I added just one 100n cap (smaller one), according to NDK datasheet (between Vcc and GND).
Just wondering, if should add the other extra 100n and 1u caps.
@wlowes
You seriously still use an el cheapo ethernet cable?
**********************
Another subject @all
Just read a master theses about vibration impact on clock crystals.
It wasn't about audio btw!
The guy measured 30dB! higher sideband phase noise under
random vibration impact in one case. That's a lot.
Bottom line. Vibrations can have quite an impact.
Even if the clock spec looks great, in a real world environment things can get messy.
If you all know that, sorry for bugging you. I didn't.
Temperature drift also impacts the performance of course.
Can we do anything about it? I do recall that some people were gluing whatever stuff on the crystals.
I was thinking now on of loosening the nylon spacers to keep the reclocker kind of floating and perhaps to put some plasticine on top of the crystals.
Any other ideas?
You seriously still use an el cheapo ethernet cable?
**********************
Another subject @all
Just read a master theses about vibration impact on clock crystals.
It wasn't about audio btw!
The guy measured 30dB! higher sideband phase noise under
random vibration impact in one case. That's a lot.
Bottom line. Vibrations can have quite an impact.
Even if the clock spec looks great, in a real world environment things can get messy.
If you all know that, sorry for bugging you. I didn't.
Temperature drift also impacts the performance of course.
Can we do anything about it? I do recall that some people were gluing whatever stuff on the crystals.
I was thinking now on of loosening the nylon spacers to keep the reclocker kind of floating and perhaps to put some plasticine on top of the crystals.
Any other ideas?
There is tight bond due to the GPIOs, so floating connections might not have the intended impact. Probably the whole PO stack should be suspended.
Cheers, Ernst
Yes sir. I even got ridiculed for claiming improved sound when I upgraded to Cat6
Thanks for the vibration reference.
No question it has an impact on sound. Prior to FIFO, I used Acko reclocker. Mac tac on the crystal and suspended the whole thing with dental floss for best sound.
With FIFOPi, I mounted the whole stack using silicone plugs in a weighty wood block. It was in turn mounted under a chassis divider which itself is on silicone inside a wooden chassis. On top of the divider (copper BTW) are zip lock bags of sand. The sand really dampens the vibration and the sound impact is easily heard.
Without all of this there is clearly a feedback loop particularly in the bass notes causing an image smear. With vibration controlled, the sound is tighter and pin point accurate.
My DAC, with wood chassis, 7 transformers and many chokes, 2 autoformers and sand weighs in at around 80 pounds BTW.
Last edited:
Vibrations
If you check back to the older group buys for the standard FIFO, the clockboards were designed as a board withing a board so you could cut the tabs joining the inner board to the outer board and suspend the inner board with o-rings to cut down on vibrations. Maybe the whole RPI/isolator/FIFO/dac stack can be mounted on a board a bit wider with more "hooks" to connect o-rings and suspend the whole stack? It's heavier than just the regular Ian clockboards:
Ian asynchronous I2S and S/PDIF FIFO KIT group buy
Also, when attaching this "outer" board to the chassis, we could use the vibration damping screws/washers that Ian recommended for the boards in the old group buys (listed "anti-vibration screw grommet set" in the group buy, look at the photos at the end of the posting):
Ian asynchronous I2S and S/PDIF FIFO KIT group buy
The temperature effect on clocks I think Andrea and others mentioned was more pronounced over the long term stability of the clock more than short term? That said, he does have clock boards with ovens to compensate for this I think?
If you check back to the older group buys for the standard FIFO, the clockboards were designed as a board withing a board so you could cut the tabs joining the inner board to the outer board and suspend the inner board with o-rings to cut down on vibrations. Maybe the whole RPI/isolator/FIFO/dac stack can be mounted on a board a bit wider with more "hooks" to connect o-rings and suspend the whole stack? It's heavier than just the regular Ian clockboards:
Ian asynchronous I2S and S/PDIF FIFO KIT group buy
Also, when attaching this "outer" board to the chassis, we could use the vibration damping screws/washers that Ian recommended for the boards in the old group buys (listed "anti-vibration screw grommet set" in the group buy, look at the photos at the end of the posting):
Ian asynchronous I2S and S/PDIF FIFO KIT group buy
The temperature effect on clocks I think Andrea and others mentioned was more pronounced over the long term stability of the clock more than short term? That said, he does have clock boards with ovens to compensate for this I think?
@palmito
Interesting that potentially good solutions just get forgotten.
Or maybe Ian had his reasons !?!?
@wealas
These special projects are IMO nice to prove a point. But many of these "solutions" to whatever problems just increase the complexity of a system.
And that usually causes other issues.
That's why I'm not a big fan of hooking up 4+ power supplies to a HAT tower.
If 4+ PS sound better then 2 it'll just show that there's a weakness on the HAT. That should trigger the designer to look after it.
The same applies to attaching Farrads of capacity to a (or any) power rail.
As long as nobody understands what's actually going on all this here remains a wild fishing exercise and will keep us tweakers busy until EOL.
I prefer to keep it simple. And no. I won't buy a clock at $50 or 2@100 if there are numerous other flaws in a setup! If I see setups with noisy ethernet cables or 2.5$ RCA cables hooked up to 1.5$ RCA jacks discussing -140dB phase noise I think people got their priorities wrong.
Enjoy.
Interesting that potentially good solutions just get forgotten.
Or maybe Ian had his reasons !?!?
@wealas
These special projects are IMO nice to prove a point. But many of these "solutions" to whatever problems just increase the complexity of a system.
And that usually causes other issues.
That's why I'm not a big fan of hooking up 4+ power supplies to a HAT tower.
If 4+ PS sound better then 2 it'll just show that there's a weakness on the HAT. That should trigger the designer to look after it.
The same applies to attaching Farrads of capacity to a (or any) power rail.
As long as nobody understands what's actually going on all this here remains a wild fishing exercise and will keep us tweakers busy until EOL.
I prefer to keep it simple. And no. I won't buy a clock at $50 or 2@100 if there are numerous other flaws in a setup! If I see setups with noisy ethernet cables or 2.5$ RCA cables hooked up to 1.5$ RCA jacks discussing -140dB phase noise I think people got their priorities wrong.
Enjoy.
As palmito's post demonstrated, they did not get forgotten. Ian along with others came up with interesting solutions to the same problem. Clocks perform better when isolated from vibration.
His floating pcb on an elastic band was taken further by other experimenters who found suspension with dental floss to be superior. Where possible I use both approaches.
His rubber grommet mounting was also effective. It prompted me for example to set the standoffs in 2cm hole filled with silicone as a cheap but effective proxy.
Everyone has their priorities. But if you are suggesting that buying advanced ethernet cable (or any cable) should be a priority over perfecting PS or clocks, then be clear that priority is not driven by sound quality.
I agree that cables make a difference, and in the case of IC, power and speaker cables I build my own to my satisfaction dirt cheap, but that is another topic. In the case of ethernet cable I can't help thinking that gagabit spec'd cable will get the bits to the Pi with sufficient accuracy. Sure there may be noise from the cable, but it is connected to the Pi which is an open board computer generating all kinds of noise. So I chose to put the ethernet cable, the Pi and all its digital friends in a Faraday cage separate from the DAC and analogue outputs. Might not be perfect, but it was cheaper than buying an exotic cable, and it works for me.
Similar approach with the supercaps. 10+ years experience taught me that sound improvements GregS hears can be duplicated in my system. So when he said that the Uptone Audio supercap supply delivered top performance, it got me thinking. So I checked out their site and realized they use Nesscaps. I happen to have a bunch of these things that also cost me nothing, so why not emulate the Uptone approach and see if there was anything to it. My experience was identical to review after review, so I reported it here in case anyone wanted to try it. I see speculation on this site as to what may cause the improvement. But speculation without scientific proof is not well received here for good reason. I'll leave it to the engineers to figure it out. In the mean time I will enjoy great sound at low cost and be content.
Last edited:
I'm not questioning that you or anybody else experience differences. I do too.
What I'am saying is you can't answer the "Why? What's causing it?" question.
And that leads to endless and very expensive trial and error projects.
Yep. Greg is the fishing master. I'd agree.
But he's simply following a basic recipe that's known for years.
He obviously brought the device and part swapping tactics to different, close to IMO insane, levels.
Of course I do understand the background behind all this.
1. Most people want quick results
2. People lack skills to approach a subject like an engineer would do
3. Even if you'd have the knowledge people lack measurement equipment
4. Most designers stick to textbook designs, they need to accept compromises - and that's where the tweakers job start.
5. What the designers can't measure won't be addressed.
6. What the competition is lacking, doesn't have to be done either.
7. ...
All that is not what I'd call satisfactory.
Anyhow.
Most of our systems sound great. And that's what counts.
Enjoy.
PS:
Advise: Just go wireless with a PI3B+ and turn the PI USB stack off altogether, which saves about 40% on the power side and thousands of interrupts per minute. Go for the easy wins first!"
What I'am saying is you can't answer the "Why? What's causing it?" question.
And that leads to endless and very expensive trial and error projects.
Yep. Greg is the fishing master. I'd agree.
But he's simply following a basic recipe that's known for years.
He obviously brought the device and part swapping tactics to different, close to IMO insane, levels.
Of course I do understand the background behind all this.
1. Most people want quick results
2. People lack skills to approach a subject like an engineer would do
3. Even if you'd have the knowledge people lack measurement equipment
4. Most designers stick to textbook designs, they need to accept compromises - and that's where the tweakers job start.
5. What the designers can't measure won't be addressed.
6. What the competition is lacking, doesn't have to be done either.
7. ...
All that is not what I'd call satisfactory.
Anyhow.
Most of our systems sound great. And that's what counts.
Enjoy.
PS:
Advise: Just go wireless with a PI3B+ and turn the PI USB stack off altogether, which saves about 40% on the power side and thousands of interrupts per minute. Go for the easy wins first!"
Last edited:
I like these suggestions. Not sure how to turn off USB, but I am sure I can find it.PS:
Advise: Just go wireless with a PI3B+ and turn the PI USB stack off altogether, which saves about 40% on the power side and thousands of interrupts per minute. Go for the easy wins first!"
I did go down the path for years of reducing CPU overhead. First with Win98 where I met Greg & others as we pruned the OS to the min. Then with Alix, later with BBB with realtime mods and all kinds of Linux scripts to shut down processes and reduce OS overhead. At some point I felt that the FIFO/isolators were finally good enough that the processor was enough out of the equation to ignore. Sounds like I should revisit that assumption. Not sure these are the easiest endeavors. I can build a linear PS from scratch in an hour. It takes me that long to recall how navigate Linux command line.
You should be ok. Thanks to the on board LEDs, you can see if it powers up and locks onto signal.. ie has enough power. Hook onto the hole 3.3v out plus ground.
Sound quality may be compromised as you get more noise from the new load. The i2s2PCM really responds well to good power. I attached a supercap crossed with a BG HiQ to the same holes and picked up a lot of definition and clarity in the upper registers. That means the onboard ldo has protection as it runs into a dead short for about a minute charging the supercap. I2s2PCM is a good candidate for a battery source + supercap. If you think about it, it is the i2S2PCM that you are actually listening to.
Last edited:
Advise: Just go wireless with a PI3B+ and turn the PI USB stack off altogether, which saves about 40% on the power side and thousands of interrupts per minute. Go for the easy wins first!"
Or use a Samsung MSATA SSD's and the adapter board which fits the Pi stack, the power draw is negligible, or you can power it separately from the USB on the LifePO4.
X850 mSATA SSD USB 3.0 Cable Hard Disk Storage Expansion Board for Raspberry Pi | eBay
Other HDD's and thumb drives into the USB get very warm, whereas this thing barely rises above ambient.
I'm not making any claims for it as such, but it is the best sound I have had so far.
I currently have 160 GB on it which took some 20 minutes to copy and it does a refresh in Volumio in around 60 seconds.
Last edited: