LEAF Dac Dual 9038Q2M VS Gustard X16 ! Pictures inside )

[Mark K]

Good news is I did receive this unit (finally)! Quick check and my Windows 11 PC via supplied USB cable connected with no manual driver installations required. My audio control panel shows I have 3 Output options: nVidia, Realtek Digital, and Speakers HiFi DSD. If I choose Realtek, then I can only go to 24/192, but if I choose HiFi DSD then I can go to 24/384! Remote works well (and FYI takes a 2032 button battery).

OK, what is the preferred filter - "fast roll-off minimum" ? I don't see a "fast linear" but there is a "hybrid fast"?

I thought all digital outputs are the same.

 

[Mark K]

Browsed their Taobao Shop in China. The unit is around 100 and we have to buy power supply and a balanced to single ended 3.5mm converting cable.

 

[JoeSVK]

Browsed their Taobao Shop in China. The unit is around 100 and we have to buy power supply and a balanced to single ended 3.5mm converting cable.

The linear power supply is aroung 40Euro from Taobao, so not a big deal. The balanced to 3.5mm we would not recommend to use, as it could harm the DAC.

And no, not all inputs are the same.

 

[deadwood83]

Can anybody with the 9038Q2M version measure VCC on the XLR and I/V opamps? I just ordered the remote version for USD $104 but I would like to bump the XLR to 4V (will never use RCA nor headphone) to match my DSp expected input.

 

[deadwood83]

Ain't nobody did anything, so I did it myself.

My Leaf will feed a DBX Driverack PA2 (which measures better than the Venu 360, but ONLY on the +4dBu setting)

a Monolithic MP3426 boosts the 5V to +13V. The nearest two opamps (IV Stage) are fed direct from the 9038Q2M. _b outputs are merged and feed one side of the dual op-amp the other two feed the other side.

Second pair of opamps is configured as a shoddy voltage follower. No FB resistor at all. Just 0-Ohm copper between - input and output.

Converting to 3VRMS (8.4VPP) was easy, and not at all easy. The integrator I/V stage needed a resistor swap for each opamp (2 per socket, 4 total).
ANyway, it;s done. I used a T12-K tip and really awful tweezers. No, I am not a machine and no, they are not perfect. I swapped the 390 ohm oddly sized resistors for 1k (which I had on hand as opposed to 'ideal' 820-Ohm)

But the measurements speak for themselves. Each pin on each XLR is now running 8.4VPP and this was just a 1kHz sine generated in Windows.

Not bad. Not bad at all for a completely untested modification. Perhaps removing the HP board helped? I just stole the 1k resistors from the HP board to swap into the 390 slots.

 

[JoeSVK]

@deadwood83 this is some next level do you have the before/after measurements? so the output voltage was boosted from 2V to 3V on the XLR? Would be nice to see some SINAD, THD+N measurements as well

Can you post a photo of the exact location of the resistors?

 

[deadwood83]

I have nothing even close to real measuring equipment. My scope is only accurate for comparing A vs B. It is possible that the noise floor of the DAC has increased from my changes (but the signal level is 4x higher so it would have to increase more than 4x to hurt).

I cannot recommend the DAC in really any but specific circumstances (e.g. pretend that an overseas shopping agent has had just enough money for a few years but they refuse to refund you). The PCB is made..... poorly and the traces will lift at the slightest provocation. I am pretty sure it is <1Oz copper.

Power is provided directly from the external 5V supply without any real filtering EXCEPT for the opamp rails where it matters less. The DAC chips are fed straight from the wall wart.

Opamp staging is pretty simple and follows the 9038Q2M reference design somewhat. Mirrored topology.

Basically ES9038Q2M is in current mode (good for SINAD) and gets resolved to voltage via opamps. I swapped those for some OPA1612s from an old SMSL MA8 board with a dead screen. The buffer opamps are set in unity gain voltage follower mode. On the I/V stage, the non-inverting input is fed with V_AVCC which goes through a voltage divider with two 10K resistors so roughly 1.65V on the non-inverting input. ESS specs the 9038Q2M as outputting about 7.3mA per pin in current mode so the 8Vpp result makes sense as I had 1k resistors on hand. Forgive my mouse-writing chicken scratch.

It;s sort of a bad DAC (especially for the $150 USD price vs competitors) unless you need specific features such as LVDS input (but then the 5VDC input runs directly under the connector with no filtering and no shielding) or a headphone amp (which is very mediocre) and ultimately it will suffer the same fate as the SMSL MA8 1st revision/early FW because the OLED has no screen saver so it will die a slow death.

I only modded it for 8Vpp because it's going direct into pro audio gear which expects +4dBu (3.4Vpp) to +20dBu (21.9Vpp) signals. Balanced passive attenuators (U-pads) on the output of the pro gear feeds consumer level amps (Topping PA3s) which take up to 1.8Vpp signals. But running the pro gear as a gain stage was far noisier; and running low-level signals into the pro gear resulted in touching the noise floor which gave a sort of lifeless and compressed output to the amps.


Anyway, if I had to guess based on SINAD of devices I own (Gustard X20, DX7 Pro, SMSL M8, SMSL SA-S1, SMSL Sanskrit 10th, NuForce HA200) I would be surprised if the sinad of this device as-delivered was over maybe 90dB? They ship it with a very cheap wall wart, economy-tier LDOs, and the power scheme means that any real noise rejection from the PSU is handled at roughly nowhere. THey took the time to isolate the XMOS and LVDS inputs via TI isolation ICs but then left the STM32 directly coupled to the DAC chips and fed them direct from the wall wart with only micro-cost ferrite beads and cheap LDOs in between (oh, and 2 caps I guess). TO make it competitive with a DX3 or D10s you would need a very clean power supply and some component adjustments at a minimum and then you are grossly over the cost of a Topping/SMSL solution. The other options will also already implement OPA1612s or ADA7XXX opamps in the I/V stages and both are either unobtanium or disgustingly expensive right now.

 

[JoeSVK]

so you would not recommend buying this DAC? The op amp swap is a must, it helps a lot. But yes, it is an additional cost.

 

[Veri]

so you would not recommend buying this DAC?

I don't think he would

 

[deadwood83]

so you would not recommend buying this DAC? The op amp swap is a must, it helps a lot. But yes, it is an additional cost.

Yeah not for the money these are asking. Other than 5V the topology really is not bad. But build and PCB quality is lacking and long term there are way better options. I'm still over the moon with my DX7 Pro. In a similar price range there's the DO100, the D10 Balanced, SU-6, DAC-A6, DX3 Pro+, Atom DAC+, Tone2 Pro, etc etc. The Leaf is fun if you want to tear it apart, but quite poor as a finished product.

 

[deadwood83]

A further development!

So I could not get DSD to work with the LEaf drivers on Windows 11 nor could I get it to work at anything above 24/192 with XMOS drivers installed. All yests were done using Jriver MC30. DSD64 test tracks are a Chopin Nocturnes SACD / Miles Davis Big Fun SACD and DSD128 tracks are the Yamina Love Letters EP.

I have learned how to 'hack' the .inf files for driver software to look for the proper PID and VID of this DAC (VID20B1&PID0008). So I started playing around with drivers I have on hand for XMOS XU208 boards. I bought an XMOS board from DIYINHK back in 2016 but I believe that is an XMOS XU216 board because the receipt talks about 768KHz and DSD512. It might be an XU208 but well.... I also cannot find it.

All DSD playback was tested using first the direct driver output in WASAPI Exclusive and ASIO (in that order) via JRiver, then attempted via FlexASIO.

Singxer SU-1 v 5.12 drivers: Will recognize but throw code 37 (cannot initialize)
DIYINHK v 3.38 drivers: Seems to work okay, but still limited to 24/192 and no ASIO/WASAPI DSD.
Khadas Toneboard drivers v4.85: Same as Singxer (no DFU in installation directory)
Aiyima/SUCA Audio XMOS drivers for A6 DAC v4.XX: Never sees the device, even with altered VID/PID. (No DFU included)
Leaf/Nobsound drivers: Same as DIYINHK. Probably similar behavior due to these being driver version 3.00.4. (no DFU included)
Holo Spring1 v5.00 drivers: (only after flashing FW) (not recognized, inoperable)

SO then I decided to roll the dice. But I decided to roll the dice in a really big way. I started playing with firmware. Kitsune Hifi has the 2.02 and 2.22 versions of the SU-1 firmware, and Khadas has their firmware available for the Tone1. I know there was previously an issue with the Tone1 having default VID/PID parameters in its flash and Khadas noted that going through the FW updates assigned a proper/proprietary PID/VID.

First I stepped through the Khadas firmware noting that at each step, including a 'successful' flash of the 2.0 firmware left the VID and PID alone. Then I decided since it all seemed to be a soft fail, maybe I would try the 2.02 firmware for the SU-1. This firmware matches the I2S pinout to Gustard which is the same pinout that Leaf uses. THis is where the journey mostly stops. DAC never locks, DAC is unable to be seen by any DFU app included with driver, etc. I broke out my closet Gustard X20. I have a SIngxer Su-1 around because the Gustard came with no USB card. So now I am faced with the question of do I try and resurrect this cheaply made overpriced DAC or do I just succumb to needing a larger power strip for my HT and hook up the Gustard and Singxer. I would really love to buy just the Gustard USB PCB but nobody has it as a separate item (or if you can find it as a separate item, please please let me know).

I guess another option would be to find my DIYINHK board and see if I can wire it direct to one of the DACs.

 

[deadwood83]

Well, post is too old to edit so here's an update. The ST eeprom holding the firmware seems like it may have been used or just super fragile. Flashrom on an SPI breakout won't see it, but will see my winbond and atmel chips. Maybe not a real part. Also it's marked as a 5M eeprom but takes a 30k bin file so..... now I am doubting a lot of the parts. I looked on the junk PCBs I have and only had a 1k, 2k, 10k, and 16M eeproms in SOP8 (and SOP-W8) format.

Leaf is living up to their infamy. Here's a fun design change compared to Daniboun and others' photos. XMOS isolation? Naaaaaaah.

XMOS is legit, but they just took the last free 'basic' driver and got a cert for their installer package.

SPDIF, COAX, and LVDS are not impacted since they bypass XMOS. But I'm using a Lenovo M75q Gen 2 (Gen 2 Ryzen 7) as my HTPC so USB is a must.
So... is this xmos implementation worth saving? Well, one nice part is that it's sort of almost physically isolated. Not that it makes a lick of difference since they eliminated the isolation IC and then stacked BT right on tip... a BT card which bridges the XMOS isolation from the PSU part of the PCB and GND plane of the analog section.

Out of curiosity, I looked at the clocks. Taiten low-spec +-50ppm on the DAC.

XMOS Section:

SCTF 22.5792MHz clock. It's oddly cheap and made in China and claims to be +-20ppm.

A crystal Marked HHKJ which I can find no data on.

And an even less marked 24MHz crystal that the maker was so proud of, they never branded it.

Anyway, DAC runs in async only mode since the only lines going to the DAC ICs are BCK, DAT, LRCK.

I think an amanero could be manually fitted into the case (with the removal of factory BT) and the dac could actually work to the limitations of its circuit. Keep in mind, the factory regulators as fitted from Leaf have 100-200nV of noise as per Richtek specs. On my unit, I hand soldered in some Ti TPS7A20 LDOs which showed a ripple about 10x less than the Richteks. the richteks again, are 100-200nv/sqrt(Hz).

The entire spec sheet from Leaf must have been dreamed up based on the competition. The opamp layout is actually quite decent and from the dacs to the output relays I can;t fault it much. But the XMOS, BT, 5V layout (the way it directly crosses the LVDS beneath the shield) and all the LDOs in this device are just garbage. They did something really nice with the I/V stage, picked a good DAC IC, then threw in the towel.


Anyway, I was left with a choice. Do I order a few STM 5M eeproms and a couple 32k eeproms? Or do I just get a different DAC? Maybe rehab my closet Gustard X20? It doesn't have the USB card though, and when I tried to use a DIYINHK XU216 card in its place, it was bad. Maybe rehab the closet Gustard AND the closet Singxer Su-1? Feed it all through USB -> Singxer -> LVDS -> Gustard? But then how would I handle the LVDS from my Marantz receiver (pre-DAC, post-audyssey)?

Anyway, Amazon had a Loxjie D40 in stock. It should be here Wednesday. Maybe I'll try and jerry rig this thing and make it a bedroom DAC or machine room DAC? SO it would be:
Office: PC + DX7Pro
Bedroom: Surface Pro + Leaf Audio
Machining Room: {Raspberry Pi?} + Singxer SU-1 + Gustard X20
Front Room: Lenovo M75q + Loxjie D40
Kitchen: M75q + Zone 2 Marantz NR1711
Laundry: Stereo Amazon Echos

Tying it all together: JRiver MC30 on PowerEdge server + Homeassistant + Echo integration.

Back end of DAC is pretty decent with some changes. Front end of DAC is junk. Remote is a gimmick. Would have been fine except I have some DSD and I demand bitstreaming.

 

[daggernc]

@deadwood83 - thank you and appreciate your efforts and inputs on this Leaf DAC.... Downside is that for the price about 8 months for a dual ES9038q2m balanced DAC and the few positive reviews it received, I went and bought one but now I am having buyer's remorse after hearing how poorly it seems to be designed...still would be nice to have someone perform real testing of this unit just to see what it manages to get.

 

[deadwood83]

@deadwood83 - thank you and appreciate your efforts and inputs on this Leaf DAC.... Downside is that for the price about 8 months for a dual ES9038q2m balanced DAC and the few positive reviews it received, I went and bought one but now I am having buyer's remorse after hearing how poorly it seems to be designed...still would be nice to have someone perform real testing of this unit just to see what it manages to get.

Ah! But if you liked it before, then you can still really like it. If you have been happy with the sound for some months then do not let my criticisms dissuade you from enjoying. your purchase.

As I said in my closing line, I would probably be rather happy with it if I did not have DSD music and just run it without the leaf audio driver. I think with the headamp section disconnected it is quite a decent I/V layout, and it would look a whole lot like a Topping or SMSL dac if the sockets were replaced with SOIC opamps (excepting the LDOs and unfiltered 5V) and the rail voltages are derived from a decent monolithic power IC followed by a substantial filter.

The advertising around the DAC is highly deceptive and my ire is mostly aimed at Leaf themselves and not the device because the framework is all there. I would be super curious though if your device has the TI 7240 chip between the xmos and the i2s multiplexers. I wonder how long they have been shipping without.

 

[daggernc]

I have been "planning" a 2-channel audio system for about a year now and started buying components as my budget allowed (the LRS speakers first, then this DAC, and now I'm looking for a decent amp with balanced in that can drive these speakers) but I haven't set up the room/system yet. All I've done is powered up the LEAF DAC to ensure it works, so it is basically new... I could open my unit up and look for the above components/take pictures, though I'm wondering if I should just sell it as "almost new" and then buy a Topping of SMSL DAC. I'd even consider some "easy" upgrades for a novice if this unit could be upgraded by someone without your level of knowledge. Cheers.

 

[deadwood83]

I have been "planning" a 2-channel audio system for about a year now and started buying components as my budget allowed (the LRS speakers first, then this DAC, and now I'm looking for a decent amp with balanced in that can drive these speakers) but I haven't set up the room/system yet. All I've done is powered up the LEAF DAC to ensure it works, so it is basically new... I could open my unit up and look for the above components/take pictures, though I'm wondering if I should just sell it as "almost new" and then buy a Topping of SMSL DAC. I'd even consider some "easy" upgrades for a novice if this unit could be upgraded by someone without your level of knowledge. Cheers.

I do think swapping the opamps closest to the DAC chips to OPA1612 (pre-soldered swappable units are available on fleabay) is worthwhile. The headphone opamps are on all the time, as is the RCA opamp. Since LDO noise increases with load, removing those (if you do not use them) is easy and I think worthwhile.

I can;t stand to throw it away. I broke down and ordered a Xingcore U30 (USB interface). I'll see if I can hack it together. I have a mill so maybe I'll even mount it properly XD (yeah, right).

I'm also not sure if I would call myself an expert at anything. Most of the good bits are already in datasheets and reference/eval designs. It's difficult to say if a topping, smsl, loxjie, soncoz, etc would be audibly better. Peace of mind? Sure. Anything you could actually hear (assuming a listener over the age of 15 who has been outside at least a few times in their life) more difficult to say. If it were me, I would just run it until it dies or the OLED burns out or you actually need something else. I only got deep into it because I needed more gain at the I/V stage to feed my DSP (pro audio gear b/c it was cheap). Assuming good speakers and decent stuff upstream, the weakest component in a modern audio system is probably the potato sitting in front of it all

 

[daggernc]

"the weakest component in a modern audio system is probably the potato sitting in front of it all"....LoL!

I've read up on swapping the I/V opamps (I believe those are the ones you are referring to?) to OPA1612s and will try that. When you say the "headphone opamps" and "RCA opamp" are on all the time and removing those could be beneficial, I take it you mean just remove/pull the stock JRC opamps? I may use the RCA input...I have a turntable that I'm also thinking of connecting to it (I know, heresy using a device with such terrible noise and distortion levels and analog encoding at that! but some of my albums I actually enjoy listening to!) Thanks for the suggestions!

 

[Veri]

I can;t stand to throw it away. I broke down and ordered a Xingcore U30 (USB interface). I'll see if I can hack it together. I have a mill so maybe I'll even mount it properly XD (yeah, right).

Xing makes good interfaces

 

[deadwood83]

"the weakest component in a modern audio system is probably the potato sitting in front of it all"....LoL!

I've read up on swapping the I/V opamps (I believe those are the ones you are referring to?) to OPA1612s and will try that. When you say the "headphone opamps" and "RCA opamp" are on all the time and removing those could be beneficial, I take it you mean just remove/pull the stock JRC opamps? I may use the RCA input...I have a turntable that I'm also thinking of connecting to it (I know, heresy using a device with such terrible noise and distortion levels and analog encoding at that! but some of my albums I actually enjoy listening to!) Thanks for the suggestions!

Correct. If you are building a balanced system then just pull the single opamp at the far right (same side as volume control). If you do not use the headphones, the entire headphone board will pull out of two sockets.

The RCAs on the DAC are not inputs. If you want to play with vinyl in the digital realm (I also do this because I really want to try and make a rumble filter and de-click in camilladsp but that is a whoooooole other project) then the E1DA Cosmos ADC is fantastic. I go TT arm direct to Emotiva XPS-1 (any vinyl pre will do) -> E1DA Cosmos ADC -> HTPC.

Xing makes good interfaces

I hope so! It's about 2-3x the cost of a generic amanero but I emailed Xing to see if they would consider a custom firmware to alter the DSD and PCM pin assignments since the output is fully handled by the FPGA. It seems it should be much easier for the Xing compared to Amanero or XMOS. That way the card would work for Gustard X20/X22 PCIE internal slot as well as direct to a DAC such as the leaf if you solder it properly. I'm not super hopeful to get a custom firmware just from an email but it never hurts to ask. I emailed TRanslogic about their quickshift because it was not compatible with my bike, then when I shared my modifications they asked if they could release the modified product as compaitble with my MY (model year) so who knows? More options is more power to the consumers.

Quick edit, I broke out the Singxer Su-1. I could not find my 1 foot mega thiccccccc HDMI cable so I used a $4 4' unit. For ultra high sample rate stuff, the CMOS is what holds this dac back (or at least, my slightly modded one with OPA1612s and good LDOs).

It will not play 768khz samples nor 705khz samples and instead locks on 192khz and outputs pure, unadulterated distortion. That may be partly due to the Singxer; which, in audio equipment years, is fairly outdated compared to the 9038q2m... and the Singxer only claims to support 384khz anyway. That said, the Singxer also plays DSD512 but claims it can only do DSD256 so....... No idea. But I don;t have any 768khz or DSD512 tracks so it's all upsampled anyway for testing.

But it will exceed the published Leaf DSD spec with ease (via LVDS IIS). This makes me hopeful that maybe it can hit 768k with the Xing. 384k is no problem. All tracks sounded good, even when the volume was lowered. As I have said a few times: the I/V stage of this DAC is really decently done. I'm too tired and it's too late to do an FFT on a 1khz tone; especially since the D40 arrives tomorrow. It has no selectable DSD filters that I could find using the knob. I would guess it's a 50k or 70k filter.

As you can see, my DSP implementation expects MEGA voltage peaks since at 35 on the volume I was worried about the neighbors and it;s barely lighting up the first two lights on the VU meters. That's 35 where full voltage (100 volume) is 8Vpp.

That poor NR1711 is basically a fancy HDMI switch now. Maybe I will add atmos or something on the main amp. Even my sub goes through the DSP.

 

[daggernc]

Dang, my bad deadwood, yes the LEAF has no RCA inputs. I have a no-name ADC that I may try but thanks for the Cosmos recommendation. I have an older Sony BluRay player with SPDIF-coax and HDMI out that I'll be using for my CD collection. I hope the Bluetooth receiver on the LEAF is decent as I'll be using that to stream. I have an old Marantz 18EX receiver beast that I may use as a temporary amp until I figure out what amp with XLR-in I can (afford) use that can drive my inefficient LRS speakers. So next question: should I just use the RCA out to feed this amp or would it be better to get XLR-to-RCA adapters??