IanCanada's Latest RPi GB Goodies Impressions... and your tweaks, mods and hints...

Disabled Account
Joined 2019
I don't know if it's a great info in your system, but it's easy to try before soldering the little smd ceramic caps.
Because the FifoPi and the NDK-SA is not a CCHD-957, your mix may be greater.


Halas, one has to try to know... but the idea is it's easy to test it step by step if the adaptators are unpopulated cause not assembled yet... and even easier as you test the ceramics which are littlier only at the end of this process. Only bad side : one has to be patient and test with music each step.


Hope it may help someone or a phanatic as I am (shame on me), I have halas no BG NX of this value to test by myself // with X7R or NPO. SO sure take it with a pintch of salt...
 
Last edited:
Thanks TioFrancotirador,

So if I understand you correctly the current output (per chip) is like this:
9038Pro = 4x 9028Pro = 16x 9038Q2M

What about the 9028Q2M and the venerable 9018?

9018 has around as much as 9028. and 9028q2m around as much as es9038q2m. Keep in mind that chips can vary couple of % to each other. So 480R is something you should start and then look for value you are looking for. Something between 460R-500R maybe ...
 
Lots of people seem to be using Roon. Should I interpret that as a recommendation? Sorry, a bit OT I know . . .

The current version of Roon's UI is fairly good and the sound is (finally) competitive with the other well-known players. The extra info on the music you have is nice and it even has lyrics now.

I also have JRiver and Audirvana Plus but now mostly use Roon and Spotify. If Roon can support Spotify it will be perfect.
 
Experiment with Clocks

The fifopi can use different clocks which is a great way to tune the sound. I did some quick experiments recently and would like to share some observations here. Please understand that the system is currently running-in so things could change, thus please take my observations with a grain of salt.

First off, I have both versions of Ian's clock adapter boards. The old one that can mount both NDK and Crystek, and the new Crystek-specific version. I have Crystek mounted on both versions of the adapters and they sound different!

Generally speaking I found the old one to sound a bit more spacious with more high freq components while the new one is smoother and fuller. This could be caused by the different number of caps in the adapters (3 in old, 7 in new).

Next I tried to use Pulsar Clocks but they don't fit as they are too tall. I ended up buying a GPIO stacking extender (see photo) to increase the height between the fifopi and the dac hat. With that there is sufficient space to install the Pulsar. Note that the Pulsar does not support enable/disable which the fifopi requires so only one clock can be used (mine's the 22Mhz one).

Comparing to the Crystek on new adapter the Pulsar has stronger mid-bass and perhaps more extended highs. The whole presentation sounds more tidy and controlled. The Crystek has more treble presence and thus sounds more spacious. One is not necessary better than the others and it will be up to system matching and preferences. In my system for my taste I prefer the Crystek over the Pulsar.

And then I thought perhaps the extra stacking extender may affect the sound so I put back the Crystek and indeed it does not sound as good as before. Well, off the extender it goes and then the Crystek clearly sounds better w/o it, with better resolution and micro-details.

This means the Pulsar could have been handicapped by the stacking adapter but alas it cannot be used w/o it.
 

Attachments

  • IMG_2143.jpg
    IMG_2143.jpg
    634.5 KB · Views: 587
  • IMG_2144.jpg
    IMG_2144.jpg
    662.9 KB · Views: 604
Last edited:
Disabled Account
Joined 2019
Hi Jacklee,


Thank you, great tests ! If you try the old adaptators free of the ceramic caps but one acrylic cap instead on the bigger cap area only : you should hear a totally different sounding experience : more articulation, fuller & flesher while having no loss of resolution. I believe what people don't like in the CCHD-957 is the extra ceramic decoupling around it.
 
Thank you, great tests ! If you try the old adaptators free of the ceramic caps but one acrylic cap instead on the bigger cap area only : you should hear a totally different sounding experience : more articulation, fuller & flesher while having no loss of resolution. I believe what people don't like in the CCHD-957 is the extra ceramic decoupling around it.

Thanks for the suggestion. I'm not familiar with acrylic caps. Do you have a source or model I can lookup? Thanks!
 
ceramic vs FCA

Hi diyiggy
I did a little further digging wrt your FCA cap suggestion. Just seeking confirmation of my understanding. Looking at the adapter pcb I believe the 3 ceramic caps are all connecting power to grnd in parallel. They measure .1uF, so it is .3uf in total.

I have a supply of CDE FCA caps that are .68uF but they are the next size up.(1206?) It appears it would be possible to solder one by bridging across a pad. The 3 square pads are +3.3v and the pointed pads are grnd. So it would not be too tough to solder one end to grnd and pick up the 2nd + down the row with a larger sized cap. In some ways it would be an easier smt solder than the 3 tiny caps supplied. Just need to be really careful to not cause a solder bridge resulting in a short. One would end up with .68uf vs the .3 specified. It might actually be a good thing. It appears there are also CDE FCA caps available at digikey in 1206 size that are .33uf. This would produce an almost exact same result as the 3 little 805 caps.

I confess that I am hesitant to do this to working clocks. After messing up several things, it is nice to just have a system that works reliably and the sound is very good now that things are settling in. Probably will try it after a few days when I get the itch to mess with it again.
 
For instance this Cornell Dublier FCA cap in 0805 size &16V (for the iancanada smd adaptator on the bottom biggest area which is made for this size)


Cornell Dublier : http://www.cde.com/resources/downloads/fca_mcd.pdf

Just a word of warning on surface mount film caps.
I've used them, and found them to be fairly easy to damage with heat. These caps were in the signal path in a ladder dac so it was very obvious if you damaged one.
I got the best results if I heated quickly, and only once. I don't try to touch up solder on SM film caps.
When I used them, I'd buy extra, in case you do damage them.

Randy
 
Disabled Account
Joined 2019
@ Wolwes : try the 0.68 : it's in between the 0.47 to 1 uF I advised ... at least with the CCHD-957. If you have not 2 others smd adaptators on hand, indeed why to bother ! And I told you the FifoPi having at the bottom an acrylic cap already, you perhaps don't need it... There is a chance iancanada used it cause it readed it somewhere... x7r or tantalum are less expensive. I'd like to know how the FifoPi compares vs the FiFo II board or the MC board.:)

I really think you should try first: one adaptator with the other unpopulated of any caps the day you try. Notice the difference you will have is also related to how you feed the board and the PCM board with its powersupply cap.

So enjoy first your good dicscover, you're FifoPi is not burned in yet :)

@ randystuch : that's very true : you need rapid soldering, not too much heat so rapid soldering skill, and most of acrylic references are only 20% precision but the Panasonic PPS which are bigger ! But here the idea is not to use them with a very precise value. I like the CDE more than the Panasonic, but didn't try the Rubycon iancanada uses at the bottom of the FifoPi. However, anyone having a Clock II or a MC Clock and those adaptattors with CCHD-957 should jump on that tip.

Enough said, it's a FifoPi thread here and the Powerboard made of LiFePo4 should maybe changes the game as well (though I really didn't like a123 cell with clocks... certainly because there was a LDO in between)
 
I just ordered me some sparkos labs and burson v6 vivid opamps. I can’t wait!

I soldered some 0,47uf black gate NX caps to the power pin of one of the clock positions. The NX series work at their best when paired in anti-parallel configuration, this is the preference by black gate called super-E cap and it will lower the esr considerably.
It is running in now and I will report the outcome later.
 
Crom,
Interesting feedback re sound quality.

Where exactly are you going to place your super caps? Right at the clk?

I have some 5v super caps. I was thinking of putting one right at the input to the +3.3v clean power on the FIFOPi. So it would be right at the output of the 3.3v regulator/input to the FIFO. I need to do some digging and find Greg's suggestion.

Maxwell has modules up to 48v. Probably very expensive. As I understand it, they series connect individual cells and balance the current flow until they get the target voltage.

Interested in your report.

Thanks. I will look for the maxwells. I just put them at the main supply input of each card. Initially one in the 2nd clock board in the DAC (mcFIFO) and then I added 2 more, one on each of the supply inputs to the twisted pear dac boards. I’m not overly enamoured with any of them so I may move them. I do have a longer term plan to replace the avcc regs in each dac board with direct lifepo power and my current on board regulation is pretty good so maybe that part of the system is already optimised. More experimentation needed. I do urge people to try double-reclocking though!!
 
Hi there.

I have a FifoPi Ultimate here for test.

Do you guys think a clock upgrade is a must? The manual somehow implies it. And at least some folks over here seem to confirm it.

From what I understand I would have to order the adapter boards from Ian and the clocks from DigiKey/Mouser. And then I'd have to put it all somehow smd-style together. Sounds very much like a DIY :wrench: challenge.
And the cost for it will at least be another $80. :eek: Almost as much as the board.

I don't understand why Ian is not offering a good set of clocks by default!?!?
Or did I miss something?
DIY activities in the clock department without having top measurement equipment might become a bit tricky. Working with adapters on clocks is probably also not the best of all options. Hmmh!?!?


Are there any measurements (jitter/noise) about The FifoPi Ultimate available btw?
What's the current consumption dirty/clean?
Didn't see anything in the manual. Probably I missed it.

Thx.
 
Last edited:
Transformer I/V Outputs

@Greg,

I read that you have experimented with various transformer output options. Could you share your latest thoughts on this?

I am now using the transformer output HAT and really likes it. In fact I likes it better than the Std I/V HAT with Sparkos opamps. Thus I want to explore if there are even better transformer choices available.

I tried to read up on current-output DAC and honestly I am not sure I fully understand it. So far my understanding is:

- The optimal load for current-output DAC is a virtual ground but this doesn't work in transformer output as there won't be any voltage generated.

- This is why a resistor (510ohm in the HAT) is used to generate the necessary voltage from the output current. This is not ideal as the voltage generated will interact with the current output source so minimizing the resistance of the load is desired.

- With a line transformer primary in parallel with the resistor part of the output power is then transformed to the transformer secondary. The exact current passing through the primary is then determined by the impedance reflected from the load of the secondary. Given the high input impedance of the next stage (i.e., preamp) this will likely be high so the current passing through the primary will be small compared to the resistor.

- Thus it would be desirable to choose transformer with step-up ratios so that the reflected impedance will be lower. Another benefit is that the step-up will allow smaller value resistor to be used at the primary as the voltage needed can be lower. However, the tradeoff is that the step-up will also multiply the source impedance which could hamper the secondary's output capability, i.e., leading to freq roll off when coupled with the next stage's impedance characteristics. So a balance must be strike here.

- Finally, ideally, if the impedance of the next stage is really high (e.g., input grid of a tube) then perhaps one can use a very small resistor value, e.g., 1 ohm with a very high-gain step-up transformer (e.g., mic or MC). In this case then the DAC can operate in near ideal current output mode. This is just some thoughts I have for now. One will need more precise calculations to determine the operating points or whether this even works or not.

Any comments and thoughts are welcome. Any suggestion for potentially better transformer options are most welcome as well!
 
Last edited: