MiniDAC - AD1955 DAC with active I/V
Jun 16, 2008 at 11:18 PM Post #18 of 33
Quote:

Originally Posted by holland /img/forum/go_quote.gif
I was thinking the same thing.

Can you do an RMAA for comparative purposes?


Unfortunately I don't have a reasonable capture device at all! It's on the list and I'm hoping to get an Audiophile 2496 or something similar soon, but at the moment all I have is a broken PCM2900-based thing with serious crosstalk and frequency response issues. I will post results when I can get them, I might be able to run the test at the girlfriend's place or at the upcoming Vancouver meet. Or I could just buy the hardware that I need anyway.
 
Jul 11, 2008 at 2:49 AM Post #19 of 33
Finally I've obtained a nice capture card so I can publish some RMAA results. Capture device is an e-mu 1212m (original revision), unmodified (though I do have plans to upgrade the input and output stages). Unfortunately I don't have a pair of 1/4" TS->1/4" TS cables right now to do a loopback measurement. I might try to get some 1/4" to RCA adapters soon. Test setup has the DAC powered by a MiniPow at +/-15V and connected to a PCM2707-based receiver mentioned earlier in the thread. It's connected to the e-mu 1212m by about 6ft of StarQuad with the shield unconnected. Results at 48KHz here: RightMark Audio Analyzer test : MiniDAC 48KHz

I believe the poor THD+N result is due to a long ~6ft unshielded interconnect running behind my computer near DVI, AC power and other signals. I'll try to come up with a better test environment later.

The frequency response is a bit more puzzling, as far as I understood my circuit, it should have flat response out to around 100KHz - but I could be misunderstand it as it was ripped off from Analog. Anyone have any ideas about this?

Noise is low aside from the strange appearance of the relatively strong signals at 1KHz, 2KHz and 3KHz. I think this could also be improved with a better PSU to push 60Hz further down towards -120dB. I'm guessing the other fundamentals are coming from the IC's proximity to other signals.

Crosstalk is a weird one too. My guess is that this is primarily caused by coupling in the interconnect since I'm using one length of StarQuad for both signals. I might build some shielded cables that are set up with separate wires for each signal to test further. The log increase seems telling - but I'm not sure of what - is this just the result of higher frequencies coupling more readily?

I might obtain results at higher sample rates later using an S/PDIF receiver instead. I'd then have to actually write the firmware to configure the AD1955 :p.
 
Jul 11, 2008 at 7:58 AM Post #20 of 33
Hi error401,

I found Rightmark great tool for evaluating dac-s and amps. I also had some issues when I have started using it. I have custom ADC for measuring and it has fairly low input impedance ( cca 500 Ohms ). So resistance of contacts becomes an issue if ground is shared between channels. But this type of crosstalk is resistive and has same amplitude for all frequencies. Yours is not, it rises with f.
So my guess is, that it originates from shared power supply, but that is just a guess.

If 0.5 dB roll-off at 20 kHz bothers you, you can always move your filters up in frequency.

With 16-bit measuring, you are limited to about 98dB SNR/DNR measurements, so you can try RMAA at 24bits. Changes, that you made to your circuits, will be more apparent.

If you want, I can send you my measurements of AD1955 dac for comparison purposes.
 
Jul 11, 2008 at 8:24 AM Post #21 of 33
Interesting idea about the crosstalk, I think e-mu 1212m has 47K input impedance (though it's not really specified anywhere that I can see). I'm pretty sure the almost-perfect log response points to something obvious, but I'm not knowledgable to know what. You could be right about the power supply, looking at the LME49720 datasheet I do see that PSRR degrades along a similar curve, down to about -75dB at 10K. I'd think the decoupling would be sufficient to keep this down though. I really don't know enough to say for sure, and I'm not sure what power supply topology would help. What power supply setup are you using for your I/V stage?

I will leave the filters for now, it doesn't bother me so much as I can't explain it. I may have simulated the filters incorrectly, but I believe their response extends out to 100KHz before rolloff.

Is it possible to set up RMAA to measure at 96/24 while playing back at 16/48? My USB module can only handle up to 16/48 and I haven't made any measurements with S/PDIF input at higher rates yet. That's next on the list when I have time
smily_headphones1.gif


I would be interested to see more on your DAC. Do you have a website or such for it?
 
Jul 11, 2008 at 9:47 AM Post #22 of 33
Hi,
I am using Jung regulators for I/V and for analog portion of dac.
Yes, with rightmark 5.5 you can play 16bit/48kHz and record with 24bit/48kHz, but I guess it is O.K. if you set all to 24bit/48khz and usb drivers will take care off extra bits.
I do not have a website, so please tell me, what are you interested in. I have some sort of "lego" modular dac. I have usb and spdif input moduls, different dac modules ( AD, TI, AKM, Crytal) and different IV converters.
 
Jul 11, 2008 at 10:45 AM Post #23 of 33
I'm mostly interested in your power supply and I/V. I assume you're using separate Jung regulators for Avdd and for the I/V. My power supply is LM317/337 @+/-15V for the I/V, and separate LP2985 off the V+ supply for Avdd and Dvdd.

I/V is taken from the AD1955 evaluation board schematic.

I like the idea of a modular DAC, I'm considering it as a future project if I turn out to be unsatisfied with AD1955. I've also run into a couple of 'issues' in that I didn't think about getting sample rate information from outside (ie. S/PDIF receiver) to configure AD1955 properly since it doesn't have an auto mode. Oops. I will see if I can jumper LRCLK into the microcontroller to detect. So I should rebuild it anyway to integrate into my future preamp (aka HiFi receiver) project, though I may just use WM8741 for this since I have a couple (thanks Jambo) and it's much easier to use.
 
Jul 11, 2008 at 11:23 AM Post #24 of 33
I am using three separate windings, greatz-s and jung regulators for AVDD and for +/- supply of I/V converters. I am getting 119.7 SNR measured with 24bits/48kHz.
Yes, sample rate detection and configuring registers is problematic. You can get pass this issue if you use ASRC before DAC chip. Anyway, AD1955 will work with default settings with 44.1/48/96 and sometimes even with 192 kHz.
PCM1794A can also be your choice, it has current output like AD1955 and works with anything you feed it with, as long you choose I2S/LJ/RJ format right.
WM8741 is only DAC, that I did not try, but I would like to, if I get one.
 
Jul 11, 2008 at 10:25 PM Post #25 of 33
error401, looks like I missed the earlier posts where you finished your amp. Sounds like a nice job, despite the measurements. I'm curious, what parts did you use for op-amp decoupling in the end? Still just the 0.1 uF?

And is there a sweet spot for the sample rate settings fof the AD1955?

Me, I'm still stuck trying to layout my power supply.
frown.gif
 
Jul 11, 2008 at 10:39 PM Post #26 of 33
Quote:

Originally Posted by FS2 /img/forum/go_quote.gif
error401, looks like I missed the earlier posts where you finished your amp. Sounds like a nice job, despite the measurements. I'm curious, what parts did you use for op-amp decoupling in the end? Still just the 0.1 uF?

And is there a sweet spot for the sample rate settings fof the AD1955?

Me, I'm still stuck trying to layout my power supply.
frown.gif



I think it still measures very well. It wipes the floor with my AlienDAC, at least.

Decoupling is 0.1uF directly below the opamp V+ and V- pins to ground, and a shared pair of 100uF OSCON-type caps for each power rail nearby. I should've made it an LC filter, and don't really know why I didn't, but it's too late for that now
wink.gif
.

I haven't fooled with the AD1955 registers yet, there's really only one relevant setting that controls the oversampling mode. At least according to the datasheet, only certain sampling rates work with each OS mode, and I've only been using it at 48Khz, so I haven't even written the firmware to change it yet. I think I will write firmware that should be able to read LRCLK and detect the sample rate and set the register appropriately. Also, I'm wondering about what boris mentioned, and maybe the DAC will run in 8x mode at higher rates.

One thing I didn't really mention elsewhere is that with the active I/V this lets me go for a completely DC-coupled signal path from source to headphones. No caps here!
 
Nov 6, 2008 at 1:46 AM Post #28 of 33
I had some PCBs made, but only enough for my own use. It's a very challenging build and I wasn't even sure it would work so I didn't make more.

If you'd like the gerbers to have your own made, I think it'd cost you about $30 for qty 1 from BatchPCB where I had them done. I can't stress enough how difficult stuffing the board is though, and I'm not certain the silkscreens are correct (or obvious due to part density), and I don't have a placement guide for you. You'd be pretty much on your own.
 
Nov 6, 2008 at 1:53 AM Post #30 of 33
Not that I know of, hence my motivation. You can purchase the ready-made AD1955 evaluation board directly from ADI for about $500 that implements a similar circuit. I know lots of DIYers have used this DAC, but I guess not many have released their designs.

If I were going to do this again I think I'd not try to cram it in such a tight space. Maybe a future project - reimplement the AD1955 evaluation board for less money as it's said to sound extremely good
wink.gif
.
 
You must log in or register to reply here.

Users who are viewing this thread

Total: 3 (members: 0, guests: 3)
Back
Top