XHRA-2HPA-TQ64 custom project - debugging advice

New to XMOS and XCore? Get started here.
Post Reply
Thorp
New User
Posts: 3
Joined: Sun Oct 22, 2017 4:47 pm

XHRA-2HPA-TQ64 custom project - debugging advice

Post by Thorp »

Hello everybody,

I've managed to finish my prototype DAC based on XHRA-2HPA reference platform hardware manual.
First of all, the sound is amazing, as I expected.
Second, I've made some changes, so I got troubles.
The first problem was the heating of LDOs, since I didn't want to have PWM buck converters in the circuit, but I fitted some heat sinks on them, as you can see in attached picture. Problem solved.
The second issue that I'm trying now to figure out how to fix seems to be a big problem: for the first 4-6 minutes from power-on, so when the circuit is cold, everything is perfect, but after that all the sample rates based on 24.576 MHz clock are scratched.
All 44.1, 88.2, 176.4 and DSD files sound excellent, I've been tested them for hours.
After few minutes from power on, all 48 kHz files have minor and rare scratches, 192 kHz files have minor and often scratches, and 96 kHz files are practically impossible to listen to. All these symptoms start gradually as much as the circuit is warming.
I thought there are some problems with oscillators, so I replaced them with other brand, but problem remains. After that I thought that the problem is the location of oscillators, near the DAC chips - where they should be, on the clean side of the circuit, but 22.5792 MHz clock is in that location too, and it works perfect.
What is happening when the DAC is warmer, and why only 24.576 clock based files have scratches?
Where to start the debugging?

Thank you in advance!
Adrian

Image


User avatar
mon2
XCore Legend
Posts: 1913
Joined: Thu Jun 10, 2010 11:43 am
Contact:

Post by mon2 »

Hi. Not an audio developer but..

1) which component(s) is / are heating up? LDO and DAC?

2) which LDO is being used?

3) what is the input voltage to this LDO?

4) does the LDO supply ample current to support this project?

5) if you apply some of the freezy spray onto the heating component, the noise goes away?

http://www.newark.com/c/chemicals-adhes ... eze-sprays

6) Can you post more details on the schematic of your power supply and DAC for a review?

7) Which 24.576 Mhz clock oscillator is being used?

Some logical stabs at the issue:

a) LDOs will emit heat as they drop down the input voltage to your desired voltage. Larger the input to output delta, more the heat. Perhaps the LDO is not suitable for the current draw of this design? It is possible that the IP consumes more current @ 24.576 Mhz as compared to the other clocks.

b) DAC requires improved PCB layout to ventilate / dissipate the heat? Often such parts feature a metal belly that requires support of vias in the PCB and proper SMD soldering to allow for this region to act like a heat sink. Assembly ok?
Thorp
New User
Posts: 3
Joined: Sun Oct 22, 2017 4:47 pm

Post by Thorp »

Hi mon2, thank you for trying to help me.
I never thought about heating related to scratched sound, but you gave me an idea. My schematic have separate transformers, so I only powered up the XMOS side for approximate 15 minutes, until it was warmed to about 55-65 Celsius - this is the temperature at which the circuit is thermally stabilized, then I powered up the DAC side, and NO scratches! The analog part of the DAC is out of the question, since it perfectly works on 22.5792 MHz clock, so I've isolated the problem: the digital part of the DAC. I also observed something important: scratches start in the right channel first, and when they start, XMOS chip is still at 55 Celsius and LDO's are still at 60-65 Celsius.
I already verified the circuit, there are no excessive heating parts on this side, except i/v converters which are slightly warm - but is normal, ADM 7150-3.3V for digital is OK, ADM 7150-5V for analog is OK.
What I have to do from now:
2 - to replace the right DAC chip, even the left DAC chip, but I feel I will not get to do that, that's why I numbered this step "2".
1 - I2S_DATA and/or I2S_LRCLK might be corrupted: I use AD1955DAC 2 chips, one for the left channel, and other for the right channel. I like all the DAC's in that class, PCM 1794, WM 8741, but there is something magic with AD 1955 sound, I love it. It is no so popular in the diy-ers community, since it doesn't have sample rate auto-detect, or pin-configuraton utilities, you need to write data in its registries. Well, for its DSD capabilities, there are no to much registry settings-you cannot adjust digital volume or mono-left/right set. So, short story, my left channel DAC chip has DSD left input pin tied to I2S_LRCLK input pin, and right channel DAC chip has DSD right input pin tied to I2S_DATA input pin. If you remember XMOS I2S_DATA becomes DSD right, and I2S_LRCLK becomes DSD left when DSD mode is on. All I know about DAC DSD input pins is they need to be tied to a logic level when not used, and they have a very weak pull-up = 0.6 mA. Probablly, somehow, I2S_DATA, or I2S_LRCLK, or both became corrupted when the DAC chips are slightly warming. I used this configuration many times, but now seems to be some incompatibility, I have to check that next week-end.

I will come back with details.
Post Reply