Asynchronous I2S FIFO project, an ultimate weapon to fight the jitter

Battery supply & LDO again

Hi Ian,

I'm going to test my new segmented R2R discrete DAC with your stuff, almost all devices will be batteries powered.
I would use your devices in this way:

- FIFO Buffer, powered with 5VDC linear regulator
- Isolator Adapter, input stage powered with the above 5VDC linear regulator, output stage powered with 3V3 LiFePo4 batteries
- Dual XO board with external oscillators, powered with the same 3V3 LiFePo4 batteries
- I2S to PCM converter, powered with the same 3V3 LiFePo4 batteries

I would exclude all LDO of the devices that will be batteries powered.

can you provide instructions to bypass or better to remove unnecessary LDOs?

Thank you
Andrea
 
Thank you very much Greg,

Have I to understand I must remove the on board ldo on the I2StoPCM board to try a cleaner 3.3V ? Seems Wolwes didn't bypass or didn't remove the ldo... at least my understanding !


I2StoPCM manual seems to tell : you can bypass the ldo if V is below the voltage needed to drop at the ouputt of the reg...but on the I2StoPCM board one needs to remove the ldo to power a cleaner 3.3v through the sided vias marked In/Gnd/3.3V vias ? And in the same time it is advised to feed the main 5V to 6V power input with 3.2 to 3.3 V LiFePo cells.


So what's the best way for the LiPoFe or any 3.3V power supply ? Main power supply plug or 3.3V free vias ? Do we need to remove the on board ldo if powered with 3.2V to 3.3V :
- through the main power supply
- through the 3.3V vias

- need to remove the onboard ldo as the manual says or is it ok with the on board ldo if power is equal or less then 3.3V (ldo doesn't work but allow to passv 3.3Vwithout regulating because not rnough voltage to drop ?


Could we have an input from iancanada about that ? One MUST to remove the onboard LDO or not for feeding 3.3V through the 3.3V input vias ??
Note point 9 in manual.
9. You can also run this I2S to PCM board by a 3.2V LifePO4 battery cell. By shorting VIN and VOUT with a jumper, you can feed the battery power directly from DC input J6.

As you note, I did not bypass on board LDO. I feed 5v from salas shunt to the onboard lp5900 ldo. I then use the 3.3vout/grnd as a convenient place to solder a supercap onto the 3.3v power rail.
 
Disabled Account
Joined 2019
Thank you Wolwes,

The more I read point 8 & point 9, the more I understand point 9 as the 3.3V go through the onboard ldo but due to the low voltage it doesn't regulate.

And from point 8 : better to remove the onboard ldo for a better sound; but J6 stays mandatory while you need this way and despite Vin& Vout is shorted you also need to solder a bypass somewhere on the pads of the removed onboard ldo (the same way if the vias Vin/Vout are not shorted but used with an external 3.3v reg or source . Both case J6 needs a powersupply as well for the power of the others circuitries....

That's what I understand from Greg when he talks about the FifoPi to illustrate that.

8. The on-board low noise LDO is LP5900SD-3.3/NOPB. VIN, GND and 3.3V VOUT pins are reserved for external regulator. Please disassemble the LP5900 in advance if you want to use the external low noise regulator.
9. You can also run this I2S to PCM board by a 3.2V LifePO4 battery cell. By shorting VIN and VOUT with a jumper, you can feed the battery power directly from DC input J6


Would like to know what iancanada loves the best with this I2StoPCM and the way to power it (if he already tried all and not only clean 3.3v through J6).


I also plane to test FCA cap in the 0805 free area after the onboard ldo though I'm not sure if it's for decoupling before or after the ldo reg ?!
 
Last edited:
diyiggy
Best to get Ian to comment. The way I read these points is the ldo can be completely bypassed by shorting 3.3in to 3.3out. It completely takes it out of the power path so you can use the normal power in jumper to supply your 3.3v without having to remove the tiny ldo.

I have had some situations where ultimately removing the IC made a slight improvement even when it was electrically bypassed. I doubt there is much to be gained here however.
 
Member
Joined 2003
Paid Member
@diyiggy,

I think @wlowes summarized it well... bypass the I2StoPCM onboard regulator per point 9 and feed the battery supply into J6. I doubt (but haven't looked at the one I have here OR the rest of the manual) that any of the onboard circuits use the J6 power directly, but are all fed from the regulator output point. AND as Walter said, even leaving the regulator chip in, once you bypass it you will get the bulk of the improvements of the battery power yet you can later try other 5V options just by cutting the jumper.

Greg in Mississippi
 
Disabled Account
Joined 2019
Thank you for that input Greg, I didn't understand there was only this ldo reg on board. Makes more sense now.


@ wolwes,

I tried the LiFePo4 A123 on J6 and an external TPS7A reg board performed better subjectivly at ears in my system. Just had to find the main good cap to power the TPS7A reg!
I suspect the on board main lytic or polymer cap // with the smd cap after J6 is dictating its "sounding signature".

My conclusion I didn't test : a LiFePo4 A123 cell will perform better both without this main lytic cap. (and also from all the other posts everywhere: without chip regulation before the load.


From Greg's input just above the only way to avoid J6 (because the cap after) without desoldering the ldo is to bypass it by the 3.3V in vias. As your supercap has a lower impedance and "infinitly curent tank" I assume the onboard main lytic cap becomes "transparent" (doesn't dictate its "sound" anymore).

So if one would use a LiFePo4 cell instead the Supercap and still avoiding ldo surgeon sequence, is it ok to proceed like this on the I2StoPCM board ? :

- negative side of the A123 cell connected to GND via
- positive side of the A123 cell connected to Vout via
- Vin & Vout via shorted (cause no powersupply connected to J6 input)

Only way to avoid the lytic cap behind J6 and to avoid removing the onboard ldo while the circuitry sees only the very low cell impedance (instead the impedance of the lytic // cap after J6 if the connector had a A123 cell directly connected to it) !

It calls for a manual revision imho if people can confirm it could work this way (A123 cell directly feeding J6 is not a so good idea after hearing tests from several sources and my own experience,I also surmise supercap on inputs vias a la wolwes way performs better than a supercap on J6 connector instead).


free smd cap pads above and bellow the pcb has to be tested with a DVM to understand if a tweak by adding a FCA cap will perform even better (my idea is most of the time acrylic performs better than class I and II ceramics for digital circuitry, at least subjectivly in my system hoping no self biass - but external no biased ears confirmed that in my system.


The simpliest way and the most tricky seems to use the supercap as you did wolwes : no need to use A123 cell and certainly close result if not better. Could be interresant to know if the sound change if you change a lytic cap with your linear reg before J6 despite you use a Super cap at Vin/Vout/Gnd vias ?!


thnak you again all for your inputs, point 9 in relation to point 8 in the I2StoPCM manual was tricky to understand the way it was written (hence all the questions, Andrea Mori, mines, etc)
 
Last edited:
I tried the LiFePo4 A123 on J6 and an external TPS7A reg board performed better subjectivly at ears in my system. Just had to find the main good cap to power the TPS7A reg!
I suspect the on board main lytic or polymer cap // with the smd cap after J6 is dictating its "sounding signature".
When I use them, I too use Ian's TPS7A reg with cap mods described earlier. I did see improved sound but have no way to compare them with battery PS. FYI, I do not have a TPS7a reg on the I2StoPCM. I run 5v salas to I2StoPCM using its on board reg.

In your test of LifePO4 vs TPS7a, what voltage is the TPS7a producing? If it too is 3.3v then its an interesting test essentially supplying 3.3v to the ldo which is in the circuit but not starting.

The next interesting test would be to try LifePO4 vs TPS7a 3.3v with the onboard ldo bypassed using method in #9. It would be good to note what PS you have behind the TPS7A.

From Greg's input just above the only way to avoid J6 (because the cap after) without desoldering the ldo is to bypass it by the 3.3V in vias. As your supercap has a lower impedance and "infinitly curent tank" I assume the onboard main lytic cap becomes "transparent" (doesn't dictate its "sound" anymore).
IMHO there are multiple ways to supply power to the brd.
1. J6 at 3.3v or something above the ldo dropout voltage.
2. Vin and ground. Same as J6. It uses on bd ldo.
3. Vout and ground at 3.3v. This bypasses the ldo as it is at the output of ldo.
Jumping Vin to Vout shorts out the onboard ldo. ONLY 3.3v can be used, and it makes little difference where the voltage is applied.
My use of a supercap resulted in significant improvement in sound. Clearer, more defined. Particularly noticed in the sound of cymbals for eg. People have speculated it is the low impedance at low frequency that results in the improvement around clocks. I do not know why it also worked well on this board. What would be interesting to me is if someone with a LifePO4 made the comparison of battery with ldo bypass to battery with ldo bypassed plus supercap. Logically it should be best because the battery provides full isolation from the mains and the supercap provides whatever it has that the battery misses to make that next level of sound clarity and dynamics.

So if one would use a LiFePo4 cell instead the Supercap and still avoiding ldo surgeon sequence, is it ok to proceed like this on the I2StoPCM board ? :
- negative side of the A123 cell connected to GND via
- positive side of the A123 cell connected to Vout via
- Vin & Vout via shorted (cause no powersupply connected to J6 input)
This would work correctly. Note, IMHO, with Vin shorted to Vout, it makes no difference if 3.3v is applied to any point. It is all the same wire.
…free smd cap pads above and bellow the pcb has to be tested with a DVM to understand if a tweak by adding a FCA cap will perform even better (my idea is most of the time acrylic performs better than class I and II ceramics for digital circuitry, at least subjectivly in my system hoping no self biass - but external no biased ears confirmed that in my system.
No idea if changing on board caps would make a difference. This is beyond my interest in experimenting as I have reached the point where I am satisfied with the sound and not motivated to risk the board for marginal return. 😊
The simpliest way and the most tricky seems to use the supercap as you did wolwes : no need to use A123 cell and certainly close result if not better. Could be interresant to know if the sound change if you change a lytic cap with your linear reg before J6 despite you use a Super cap at Vin/Vout/Gnd vias ?!
I think what you are saying is it would also be interesting to know in the case where a supercap is used on the I2StoPCM board, what if anything is the impact on sound of:
Battery vs linear power supply, and
Linear powersupply regulators. TSP7A stock, with cap mods, or other such as Salas etc etc.
Again, answer not coming from me as I have for the time being reached point of marginal return reducing motivation to make changes for sake of filling out the decision table.
But I will read with interest experience of others.
 
Disabled Account
Joined 2019
to answer your question, the TPS7A I use is not a 3 pins board from iancanada or someone else but a standalone reg board solution with that chip. It's connected to J6.

I remember having both tried 5V and 3.8V with this reg as the Vout can be setuped to see if the onboard ldo of the I2StoPCM will react better with a low voltage nearer from its dropout point (less dissipation heat for the ldo reg chip). Was the same result to my ears.

The best difference was when I experienced with the main cap (3300 uF after diodes rectifier bridge) of the TPS7A reg board I have. Big difference despite this modern chip !!!! I believe one has to experiment at this point in relation to the whole sounding signature of his hifi to see if needed or to adapt as it's changing the global EQ.

I do think (while not really tested in real life) more significant result is to avoid the main cap of the I2StoPCM board to have subjectivly at ears more difference when experimenting with the power supplies tweaks!

I will try a123 cell bypassing the ldo embeded reg with the sided vias tip (now I understand no one needs really to desolder it to bypass it). But knowing me, it will take a long time.

I really think with all those possibilities, the manual tips section should be completed (iancanada if you read me... too much confusion despite the great versality to power your I2Sboard, free caps pads for tweaking purpose should be documented as well: your work for the diyers worths it imho)
 
Last edited:
The latest FIFO Manual is here. (30/12/2018)

https://www.diyaudio.com/forums/gro...us-i2s-pdif-fifo-kit-buy-318.html#post5726311

Page 7 says to bypass the LDO's by using jumpers across the pads of R58/59. which are the back of the board in the middle. I used a short piece of wire, having removed the probably lead free solder on those pads.

Though I see there is also an I2S - PCM board available which I may be getting confused with.
 
Last edited:
Hi.

After more then a week of burning-in my new NDKs (soldered right to the board), things have been smoothed out.

I'm quite happy the way it sounds now. I'd agree to other peoples conclusion that changes are rather subtle. I mean. It's a reclocker. And not the clock-master or DPLL.

I do think that the Allo I2S-master/DPLL-off approach is much preferable on ES9038 DACs over the FifoPI/I2S-clock-slave approach.


Just to mention it. I run the linear-slow filter. Not these smoothy mega-rolloff minimum phase as some other fellow inmates. This might lead to different conclusions.

This exercise has shown once more that all this wild short-term part swapping is really nonsense and just leads to wrong conclusions.


I did one more thing to fight the earlier discussed vibrations. The best clocks
won't help if you potentially increase 30dB phase-noise simply due to poor implementation or surroundings.

I now took a huge chunk of soft kneaded Tereoson Kneading Mass and attached it to the now very spacey clock area.
My calculation. Adding mass, plenty of soft mass, might lower the overall vibration impact. Another idea was that by having the clocks and the whole area covered I'm getting a more stable environment in terms of temperature.
I do think that this exercise was well worth the very low effort. Micro details
and low end details seem to have gained a lot from it.

For know I'm more then happy with the overall result. It's time to get back to my software domain. ;)

Enjoy. And thx again for all the advise.
 
For now I'm more then happy with the overall result. It's time to get back to my software domain. ;)

Enjoy. And thx again for all the advise.
Happy to hear it worked out for you. I agree with your advice BTW for a fine sounding system.

This exercise has shown once more that all this wild short-term part swapping is really nonsense and just leads to wrong conclusions.
On this point, I'd like to offer another opinion aimed a folks that might be new to this game and to give them the range of options.

My comments apply to the FIFOPi as a source for your downstream DAC of choice. DAC's need their own sorting and people have a range of preferences.


FIFOPi is a well sorted I2S source for DIYers and offers 3 ranges of price/performance.

1. HiFi - use it out of the box with supplied test clocks and any smp supply.
It will sound like a good CD player. Surprisingly good, but digital edge that almost killed digital CDs.

2. Audiophile - add good power, clocks & definitely a little vibration control and it will rival the really good stuff at dirt cheap price.

3. Reference quality - This is where you get into some level of parts swapping, optimize power to the max, and take a lot of care to eliminate vibration and EMI/RF noise and your system can take on the best out there.

#2 is a great place to be.

#3 does not have to be much more, but it does take some knowledge and skill at execution. If you look at what a Benchmark or and Esoteric or a top of the line Lampizator does, it is not a big deal to pull off most of what counts in the DIY realm at realatively modest cost. And I can testify that Ian's FIFOPi, I2StoPCM and regs can be essential building blocks in such a system. I can't speak to the GB DAC's and leave that to others. ;)
 
If you look at what a Benchmark or and Esoteric or a top of the line Lampizator does, it is not a big deal to pull off most of what counts in the DIY realm at realatively modest cost.

Have you ever listened to a Benchmark DAC-3? Have you directly compared what you consider to be a #3 Ian dac with DAC-3?

I strongly suspect you are making stuff up that you imagine might be true, but you really don't know for a fact.

In reality, it is daxn hard to make a dac in the class of DAC-3. For one thing it has an external interpolation filter in FPGA. The built-in PCM filters in stock Sabre dac chips don't compare at all, none of them sound right, but that's all the DSP power the poor ESS chip has. How are you doing the equivalent of the FPGA filter with Ian dac?
 
Have you ever listened to a Benchmark DAC-3? Have you directly compared what you consider to be a #3 Ian dac with DAC-3?

I strongly suspect you are making stuff up that you imagine might be true, but you really don't know for a fact.
I have been clear that I have not heard a DAC-3 and welcome anyone in Toronto to bring one over to provide that experience. Or other well regarded commercial products. I'll enjoy it and likely learn a few things.

I do visit the annual trade show in town to audition stuff I would never want to pay for, and host/participate in local DIY events. So opinions are not based solely on imagination. I also know for a fact you have never heard what I consider to be #3, so we'll just have to leave it there for now. Peace.
 
Disabled Account
Joined 2019
A fair benchmark would assume to not use the embeded pre and digital volume of the ESS Benchmark 3 DAC. I assume wolwes has a classic analogic pre so hard to sort out apfels with potatoes !


But if the benchmark 3 also has the option to bypass pre and embeded digital volume attenuation for a full 2V voltage outputt, an even fairer benchmark should be made with an external transparent volume atenuator as the Muse volume chip Pass uses in its High End preamps !


Without such experiments made in blind tests environment I'm not sure both of you can claim which of the Benchmark 3 or wolwes's nos old cmos chip is better.


Where the B3 can a winner is the cost/SQ ratio. Cause DIY Dacs needs long and costly tweaking (but are tailor made in relation to the whole diyer's hifi, a plus imho! Often ready made dacs as good they can be, make some money saving on the power supply section with a lot of trade offs diyers haven't... I learn nothing to both of you here.


Without such fair benchmark, claims are like preaching in the desert imho...:eek:
 
Last edited:
Have you ever listened to a Benchmark DAC-3? Have you directly compared what you consider to be a #3 Ian dac with DAC-3?

I strongly suspect you are making stuff up that you imagine might be true, but you really don't know for a fact.

In reality, it is daxn hard to make a dac in the class of DAC-3. For one thing it has an external interpolation filter in FPGA. The built-in PCM filters in stock Sabre dac chips don't compare at all, none of them sound right, but that's all the DSP power the poor ESS chip has. How are you doing the equivalent of the FPGA filter with Ian dac?

Hi Markw4,

Which Benchmark DAC-3 do you have? Can I have a link?

Thanks,
Ian