- Thread Starter
- #21
I see you are not fond of the current limiter solution
-
Welcome to ASR. 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!
"Small form factor" modular multichannel TPA3255 amp build
- Thread starter MCH
- Start date
- Thread Starter
- #23
Will do like this: leave the footprints for a power input that goes through the current limiter IC. If necessary, I can populate and use it. If not, I can ignore it and use any of the other connectors that are directly connected bypassing the current limiter.
(Now that I've read the datasheet I want to have to use it)
- Thread Starter
- #24
Made some advances, and i think i am almost ready to lay down the traces (changes will be more difficult after that).
I implemented the in-rush current limiter circuit discussed above. It will be optional: will leave the footprints at the bottom side that i will populate only if necessary. In the case i need to use it, i just need to solder the necessary parts and use the new connector (below the 5 others) for the power input.
I have also decided that the amp PCBs will be printed with 2oz copper but the control board will be 1oz only, to reduce cost. This PCB will be the one carrying more current, but only in one single trace that is very accessible and with no components on top, so i decided to leave the power bus trace without solder mask and tin it manually. Will be the first time i do this, i hope it works. I don't know if there is a problem with this approach
Another thing i have been dealing with is how to connect the heat pipe to ground:
TI explains that the heat sink (in my case heat pipe) must have a good connection to the PCB ground, the reason being it is a noisy device and could increase the noise in the amp. It is not enough to leave the connection to ground depending on the contact of the heat sink with the pad on the chip (that is also connected to ground). Their recommendation is to use heat paste between chip and heat sink + screws between heat sink and PCB ground.
Obviously, i cannot do that with a heat pipe.
At first i thought that i could use liquid metal instead of heat paste between chip and heat sink to grant a good electrical contact (I would have to coat the the copper pipe, but that should be no problem). But i could not find any information about the material/compatibility of the pad on the chip nor its exact shape, so i abandoned the idea.
But this morning i thought about using a couple of spring load contacts. The ones indicated with arrows in the picture below have a height of 1.5mm when free and are aimed to work at 1.2 mm when pressed, that is the exact height of the chip. Let's see how they work.
Another detail i need to deal with the is, potentially, impedance matching of the oscillator pins. I will be connecting the clock of the master board to the other boards through cables. I will have to find out what will be the best way to avoid potential issues beforehand. The frequency is not very high, but the distance between chips yes. TI does not make any provision about this in their evaluation board, maybe it is a non issue. I am not sure if the 3E audio boards (that expose the connectors to set their boards in master/slave mode) have some sort impedance matching. Will have to see what i can find out online....
I implemented the in-rush current limiter circuit discussed above. It will be optional: will leave the footprints at the bottom side that i will populate only if necessary. In the case i need to use it, i just need to solder the necessary parts and use the new connector (below the 5 others) for the power input.
I have also decided that the amp PCBs will be printed with 2oz copper but the control board will be 1oz only, to reduce cost. This PCB will be the one carrying more current, but only in one single trace that is very accessible and with no components on top, so i decided to leave the power bus trace without solder mask and tin it manually. Will be the first time i do this, i hope it works. I don't know if there is a problem with this approach
Another thing i have been dealing with is how to connect the heat pipe to ground:
TI explains that the heat sink (in my case heat pipe) must have a good connection to the PCB ground, the reason being it is a noisy device and could increase the noise in the amp. It is not enough to leave the connection to ground depending on the contact of the heat sink with the pad on the chip (that is also connected to ground). Their recommendation is to use heat paste between chip and heat sink + screws between heat sink and PCB ground.
Obviously, i cannot do that with a heat pipe.
At first i thought that i could use liquid metal instead of heat paste between chip and heat sink to grant a good electrical contact (I would have to coat the the copper pipe, but that should be no problem). But i could not find any information about the material/compatibility of the pad on the chip nor its exact shape, so i abandoned the idea.
But this morning i thought about using a couple of spring load contacts. The ones indicated with arrows in the picture below have a height of 1.5mm when free and are aimed to work at 1.2 mm when pressed, that is the exact height of the chip. Let's see how they work.
Another detail i need to deal with the is, potentially, impedance matching of the oscillator pins. I will be connecting the clock of the master board to the other boards through cables. I will have to find out what will be the best way to avoid potential issues beforehand. The frequency is not very high, but the distance between chips yes. TI does not make any provision about this in their evaluation board, maybe it is a non issue. I am not sure if the 3E audio boards (that expose the connectors to set their boards in master/slave mode) have some sort impedance matching. Will have to see what i can find out online....
- Thread Starter
- #25
hey @mcdn i need your help.
As mentioned in the previous post, i want to find out what is the best way to connect the oscillator pins (pins 9 and 10) between chips for master/slave mode. This is something that the TI documentation does not cover (or maybe they are suggesting not to use any resistor in between?).
In any case, there are a bunch of boards out there that have more than one chip, and are wired in master/slave mode, but the only one i could find complete pictures without heat sink is the one you measured here in ASR (i believe it is an Aiyima).
According to the pictures, following the traces, the oscillator pins seem to be connected between chips through 33R resistors (see picture below), but i would love to have it confirmed. If you happen to have the board accesible, is this something that you could check? If you don't fancy messing there no worries, thank you anyways!
the two 33R resistors indicated with a red arrow:
As mentioned in the previous post, i want to find out what is the best way to connect the oscillator pins (pins 9 and 10) between chips for master/slave mode. This is something that the TI documentation does not cover (or maybe they are suggesting not to use any resistor in between?).
In any case, there are a bunch of boards out there that have more than one chip, and are wired in master/slave mode, but the only one i could find complete pictures without heat sink is the one you measured here in ASR (i believe it is an Aiyima).
According to the pictures, following the traces, the oscillator pins seem to be connected between chips through 33R resistors (see picture below), but i would love to have it confirmed. If you happen to have the board accesible, is this something that you could check? If you don't fancy messing there no worries, thank you anyways!
the two 33R resistors indicated with a red arrow:
mcdn
Major Contributor
- Joined
- Mar 7, 2020
- Messages
- 1,021
- Likes
- 1,773
Last edited:
- Thread Starter
- #29
Interesting piece of information that answers the questions I had not covered by TI eval board documentation: the service manual of the Denon Heos AVR that uses 3 TPA3251 for LR, C and rear channels. In between other things one can find in the schematics:
How to wire the master slave pins - what i was looling for:
Nothing special at the master side. At the slave side, two 10 ohm resistors and two 10pF caps, all this to be placed close to the input pins. Exactly what i wanted to know
How to add relays to the speaker outputs to avoid on/off pops. I don't plan to use relays in my boards, i am going to trust the anti-pop system of the eval board, but I see this approach in most aiyima/fosi/etc. TPA3255 little amps. Good to know how to do it correctly, and with part numbers to make things even easier. The MCU signal switches a transistor that cascades two other transistors that energize the relays using PVDD (red line).
How to wire the master slave pins - what i was looling for:
Nothing special at the master side. At the slave side, two 10 ohm resistors and two 10pF caps, all this to be placed close to the input pins. Exactly what i wanted to know
How to add relays to the speaker outputs to avoid on/off pops. I don't plan to use relays in my boards, i am going to trust the anti-pop system of the eval board, but I see this approach in most aiyima/fosi/etc. TPA3255 little amps. Good to know how to do it correctly, and with part numbers to make things even easier. The MCU signal switches a transistor that cascades two other transistors that energize the relays using PVDD (red line).
- Thread Starter
- #30
i decided that i am going to go for the relays to avoid power off pop, the reason is the following:
On the contrary to the Denon AVR, none of the popular recent TPA3255 that I know uses a relay to mute the speakers, not even the TI eval board uses one, and many uses report a more or less noticeable pop, most commonly at power off. this is what we have:
fosi V3 stereo; fosi ZA3; Aiyima A07 max; 3E audio board stereo (this one the worst offender (note that one of the reasons wiim claims for not having implemented PFFB in their AMP is precisely the pop on reseting the chip. No idea if it is related, but i believe worth mentioning)
But the new Fosi V3 mono is the first of the breed to use a relay to mute the speaker (the bigger black relay near the thermistor). The implementation is a bit different to the Denon, as it only mutes one of the speakers terminals (i believe the other contact bypasses the thermistor, don't ask me why)
and look at the result, perfect!!
Still has a spike at "on", but that is not so bad on my book as it feels much worse when your listening session has finished, everything is in silence, and a few minutes later, when the amp goes to sleep, "POP!" I hate that.
I can imagine they needed to do it as a result of having the the audio sense option, but anyways, well done Fosi!
I only wonder why the mechanism does not apply on power on, but maybe it is a matter of programming the microcontroller differently, who knows.
On the contrary to the Denon AVR, none of the popular recent TPA3255 that I know uses a relay to mute the speakers, not even the TI eval board uses one, and many uses report a more or less noticeable pop, most commonly at power off. this is what we have:
fosi V3 stereo; fosi ZA3; Aiyima A07 max; 3E audio board stereo (this one the worst offender (note that one of the reasons wiim claims for not having implemented PFFB in their AMP is precisely the pop on reseting the chip. No idea if it is related, but i believe worth mentioning)
But the new Fosi V3 mono is the first of the breed to use a relay to mute the speaker (the bigger black relay near the thermistor). The implementation is a bit different to the Denon, as it only mutes one of the speakers terminals (i believe the other contact bypasses the thermistor, don't ask me why)
and look at the result, perfect!!
Still has a spike at "on", but that is not so bad on my book as it feels much worse when your listening session has finished, everything is in silence, and a few minutes later, when the amp goes to sleep, "POP!" I hate that.
I can imagine they needed to do it as a result of having the the audio sense option, but anyways, well done Fosi!
I only wonder why the mechanism does not apply on power on, but maybe it is a matter of programming the microcontroller differently, who knows.
- Thread Starter
- #31
PCBs finished, time to order the parts. There are 74 references in the basket and that does not include everything.
Completing your BOM and ordering also offers some unexpected learnings:
The PVDD caps. The only recommendation from TI, besides of enough capacitance and voltage margin, is that they are low ESR. I had a look at what was around and available to me and i landed on what Fosi (ZA3, V3) and Aiyima (A70) use, Chemi-con KY series 63V 2200uF.
KY series is for low impedance, size fits, perfect!. The curious fact is, that this cap should not exist! Following the nomenclature of this series the reference should be EKY-630ELL222MM40S. But this reference does not exist in Chemi-con website, and the datasheet of the series shows the 1800uF as the largest of the 63V rated:
???. But the fact is that there is one - and only one - reputable supplier that does offer the 2200uF: LCSC, and they are relatively cheap. Curiously, even the datasheet that LCSC offers for this reference is the same that Chemi-con itself and Mouser Digikey etc. have that does NOT include this value.
After evaluating other possibilities, this ended up being the most convenient and economic for me so i am going to go for the mysterious cap, let's see what the multimeter says.
Second thing that did not go according to plan is the power connectors. I went for JST VL series, same as in purifi amplifiers. The 2 circuits cable to board assemblies are rated for 20 amps if 12 AWG are used. There is a catch though. To be able to use 12 AWG cables you need to use the biggest pin contacts they offer (SVM-81T-P2.0). Those big boys can take up to 3.5 mm^2 of copper, but they are way too big for my cheapo crimping tool. I don't fancy investing big money on a bigger crimping tool, so i am afraid i will have to downgrade to the next smaller contacts and be happy with 18AWG and 8Amps rated connectors. The datasheet isn't clear to me and could be that these smaller ones can take 16 AWG as well, but I am not sure.
Look at that, can take 4.1mm diameter cables. If anyone knows a cheap crimping tool that can do the big 4.1mm JST contacts, i am all ears, but needs to be very cheap.
Completing your BOM and ordering also offers some unexpected learnings:
The PVDD caps. The only recommendation from TI, besides of enough capacitance and voltage margin, is that they are low ESR. I had a look at what was around and available to me and i landed on what Fosi (ZA3, V3) and Aiyima (A70) use, Chemi-con KY series 63V 2200uF.
KY series is for low impedance, size fits, perfect!. The curious fact is, that this cap should not exist! Following the nomenclature of this series the reference should be EKY-630ELL222MM40S. But this reference does not exist in Chemi-con website, and the datasheet of the series shows the 1800uF as the largest of the 63V rated:
???. But the fact is that there is one - and only one - reputable supplier that does offer the 2200uF: LCSC, and they are relatively cheap. Curiously, even the datasheet that LCSC offers for this reference is the same that Chemi-con itself and Mouser Digikey etc. have that does NOT include this value.
After evaluating other possibilities, this ended up being the most convenient and economic for me so i am going to go for the mysterious cap, let's see what the multimeter says.
Second thing that did not go according to plan is the power connectors. I went for JST VL series, same as in purifi amplifiers. The 2 circuits cable to board assemblies are rated for 20 amps if 12 AWG are used. There is a catch though. To be able to use 12 AWG cables you need to use the biggest pin contacts they offer (SVM-81T-P2.0). Those big boys can take up to 3.5 mm^2 of copper, but they are way too big for my cheapo crimping tool. I don't fancy investing big money on a bigger crimping tool, so i am afraid i will have to downgrade to the next smaller contacts and be happy with 18AWG and 8Amps rated connectors. The datasheet isn't clear to me and could be that these smaller ones can take 16 AWG as well, but I am not sure.
Look at that, can take 4.1mm diameter cables. If anyone knows a cheap crimping tool that can do the big 4.1mm JST contacts, i am all ears, but needs to be very cheap.
Last edited:
- Thread Starter
- #32
Making some progress...
I assembled what i call the "control board" that controls and provides 12V and 3.3V to up to 4 amplifier boards (connectors only for one at the moment):
the part to the left is the audio sense circuit, that i have not tested yet, but an initial single try the sensitivity wasn't great. I changed the input cap from electrolytic to ceramic from the original design, i hope it wasn't a mistake.
The back side has a second set of low voltage supplies and overcurrent and inverse polarity protection - both not populated yet -.
Also assembled one amp board just for testing, set to mono and with both chip and inductors from aliexpress to minimize loses knowing that most likely will not survive the tests:
(still swimming in flux)
The thing is tiny... The cap is like that to leave space for the header. Once i build the definitive version it will be cable to board and the cap will sit in the intended position.
The jumpers for mono operation are diy from a copper tube, didn't want to wait for the ones in order. But i like the result, might prepare more.
Frequency response tests look promising. At first i was disappointed by the almost 1 db down at 20kHz with 4 Ohm, but with 8 Ohm it is pretty straight! PFFB is doing its job quite ok for a first try, will try to find the way to improve it a little bit:
Now comes the most complicated but interesting part for me, the heat management with heat pipes. I hope i will have something to show before the end of the year
Casualties so far:
1 tpa3255 chip: connected 12V to a logic pin by mistake
1 LDO: very strange, just by plugging. The regulator works, but it must have gone super noisy because if I use it the measurements get all messed up.
I assembled what i call the "control board" that controls and provides 12V and 3.3V to up to 4 amplifier boards (connectors only for one at the moment):
the part to the left is the audio sense circuit, that i have not tested yet, but an initial single try the sensitivity wasn't great. I changed the input cap from electrolytic to ceramic from the original design, i hope it wasn't a mistake.
The back side has a second set of low voltage supplies and overcurrent and inverse polarity protection - both not populated yet -.
Also assembled one amp board just for testing, set to mono and with both chip and inductors from aliexpress to minimize loses knowing that most likely will not survive the tests:
(still swimming in flux)
The thing is tiny... The cap is like that to leave space for the header. Once i build the definitive version it will be cable to board and the cap will sit in the intended position.
The jumpers for mono operation are diy from a copper tube, didn't want to wait for the ones in order. But i like the result, might prepare more.
Frequency response tests look promising. At first i was disappointed by the almost 1 db down at 20kHz with 4 Ohm, but with 8 Ohm it is pretty straight! PFFB is doing its job quite ok for a first try, will try to find the way to improve it a little bit:
Now comes the most complicated but interesting part for me, the heat management with heat pipes. I hope i will have something to show before the end of the year
Casualties so far:
1 tpa3255 chip: connected 12V to a logic pin by mistake
1 LDO: very strange, just by plugging. The regulator works, but it must have gone super noisy because if I use it the measurements get all messed up.
Last edited:
- Thread Starter
- #34
turns out that i had one of the pffb not correctly wired, that is, the feedback to the positive input was connected but the one to the negative input was not connected (don't ask 
), and this was, apparently, the reason behind the decay of the signal past 10kHz with 4 Ohm shown previously. Wired both correctly, and i get ca. -0.35 and +0.35 dB at 4 (blue) and 8 (orange) Ohm respectively at 20kHz. Not too bad i would say, TI suggested values are spot on.
(note that the 0.2db decay at 20Hz is my ADC, doesn't come to the DUT)
), and this was, apparently, the reason behind the decay of the signal past 10kHz with 4 Ohm shown previously. Wired both correctly, and i get ca. -0.35 and +0.35 dB at 4 (blue) and 8 (orange) Ohm respectively at 20kHz. Not too bad i would say, TI suggested values are spot on.(note that the 0.2db decay at 20Hz is my ADC, doesn't come to the DUT)
Anyone up for a challenge? Why using SER2918H-103KL coils instead of the toroid MA5172-AE (TI eval user guide) would produce about 20F degrees more at the TPA3255 heatsink. Verified by replacing all 4 coils with MA5172 on the same design. FLIR camera also shows ~100F with MA5172 and ~120F with SER2918. This is ambient room temp, no fan to remove as many variables as possible. Idle wattage also goes up, which is expected. Makes no sense as 2918 is a lower DCR coil (same uH rating). I also noticed other DYI designs using 2918s - if you have an amp running with those in BTL mode (shunted cap) - what is your temperature at the heatsink without the fans at room temp? Note this is idle (silence) TPA3255 is out of reset; heatsink is cold if TPA3255 is in reset, which is also to be expected.
- Thread Starter
- #36
I have no idea but I might give it a try. For the moment I have done all my tests without heatsink at all and I didn't notice the pad very hot. Not even hot at idle.
- Thread Starter
- #37
It has been a long while since I don’t progress this project, the reason being that I was expecting to set up my 3D printer this summer again but it seems it is not going to happen until next year. And without at least a prototype of enclosure to keep things in place, most importantly fans and heat pipes, it doesn’t make any sense to start measuring temperatures :-/
In the mean time I will probably redesign the power section that supplies the opamps and logic. I am afraid the 1A max that the original circuit copied from the evaluation board won’t be enough for 4 amps so I will most likely switch to a SIC462 buck (up to 6 amps) and then split the duty between various LDOs.
Also, to keep a bit my interest in the project alive, yesterday I decided to have a look inside the TPA3255 chips I got from Aliexpress:
As I knew I was going to burn a few during the tests, and Mouser and Digikey are charging an unbelievable markup since some months ago (I see mouser price has gone a couple of euros down now) I got a few from Aliexpress, at ca. 2-3 euros each. I often buy parts from Aliexpress that in my opinion are unlikely to be fake (i.e. unique package, pinout or functionality, niche products…) and my experience has always been good, but one never knows. I am always interested in knowing if what I get from them has any chance to be the real thing, and the associated detective work. It is part of the hobby I guess.
I could not find any picture of the TPA3255 decapped and I am not willing to decap one of the ones I got from mouser, but I thought it could be helpful to share my findings and maybe someday someone post pictures of a known authentic one to compare, or maybe I burn one myself.
I used the technique of thermal shocking the epoxy (hot air at 450C/cold water cycles) and the encapsulation came out pretty easily.
I was expecting that the thermal pad (bigger grey rectangle) more tightly joined to the die, but it came out pretty easily, there is a sort of thermal paste in between, I will try to look later what sort of adhesive it is, but it is for sure filled towards thermal transfer.
The action was obviously on the other side of the die, one can clearly see the 4 pairs of ourput mosfets at naked eye, they are huge. To be able to output the amount of current the chip is capable of (I can imagine), there are 3 to 6 wires coming out of each of the sections of the mosfets.
Unfortunately, I don’t have means and patience to take better pictures, but you get an idea. What you see at the top right is some epoxy residue that I later took out with acetone.
Then I moved to the microscope to try to find markings, that was the whole point of all this. From what I read online, many TI parts have factory ID numbers etched on the die, that don’t necessarily match tha part number and change with the different revisions of the design. I’ve also read in their forum, that as dies are getting smaller and more cramped, sometimes they don’t even put markings. So I was not expecting to find “TPA3255” but I was hoping to at least find a number.
It took me a while to inspect the die, it is pretty busy when looked under the microscope and the details that become visible depend a lot on the angle the light hits the die. Also I have the strong impression that if one etches the mosfets, there is stuff underneath, but I was not going to do that.
But when I was about to abandon, I saw this in one of the corners:
What I see there is a M in a circle and what seems to resemble TI’s logo. I know it is a stretch, I have seen much better defined TI logos on dies, but always on much larger spaces.
Well, I leave it there for now, if someone knows something about TI markings or knows the TPA3255 die, I wo
uld be pleased to read his comments.
In the mean time I will probably redesign the power section that supplies the opamps and logic. I am afraid the 1A max that the original circuit copied from the evaluation board won’t be enough for 4 amps so I will most likely switch to a SIC462 buck (up to 6 amps) and then split the duty between various LDOs.
Also, to keep a bit my interest in the project alive, yesterday I decided to have a look inside the TPA3255 chips I got from Aliexpress:
As I knew I was going to burn a few during the tests, and Mouser and Digikey are charging an unbelievable markup since some months ago (I see mouser price has gone a couple of euros down now) I got a few from Aliexpress, at ca. 2-3 euros each. I often buy parts from Aliexpress that in my opinion are unlikely to be fake (i.e. unique package, pinout or functionality, niche products…) and my experience has always been good, but one never knows. I am always interested in knowing if what I get from them has any chance to be the real thing, and the associated detective work. It is part of the hobby I guess.
I could not find any picture of the TPA3255 decapped and I am not willing to decap one of the ones I got from mouser, but I thought it could be helpful to share my findings and maybe someday someone post pictures of a known authentic one to compare, or maybe I burn one myself.
I used the technique of thermal shocking the epoxy (hot air at 450C/cold water cycles) and the encapsulation came out pretty easily.
I was expecting that the thermal pad (bigger grey rectangle) more tightly joined to the die, but it came out pretty easily, there is a sort of thermal paste in between, I will try to look later what sort of adhesive it is, but it is for sure filled towards thermal transfer.
The action was obviously on the other side of the die, one can clearly see the 4 pairs of ourput mosfets at naked eye, they are huge. To be able to output the amount of current the chip is capable of (I can imagine), there are 3 to 6 wires coming out of each of the sections of the mosfets.
Unfortunately, I don’t have means and patience to take better pictures, but you get an idea. What you see at the top right is some epoxy residue that I later took out with acetone.
Then I moved to the microscope to try to find markings, that was the whole point of all this. From what I read online, many TI parts have factory ID numbers etched on the die, that don’t necessarily match tha part number and change with the different revisions of the design. I’ve also read in their forum, that as dies are getting smaller and more cramped, sometimes they don’t even put markings. So I was not expecting to find “TPA3255” but I was hoping to at least find a number.
It took me a while to inspect the die, it is pretty busy when looked under the microscope and the details that become visible depend a lot on the angle the light hits the die. Also I have the strong impression that if one etches the mosfets, there is stuff underneath, but I was not going to do that.
But when I was about to abandon, I saw this in one of the corners:
What I see there is a M in a circle and what seems to resemble TI’s logo. I know it is a stretch, I have seen much better defined TI logos on dies, but always on much larger spaces.
Well, I leave it there for now, if someone knows something about TI markings or knows the TPA3255 die, I wo
uld be pleased to read his comments.
Last edited:
- Thread Starter
- #38
This is good news, from what i see in google, logo + circled M is pretty standard in TI parts. Gonna deem the chips legit for now
I will need some OPA1622 for another project, and that would be an example of a part that i would never ever buy from Aliexpress, but I might now give it a go and sacrifice one to check what is in there....
I will need some OPA1622 for another project, and that would be an example of a part that i would never ever buy from Aliexpress, but I might now give it a go and sacrifice one to check what is in there....
Similar threads
