Monster DAC from AD AD1955

This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Hi, everybody!

Have you seen the latest in digital audio? AD1955 from Analog Devices. Price tag 6.78 USD!

-110 dB dist

123 dB S/N

Not bad!

"Mums" (jummy) as we say in Sweden.

Edited: When I have read the datasheets, Crystal seems to be the leader, followed by AKM but Analog is catching up. The chip has a nice performance despite the "third" place.
Very reasonable price for the capabilities; with all the functionality they're putting in a 28 pin package, it's no wonder control functions must be programmed via the SPI interface; but this makes a DIY implementation a bit harder, as it lacks the "hardware modes" of the CS devices like the CS4397 and CS43122. And I don't want to put in a micro just to implement SPI.

OTOH, it has direct current outputs, which if you want to avoid the I/V opamps of the Crystal devices is a good thing. But, even with the older CS chips, like the 4390, you can make a very nice sounding DAC with attention to the proper details- output opamp or not.


Looks interesting, but was wondering.... Off Topic.

Most modern 1-bit Delta Sigma DACs have alot better THD and S/N specs than the earlier multi-bit R2R ladder DACs. But looking at the "Accuracy, Gain Error" spec, the multi-bits have at least a magnitude lower (AD1865@ 0.2% vs. AD1955@ +/-3.0%) performance spec. Could this attribute contribute to the perceived sonic differences between these types of DACs?
Since the quoted gain error spec is a measure of the variation in full scale output voltage, compared with full scale digital input voltage, and since channel tracking is generally quite good in stero DAC's, I'm not inclined to think that's the reason.

A lot depends on the overall implementation; some people don't Crystal DAC's are truly "high end", but with the right implementation, you can get an SPIF driven 4390 based DAC to sound better than a moderately expensive integrated player (like the Sony SCD777ES)... (I know, I have both ;)

I think low level gain error and spuriae are important; evidently Crystal does, too, given the work they put into their CS43122.

Paying a lot of attention to jitter and power supply regulation is also important; jitter can throw away the performance inherent in an otherwise good DAC. Anything affecting low level linearity and gain error is probably the place to focus attention.

And comparing an 18 bit PCM DAC and a 24 bit sigma delta, which uses a totally different concept for the reference (charge based, not voltage based) is sort of an apples and oranges thing, though in the end, it's still the output that counts. The PCM DAC would certainly be the way to go if you wanted absolute minimum gain error- but this is more important in measurement ADCs, not so important (I think, but who knows) in a playback DAC.

Best regards,


Then again, if you want to have your cake and eat it too, a good choice to consider is the BurrBrown/TI 1730; a combination segmented PCM and oversampling delta sigma architecture; wide dynamic range, quiet, and you get to work with the current outputs directly.

It could be a very tempting device to use in combination with the CS8420, say at 4X upsampling/reclocking to 176.4 kHz.


It should be a real interesting part. I believe Hansen is using it in the new Ayre CX-7.

I've been simulating some different approaches to I/V design that coudl be used with it (non feedback common base or common source topologies), but I'm also wondering if it wouldn't be interesting to use a broadband line transformer to couple the differential outputs to an I/V stage; a lot of the common mode switch noise would then never be seen by the I/V stage, instead of having to be tolerated as a brief HF overload. If caps to ground at the I/V output are used to help supress spikes, they'd have to be matched, then.

It's remarkable what AudioNote has been able to patent on the concepts for using a transformer at the output of a DAC, but then I'm frequently surprised at what (obvious) concepts have been patented.

Good luck with your project, and let us know how it turns out- I've got to finish one I'm already working on with a CS8420 mated to a CS4397 before I tackle the 1730. (it's my worst habit, coming up with new project ideas before the old ones are completely finished ). ;)


The DIR1703 and PCM1730, 3 of each, just arrived (thank you TI samples!).

Both are in 28 SSOP (nano-sized) packages. They are going to be a major PITA to prototype, but hey, I'm here for the adventure. The complete PCB (DIR/DAC/Analog outputs) could be around 1" x 1". That ought to make those Gain Card folks envious.

Initially, I just intend to implement the Analog Output Section from the '1730's datasheets, except substituting AD8620s (dual AD8610s) for the recommended OPA627s.

I also have been thinking of a non-feedback common base BJT current mirror topology (designed by Dave Ingels), but it might possibly be prohibitive due to the high degree of device matching required for differential operation. I dunno, i'm going to have to give it some more thought.

The Audio Note approach you mentioned does work well. Previously, I had a well modified Audio Note DAC 1.2 set up with Sowter 8347s, and then was a beta tester for a pair of Audio Note DAC 5 Signature all silver "Magic" I/V interface transformers.

I'll report back when I get this thing up, running, and listenable. It going to take me awhile, as getting back to work is my first priority right now.
From a DIYers perspective, the DIR1703 -> DF1704 -> PCM1704 looks slightly more appealing to me. All three chips can be obtained from Digi-key (unlike the PCM1730 or PCM1738, its cousin.) The DIR1703 is a very nice receiver -- very, very low jitter (at the cost of an external clock) and can resample to 24/96. The PCM1704 has great specs and is used in very many boutique CD players as well as Mark Levinson DAC's.

The only downside is that the DIR1703/DF1704 are only available in 28SSOP. This is quite the nuisance for breadboarders or people without a wave-soldering machine. You can get 28SSOP to 28-DIP converters for $30 apiece, but you still have to solder the chips on somehow.. but at least you can do some breadboard experiments. The PCM1704 is available in a very convenience DIP package, although it's pretty expensive (~$35/ea, and they're mono DAC's..)

For those who _really_ want the 24/192 capability, the DF1706/PCM1706 pair will work. Of course, you need a high quality upsampler somewhere along the line. The AD1896 comes to mind, but that's pretty hard to get in small quantity. (Remember, S/PDIF only goes up to 20/96 or so. Any DAC doing 24 bits is pulling from the 'reserved bit' pool, and anything doing 192kHz needs an upsampled signal.)

My $0.02,
You can hand-solder SSOPs with an iron if you are careful. Tin the pads, load up a small amount of solder on the iron, and just move down the pins w/o worrying about shorts. If you go back with a fine desoldering braid you can remove only the shorts, then resolder anything that has gone open.

Using cold spray to keep the die temperature low is probably not a terrible idea.
I've finally tried this method out --

It's very clever, but... it doesn't work. At least, I tried it last night and I just couldn't get the temperature high enough to get the solder to melt properly. It just sort of solidified into a big crumbly block of powdery stuff.

If one _could_ get it to work, this would solve lots of problems for DIYers.

best DSD-input DAC?

hmm, so i'm looking to build a new analog board for my Sony SCD-C222ES SACD changer. because Sony was kind enough to put the DAC on the analog board instead of w/the digital circuitry, i even have the option to change the DAC chip entirely. so i'm looking for suggestions... the obvious choices are:

BB PCM1738
Crystal CS4397

the stock DAC is a DSD1702. i'm assumming the 3 chips above will offer improved performance over the 1702 but i'm not sure if it's worth the extra implementation hassle.

My opinion is that the "easiest" implementation would be with the CS4397; I've been working with that chip, and combining it with an 8420 as an SRC/de-jitter device; you won't need a receiver, of course. It comes down to how much of a drawback is the built in opamp and I/V converter; if you do everything else right, even the "humble" 4390 can be made to sound better than an SCD777ES on Redbook. I'm using transformer coupling and a NLFB buffer, with balanced outs. Capacitor selection is important, I think; I settled on OSCONs and stacked films for the digital and analog sections with additional small value ceramic on the digital.

I do like the sementated DAC architecture chips from BB, and I'll try that with my own I/V converter next, but a couple of other projects stand in the way. If I was starting from scratch again, I'd probably go that route first, though it will present additional challenges.

In either case, unlike the 4390, a four layer board was inevitable and necessary to get the best performance.

Happy building....

peranders said:
When I have read the datasheets, Crystal seems to be the leader, followed by AKM but Analog is catching up. The chip has a nice performance despite the "third" place.

Crystal and AKM seem to share a common development center. You will find that most Crystal parts have a very similar or sometimes even 100% functionally identical AKM sister part. The data sheets are most certainly written by different teams who seem to have a completely different philosophy of compiling documents. The measured or guaranteed values are also different. I am not sure whether this is due to different measurement equipment and procedures, different pcb prototyping or even differences in the chip production process.

What Crystal and AKM parts are you referring to that are superior on paper to an AD1955? The CS43122 is hardly on par even with an AD1853.

This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.