AliExpress AD1865 R2R DAC

Status
Not open for further replies.
Member
Joined 2008
Paid Member
Over on the Head-Fi forums, Jason Stoddard wrote a very lengthy series of posts about how Schiit Audio came to be, along with many of the interesting challenges they've faced. I haven't read all of it, but I found it quite interesting (and I don't even own any Schiit gear).

But one thing that I remember reading (and it's been a while, so I'm quite liberally paraphrasing) that parts sourcing is an on-going challenge. He said that Mouser and DigiKey are actually his last-resort vendors, because even their best bulk pricing is still significantly higher than the huge-volume-only distributors he prefers to use.

To me, it kinda makes intuitive sense that selling small quantities to hobbyists like me incurs a tremendous amount of overhead. I mean, I can go to Mouser/DigiKey and order a single $0.10 SMD resistor. Obviously I don't do that, as shipping alone is at least $5. But I could, and I can't imagine that scenario is at all profitable for them. Even my real orders are usually only $20-50, and they are typically comprised of very small quantities of lots of different parts. Consider how big their warehouse must be given the sheer number of parts they have: they have to pay someone to go pick my relatively small-dollar order (or pay for a really fancy automated picking system). And a large sophisticated database to track their inventories, their own supply chains, etc etc. It's baffling. Not to mention to mention how often I use their website without buying anything, just to see what kinds of parts are available.

There must be 10s of thousands of customers like me. Compare that overhead to a distributor who only deals with customers buying strictly huge lots (10k? 100k? minimum). That's a much smaller customer base, and presumably those people know what they want. The support overhead has to be significantly lower.

Throw currency conversions into the mix, dramatically lower living costs (compared to wealthy western nations), along with "grey market" or potentially fake parts, and I can maybe, kinda-sorta see how these "impossibly" cheap products come to be.

You'd think by now they'd have some detailed case studies of this situation to teach in business school. (Maybe they do, I've never been to business school!) I do find this situation interesting.

While I'm off-topic, another interesting takeaway from Jason Stoddard's book is just how much the case (i.e. chassis/enclosure) costs for a finished product. He spun a lot of prose explaining the lengths he went to to get those (relatively) cheap, simple, utilitarian enclosures.
 
I reckon the "not recommended for new designs" parts the seller is using like OP42 op amps and the grey skinned "out of production" Rubycon nonpolar caps would have been even cheaper than usual... to shift old stock.... while I have no issues with old silicon, those caps will come out, get tested, be over 3% voltage loss and hit the bin almost as fast as I typed this message. I noticed the AD1837 board has bypass caps for the coupling electros so it should be possible to get a fairly decent sounding coupling with big Wima MKS and Vishay MKP1837 bypass. The AD1865 has no bypass cap, only the old ectro. I also saw that the AD1865 board has no shunt resistor after the output coupling cap (to draw a little current within the PCB circuit and charge up the cap... with a slight pop?), which I found odd. I think that a small series resistor (isolate the op amp from the cap), then the coupling cap, then a large shunt resistor would create a tight current loop on the PCB so the DAC wouldn't be heavily influenced by any output cable, following circuit etc. Alternatively, the two resistors would do, with the cap after them and before whatever is connected, noting it'll be a high pass with whatever the subsequent input impedance is.
 
Last edited:
Throw currency conversions into the mix, dramatically lower living costs (compared to wealthy western nations), along with "grey market" or potentially fake parts, and I can maybe, kinda-sorta see how these "impossibly" cheap products come to be.

You'd think by now they'd have some detailed case studies of this situation to teach in business school. (Maybe they do, I've never been to business school!) I do find this situation interesting.
Definitely plausible - and a big factor would be labour costs. Another factor to consider: When you are close to large-scale manufacturing you can also pick up small volumes of leftover parts quite cheap. I've bought kits from eBay with "fancy" components like RN55 resistors etc. when the components are genuine but the values are slightly off (e.g. 2k15 instead of 2k2) and I definitely think this is what's going on here.
 
Something like AD8065 (with more than 10X the GBW) might well do better.

In the past couple of days I've lashed up an op-amp based DAC output stage using the AD8065s I have. They're 145MHz GBW opamps, and surprisingly cheap on Taobao (about $1 for 3) so I figured it was worth a shot.

I'm using one AD8065 for I/V stage (3k feedback resistor and 5pF across it). I'm biassing the TDA1387 with current sources to GND on its outputs so as to get the opamp output voltage close to mid-rail. The I/V opamp's biassed into classA with a CCS to GND. There's a 3rd order MFB filter after the I/V which does a bit of droop correction. These opamps are powered from the same 6V rail as the DAC chips themselves, albeit with an LC filter between the DACs and the ADs.

The end result is fairly good on simple music, only when there's a lot going on does it become more mental effort to try to separate the strands of musical activity. And while there's an appealing bloom or air about the presentation its not completely a 3D one (as best I can tell on headphones). Still its pretty good for the price and simplicity - only against my experimental MkIII lingDAC does it become clear it has shortcomings. Probably all of it could fit on a 5*5cm PCB. If anyone wants the schematic just ask and I'll rustle it up.
 
It would be better if the board have MSB trimmer. It must be worth to mod if it can. The seller also have another 2x AD1865 parallel DAC. Interesting, It have a digital filter, but it must be compatible upto 48kHz. The PCB pattern looks not sophisticated than the single AD1865 board.
 
It would be better if the board have MSB trimmer. It must be worth to mod if it can. The seller also have another 2x AD1865 parallel DAC. Interesting, It have a digital filter, but it must be compatible upto 48kHz. The PCB pattern looks not sophisticated than the single AD1865 board.


There is an IC before the AD1865 but I can't see exactly what it is. I assumed it was to truncate the bit depth... I am not sure I am capable of an MSB trimmer - i thought they require careful accurate measurement to get right? I don't think I have test gear good enough for this.
 
In the past couple of days I've lashed up an op-amp based DAC output stage using the AD8065s I have. They're 145MHz GBW opamps, and surprisingly cheap on Taobao (about $1 for 3) so I figured it was worth a shot.

I'm using one AD8065 for I/V stage (3k feedback resistor and 5pF across it). I'm biassing the TDA1387 with current sources to GND on its outputs so as to get the opamp output voltage close to mid-rail. The I/V opamp's biassed into classA with a CCS to GND. There's a 3rd order MFB filter after the I/V which does a bit of droop correction. These opamps are powered from the same 6V rail as the DAC chips themselves, albeit with an LC filter between the DACs and the ADs.

The end result is fairly good on simple music, only when there's a lot going on does it become more mental effort to try to separate the strands of musical activity. And while there's an appealing bloom or air about the presentation its not completely a 3D one (as best I can tell on headphones). Still its pretty good for the price and simplicity - only against my experimental MkIII lingDAC does it become clear it has shortcomings. Probably all of it could fit on a 5*5cm PCB. If anyone wants the schematic just ask and I'll rustle it up.


I expect the LC filtered 6V supply has quite a bit of impedance, esp at HF, and that won't be doing the high speed op amps any favours. I would expect that to sound far from its best. I'm sure you know how critical amplifier psu is, even for a little op amp with decent PSRR. Perhaps target that to get the sonics to improve? At the risk of teaching my grandma to suck eggs .... op amps usually perform poorly the lower their supply voltage is, and from what I've seen in datasheets, split rail is often better than single rail, even with class A bias.
 
Last edited:
Quite the opposite in reality from what you're expecting.

An LC filter has the C as shunt element, here I'm using a couple of 1000uF polymers in parallel so their ESR most likely won't be the bottleneck (its <10mohm per cap), the wiring will be. In addition each opamp (in SOT23-5) is decoupled by a 47uF ceramic directly beneath with even lower ESR than the polymers.

<edit : attached impedance curve of local decouplers>

As for poorer performance at low voltages, AD8065 isn't your typical opamp. Its fully characterized at 24V (max), 10V and 5V. The GBW is slightly higher at 5V (155MHz vs 145MHz).


A discussion over split rail vs single rail would make an excellent thread. Is there any evidence that split rail is better? I use single rail here for a number of reasons - the main one is that opamp PSRR on the -ve rail is normally poorer than for the +ve. That is indeed the case with the AD8065 - dispensing with the -ve rail means we also neatly side-step the -ve rail PSRR issue (my thanks to @KSTR for pointing this out long ago with his chipamp designs).
 

Attachments

  • TDK476.png
    TDK476.png
    16.7 KB · Views: 492
Last edited:
Another good topic - source impedance vs esr. Many moons ago Eric Juaneda did some subjective testing of this - he found even the power trafo impedance affected the sound and his conclusion was simple - use the biggest trafo because it has the lowest source impedance. A big cap can be good (if it has a flat impedance over a decent bandwidth, which, um, they don't...?) but it's no substitute for a low impedance large bandwidth supply (the regulator is often the choke point, never mind an inductor after it). You can bet the datasheet results weren't achieved with an LC on the power rails so they could conveniently power their op amp from the same supply as the DAC. It's people like us that do that eh? It's a good point about -Ve rails. My prime concern with using single rails is DC offsets, because they require an AC coupling cap. But if they are trimmed out.... then single rail might actually be an overall better solution!
 
Last edited:
Pre-amp iirc. he did a lot of subjective listening tests with LM317 and various capacitors. He used an inductor before the regulator - here's a link. High End Audio - LM317
here is the pre-amp with 300W transformer High End Audio - Preamplifier model 3
and this is the page about transformer size High End Audio - Power Supply
he claims this would hold true for DACs etc too.


I use a 1kva as an isolation transformer (220-0:220-0) to filter the mains (pre trafo filter is a huge air core L from a car subwoofer, C is Y-rated cap, post filter is LCLC with common and differential filtering) but I can't conceive of using such beasts, or even 300VA, for a preamp that uses an LM317. Yet he does. Or did - it's ages old pages, so YMMV ? It's an interesting discussion though - that source impedance idea.
 
Last edited:
His understanding of power supplies is decidedly shaky, and his ideas about how to make them better look contradictory to me.

For example he says the longer the conduction time - time while trafo is delivering current to the caps - the better the SQ. I might well agree with this (after all, a choke input PSU is about the best) but he contradicts it by saying increasing the trafo size makes the SQ better. A bigger trafo will decrease the conduction time because it'll support a higher peak current (lower internal resistances).

Given this contradiction I'm not going to be following up the other two links.

The inductor before the regulator's a great idea in principle, however it rather looks as though his implementation is flawed. He's using the primary of a toroidal transformer as his inductor. This isn't a good idea for a couple of reasons - first the DCR will be a bit on the high side. This might not be a problem though. The second issue is more serious - he's likely going to saturate that inductor without much DC flowing through it because the inductance of a toroid's mains side is phenomenally high as power supply Ls go (think in the region of tens of Henries) hence lots and lots of magnetization but no way to reduce the B due to no air gap.
 
Last edited:
My main concern was that capacitor and regulator tech have improved so conclusions of the "LESS IS MORE" kind don't hold true in some cases. But the principle of low impedance from source to device over a wide bandwidth seem self-evident to me. Add in low noise and that's the holy trinity of power which will bless everything in HiFi. I've heard the benefits of even simple changes like replacing 1n4007 with MUR860 as rectifiers in a single op amp psu so I'm not one for understating the importance of a psu.
 
Last edited:
The end result is fairly good on simple music, only when there's a lot going on does it become more mental effort to try to separate the strands of musical activity. And while there's an appealing bloom or air about the presentation its not completely a 3D one (as best I can tell on headphones).

I've tried this DAC through my dual TDA8932 amp into speakers and its clear its adding what I call 'false detail'. That's excess upper-mid range. On orchestral violins for example they are shifted forward in the stereo image because there's apparent distortion which the ear/brain interprets as coming from the speakers themselves, rather than from the acoustic space.

Seems, despite the very low impedance supplies I can't get rid of 'opamp sound' in this DAC.:eek:
 
Member
Joined 2008
Paid Member
In the past couple of days I've lashed up an op-amp based DAC output stage using the AD8065s I have. They're 145MHz GBW opamps, and surprisingly cheap on Taobao (about $1 for 3) so I figured it was worth a shot.
(...)
If anyone wants the schematic just ask and I'll rustle it up.

If it's not too much trouble, I'd be interested in that schematic. :)
 
Status
Not open for further replies.