• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

Topping A90 Discrete (A90D) - Teardown and personal thoughts

trl

Major Contributor
Forum Donor
King of Mods
Joined
Feb 28, 2018
Messages
1,995
Likes
2,616
Location
Iasi, RO
I found not many teardowns with this model, especially with the output transistors uncovered and with surrounding close-ups, so I decided to purchase one for myself (second hand) to pair it with my D90 MQA then write an article here on ASR. Looks like taking the PCB out from the case if somehow difficult, needing careful attention to the two ribbon cables that go to the rotary knob and to the LCD controller from front of the case, but also needing an unusual long and thin screwdriver that actually fits between the power supplies and side of the case. Assembling everything back is even worse, with a headlight on my head and a not-long-enough screwdriver, so I don't recommend anyone to tear apart the A90D, unless you're fine with possibly breaking it or losing its warranty.

The Topping A90 Discrete was reviewed by Amir here: https://www.audiosciencereview.com/...0-discrete-review-headphone-amp-preamp.35114/ and with it’s SINAD of over 119 dB, a max. output power of 6 Watts on low-impedance headphones and a perfect channels volume match, it was clearly measuring exceptional.

333AB8BD-A196-4F05-A635-9A5756CCF7A5.jpeg

Topping A90 Discrete (A90D) - top view of the PCB


Right beside the power plug there is a Metal Oxide Varistor @ 400 V that provides surge protection by tripping the MCB from your electrical panel in case an electrical malfunction occurs upstream your home (e.g.: overvoltage, lightning etc.). Even if just a simple MOV will not make wonders in case of small power spikes on the mains, I'm happy to see there’s a MOV inside this headamp because it should offer a decent protection for the device when getting 400 V or more on the mains and should also minimize the risk of fire due to overvoltage.

C06A3A4E-CD85-4976-9A7D-657E5EDD8FD8.jpeg

Metal Oxide Varistor (MOV) used for power surge protection


The unit gets clean power from a couple of internal AC-DC encapsulated power supplies from a reputable manufacturer, Mean Well IRM-20-15, that deliver +/-15V @ max. 1.4 A. These are isolated power supplies with very good filtering inside and the measured ripple & noise below 50 mVp-p (per https://www.meanwell-web.com/content/files/pdfs/productPdfs/MW/IRM-20/IRM-20-15-rpt.pdf. The power supplies are protected from overvoltage (in case of malfunction internally) and short-circuit too. Isolation type is Class II, meaning that these power supplies can be safely used in electrical equipment with or without grounding, customers not being affected in case a malfunction inside the supply will occur. This is a great thing that alllowed Topping to easily implement in A90D the "ground lifting" feature that helps in cases where audio hum is audible due to grounding resistance difference between interconnected audio equipment, where electrical currents are flowing into signal wires due to this.

8AA8EA4D-A933-4B69-A781-33C8B63F1A1B.jpeg

Mean Well AC-DC power supplies


We can also spot on the PCB three linear regulators ON-SEMI LM317D2T that power ON all the "logic" from the board that takes care of proper operation of the amplifier, relays and protections, but also the front panel lights and logic commands.

0AAF029C-A814-4305-B6FC-354C52020455.jpeg

LM317 linear power regulators (used for logic part)


Volume control is fully digital, with a resolution of 100-steps, and is done around a rotary knob and an array of resistors and stepped relays, a choice rather difficult find in headamps on this price range. Relays are pretty low in noise, I can hear the noise made without headphones, but comparing it with let’s say Flux Volot the A90D is making like half of the noise when changing the volume up and down.

The relays used are 2-pole OMRON G6K-2P-Y @ 12 VDC with a contact resistance lower than 100 mΩ, ensuring a virtually zero voltage drop across the 2-poles of each contact. There are also fifteen small SOT-23 case SS8050 transistors that are taking care of providing power to the OMRON relays.

FA78C0F1-915A-4CBC-A25A-3A23F7352C19.jpeg

Small SMD transistors array takes care of powering the relays from the PCB board


The rotary knob does not only controls the output volume, but it is also used to select the right audio input (single press), to choose between PRE and/or HEADAMP operation (double press) and, of course, it's used for powering the unit ON (short press) or OFF (long press).

Between the bottom of the case and the backside of the PCB there is a protective isolation film that helps the soldering from the backside of the PCB to touch the bottom of the metallic case.

F1468528-215B-4C66-8B94-08503D783215.jpeg

Protective isolation film


5E6F6330-B0A8-4431-B339-F6A3DF060A07.jpeg

Copper grounding point between PCB and outter case

The front display is controlled by a TM1638 chip that is widely used for similar tasks in Arduino compatible devices, see https://www.instructables.com/Arduino-and-TM1638-LED-Display-Modules/ and https://www.handsontec.com/dataspecs/display/TM1638.pdf on how this chip is widely used for controlling LED displays in Arduino projects. There are libraries for ease of programming this chip and it’s nice and unexpected at the same time to see it used in a headphones amplifier.

F5118A32-8EDA-427A-878E-FE09A8B58200.jpeg

Front LED controller

The output stage is created around four pairs of complementary PNP/NPN BJT BISS (Breakthrough In Small Signal) transistors: PBSS305PZ and PBSS305NZ. These are bipolar transistors with a very low Vce-saturation, meaning that heat dissipation will be lower due to a better efficiency. Transistors temperature is controlled by the PCB itself, but also by a small aluminum heatsink attached on top via double-sided self adhesive thermal pads secured with three screws by the PCB. These small and fast transistors are able to dissipate around 1.5 Watts each, max. amperage driven being up to 10 Amps (short term).

D271B6EE-CCFC-4FEF-B727-D4487C75EF11.jpeg

The eight output stage transistors (covered with thermal conductive tape)
and the corresponding goldish electrolytic
decoupling capacitors


CD331896-51FC-40D6-A84A-5A7F3284A48B.jpeg

One of the four complementary pairs of the output stage transistors


All the logic and controls are takes care by a dedicated ARM Cortex microprocessor operating up to 48 MHz STM32F072C8. It has a CAN bus, but also an ADC and a DAC inside, so it can accept both analog and digital signal inputs at the same time. This way the microcontroller is able to read all the states about A90D’s operation like stand-by, volume level, gain, inputs & outputs, inside voltages, DC output etc. If a reading goes wrong then an error code is throwen on the LCD and device might enter in Stand-by.

There is also a UC25HQ40 that is a 4 Mbits ultra low power memory that communicates on SPI. This flash memory is probably used for storing the states for the front panel commands like In/Out, Pre/HPA, volume level, the two memories C1 & C2 etc., while A90D's firmware is stored inside the flash memory from the above STM32F072C8 microcontroller.

3B18AC11-7750-4E54-8AD8-CED935585AE4.jpeg

The STM microprocessor along with the memory flash chip


For DC-output protection there is a TL072C op-amp that detects any trace of DC voltage on headphones outputs and communicates this to the microprocessor that is instantly placing the amplifier in protection mode (-5) and cuts off the power.
33628E2C-577E-4DB7-BDC6-328776E4D45F.jpeg

TL072 op-amp responsable for triggering the DC-output protection


Preamplification and gain stages are built inside four NFCA (Nested Feedback Composite Amplifier) modules, each one containing several SMD transistors and passive components too Close-up pictures will be soon revealed once the new version of A90 Discrete amplifier will appear on the market, given that Topping decided to get rid of the outter case for its NFCA modules (probably due to this). Till then, below is an inside picture of these modules grabbed from manufacturer website.

IMG_0092.jpeg


The big blue adjustable multi-turn resistors from top of the NFCA modules are there for adjusting the output DC by the manufacturer. Those adjustable resistors are sealed on top and trying to manually adjust any of them will add DC voltage on the output of the amplifier and pops and click may occur, but this will also void the warranty of the device! I measured 0 mV (zero mV!) on the XLR 4-pin plug, on both channels, with no music playing, so my A90D headamp was perfectly adjusted by the manufacturer. However, if you hear any random pops and clicks on the A90D headamp, just email the manufacturer and wait for a resolution on this, but don’t try to handle this by yourself by adjusting those resistors!

8A976425-B534-4467-B580-4217BEC63D42.jpeg

Multi-turn resistors responsable for adjusting the DC-output by the manufacturer


The inside temperatures are quite good, hottest spot was the side with the outputs from the Mean Well power supplies at around 47-48 C, while the other faces of the power supplies were having temperatures between 35 and 41 C. The heatsink from the output transistors was oscillating between 40 and 42 C and the NFCA modules were pretty contant between 39 and 40 C. With case closed the above temps will definitely raise a little bit, probably with 5 C or maybe a bit more, but given that the top of the case stays below 35 C, I consider this headamp as being on the warm side rather than on the hot side like other headlamps I own (Matrix HPA-3B and Burson 3X GT 2023 are getting to the range of 45-50 C on the sides or top of the case, depending on the output volume).

0A7C50B9-21EE-41BC-912B-2A7842EBA516.jpeg

Temperature measured directly on the heatsink


5428DEBD-24C1-46EA-A723-012379FF6EAC.jpeg

Temperature of the output stage while simultaneously driving with high SPL
two planar headphones (70 Ohms LCD-2 Fazor and 25 Ohms Fostex TH-X00 ebony)


F9B9EE37-A166-40ED-92BB-20B8C19B50D0.jpeg

Temperature on the NFCA modules​


34A0B0E6-366D-4351-A1A4-F51D879108A8.jpeg

Lowest temperature on the power supplies​


F8F00ADD-BAE6-4E37-B6AC-5B15EC2FCBDF.jpeg

Highest temperature on the power supplies


PROs:
  • SOTA headphone amplifier with a SINAD of almost 120 dB is more than ears I would ever need.
  • Matches all the planar and dynamic cans I have on the lowest gain, even my 600 Ohms balanced DT880 were driven at very high levels with lowest gain and volume set between 70...80 out of 99.
  • Perfect channel matching while zapping between volume states, thanks to the relays array that takes care of adjusting the output volume in 100 output levels.
  • The PREAMP function helps me drive active speakers and control the output level as well. I am more than happy with the PRE + HPA setting, because this gives me the option to connect the active speakers to the PRE-Out, while before that I needed to swap the XLR cables from the external headamp with the ones from the speakers to be able to listen to speakers or to use a passive In/Out XLR device. Now, since I got the A90D, all I need to do is to switch between active speakers and headphones (or have them both) from the remote control.
  • The top of the case doesn’t get hot like I was used to with most headamps I had or I still have around. I’m talking about below 35 C after hours of continuous operation in a 21- 22 C bedroom, and having a D90 DAC on the bottom of the A90D that also gets some heat generated too.

CONs:
  • The vertical angle of the remote control is just too narrow for my taste. I measured about 20 degrees, while my D90’s remote has more like 160 degrees on the vertical plane. Horizontally instead I get over 170 degrees for the A90D while the D90’s remote has almost 160.
  • Also, I would have liked it more if the remote's volume control would be more fluent, more linear, instead of jumping so fast between volume states after a long press of about half a second. So, I would like to see a lower waiting time between when I firstly press on the + or - button and when the action becomes repetitive (maybe a quarter of a second instead of a half), but at the same time a lower speed in incrementing/decrementing the volume states would be lovely too. Hope a new firmware update might be able to help with my request, although I’m aware that most customers might be more than happy with how the remote control works right now.
  • I don’t think I’m quite in love with the ease of switching to the Highest gain setting for the PREAMP, although I am aware that this is a feature implemented in many PREAMPs, but I'd prefer a harder way for passing over 4-5 V RMS on the XLR 3-pin output plugs (a long press on Gain or maybe a hard switch). I’m more...safe in this area and I’d prefer to stick with a max. output level of 4-5 V RMS for my active speakers, so I’d prefer to increase the gain not so easily for the PRE and I think for the HPA too.
 
Thanks for this. I love it! The commentary on the PCB, the components, the photos, close ups - all of it. If you went into even more details about the circuitry, I would not complain.
 
Thanks! I always enjoy your teardowns!
I have a question: you mention the volume control is digital and you mention some relays... Do you mean that the stm32 controls the relays? Or a different IC?
What sort of encoder do you think the volume knob is? Thanks
 
  • Like
Reactions: trl
I see no other microcontrollers on the board, so I would say that the STM chip handles the array of volume relays as well.
 
  • Like
Reactions: MCH
I am aggregating here a few screenshots from the Operation Manual with specs and error codes too, for the posterity. :)


2F142E33-58B3-4FA4-AE19-52CEEC381A9C.jpeg

Front panel


C1119E48-28AE-4107-B230-E8B8DA118152.jpeg

Rear panel


95F96B9E-92E6-4AB5-84C0-23F551E0FE25.jpeg

Remote operation and Specs


29911EF3-C2B6-4D6A-8250-13C5C37AF7C0.jpeg

Error codes
 
Nice detailed tear-down. I'll add that I have some equipment that uses those Meanwell switched power supplies. Their switching frequency is 65 kHz.
 
  • Like
Reactions: trl
Thanks for the comment @MRC01, below is a pic with the internal diagram and on top-right there's Fosc written there as well, same 65kHz you were talking about. Although this frequency is inaudible, as we can see on the PCB, the two power supplies are followed by LC filters that act as low-pass filters, so there will not be any significant frequency to be seen on an oscilloscope in this particular case.

07C17A8C-44E6-48BF-9915-55D1D1C71DA9.jpeg

Mean Well power supply - Block diagram
 
Indeed, that is why I mentioned it. A while back I was reviewing & testing some new prototype gear, discovered a spurious 65 kHz tone and chased it down to this root cause.
 
I see no other microcontrollers on the board, so I would say that the STM chip handles the array of volume relays as well.
With all those relays, switches, external memory, usb interface, display, inputs.... they must be using every single of its 48 pins!
 
With all those relays, switches, external memory, usb interface, display, inputs.... they must be using every single of its 48 pins!
And I wonder how that impacts longevity, compared to an old-fashioned amp having just a power switch and analog pot volume knob.
 
Very interesting teardown. Thanks for posting!

The vertical angle of the remote control is just too narrow for my taste. I measured about 20 degrees, while my D90’s remote has more like 160 degrees on the vertical plane. Horizontally instead I get over 170 degrees for the A90D while the D90’s remote has almost 160.

That seems to be an absurdly common problem with most of the SMSL and Topping devices. Probably affects Sabaj and Loxjie and all the other "sister brands" as well. It's such an annoying oversight: It's 2023, we're measuring gravitational waves. IR remotes should not be a challenge anymore. That being said, if you want to keep your A90D, you could look into modding the remote like I did on my AO200. It didn't perform any miracles, but it helped. I assume that the main bottleneck in my case is the positioning of the IR receiver in the amp, but it might be different for you.
 
@RandomEar I already scheduled to open the remote for this weekend, so many thanks for the insight on this matter. However, in my case the IR works perfectly on the horizontal plane, but much more narrow vertcally.
 
The vertical angle of the remote control is just too narrow for my taste. I measured about 20 degrees, while my D90’s remote has more like 160 degrees on the vertical plane. Horizontally instead I get over 170 degrees for the A90D while the D90’s remote has almost 160.
A70 pro and D70 pro Sabre has very good angle for remote control. No problem at all.
If anyone is wondering.

ps. I could even point on wall/ceiling in my room and it still worked fine.
 
Last edited:
With all those relays, switches, external memory, usb interface, display, inputs.... they must be using every single of its 48 pins!
I couldn't find many chips on the board, maybe worth mentioning the two OPA1612 from below pics that are probably used as audio buffers.


21DADCF1-D6FF-4F1D-807B-FD5991E3BF59.jpeg


812E9027-CA64-448F-8D12-D2784EB9EBE4.jpeg


I'm no way microcontroller programmer, but I was playing around with Microbit and Arduino a couple of years ago and remember how easy was for me to put the GPIO pins on Low or High and do my thing with multiple relays at once. Given that the STM chip from the A90D has 37 GPIOs, according to the datasheet, I would say that handling a few relays from this board would not be an issue.

4FD44796-EBEE-4210-B2AF-1E0C20B26C18.png

Microcontroller pinout

4E72269A-FFE3-4A8A-8071-5C47A42B7560.jpeg
 
any chance of a teardown guide? I want to teardown mine to drill holes on top. Thanks in advance.
 
Very interesting teardown. Thanks for posting!



That seems to be an absurdly common problem with most of the SMSL and Topping devices. Probably affects Sabaj and Loxjie and all the other "sister brands" as well. It's such an annoying oversight: It's 2023, we're measuring gravitational waves. IR remotes should not be a challenge anymore. That being said, if you want to keep your A90D, you could look into modding the remote like I did on my AO200. It didn't perform any miracles, but it helped. I assume that the main bottleneck in my case is the positioning of the IR receiver in the amp, but it might be different for you.

I grabbed out the LED a bit more, to look like in the picture from your article, given that my LED was inserted a bit deeper into the remote. I've also added reflective sheet around the LED to increase its light reflection and efficiency and now it does work a bit better, so thank you!

Comparing the Topping remote with the one from my Apple TV 3, the Topping one is definitely a winner due to a wider angle on both horizontal and vertical axes, but also being easier to press on the buttons.

E8EBA775-AB8D-4A62-B024-6B632DA4EA34.jpeg



6B67BC80-579D-46B0-A847-8008426CAD3D.jpeg
 
any chance of a teardown guide? I want to teardown mine to drill holes on top. Thanks in advance.

Everything is straight forward, but after taking off the back plate you will need to figure out a way to easily disconnect the two ribbon cables coming from the rotary knob and display.

Worst part is to get the two screws from the front plate off, given that you'll need a very long and very thin screwdriver (the one from iFixIT is shorter and thicker, even with the longer adapter included).

Chances to damage the ribbon cables and/or to not be able to put everything back are quite high, so I strongly don't encourage anyone to teardown this device!

BTW, why drilling the top case for? I measured quite low temps on top, even after driving at insane levels two planars at the same time the temps were like around 40C. My Matrix HPA-3B was hitting 60C when I was driving two planars for testing. :)
 
Thank you Kevin, so @amirm was right in his review that each NFCA module is having 39 transistors inside...I just counted them all now. :)

If I'm seeing this right, looks like the OPA1612 op-amps I was talking about earlier are the input buffers for the 4-pin XLR plugs.
 
Back
Top Bottom