32 bit embedded cores per tile and two tiles) and ill based for the task (hard to split workloads so you have to deticate entire one) and low on OPT (4KB per tile). You can do all audio processing you ever dreamed of on general purpose code and instructions with use of MTU on 4x A7 Arm cores at 1 GHz (old V4A requirements) tho with lower efficiency than on M or R cores and to actual DSP optimised for such tasks but without detail documentation (which you certainly won't get from QC) it's impossible to really program DSP.

 

[ehabheikal]

And what about USB Audio Player Pro app ?
It works with Qobuz.
Drawback: Not for Qobuz downloaded files.

Oops, that's Android only.

Isn't there something similar for iPhone ?

No eq/peq system wide app on iphone.

 

[CedarX]

I didn't know, didn't find any software EQ for XU316, I did look at dataset. It's tile based synchronous micro architecture (5 to 8 32 bit embedded cores per tile and two tiles) and ill based for the task (hard to split workloads so you have to deticate entire one) and low on OPT (4KB per tile). You can do all audio processing you ever dreamed of on general purpose code and instructions with use of MTU on 4x A7 Arm cores at 1 GHz (old V4A requirements) tho with lower efficiency than on M or R cores and to actual DSP optimised for such tasks but without detail documentation (which you certainly won't get from QC) it's impossible to really program DSP.

Seems pretty well documented to me:

lib_audio_dsp: Audio DSP Library v1.3.0

www.xmos.com

And although I’m not entirely sure, an XU316 may be required, so won’t work on the G5:

 

[ZolaIII]

Seems pretty well documented to me:

lib_audio_dsp: Audio DSP Library v1.3.0

www.xmos.com

And although I’m not entirely sure, an XU316 may be required, so won’t work on the G5:
View attachment 438552

It's not a DSP, and it's more than order of magnitude less efficient. In place Pithon writen macros are still highest level they let you.
This is same lv pre writen in QACT:

To my surprise list of filters even has tilt.

 

[CedarX]

It's not a DSP, and it's more than order of magnitude less efficient. In place Pithon writen macros are still highest level they let you.
This is same lv pre writen in QACT:View attachment 438586
To my surprise list of filters even has tilt.

May I ask where can I get this QACT tool? I have tried to register at Qualcomm before and, like anybody who is not a major OEM, my request was politely declined.

 

[ZolaIII]

May I ask where can I get this QACT tool? I have tried to register at Qualcomm before and, like anybody who is not a major OEM, my request was politely declined.

Try to find one deliberately leaked from one of such OEM's.
Edit: don't expect much from Qualcomm and if you want to see how well documented looks alike go ahead to CEVA or Tensilica.

 

[gvl]

It's not a DSP, and it's more than order of magnitude less efficient.

Why isn’t it? Sounds like you control it through Python but it must be implemented in compiled language C/C++. Or you mean it’s software and not something like FPGA or dedicated DSP hardware?

 

[ZolaIII]

Why isn’t it? Sounds like you control it through Python but it must be implemented in compiled language C/C++. Or you mean it’s software and not something like FPGA or dedicated DSP hardware?

I mean what I wrote about it's architecture, it's tile based sets of 32 bit controllers (CPU's) with deticated scheduler and you can pack it (multiple instructions in long word (up to 256 bit for 8 32 bit CPU's) but you can't split it in separate workloads. So it's Arm all do they didn't provide which cores as they aim whose performance I suspect M4 and it's really too much to study scheduler right now but I will (as I used to do that). Still general purpose CPU instruction set not a DSP.

 

[gvl]

Ok, not my field, but I get it, it’s a general purpose architecture not designed specifically for DSP. However, for the purpose of this discussion I think we can agree that it has enough power to run 10-band PEQ, and there’s a ready to use library that implements it. However it Sounds like Topping didn’t use it on the D50III for whatever reason, like the license cost perhaps.

 

[ZolaIII]

Ok, not my field, but I get it, it’s a general purpose architecture not designed specifically for DSP. However, for the purpose of this discussion I think we can agree that it has enough power to run 10-band PEQ, and there’s a ready to use library that implements it. However it Sounds like Topping didn’t use it on the D50III for whatever reason, like the license cost perhaps.

I talked about XMOS. QC is M2 and DSP. It whose always like that (with all of them), they give you DSP that does something trivial or doesn't do anything at all. Only time people beat them (and got what they want) whose with Creative Audigy reverse engineer drivers (long time ago). I will experiment with QC when I have time and if I manage to get it to work without too much hassle I will write about it so that others also can.

 

[mike7877]

Below a bit of words are some pics of G5's guts
This cadaver's untimely end was voltage related - there was too much of it where it shouldn't have been!
Unfortunately, due to shock and disbelief, after the tragedy, those around G didn't think to use the soscilloscope to gather specifics to share

Looking at the chart above, you'll see that, unfortunately, G5's gain specs are unlike the ones given for Topping's other amplifiers. Above, circled in red, are the decibel amounts of gain for the Low Medium and High settings. Usually we are given the voltage equivalent on the input for 0dB amplifier output.
I'm going to go out on a thin limb, way up-high in the sky, of a pyramidal oak nearing the end of its life and say:
if there's a -8dB setting, +8dB should be safe.
What's +8dB? It's a bit more than 2.5x the voltage.
But, now... what is full scale, then? The G5's DAC analog output is specified to 2.1V RMS, so we're going to use that:
2.1V * 1/0.3982 ? It's 5.274V
OK, so on the analog input, you should be able to apply 5.2V.

This isn't for sure for sure, but I'll say, there was a good couple hours where I had somewhere in the range of 2-5dB above 2V on the G5's input for a couple hours and things seemed fine. It could still be that over 2.1V is bad for the input and I was slowly doing damage at +2-5dB for that time. Well, unless Topping decides to fill us in, I'd say, it's best to stay under 2.1/2.5V when you can (it should be pretty easy - most sources won't exceed it anyway, except maybe preamps (for power amps of speakers). Also, most things that can give you more than 2.1/2.5V will have sufficient SNR/DNR at 2.1/2.5V anyway, so don't think it's imperative you get the source close to its full-scale output all the time. If you care about your G5 lol xD

What does a G5 sound like when he's cooked?
Oddly, a LOT like a busted woofer, but if that woofer could play full spectrum audio.
...the way that it doesn't get very loud and the way that it sounds a little rattly (even though it's not moving...), it's uncanny

I had read that op-amps have over-voltage protection, so I wasn't worried about my G5.
G5? Input? Clearly going to be an op-amp of some kind. Knowing Topping, probably an OPA1656 or OPA1612 or something..
Well... it turns out... those internal diodes aren't for an unrelenting "barely-over-voltage" signal, no... they're not. Those diodes... they're for static discharge! Or maybe connecting to a device which has a small fault, something like a high impedance source for high voltage - once connected, this 50-100V surface charge is on the input of the op-amp. But it gets sent to ground by the diodes. And then, once the two devices are fully connected, the extraneous high voltage is no-more - brought to ground by whatever


As promised, the pictures:

look at all the caps! I don't have a picture of the other side of the board - it's mostly empty, though, save for a few more caps than you can see here, the in/out jacks, volume control, and a big flat empty space for the battery to butt up against.

Lookin' good! This board is actually tiny. The blue daughterboard - the width of that 4 pin chip, is a little smaller than the width of a standard op-amp. The height is more than a little shorter... The turquoise volume pot, immediately to the left of it is an op-amp surrounded by eight black resistors of 2490 and 4300 ohms - that square op-amp's dimensions are 3.3mm wide x 3.3mm long x 1.1mm high. In literally one square inch you could fit 36 of those OPA1612s (if their pins were underneath)

There's a closer look near the DAC - you can see the 1331 ohm resistors on either side of the 1612s, marked OVII TI 131. The G5 uses a still rare ES9068AS DAC chip. It also has a decent ADC included which we don't get to know about because the 9068AS datasheet is kept from us


So yeah, don't cook your G5 like I did...
Sorry the pics aren't higher resolution. They were 6000x4000, but the forum software limits a picture's largest dimension to 1500. If your image is over 1500, it shrinks your image for you. One trick to keeping picture quality is changing the size yourself with some proper software. I like irfanview - you can choose the resampling filter - though the better ones take more CPU time, it's still only a couple seconds on your desktop. Server software is optimized to save power and time, with quality often taking a back seat. These pictures, when I first uploaded them and the site had its way with them, looked absolutely terrible. These I served, with no modifications necessary, and now they're here, and beautiful

 

[antcollinet]

What makes you think it is the input voltage that killed it? Unless you've diagnosed the actual failed components, it could be any random failure.

Normally I'd expect input resistors to protect analogue inputs from over-voltage - unless you've really gone extreme.

 

[mike7877]

What makes you think it is the input voltage that killed it? Unless you've diagnosed the actual failed components, it could be any random failure.

Normally I'd expect input resistors to protect analogue inputs from over-voltage - unless you've really gone extreme.

I didn't confirm 100%, but I know I was overvolting the dang thing at the time it broke, and until that time it had never received such voltages. And it'd been on and used a lot lot lot until then.
Whether it's the analog input, bluetooth connection, or USB to the DAC, the sound is the same, distorted, quiet, scratchy, like a fried woofer almost. Not even with sensitive IEMs at the lowest level does the signal come out proper. I'm fairly sure it was the preamp input section that got fried - whether it's the headphone output or the higher impedance "line out" output, the sound is the same: broken

 

[Robbo99999]

I didn't confirm 100%, but I know I was overvolting the dang thing at the time it broke, and until that time it had never received such voltages. And it'd been on and used a lot lot lot until then.
Whether it's the analog input, bluetooth connection, or USB to the DAC, the sound is the same, distorted, quiet, scratchy, like a fried woofer almost. Not even with sensitive IEMs at the lowest level does the signal come out proper. I'm fairly sure it was the preamp input section that got fried - whether it's the headphone output or the higher impedance "line out" output, the sound is the same: broken

How did you overvolt it & why did you do so? Were you using a power supply that was giving more than 5V?

 

[mike7877]

How did you overvolt it & why did you do so? Were you using a power supply that was giving more than 5V?

I decided to outline how its internals are arranged by describing its behaviour to answer your question.

The G5 takes 5V from USB and there's a boost converter that runs a 2 cell Li-ion charging circuit. Whenever the USB has power, the internal lithium battery is charging (until it's full, of course... it would have been a much better device with an added switch which let you disconnect the battery and run off USB power. I think they didn't do this because the boost converter they used on the USB 5V is only good for about 1.4A (at 5V, proportionally less on the other end (if it was 10V, output would be 0.7A less 10-15% for efficiency loss. Topping could have used a more powerful USB power chip, but the extra power could become a liability with some devices (some could literally break). Additionally, even if it was fine, I think the lithium battery provided even more DC filtering which wouldn't be available.
Anyway, the battery is two 3.7V cells in series, after which there is no regulation before the final stage of headphone amplification (BUF534 I believe). Two 3.7V batteries in series will provide 7.4V nominally, and the device will run with the battery anywhere between 8.4V and 5.8/6.0V. The lower the battery, the less the peak power output capability. The red low battery light is the most basic indicator - when battery voltage falls to below the low 6V range, it goes to red. When driving lower impedance headphones, bass drum/kick can cause the power light to flick to red when the battery is on the lower side.

Overvolting is technically the incorrect word to use where I did, but I used it the way I did after previous context reasonably explained its use (if you noticed, the post is very long - as much as I like to be completely accurate with what I say, and include everything I reasonably expect would be a question, things become wordy when you do. The longer something is and the more it needs to be shortened and kept interesting, the more work it is to write. Because of this reality, individual posts can only be so long or so interesting...There really is no limit, but eventually you're scrapping entire paragraphs because moving one thing to the paragraph before means that scrapped paragraph can be shortened two sentences and will flow way better. Usually before I get to that extreme I'll just remove relevant context or important reasoning, even though it's definitely going to come up later lol.

Anyway... Last thing - the speed of charging is dependent on how much the voltage on the USB charge port droops. If it stays at 5.13V or above, 1.4A flows the entire time until the battery nears full (when it slowly tapers to nothing). If you're using an old charger that's only giving 4.8V, or a thin, long USB cable that's wasting 0.6V over the 10ft run, at 4.5V only 200mA flows (about the lowest that will flow). At 4.85V about 700-800mA flows, and if you were to guess 4.98V gives you a bit over 1.0A you'd be right as well. This externally applied voltage has no impact on the running voltage of anything - just the current of the battery charge

 

[Robbo99999]

I decided to outline how its internals are arranged by describing its behaviour to answer your question.

The G5 takes 5V from USB and there's a boost converter that runs a 2 cell Li-ion charging circuit. Whenever the USB has power, the internal lithium battery is charging (until it's full, of course... it would have been a much better device with an added switch which let you disconnect the battery and run off USB power. I think they didn't do this because the boost converter they used on the USB 5V is only good for about 1.4A (at 5V, proportionally less on the other end (if it was 10V, output would be 0.7A less 10-15% for efficiency loss. Topping could have used a more powerful USB power chip, but the extra power could become a liability with some devices (some could literally break). Additionally, even if it was fine, I think the lithium battery provided even more DC filtering which wouldn't be available.
Anyway, the battery is two 3.7V cells in series, after which there is no regulation before the final stage of headphone amplification (BUF534 I believe). Two 3.7V batteries in series will provide 7.4V nominally, and the device will run with the battery anywhere between 8.4V and 5.8/6.0V. The lower the battery, the less the peak power output capability. The red low battery light is the most basic indicator - when battery voltage falls to below the low 6V range, it goes to red. When driving lower impedance headphones, bass drum/kick can cause the power light to flick to red when the battery is on the lower side.

Overvolting is technically the incorrect word to use where I did, but I used it the way I did after previous context reasonably explained its use (if you noticed, the post is very long - as much as I like to be completely accurate with what I say, and include everything I reasonably expect would be a question, things become wordy when you do. The longer something is and the more it needs to be shortened and kept interesting, the more work it is to write. Because of this reality, individual posts can only be so long or so interesting...There really is no limit, but eventually you're scrapping entire paragraphs because moving one thing to the paragraph before means that scrapped paragraph can be shortened two sentences and will flow way better. Usually before I get to that extreme I'll just remove relevant context or important reasoning, even though it's definitely going to come up later lol.

Anyway... Last thing - the speed of charging is dependent on how much the voltage on the USB charge port droops. If it stays at 5.13V or above, 1.4A flows the entire time until the battery nears full (when it slowly tapers to nothing). If you're using an old charger that's only giving 4.8V, or a thin, long USB cable that's wasting 0.6V over the 10ft run, at 4.5V only 200mA flows (about the lowest that will flow). At 4.85V about 700-800mA flows, and if you were to guess 4.98V gives you a bit over 1.0A you'd be right as well. This externally applied voltage has no impact on the running voltage of anything - just the current of the battery charge

Ok, cool, you've described how the unit works in terms of power delivery but you've not explained how you "over volted" it? I quick checked your longer post with the pics and couldn't really see how you'd "over volted" it. You don't have to go into mega depth - what did you do replace some internal components, re route something, use a different power source, what?