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

[andrea_mori]

did I say it couldnt be done? you should read more carefully

I read very carefully instead, you said "EVERYONE has trouble ...", that's not true.
Those who own the proper instrument had no problem measuring at this level.
Take a look at Agilent manual and then take a look at post #2344. As you can see is very easy measuring -95dBc@1Hz and -135dBc@10Hz, WITH THE PROPER EQUIPMENT
And now convert that measurements in the time domain.

 

[qusp]

yes, everyone DOES have trouble, measuring 1ps phase noise is troublesome, no matter what equipment you have.

you are utterly discounting the time to set up, the time to build and set up the power supply, the time to build the buffer to avoid loading the clock with the probes, shielding the whole thing

easy....

you need to study how to read context

 

[iancanada]

Same feeling here Ian, my own FIFO project two years ago set me back for 100$ worth of parts from Digikey. Parts still in my electronic cabinet nicely going toward obsolescence. On my own this FIFO project (all boards) could have cost me easily a 2K-3K$ in R&D. Equivalent performing devices on the market cost from 400$ to 1500$. Ok they are final products but you don't get the flexibility and tweaking possibility given by Ian FIFO.

I think their is also a confusion about DIY projects and open source projects. Back in the 80s Heatkit was a commercial endenvor and a DIY provider...

Jitter measurements are a important guide to make development decision but they required costly and very well calibrated equipment operated by skillfully technicians. And even then the results could be challenged.

Like some one said not so long ago :

If you believe you can make a difference, not just in politics, in public ... prepared to accept that you are not going to get 100 percent approval

Thanks Fridrik, 100% agree with you.

Ian

 

[iancanada]

Well first of all there is no such thing as a 3.2V LiPFE04 battery, batteries don't have regulation their voltage changes as they are used, from 3.6V down to 3.2V.

Second what do you think, do you want to mess around with batteries or use a LDO regulator that you never have to worry about charging or installing backwards and frying your pcb. Batteries are for OCD types like qusp

Basically if you have to ask go with the regulator.

Thanks regal

Ian
From the Si570 Clock Board guide:

>This LDO board has been determined to be one of the best ways so far to power the Si570 Clock Board.<

Do you implying that powering up the clock with the LDO board give subjectively (in your setup) a better sound then directly powering the clock from Lipefo4 3.2V Bat or simply that the LDO powering is a best solution overall, if we take convenience in account?

Great manual by the way, thanks for all the info in the tips sections

Hi Fridrik

regal gives very good point.

Battery has some disadvantages:
1, Voltage is not stable, decreasing along the discharge, so there are always some sweet point, another secret.
2, Internal resistor is not keeping constant also
3. Sensitive to the load
4, Over-charge and over-discharge can damage the battery easily.

However the battery is the easiest way letting your touch the low noise power supply.

Ian

 

[iancanada]

yeah its as easy as buying the equipment....

Demian, who has much experience with such things, has already commented on the difficulty of such measurements, if you think its easy just because you have the right equipment, you are fooling yourself. did I say it couldnt be done? you should read more carefully

Yes, both of the two things are very diffcult:
1, Measure the phase noise of a low jitter clock
2, Measure the noise of a low noise power supply.

Demian also suggested measuring them from the output of the audio signal, that is kind of finial solution for audiophiles, you just need a AP

Ian

 

[1audio]

Yes, both of the two things are very diffcult:
1, Measure the phase noise of a low jitter clock
2, Measure the noise of a low noise power supply.

Demian also suggested measuring them from the output of the audio signal, that is kind of finial solution for audiophiles, you just need a AP

Ian

Both are difficult, especially when approaching the state of the art.

However I did get close to a very useful tool using an FM tuner to look at phase noise. Start with a very low noise tuner. I used a Yamaha TX-950 which is one of the lowest noise tuners ever made. Mod it by taking a direct output from the detector (no low pass filter or eq.) Output to FFT. Input from clock will actually be a harmonic, 2X 4X etc and it works. The noise is multiplied by the harmonic multiple. The input is AC coupled so it won't screw things up. Adding an isolation transformer on the input can help remove ground loops and noise coupling into the DUT. It can be a useful troubleshooting tool. Calibrating it in absolute terms proved to be too much for me. Too many steps from SSB carrier noise to FM deviation in Hz to work through so its a relative tool.

If you can measure the power supply noise with a normal instrument its not that low. Its possible to get close to 1 nV/rtHz, which is below what you can measure with an AP directly. I built a transformer coupled preamp to get the equivalent noise down, I think it was about 5 Ohms. Huge PITA as well.

You don't need an AP to get the jitter at the output. You need a decent sound card (Juli@, EMU 0404, EMU 0202 for example) and some suitable FFT software like ARTA or this WaveSpectra and a J-Test file like the two attached.

 

[tubo]

Hi guys,

I finally got my cchd957 for dual xo working and it sounds wonderful. It improved my shigaclone transport to another level that i didnt knew was possible.

Since groupbuy 4 is on going, im planning on getting the si570 board but need clarification first.
My dac is a nos ad1865 and im mostly playing 16/44.1 through my shigaclone or sq touch.
My question is, do i need the si570 clock to further improve sound quality or just stick with cchd957 since i mostly play 44.1?

Please any suggestions

Thanks guys

 

[qusp]

no, you should just stick with the cchd957 if you are playing mainly 44.1 with a non ESS dac

 

[tubo]

Thanks qusp

 

[WhiteBull]

Ian and others hello,

some questions from me please, regarding TPS7A4700 low noise regulator board V2.5:

- from what I understand, you recommend the reg over a LiFePo4 battery for the Si570 clock board based on sonic attributes, correct? You've mentioned some disadvantages before, are there any advantages?
- I've recently connected two LiFePo4s instead of the AVCC tridents on Buffalo II and it was a great improvement. Would the regs be suited at this place as well then?
- how about a possible benefit of replacing the other two tridents: DVCC and Core 1.2V (in case it goes close enough to 1.2V)?

- You suggest battery-based 4V-6V DC power for the regs. Is 6V really the upper maximum input for the regs, because then 2S LiFePos and 6V gels fall out, only some NiMh combination would work, or maybe 4V gel?

Thanks,
Abel

 

[iancanada]

Thanks regal



Hi Fridrik

regal gives very good point.

Battery has some disadvantages:
1, Voltage is not stable, decreasing along the discharge, so there are always some sweet point, another secret.
2, Internal resistor is not keeping constant also
3. Sensitive to the load
4, Over-charge and over-discharge can damage the battery easily.

However the battery is the easiest way letting your touch the low noise power supply.

Ian

Ian and others hello,

some questions from me please, regarding TPS7A4700 low noise regulator board V2.5:

- from what I understand, you recommend the reg over a LiFePo4 battery for the Si570 clock board based on sonic attributes, correct? You've mentioned some disadvantages before, are there any advantages?
- I've recently connected two LiFePo4s instead of the AVCC tridents on Buffalo II and it was a great improvement. Would the regs be suited at this place as well then?
- how about a possible benefit of replacing the other two tridents: DVCC and Core 1.2V (in case it goes close enough to 1.2V)?

- You suggest battery-based 4V-6V DC power for the regs. Is 6V really the upper maximum input for the regs, because then 2S LiFePos and 6V gels fall out, only some NiMh combination would work, or maybe 4V gel?

Thanks,
Abel

Please see the battery disadvantages I listed above.

TPS7A4700 is very good based on my own experiment. I suggest you try both and select the one you like the most.

I use lifepo4 direct before, now I use 6V battery over TPS7A4700. They sound different in some way. But hard to tell which one is better.

I don't think big problem if you go a little bit over 6V. the limitation is the on board LDO. But have to be careful the surge and spike voltage.

Ian

 

[WhiteBull]

Thank you Ian,

but since I'm trying to figure my setup to be able to proceed with the groupbuy, I could use some more precise answers please!

- I've recently connected two LiFePo4s instead of the AVCC tridents on Buffalo II and it was a great improvement. Would the regs be suited at this place as well then?

- how about a possible benefit of replacing the other two tridents: DVCC and Core 1.2V (in case it goes close enough to 1.2V)? So are your regs suited for this positions and how low can we go with them? From the pdf it looks like it's 1.4V. Is that right, and if so, can the Buffalo take it?

- the input for the regs has to be always between 4-6V, regardless of the selected output voltage, or do we have a specified minimum dropping voltage related to the selected output voltage?

- Can we use 2 LiFePos or a 6V gel to power the regs, both around 6.8V charged? Or would a 4V gel fit better? Which 6V battery are you using?

If this works, I would exchange the tridents in order to get rid of the placid and use a battery instead.

Regards,
Abel

 

[qusp]

you will not get a precise answer to such a question, Ian deliberately left it up to you, the choice is yours. these areas of playing with clock and power supply are the only DIY left to do with this project, why do you want someone to take that decision away from you?

there are practical concerns and subjective ones, both of which he answered. the tps will not do 1.2v and no you cannot use 1.4v for the 1.2v ESS supplies

he told you hes using regulation after the batteries, he is using lifepo4. no the tps is not suitable for AVCC, its not low noise enough, AVCC voltage and noise directly effects Dac analogue output performance as well as voltage swing and offset. every single battery chemistry states the voltage at nominal voltage, so a 6v gel battery will be higher at full charge as well and lower later.

 

[WhiteBull]

Thanks, I'm trying to get some technical information to make a setup and buying decision! Nothing subjective! My questions are as precise as I can put them!

Honestly I dislike your beating, I can live without it!

Since I've picked between the lines that the regs are technically not suited to replace the tridents, I'm still not clear about the input voltage for them. If Ian is using LiFePos as you say (Ian didn't by the way), then the regs are able to handle around 7V. Ian mentioned using a 6V battery, and that's what I would prefer too, but then charged when get close to 7V again.

So all I wanted to know with my last two questions is how we can use a 6V battery without hurting the reg.

Regards,
Abel

 

[bkdog]

Thank you Ian,

but since I'm trying to figure my setup to be able to proceed with the groupbuy, I could use some more precise answers please!

I do not think that more precise answers are possible, as Ian said, this is the area of personal preference, you'l have to try it and see for yourself.

- I've recently connected two LiFePo4s instead of the AVCC tridents on Buffalo II and it was a great improvement. Would the regs be suited at this place as well then?

Hey thanks for the tip, I'll try batteries in place of the trident regs one day later this year... I see that in the time I've been writing this that qusp has already replied, and he's right, so I'll leave this here.

- how about a possible benefit of replacing the other two tridents: DVCC and Core 1.2V (in case it goes close enough to 1.2V)? So are your regs suited for this positions and how low can we go with them? From the pdf it looks like it's 1.4V. Is that right, and if so, can the Buffalo take it?

- the input for the regs has to be always between 4-6V, regardless of the selected output voltage, or do we have a specified minimum dropping voltage related to the selected output voltage?

- Can we use 2 LiFePos or a 6V gel to power the regs, both around 6.8V charged? Or would a 4V gel fit better? Which 6V battery are you using?

If this works, I would exchange the tridents in order to get rid of the placid and use a battery instead.

Regards,
Abel

 

[RayCtech]

Thanks, I'm trying to get some technical information to make a setup and buying decision! Nothing subjective! My questions are as precise as I can put them!

Honestly I dislike your beating, I can live without it!

Since I've picked between the lines that the regs are technically not suited to replace the tridents, I'm still not clear about the input voltage for them. If Ian is using LiFePos as you say (Ian didn't by the way), then the regs are able to handle around 7V. Ian mentioned using a 6V battery, and that's what I would prefer too, but then charged when get close to 7V again.

So all I wanted to know with my last two questions is how we can use a 6V battery without hurting the reg.

Regards,
Abel

Hi Abel,

If it is the TPS7A47 LDO voltage regulator then 36 volt input and output +1.4 to +20.5 volt are the specs..

For a 3.3 volt output you can feed it with +4 (+3.6 <-> +3.7) to +36 volt.

Battery are not going to give you anything but trouble..

A capacitor multiplier in front of the regulator will give you ca. 40dB extra noise reduction and better performance than any battery..

 

[bkdog]

Thanks, I'm trying to get some technical information to make a setup and buying decision! Nothing subjective! My questions are as precise as I can put them!

Honestly I dislike your beating, I can live without it!

Since I've picked between the lines that the regs are technically not suited to replace the tridents, I'm still not clear about the input voltage for them. If Ian is using LiFePos as you say (Ian didn't by the way), then the regs are able to handle around 7V. Ian mentioned using a 6V battery, and that's what I would prefer too, but then charged when get close to 7V again.

So all I wanted to know with my last two questions is how we can use a 6V battery without hurting the reg.

Regards,
Abel

You probably should ask these questions over at the TP thread, or on their forum. I imagine that Ian cannot really offer that sort of advise on another manufacturer's product...

 

[qusp]

whitebull, the reason I get frustrated, is because questions like input voltage are covered quite well repeatedly in this thread and hochopeper has taken a lot of time to collate the information in the wiki, yet we still get elementary questions posted in this thread. the dropout voltage limits are also covered here and clearly in the datasheet for the regulators.

Ian and others in this thread should not have to answer questions that could easily be answered if users took a few minutes to look some of this stuff up for themselves. the differences in applications for batteries are as varied as the different types of batteries and are also effected by factors such as your charge program. those questions are for you to answer. then youve also added in some questions about the application of the ESS dac, which has nothing to do with Ian.

 

[RayCtech]

Since I've picked between the lines that the regs are technically not suited to replace the tridents

You probably should ask these questions over at the TP thread, or on their forum. I imagine that Ian cannot really offer that sort of advise on another manufacturer's product...

TPS7A47 with properly decoupling and a capacitor multiplier in front should beat tridents - and why ask questions where he most certain would get biased answers

 

[qusp]

Ray, its misleading to say you can realistically input +36v for a +1.4v output and you have not even covered load current. this application of the TPS reg does not have much copper for heatsink, its designed for a low dropout voltage at 100-150ma in this application. the TPS gets best performance with ~1V dropout to work with.

a cap multiplier will negate or complicate the use of the load sense function, load regulation and transient response suffers, so not sure of the improvement

the question was directly feeding the board with batteries, or regulate beforehand, therefore the dropout of the onboard regs was the dropout under discussion, not the dropout for the TPS7A4700