• 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!

ISO: 5V Low Noise/Ripple 3A+ SE supply.

deadwood83

Member
Joined
Sep 1, 2022
Messages
91
Likes
160
I am not a fan of linear supplies. I will get that rout of the way now.

However, I find myself in need of something approaching such a supply. I have a janktastic DAC (with some promising attributes) which takes 5VDC because it has basically no power conditioning whatsoever. The whole path to AVCC is Wallwart -> Barrel Jack -> Cap -> Relay -> Cap -> Ferrite Bead -> LDO.

See below for the map I made.

1665201800968.png


I'm more used to seeing a buck/boost chip somwhere in there operating in the hundreds of kHz or even MHz, making a linear or otherwise exceptionally clean supply a moot point. In this case, no such thing exists. The wall wart goes through two caps and a ferrite bead before hitting LDOs for AVCC, VCCA, and DVCC. It also travels directly underneath the LVDS port which defeats half the point of having the freaking shield attached to the PCB.

I have seen linear CN kits on ebay, MeanWell medical rated PSUs (though not sure if that actually has any stricter standards for PSU noise), and some DIY ' Denoiser' boards which I do not fully understand because they use older parts with lower switching frequencies. I have some TI TPS7a20 LDOs on order (to swap for the step from 5V to 3.3V) for their 20+dB higher PSRR compared to ST/Richtek SOT-23-5 but that won't help my poor LVDS which I actually plan to use for stealing I2S from my AVR.

I am looking for recommendations which are not both stupidly large and stupidly expensive.

Basic DAC spec is an XU208 and LVDS de-plexer passing through some TI isolators into the ESS chips; initialized/controlled via an STM32 which, oddly, is not isolated. OPA1612's for I/V stage and parallelled Nichicon 100uF KW (SMPS) caps (with .47uF film paralleled on board bottom). Lots of inductors and caps after the Monolithic boost for the opamp rails but jack squat for the AVCC, VCCA, and DVCC. I removed the useless HP amp board and BT; which can be handled by my AVR since it pipes data through the HDMI flow so I can steal its I2S.

This will be feeding a driverack PA2 (already adjusted I/V stage gain) which will in turn feed bi-amped (one stereo amp per channel) DIYSG Fusion 10 Pure speakers. The 'Easy' answer would have been JBL 708Ps on a Topping DX7 Pro/+ but I've had these speakers for the past decade + and I like them well enough (though I need a sub after trying the JBL 308P MK2s in my home office).

Can anybody make some sub-$200 USD recommendations? Needing to build with a kit is a bonus, as I am always a fan of hot, molten metal gluing legos together. Needing to source individual components and order PCBS from a MFG.... not as much of a bonus.

This is just for my HTPC which might be a tad overkill but hey, this money was stuck with a taobao agent who always found an excuse not to refund me for.... over two years. I do have a closet Gustard X20 but it's just sooooo honking big and would eat up valuable space in my low-profile multimedia cabinet. Still not sure where to use said Gustard since it lacks the USB card but that's a different topic.

Thanks in advance!
 

sq225917

Major Contributor
Joined
Jun 23, 2019
Messages
1,361
Likes
1,613
What makes you think there are gains to be had from a lower noise PSU?
What's the spec of the LDO's feeding the digital and analogue rails?
Is the wallwart a smps?

Might you not just be better off with a decent DIY two-pole SMPS post filter and save the money for better LDO's ?
 
OP
D

deadwood83

Member
Joined
Sep 1, 2022
Messages
91
Likes
160
I swapped the LDOs (unfortunately SOT-23-5 so choices are limited) with TI TPS7A2033 units which helped a little. But the board layout is.... yeah.

I have read a lot of mixed things on the filters, and the only one I have really seen without making a custom PCB is the one on the diyaudio store. But I would be super open to learning more if you have some example links!

MEasuring the PSU output gave me a lot of 'stuff' in the 30k-66k range which I believe can be seen here:
1667493770341.png


I have a 100% unfounded hunch that the wall-wart noise is bleeding through to the AVCC and AVDD LDOs on the DAC section of the layout, and also into the opamp rail power supply which is a Monolithic MPQ3426. The fSet resistor is 27k so by monolithic's datasheet, it should be operating about 1.35MHz. Looking at the factors of 1350000, I can see 30k, 33.75k, 37.5k, 45k, 50k, 54k, 56250, 67.5k. I have a hunch the PSU is hitting a point of constructive resonance with the opamp rail boost converter.

Oddly, swapping the I/V follower opamps from NE5532 (genuine, from Mouser) with LME47920 (also genuine, from Mouser) dropped the THD+N in the 1kHz by about 8dBc. But the register settings from the MFG are all sorts of wonky because the 9038q2m has quite a few tweaks for 2nd-3rd harmonics available which are very clearly not used.

1667495368827.png




Anyway, power seemed like an easier avenue to address as opposed to reverse engineering the STM32 circuit and onboard firmware to try and preserve the remote / display / input select while altering the register values. Also it's just being wonky when I try and dump the FW via linux.

1667495339647.png
 

kchap

Addicted to Fun and Learning
Forum Donor
Joined
Jun 10, 2021
Messages
579
Likes
566
Location
Melbourne, Oz
All very interesting but the offending signal is less than -80dB down and well above the audible range.
 
OP
D

deadwood83

Member
Joined
Sep 1, 2022
Messages
91
Likes
160
yes sir. I do not expect to be able to hear a difference, this is just a funsies project. I re-implemented the I/V stage with some different value and tolerance components in addition to the LDOs and it was quite fun. I have this and a Gustard X20 which I can 'surgery' on, but this is much easier/cheaper to replace and probably has greater potential due to the SW controls in the 9038q2m compared to the 9018.

I guess if I wanted to go full stupid I could try to use reflow to swap the X20 9018s with 9028s then rebuild firmware for the new register options but the Gustard also lacks a USB input so.....
Firmware is a little scary.
 

restorer-john

Grand Contributor
Joined
Mar 1, 2018
Messages
12,588
Likes
38,291
Location
Gold Coast, Queensland, Australia
What do you need 3A for? The DAC surely isn't consuming 15W off the 5V rails. What is the actual current draw via the input jack (measure it)?
 

sq225917

Major Contributor
Joined
Jun 23, 2019
Messages
1,361
Likes
1,613
That depends, is there a display and wireless adaptor in the case? If not then its likely not even drawing a full amp.

I like a shits n giggles project from time to time. I'd just replace the wall wart with a Chinese ebay 5v 15va psu and be done with it for £40. ( if I didn't already have loads of linear psu and cascaded regs to play with).
 
OP
D

deadwood83

Member
Joined
Sep 1, 2022
Messages
91
Likes
160
1.3" OLED, IR receiver, and BT5.0 board. W/o the BT it draws ~ 9W and with the BT it's about 12. I think the 15 was so they could have somebody use RCA out, XLR out, and HP out (no way to switch between them) simultaneously.

Also, I wonder if STMicro just has a broken linux driver. I got the firmware off with OpenOCD and it loaded in Ghidra. Memory was properly mapped by SVD-Loader. This is why playing with power seemed easier :D

1667574423558.png


One of the many functions crammed into this 16k of programmable memory.
 
Top Bottom