AliExpress AD1865 R2R DAC

Status
Not open for further replies.
Can't see the image - but I've not abandoned the opamp DAC idea yet. I'm experimenting with band-limiting the DAC's output before feeding it into the I/V, getting more promising results...

I don't know what size the inductor is before the 2,000uF, but if it is 1uH or larger, this could create significant/rising psu impedance in the audio band immediately before a high speed amplifier. That is asking for an imbalanced sound signature. This LC may even have created a very demanding load for the regulator (what is it?) so it might be close to oscillation. And this rail is feeding a DAC too. Is this a "low impedance supply"? Seems that is a mis-characterisation. It is a low impedance capacitor bank after what might be a high/rising impedance supply shared with a DAC. Why not put in an ldo? It's somewhat ironic to me that you dismissed Eric yet justify this. Why not just use an ldo and make sure your psu is not part of the audible problem? To me, it looks like this is a bad experiment and is a waste of time. Hard to take the results seriously. Sorry to be harsh but why waste time on something obviously inferior - because it is convenient? Um.

A ferrite and cap before the ldo will limit DAC noise passing through above the ldo's bandwidth.

A few pages back it was suggested the AD8065 was better than the OP42 . Is not an ldo better than an LC? Even better - they each get their own regulator.
 
Last edited:
If it's not too much trouble, I'd be interested in that schematic. :)

Since I already simmed it in LTspice, here is the filter stage which is the 2nd opamp (the first one is the I/V which I'll draw up next).

The AD8065, while it has a rail-rail output stage (meaning it can deliver the full 2VRMS CD-standard output signal while running from a single 6V rail) doesn't have a rail-rail input stage. So we need care in arranging the input bias, the highest it can go is 3V on a 6V supply. This kind of filter (called multiple feedback or MFB) works very well in situations where the input common-mode range is limited, it uses the opamp in inverting mode. The caps chosen create a slight lift in the HF but not a full NOS droop correction, then roll off at 18dB/octave (due to there being 3 of them). It has a gain of almost 6dB so allows the I/V stage to operate with just 1VRMS at its output.
 

Attachments

  • AD8065MFB.png
    AD8065MFB.png
    8.8 KB · Views: 542
I don't know what size the inductor is before the 2,000uF, but if it is 1uH or larger, this could create significant/rising psu impedance in the audio band immediately before a high speed amplifier. That is asking for an imbalanced sound signature.

Its 100uH and no, the audio band impedance is in the main determined by the capacitor not the inductor. At the lowest freqs (up to 200Hz or so) the DCR of the inductor dominates (about 0.25ohm), beyond that its the capacitor doing the work until around 20kHz we're left purely with the ESR of the caps (5mohm or so).

I suggest you learn how to use LTspice and simulate it for yourself.

I'm not using an LDO because that'll be an unknown factor, I already have a working shunt reg of my own design. If you want to educate yourself about that, check out my lingDAC thread, the schematic for it is on there.

It doesn't bother me at all that you find it hard to take the results seriously - in fact I've gotten it sounding a lot better now without changing the PSU in any way. Not to say that its already perfect mind...
 
OMG. 100uH ! That's ******* huge impedance, starting to rise above 350Hz. The transient response is dire. Which you've "solved" with capacitors? Even crap stuff on Ebay uses separate regs. The cap bank will dominate the sound signature - I know because I tried this years ago. Never again - total waste of time to try except to learn not to do it. So of course, you can keep working on it to improve the sound. That can go on and on.

Well, good luck. We can agree to disagree and I'll start a thread on the AD1865 if/when it arrives (still hasn't shipped). Oh, but wait, this thread is supposed to be about AD1865. ;-)
 
Last edited:
(the first one is the I/V which I'll draw up next).

Here's the I/V stage - ignore the LT1364 label in this and the previous schematic - its definitely using AD8065 :)

This is the schematic of what I was running which still sounded to my ears distinctly of opamps. That being an artificial brightness to the sound - which some listeners do like, calling it 'detail'.

I've since made a few changes which reduce the brightness but if you're a 'detail' lover then this design is for you :D
 

Attachments

  • AD8065_IV.png
    AD8065_IV.png
    11 KB · Views: 537
Member
Joined 2008
Paid Member
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).

This is off-topic for this thread, but... I don't really understand bandwidth of opamps. I'm familiar with bandwidth in a computing context, that's always expressed as something-hertz, e.g. bytes per second or transactions per second or operations per second. But for opamps it's just "per second", I don't understand what that means.

Also, a related question: how do we determine the minimum needed bandwidth for an application like this?

Furthermore, in the AD8065 I/V stage schematic above, why not have a capacitor to ground right at the tda1387 output? I know doesn't filter everything, but if we can trivially get rid of some junk early on, why not?
 
I notice @Hugh Jazz has updated the first post to indicate the topic has digressed - shall we ask a mod to create a new thread on the TDA1387 DAC from the same designer as his AD1865 board? Then all the posts about opamps and TDA1387 can be moved out of his thread. Or you Matt can start one yourself as you're the only one with the PCB so far, then get the relevant posts moved across. With your own thread we can go OT as much as seems fit.... :p

So bandwidth of opamps - actually its not pure bandwidth (even though yes, the units are Hz) its gain * bandwidth. For a voltage feedback opamp the gain bandwidth product is a relatively constant quantity - its compensated with a single pole which decreases the gain at 6dB/octave - the freq when the gain becomes unity is roughly that of the 'bandwidth' of the device.

As for your second question that's much more difficult to give a precise answer to in an I/V application. In a general audio application its much easier to answer it because the signal bandwidth is well defined (20-20kHz). But coming out of a DAC, who knows what the bandwidth is? Lynn Olson put a spectrum analyser on the output of his PCM63 and found a spectrum that went into the 10s of MHz.

As for the cap to GND question - that's what Scott Wurcer suggests in the datasheet for AD797. Yesterday I did simulate a cap to GND and got a very interesting result for the frequency response. I shall post up the plot later.
 
I suggest you continue to discuss the TDA - 2 reasons

-the first post makes it clear that this is what the thread is about. maybe ask admin to change the thread title too.
-the seller still hasn't shipped - 10 days now - so I cancelled my order. I won't be posting about the AD1865 on diyaudio.

For bandwidth - look at the mclk for an idea of the MHz noise in the DAC.
 
Last edited:
Thanks, I'll stay on the topic of filtering the DAC's output then as Matt asked....

Here are a couple of plots made in LTspice of the I/V circuit I posted up earlier. The first one is what happens with a 2n2 cap to GND from the DAC's output. I chose 2n2 as that's about what I recall Scott Wurcer using on the AD797. As you can see there's a very large peak in the FR. We can quite easily get rid of that though by putting a feedback cap across the 3k (opamp feedback resistor). With 150pF we get a very nicely damped response which is shown in the second plot.

Note I'm using LT1122 for this - chosen as the LT opamp most similar to OP42 (14MHz GBW, 60V/uS slew, JFET input). Changing to the AD8065 would increase the freq of the peak as its a faster part.
 

Attachments

  • CaptoGND.png
    CaptoGND.png
    7.2 KB · Views: 481
  • CaptoGND150pF.png
    CaptoGND150pF.png
    7.4 KB · Views: 489
Eyeballing the second plot the resulting circuit has roughly a 500kHz bandwidth, -3dB. That's 25X the audio bandwidth. This observation gives us a starting point for getting an answer to Matt's earlier question - how to choose the GBW for an I/V opamp?

One way to go about it would be to take the venerable NE5534 as a starting point as its used in so much audio circuitry - it has 10MHz GBW. If that's the benchmark then for 25X the bandwidth we should be looking at 250MHz GBW for this I/V application. Even the AD8065 doesn't make the grade there (155MHz) but it does have a sister part the AD8067 which has even higher GBW (540MHz). The trouble with that part is its not unity gain stable so can't work in a normal transimpedance stage as we have here.

I do have an idea for how to use it though, I'll not write about it until I actually have some of these chips in my hands to try out. Do stay tuned :)
 
I Lynn Olson put a spectrum analyser on the output of his PCM63 and found a spectrum that went into the 10s of MHz.

I've read that article by Lynne Olson. After years of reading people say that op-amps are no good for IV, and trying valve IV stages and discreet transistor IV stages, and op-amp IV stages Ive come to the conclusion that op-amps are perfectly fine, in fact best for IV.

I don't like valves because each valve sounds different and consistent performance cannot be obtained. Discrete sounds fine, but no better than op-amps. I recently found a Yamaha CD-500 CD player with PCM54 DAC and slow old NJM op-amps and it sounds a little dark, but very smooth. Hard to say if the sound is due to the DAC or output stage, but I like it none the less.

I'm not convinced that op-amp bandwidth matters as much as some people say. My best sounding DAC uses AD711. I haven't tried the 844 or the Sen/zen IV stage. Perhaps they are markedly better?
 
I just came across this article in my LI feed, its worthy of wider dissemination as I've gotten a whole lot of benefit from KCP's writings : Push Me, Pull Me—When Supply Currents Go Rogue | Electronic Design

Your journey through the realms of I/V seems to not quite match up with my own @erin. I originally (wearing my EE's hat) considered opamps the best solution and didn't hear any problems with them. Then when I got into modding active speakers (the opamp-based XOs and chipamps) I discovered stuff which impacted the SQ out of opamps (and that's related to KCP's article). Which took me in the direction of discrete designs so that I could make everything classA. Now I'm toying with opamps again to see if they can offer some benefits compared to discretes - at this early stage in my exploration I've not gotten to the level of satisfaction obtainable with discretes but there is a certain impressiveness about the sound which initially is appealing but likely won't stay the course.
 
@abraxalito, thanks for the interesting link to the opamp article.

I suppose for anyone to speak with authority on the subject of IV stages, a person would have had to have tried every single variety.
If a person has tried every type, what's to say that their favourite isn't just a personal preference, rather than being definitively better in all areas?

The thing I find curious is how some CDs or digital files can sound awful, and I used to think there was a problem with my DAC, or IV stage, or source, and then I played a different recording and the sound would be beautiful and faultless. This lead me to conclude that opamps are good because the opamp delivers more detail than either the discreet stage or the valve stage.

I think majority of the problem is that digital audio is so accurate and revealing of problems with the recording that if there is any issue whatsoever, it is revealed by a good DAC warts and all. The fact that some recordings sound wonderful with smooth, clear highs without a hint of fatigue demonstrates to me that it isn't the opamp causing the issue.
I personally attribute bad sound to the recording or mastering.

I listen to classical music and I find that so many analogue recordings used very shrill microphones and mic preamps, so they translate very poorly to digital.

The main point I want to make is that if opamps sound bad, then all music should sound bad, fortunately good recordings show how good opamps can sound. When I had my valve IV stage, the bad recordings still sounded bad but no music sounded better through the valve.
But for a while I talked myself into thinking it was better. Glowing bottle fever I suppose!
 
I'd not wish to speak with any kind of 'authority' on I/V stages, merely share experiences with them. If a person has listened to every type there would be a myriad of other circuit details which may have been uncontrolled (decoupling is one major one, grounding another) so any pronouncements would have to be nuanced.

As regards personal preference vs 'better' I view an audio system as representing a piece of glass. Certainly there are people who prefer tinted glass - but more transparent glass is going to be more revealing of recording details than anything tinted.

True enough that recordings differ a lot - a more transparent system though I've found renders poor recordings more enjoyably than a lesser one. Yep opamps deliver more 'detail' (at least in my experience) but I don't find that detail is really on the recording, rather its a slightly unnatural emphasis on certain upper mid-range frequencies.

Ultimately its about enjoyment and long-term listening satisfaction, which is why I hesitate to come to any conclusions without living with a sound for a length of time (weeks rather than days). My diet is 90% classical music and I tend to agree with Lynn Olson that its more challenging for a system to reproduce accurately than the more mainstream audiophile genres like 'girl with a guitar'.
 
Yep opamps deliver more 'detail' (at least in my experience) but I don't find that detail is really on the recording, rather its a slightly unnatural emphasis on certain upper mid-range frequencies.

How could you know it isn't on a recording? Maybe not on the tape or other media the master was mixed down to, but how can you know the mastering ADC was perfectly transparent?
 
Status
Not open for further replies.