RTZ SDM (DSD) DAC

[iansr]

A huge thanks to Tony for doing all that tricky SMD stuff and putting this DAC together for me. Over the next few days I’ll finalise the power supplies and hook it up for a listen. Really looking forward to hearing it

[Tony Moore]

I am back to the first RTZ DAC again now. I just cannot leave well alone...

Latest stage was to bring DAC1 up to the status of DAC2 regarding supplies and power distribution:





Basically LDO regs / superregs on all separate supply lines. (double-decker) At the moment this dac is using the Squeezebox Receiver as source of I2S but that will change in the next experiment!

The plan is to do some experiments with better master clocks and re-clocking of the DSD data coming out of the PCM2DSD converter. To do this I can't use the SBR as I cannot easily pick up on the MCLK. Or at least don't want to. However, if I switch over to the I2SoverUSB (JLSounds) board that I have then I can inject an external MCLK at double the frequency of the BLCK of the DSD. This is going to take a bit of fiddling as there is not sufficient room in the case fit much else and at this point I don't want to bring it all out onto a wooden box again. So plan B is I have an I2SoverUSB already mounted on a pcb I made which has LVDS transmission over an HDMI cable and I have a receiver board I made to hook up to the PCM2DSD so if that works it will provide a test-bed to try out reclocking and assuming I come up with a better result then I can build a dedicated pcb to carry the I2SoverUSB and PCM2DSD together with 2 x XO modules, RF relay, LDO psu and isolation/reclocking for the DSD signals.

Watch this space!

[iansr]

So my RTZ DAC now has 5 separate PSs attached and is up and running. I’m using LT3045 / 3094 regs almost everywhere but instead of just using trafos I’m using some Calex LPSs as pre-regulators. I bought the Calexi at knock down prices when they went EOL a few years back, so I figured why not use them. I’ve no idea if the pre-regulation is adding anything (or should that be subtracting !) but I can tell you the DAC sounds fantastic.

Huge thanks again to Tony for the SMD work and testing the boards. Will post a pic once I’ve reduced the file size.

[Tony Moore]

In an attempt to squeeze even more performance from the RTZ DAC I have been prototyping, with help from Mark over on DIYAudio ways to improve the clocking of the JLSounds / PCM2DSD / RTZ combo.

Mark started a thread based around his design of a set of prototyping pcbs that allow for various building blocks of clocks, PLLs, logic and re-clocking/isolation that can be used to try various combinations and listen to the changes. He kindly sent me pcbs to try out.

With his boards I have experimented and then decided on a compact version, based on the same component choices and topology and this pcb is now up to version 9! (Although not all versions were constructed)

This is the version 9:



Essentially this board has sockets for 2 x XOs of either 22/24Mhz or 45/49Mhz, an RF relay to switch the XOs output, then clock buffer chip and D flip-flops to re-clock the DSD output from the PCM2DSD unit (mounted underneath the board) The I2SoverUSB fits on top. If using 22/24Mhz XOs then the 45/49Mhz doubled clock is created by a PLL. LDO power supply is based in LT3045.

Advantages is that the DSD is re-clocked by a very low phase noise clock, plus the co-ax delivery of the DSD to the RTZ board is superior also to the IDC header, where the pins are close enough to be interfering with each other. (The IDC connections can be disconnected on the RTZ by removing the links Marcel provided)

Many thanks for Mark for providing many of the ideas. component choices and general experience of what works well.

[MistyBlue]

Awesome stuff, Tony.

I was recently trying out some manual SMD soldering with some paste and a practise board from aliexpress, to pave the way for doing the Spartan6 on the PCM2DSD board. Surprisingly, it was easier than I originally thought with the QFP packages, and it was the 3 pad SOT transistors that gave me the most trouble with the initial tack down and positioning, so I got some 0.5mm SAC305 solder wire to do that part.

I was just starting to feel more confident about it and the Sugden A21 I’ve been using to drive the LF section finally gave up and developed a buzz which starts after it warms up to normal operating temperature. I could investigate and fix it, but I’ve had the MoFo parts gathering dust for a bit too long, so I’m going to have to divert efforts to do that.

I’m a bit gutted, but the situation was hardly surprising given the Sugden’s age.

[Tony Moore]

Actually, the Spartan6 is the one part I don't use paste with. I find it is easier to use clear flux paste and then line up the chip, hold it down firmly and pin the corners with soldering iron. Then drag solder each side in turn. Finally microscope to check for bridges and use solder wick on them.

My problem with solder paste is that it is very difficult to get the correct amount, even with a mask. Then once the paste is on aligning the pins on that package is a nightmare as you can't see properly anymore. Also the solder tends to turn into tiny balls too which are ok on the bigger pin separation packages but the Spartan pins are so close it's really easy to get some stuck underneath that the wick can't shift.

A friend of mine says he glues down big chips like that prior to drag soldering and I did try that on one but my feeling is it might raise the chip marginally and then you're not certain the pins are in good enough contact when soldering.

It's all down to personal preference in the end. Whatever works.

It's a reasonably expensive chip so I'm _always_ nervous when soldering one. One thing that can be disastrous is when wicking off any bridges be very careful not to pull the wick without enough heat as the pins will bend very easily. Then it's game over.

[Tony Moore]

In situ (temporarily) in the 2nd DAC case:



Not yet working, I seem to have a dry joint somewhere...

Close up:



24Mhz XO missing for now as Mouser have back-order.

[Tony Moore]

Working now!

[MistyBlue]

Interesting stuff. Thanks for sharing.

I did try the technique similar to what you describe at first, as it was featured by the Australian guy who does EEV blog. I think what put me off was the way the solder was behaving when I tried – it just wouldn’t transfer and kept stuck in a ball against the iron tip. However, I did try it with a reel of 0.7mm lead free stuff I bought many years ago which I had left on the shelf before remembering that I stopped using it for a good reason.

worst solder ever.jpg (84.88 KiB) Viewed 7170 times

Worst solder ever.

Not that I’m against lead free solder, I actually prefer to use it – it’s just this particular solder was, how can I put it, crap.

So recently, I got some samples from Ebay of Qualitek SAC305 NC601 and Kester SAC305 285 for evaluation before buying a reel, as I need some more 0.5mm for general use. These have both been good for the super reg boards I’ve been doing and seem OK size wise for most of the SMD stuff. I think the qualitek stuff has the edge with the wetting.

The solder paste has been working quite well. I tend to splodge the pads with the dispenser, leaving some parts of the pads still visible, place the component and align it under the microscope and then tack it down at opposite ends (pads) and go from there, weaving the soldering iron tip between the pins and allowing the surface tension of the melted solder to wet onto the pads and pins, while watching/coordinating through the microscope. I do worry about some of the residue (mostly flux, but occasionally there is solder powder remaining that has failed to agglomerate upon melting), and of having a conductive bridge forming under the device where I cant get to it. The smaller TSSOP packages are not so easy due to the fine pitch and require solder wick to get rid of the bridging.

When I did stuff with the reflow oven and the pick and place/pcb prototyper machine development, I used to dispense the paste by robot, doing each pad with a kind of programmed smear motion, which worked quite well. I’m always shocked by how much solder paste ends up on pads when using a stencil and am surprised that it doesn’t just bridge all the pins. Under reflow/hotplate conditions the components tend to self align with the surface tension, which was always satisfying to watch. Of course, doing things manually is a bit different.

Originally, I used Multicore LF318, but it’s expensive. I’ve actually still got the tub in the bottom of the fridge, but it’s long out of date. Another one I was very impressed with was Chipquik SMD291SNL, which is available in small quantities (ie just a small syringe full). I could then just load it onto the machine without having to scoop it out of the tub into a syringe barrel and deal with the inevitable mess and put it back in the fridge when I was done. I recently tried some unbranded stuff from aliexpress which seems OK. The Chipquik is my favourite though. Must get some more.

I did some searching around today after concluding that I really need some sort of clamping mechanism to hold down SMD components while tacking them on. Typically, I had designs forming for all sorts of over engineered and elaborate stuff to do this, but found this: http://vpapanik.blogspot.com/2015/02/the-smd-beak.html

The other member of the household has been hinting at me about welding aids like this for a while. So I rushed off to the shed to build myself one. No doubt I will be building a larger one for her welding antics soon.

SMD solder beak.jpg (111.22 KiB) Viewed 7170 times

I will give the solder dab/drag technique another go, as it seems to be the most recommended, and I’ve now got some decent solder and flux pens to try.

[Tony Moore]

Well, moving on with this project I have now got to version 11 of the re-clocking board I am working on. Eleven revisions sounds like a lot (it is! Oops) but only 4 were constructed, the others are scrap.

To give some idea:



and this is it lashed up into the DAC I brought to Owston last.



Basically, this board hosts the JLSounds I2SoverUSB and PCM2DSD with XO sockets for the likes of NDK, Crystek, IanCanada, etc XOs, an LDO to power the XOs and logic, a clock buffer, D-regs to re-clock the DSD outputs and a PLL to double the clock to feed the JLSounds external master clock.

The correct XO is selected based on the JLSounds sample rate. The board will pass through native DSD or handle PCM with correct setting of the clock.

The co-ax connections also improve on the cross-talk that would normally be seen on the DIL header. (ie remove the zero ohm links on the RTZ, which I've not done yet but will do and compare)

[Ray P]

Interesting developments Tony, I'll be interested to try some of your boards - I'm starting to get back into audio projects now summer is over!

[Tony Moore]

Assuming this v11 works out ok I will send you a pcb Ray.

[Ray P]

Assuming this v11 works out ok I will send you a pcb Ray.

[Tony Moore]

Update. Now re-cut the rear panel of my 2nd RTZ DAC to allow for the USB socket of the JLSounds. Also re-printed new sub-structures to hole the extra power supplies and reclocker board.



The main additions now are separate supplies for the JLSounds USB side and a pre-regulator on the clock supply. Both these things seemed to offer improvements.

There are 8 supplies in there as of now. No more room so I'm calling it done! (I know, I've said that before!)

I should've said, it sounds very nice! It is as identical to the 1st RTZ DAC as I can get it and they sound the same to my ears. I now have to complete the 1st DAC which has more internal room in the case, however the rear panel does not come off so cutting the hole for the USB is going to be a nightmare.

[Tony Moore]

And....this is the other DAC. Some people might recognize this one from days gone by? Now in another iteration!