Its possible, but distortion may go up quite a bit.
Some info attached on ESS dac output stage considerations.
Also attached is some information on the MFB filter that MKHunt suggesting modifying (starting on page 7)
Also: https://www.ti.com/lit/an/slyt343/slyt343.pdf?ts=1665638973509&ref_url=https%3A%2F%2Fwww.google.com%2F
On balanced version of MFB filter as commonly used in dacs: https://d3uzseaevmutz1.cloudfront.net/pubs/appNote/AN048Rev2.pdf
Some info attached on ESS dac output stage considerations.
Also attached is some information on the MFB filter that MKHunt suggesting modifying (starting on page 7)
Also: https://www.ti.com/lit/an/slyt343/slyt343.pdf?ts=1665638973509&ref_url=https%3A%2F%2Fwww.google.com%2F
On balanced version of MFB filter as commonly used in dacs: https://d3uzseaevmutz1.cloudfront.net/pubs/appNote/AN048Rev2.pdf
Last edited:
I finally actually listened to my stupid DAC today. At first I thought everything was wrong because the tweeters were not making noice, the mids were trying to hit the deepest bass, and the subwoofer was silent.
Turns out I connected tweets to mids, mids to sub, and sub to highs on my PA2. Whoops. Sorted it all out and I can actually get almost near clipping at 100% volume! The 49720's dropped the output from 8Vpp to 6Vpp somehow. I am guessing they present a different load to the OPA1612s. I guess I could always solder up the OPA1656s and use them as voltage followers. Our TI guy on here in the 1656 thread said they would work well in that role.
It sounds really nice though. I would guess it is better than the onboard PCM5101A in my AVR but it is difficult to say since the output is so much hotter from this little 9038Q2M dual thing.
I absolutely LOVE having full use of the output range though! Hoooooly smoke it;s nice to have a truly dynamic volume adjustment that allows fine control over a super wide range. I have been missing that in my HT setup since like..... 2016.
Turns out I connected tweets to mids, mids to sub, and sub to highs on my PA2. Whoops. Sorted it all out and I can actually get almost near clipping at 100% volume! The 49720's dropped the output from 8Vpp to 6Vpp somehow. I am guessing they present a different load to the OPA1612s. I guess I could always solder up the OPA1656s and use them as voltage followers. Our TI guy on here in the 1656 thread said they would work well in that role.
It sounds really nice though. I would guess it is better than the onboard PCM5101A in my AVR but it is difficult to say since the output is so much hotter from this little 9038Q2M dual thing.
I absolutely LOVE having full use of the output range though! Hoooooly smoke it;s nice to have a truly dynamic volume adjustment that allows fine control over a super wide range. I have been missing that in my HT setup since like..... 2016.
Hi, does anybody tried this board? It's cheap, $23
From first view it lacks option for negative power input, but it's different design, so maybe worth trying?
ES9038Q2M Decoder Board Input ES9038 Asynchronous USB Module Can Be Used With Italian Interface
From description it states:- Have its own development program, complete optimization of the decoding chip, better signal integrity, less signal distortion
- Dual op amps are used, and the op amps are powered by positive and negative power supplies.
From first view it lacks option for negative power input, but it's different design, so maybe worth trying?
Last edited:
Got one of those dac boards some time ago. IIRC it has a little DC/DC converter on it for the negative opamp rail. The thing I remember the most about it was that I thought it sounded pretty bad and that the board was too tightly packed to be able to fix very much of what's wrong with it. The other thing is that it needs an I2S source, which will probably cost more than the dac board. My recommendation would be to avoid this one.
The 'original' Lusya single socketed opamp board has I2S input:
https://www.aliexpress.us/item/2251832702899096.html
There is also one with a few more discretes:
https://www.aliexpress.us/item/2251832666350666.html
Sky song has I2S input. Layout is similar to the others. and was shown a few posts back. Larger components means easier modifications. Still layout limited.
https://www.aliexpress.us/item/3256804206636076.html
A bit more expensive, but actually has a 3-opamp I/V stage so might be in current mode with some noise rejection. Nice crystal on it. Claims to use OPA1612 which markw likes and I have no complaints about 1612 on my dual 9038 board. Convenient form factor and connection for a Pi, but you won;t be making many mods without hot air or super careful manual work due to cramped layout.
https://www.aliexpress.us/item/3256804246442991.html
Another PI-FF option but seems to be voltage mode owing to single SMD package opamp
https://www.aliexpress.us/item/3256802309885030.html
I think at 30USD you will find many sort of 'toy' DACs which technically work, but aren't actually any good. But if you are using it in an environment with some sort of noise from PC fans, a projector, an HVAC system, etc, it might be totally fine.
The tricky part is that if you are feeding from a Pi, you can also use USB output which means you can get say, a Topping D10 Balanced (130USD, balanced ONLY) or D10s (110 USD, SE ONLY) for not much more than the $70USD hat board. The D10S is USB powered so life gets even easier. Both measure well (118dB SINAD and 112dB SINAD respectively). A project board may never reach those performance levels without more than doubling the initial cost. A $30 USD board is almost guaranteed to never reach those levels without a new PCB and circuit.
But playing around IS a LOT of fun.
https://www.aliexpress.us/item/2251832702899096.html
There is also one with a few more discretes:
https://www.aliexpress.us/item/2251832666350666.html
Sky song has I2S input. Layout is similar to the others. and was shown a few posts back. Larger components means easier modifications. Still layout limited.
https://www.aliexpress.us/item/3256804206636076.html
A bit more expensive, but actually has a 3-opamp I/V stage so might be in current mode with some noise rejection. Nice crystal on it. Claims to use OPA1612 which markw likes and I have no complaints about 1612 on my dual 9038 board. Convenient form factor and connection for a Pi, but you won;t be making many mods without hot air or super careful manual work due to cramped layout.
https://www.aliexpress.us/item/3256804246442991.html
Another PI-FF option but seems to be voltage mode owing to single SMD package opamp
https://www.aliexpress.us/item/3256802309885030.html
I think at 30USD you will find many sort of 'toy' DACs which technically work, but aren't actually any good. But if you are using it in an environment with some sort of noise from PC fans, a projector, an HVAC system, etc, it might be totally fine.
The tricky part is that if you are feeding from a Pi, you can also use USB output which means you can get say, a Topping D10 Balanced (130USD, balanced ONLY) or D10s (110 USD, SE ONLY) for not much more than the $70USD hat board. The D10S is USB powered so life gets even easier. Both measure well (118dB SINAD and 112dB SINAD respectively). A project board may never reach those performance levels without more than doubling the initial cost. A $30 USD board is almost guaranteed to never reach those levels without a new PCB and circuit.
But playing around IS a LOT of fun.
Agreed that RPi > USB > Topping dac, would more or less probably be the best option in terms of value (that is to say, good performance for the cost).
OTOH, if you want to have some fun modifying dacs, maybe to see what you can learn from it, then one of the Chinese ES9038Q2M dacs recommended in the thread could work for that. However getting one of those dacs to sound decent might take a lot of time and bit of money.
OTOH, if you want to have some fun modifying dacs, maybe to see what you can learn from it, then one of the Chinese ES9038Q2M dacs recommended in the thread could work for that. However getting one of those dacs to sound decent might take a lot of time and bit of money.
I finally ran a 1k FFT through my E1DA Cosmos ADC on my modded cheap dac / socket board thing. The LDOs definitely helped, but it desperately needs some sort of PSU which is better regulated.
A few harmonics but otherwise not awful.
For comparison here are a few more DACs.
Gustard X20:
Loxjie D40, This dac is impressive.
A few harmonics but otherwise not awful.
For comparison here are a few more DACs.
Gustard X20:
Loxjie D40, This dac is impressive.
It is possible something is wrong with my measurements. I just figured out how to measure (I think) mostly on my own about 2 hours before.
The setup looks like:
Single grounded outlet (only 1 of 2 on a dedicated 20A breaker) -> Lenovo M75Q Gen 2 Ryzen 7 Pro 1L PC + DAC. I do not have a functional USB input on the 9038q2m because the spi flash eeprom pooped the bed so USB output looks as below:
Lenovo M75q USB -> Singxer SU-1 -> I2S LVDS to DAC
Connection to the E1DA Cosmos ADC is via single USB-C Cable from Lenovo M75Q Gen 2
DACs are directly connected from their XLR outputs to the ADC input using independent single XLR carrier cables. No snakes.
I think REW is reporting 32-bit mode. The Gustard and LEaf DACs both have a really hard rolloff at 22k.
When multiple low-wattage devices need to be connected in a single 'chain' I will use a multi-headed IEC cable like the below to ensure that each piece of equipment is basically star grounded to the same outlet (as the PC).
I am more than open to tips on how to improve, as this is a very new capability for me.
The setup looks like:
Single grounded outlet (only 1 of 2 on a dedicated 20A breaker) -> Lenovo M75Q Gen 2 Ryzen 7 Pro 1L PC + DAC. I do not have a functional USB input on the 9038q2m because the spi flash eeprom pooped the bed so USB output looks as below:
Lenovo M75q USB -> Singxer SU-1 -> I2S LVDS to DAC
Connection to the E1DA Cosmos ADC is via single USB-C Cable from Lenovo M75Q Gen 2
DACs are directly connected from their XLR outputs to the ADC input using independent single XLR carrier cables. No snakes.
I think REW is reporting 32-bit mode. The Gustard and LEaf DACs both have a really hard rolloff at 22k.
When multiple low-wattage devices need to be connected in a single 'chain' I will use a multi-headed IEC cable like the below to ensure that each piece of equipment is basically star grounded to the same outlet (as the PC).
I am more than open to tips on how to improve, as this is a very new capability for me.
also in REW via FlexASIO from the DAC I am measuring. Generator frequency is also locked to RTA FFT in REW.
ETA: Settings:
They might just be that noisy. YOu can see in the image posted in https://www.diyaudio.com/community/threads/es9038q2m-board.314935/post-7159677 which is the Leaf 9038 Socket that the averaging shows a hard bottom line at -130dBFS but the peaks of the noise go to -120dBFS.
I also keep averaging on, which will naturally remove a lot of the ultra low measure noise. Example:
Loxjie D40 with averaging:
Without:
I think part of the issue might also be the 9038 board and my Gustard both output rather hot voltages. I just realized my ADC was set to 1.7V sensitivity. So that definitely skews things by a decent margin. Trying the Loxjie at 4.5V sensitivity on the ADC shows that it does not reach a full 4.5V output at 0dBFS digital setting but the noise floor is further reduced.
So Now that I have typed this all out and thought and played and measured while typing, you are definitely correct. The Gustard and 9038 boards for sure put out a much higher voltage which will make the noise floor appear higher. I can re-measure them. I am actually a little curious about the 9038 at a different input sensitivity since I did that mod to increase I/V stage gain for 6-8Vpp (6Vpp with 49720 followers, 8Vpp with NE5532 followers which seems odd) output. My gut says it really won't be much different but I am always open to be surprised. Measurements (at 1.7V input sensitivity) of NE5532 and 49720 showed a 10dB higher noise floor with the NE5532.
Gentlefolk! I have overcome my learning disability! It took some time and trial and error, but I think I have the good stuff now. L channel is in blue in all traces.
Loxjie D40 :
Gustard X20:
Leaf Audio Dual 9038q2m:
I was hoping the Gustard would do better. If it could just shake those harmonics, it would be near SOTA, almost 10 years after its launch.
The leaf ... well, the power supply noise almost dominates the second harmonic. The power supply noise DEFINITELY dominates the multi-tone test. It also suffers from adverse harmonics, but the 9038q2m has config registers specifically to address those. There's even an STM32 in the DAC. They could have fixed it really easily (though PSU would still cause problems). The Gustard can also output a BUTTLOAD of voltage on XLR. Holy smokes. It's over 6Vpp by a decent margin. I still think the D40 is pretty incredible, even if it has that really annoying 60Hz hump.
Loxjie D40 :
Gustard X20:
Leaf Audio Dual 9038q2m:
I was hoping the Gustard would do better. If it could just shake those harmonics, it would be near SOTA, almost 10 years after its launch.
The leaf ... well, the power supply noise almost dominates the second harmonic. The power supply noise DEFINITELY dominates the multi-tone test. It also suffers from adverse harmonics, but the 9038q2m has config registers specifically to address those. There's even an STM32 in the DAC. They could have fixed it really easily (though PSU would still cause problems). The Gustard can also output a BUTTLOAD of voltage on XLR. Holy smokes. It's over 6Vpp by a decent margin. I still think the D40 is pretty incredible, even if it has that really annoying 60Hz hump.