advertisement


Got Ticks and Fleas?

Yes; you should expect that!

One should also be able to explain complex ideas in language simple enough that anyone can get it - so with the crayons out, here is a little illustration of the only quality that matters in the clock output, down to 10s-of-pS deviation as I wrote above: this is what 'jitter' is in a clock signal:

jitter.gif


Any o'scope internal timebase performance is c. at least 10^4 x worse than the clock in this regard, so it cannot be seen on a scope directly. The crests of the clock signal, high and low, are way above/ below the slicing level where the dac converts the signal (which are usu c 0.6v from each of high and low voltages. Which is what the shape at the extremes, really doesn't matter.

- Now go read the earlier post above, again, with this illustration in mind, until you get it.
 
martin clark said:
Yes; you should expect that!

One should also be able to explain complex ideas in language simple enough that anyone can get it - so with the crayons out, here is a little illustration of the only quality that matters in the clock output, down to 10s-of-pS deviation as I wrote above: this is what 'jitter' is in a clock signal:
jitter.gif

Any o'scope internal timebase performance is c. at least 10^4 x worse than the clock in this regard, so it cannot be seen on a scope directly. The crests of the clock signal, high and low, are way above/ below the slicing level where the dac converts the signal (which are usu c 0.6v from each of high and low voltages. Which is what the shape at the extremes, really doesn't matter.
- Now go read the earlier post above, again, with this illustration in mind, until you get it.
Click to expand...
Timebase performance can be somewhat sidestepped using FFT and freq. domain analysis ... i.e, SA. One can squeeze out pS jitter data ... and audio engineers have been do so for decades now.
I can even do it with a cheap TinySA. The speed isn't as great as an old, analog HP or modern APx555. But one can watch the spikes move up and down in real time. If anyone wants, I can shoot a video and let you see with your own eyes.
Meanwhile ...
The tiny osc, using 4.00v TekPower.

Philips 11.289 xtal:
image.jpg



Generic 11.289 xtal:
image.jpg
 
13mh13 said:
Timebase performance can be somewhat sidestepped using FFT and freq. domain analysis ... i.e, SA. One can squeeze out pS jitter data ... and audio engineers have been do so for decades now.
Click to expand...
NO! - a calculated 'result', does.not. mean it is a real result. It is is far far too influenced by the quality of the 'scope timebase, which is NOT optimised for Low Phase Noise. The FFT etc calculation can be as nice-as-you-like, at 64bits totally beyond-contention -

Yet if the reference for that calc is crap, the output uncertainty is far worse. Peril of a digital age - people believing a delivered number, without understanding, let alone questioning what lies behind.

An analogy : I can buy good-priced DVMs that will return single-digit uV measurements. I have a few, in fact.
Are they accurate, or precise (two different things, BTW) - no! They can be made better - but that's a different, if allied, tale.


Thanks for the link to the datsheets etc - will red tomorrow.
 
I tested my osc w/ and w/o Flea reg. using my SA.
You have to WATCH the phenomenon in real time, so I may follow up with a video.
I saw no difference in noise floor of time domain “forest “ graph shown earlier.
 
Cool - please do that!

ETA: does your scope have a 50-ohm input? I don't mean just 50-ohm BNC, but a proper, '50ohm' designated input..?
 
martin clark said:
Cool - please do that!

ETA: does your scope have a 50-ohm input? I don't mean just 50-ohm BNC, but a proper, '50ohm' designated input..?
Click to expand...
My o'scope or SA?
I have Tek 465 analog and Siglent 1202 digital o's. Not sure what they have built in? The 465 is 1 M-ohm. But I see no evidence of reflection or ringing when probing digital signals.
The SDS-1202 automates a lot of things, so I assume it automagically its tuning for what it detects on the probe lines. And, again, I see no evidence of reflection or ringing when probing digital signal w/SDS-1202.
I have the std 10x probes. The 1202 digital 'o came with extra stuff, but not sure about 50 adapters ? I bought and built some adapters as well, mostly for my TinySA and NanoVNA.
 
13mh13 said:
The SDS-1202 automates a lot of things, so I assume it automagically its tuning for what it detects on the probe lines. .
Click to expand...

My emphasis, but - Oh ffs - you must KNOW exactly-what and how you measuring, and the limits of those processes (including the limits of one's own understanding - yeh, the humility bit we all struggle with -) to even begin, to assert outcomes ..!

Have you actually read, and absorbed the appendices in Linear Tech AN-47A I recommended above yet?

Let's take this conversation up - in a week or two. When you might have.
 
martin clark said:
Have you actually read, and absorbed the appendices in Linear Tech AN-47A I recommended above yet?

Let's take this conversation up - in a week or two. When you might have.
Click to expand...
132 pages?!!! What a deflection!
But typical of Magical Mystery Man Martin!
Frankly, dude, I can't figga ya out ....
But I did go this far ... which is a great summary ...
And still .... I have no clues as to what you're talkin' about. Why don't you take one of my graciously-contributed photos ... and point to the mistakes? MSDraw works well!
 
Why can't you ever, ever, be bothered: to try to actually learn new things, for yourself, from reading detailed resources, long-suggested? If you want to get anywhere you will need to put some effort-in, to apprehend and understand - not cite endless superficial video gloss you still don't comprehend.

Want to talk about ..'deflection?' Mostly, what you have long-posted is about a decade & a half of whining 'But why wont you give me the good stuff now? It's your fault its gone Wrong!' across many forums;

- when if you put half as much effort into ... - oh; nevermind.

I recuse myself.
See ya in a fortnight.
 
Martin … your last post was one of your finest hours. Put the Dewars on my tab tonight … looks like you need quite a bit.
About that 50 ohm biz. Well, the Siglent has that as an internal / selectable termination (1M ohm is std). Don’t see much change. What am I supposed to see?
 
13mh13 said:
132 pages?!!! What a deflection!
But typical of Magical Mystery Man Martin!
Frankly, dude, I can't figga ya out ....
But I did go this far ... which is a great summary ...
And still .... I have no clues as to what you're talkin' about. Why don't you take one of my graciously-contributed photos ... and point to the mistakes? MSDraw works well!
Click to expand...

Oh my god what a Rodney, they say that some have lost a lot of social skills because of lockdown, sounds like someone has been on a desert island for a decade. You clearly have no respect for Martin so why piggyback off his designs and slag him off? Just leave it out and get banned somewhere else.

Rgds
Stuart
 
13mh13 said:
My Flea became a tick because:
-- a tick is an arachnid (8-legs)
-- the orig. Veroboard Flea I hacked in 2007 now has two 4-pin oscillators (2x4=8), each can be quickly switched in between.
-- The Dream Project is based on an original, strange NOS Philips/Magnavox CDP that was made for 8-pin tda1543

The photos below are a continuation of another exterior thread that was de-serviced by "those in charge".
Some of this the following content is also based on postings here on PFM. E.g. see:
https://pinkfishmedia.net/forum/threads/flea-to-saa7210-remove-22k-resistor-for-x_out-x_in.280206/

Some of what you're looking at in the photos below:
Orig. Magnavox CDB2000 (Dream Project, in situ) --> CS8421 + TDA1545a --> Rudolf B/Jocko Homo IV (Dream Project, in situ).

The LARGE FONT stuff is per a "plug-n-play" feature I had wanted for a very long time. It allows one to keep as much of the upstream and downstream (in situ) stuff in place while still allowing me to rapidly switch out, say, dac chips. The plug-n-play is not a hot/live feature (to protect electronics), but does allow one to swap out DACs (or DF+DAC, ASRC+DAC, etc) in under a minute. This should please some objectivists.
In the exterior thread, you'll note I had experimented with several other plug+play modules, all built on Vero.
The CS8421 + TDA1545 was an important module experiment as it allows one convert from I2S to EIAJ with a single SSOP device. All prev methods that I'm aware of (including ones that I have prototyped) have been messy glue logic using multiple 74hc devices. Various CS841x receiver chips -- like classic CS8414 -- also allow for I2S to EIAJ conversion. However, the CS841x receiver chips do not allow for I2S input. Only SPDIF.
With the CS8421 + TDA1545, using quite a bit of resistor-based pin-hardware logic control/hacking of the CS8421, I've got both I2S-to-EIAJ conversion and the goodness of ASRC -- all on that small SSOP device. Cool. Alas, for NOS purists, the CS8421's minimal output is 2x. Ahh ... but sonics ....
As reported in ...
https://pinkfishmedia.net/forum/threads/flea-to-saa7210-remove-22k-resistor-for-x_out-x_in.280206/
... the sonics are very, very good.

And it was resourceful use of the orig Flea that sat retired since 2009.
And Martin may recognize one of his LM317 tracking -pre-regs (it's a dual +/- 15v version bult in 2007; currently using just +15v for the Rudolf IV)








BONUS:
Below is a TDA1387 module, rapidly swapped out, and now playing as a NOSer (under 30sec).

Click to expand...

And I thought 'Children Of Time' was a work of fiction, however consummate work from a Salticidae. How are the Mantis Shrimps getting on?
 
Found an opamp-based regulator stage in a commercial audio product. A 1988 Kenwood DP-1510 CD player. I just purchased this unit second hand.
The opamps used for power is a jellybean JRC 4558D.
Martin, what do you think?
Room for improvement? Maybe a better opamp?

In the schema image below (IC12) is the opamp that is part of the Kenwood's DP-1510's "AVR" power section, supplying +/- 5 volts to: digital filter; two PCM56 dacs and two output (LPF) NJM4565D opamps; and xport servo/coil IC that are also 4558D.

So they, in the total AVR ckt, are supplying quite a lot of power/current. This is what I was hoping for with the Flea, and ran into all sorts of problems described in this thread.

EDIT: There is quite lot to that AVR ckt. E.g., Q11. It is a big, chunky D1944 (TRANS NPN 60V 3A TO220F-3) for a +9v leg. The mirror (-9v leg) is a mere 2SA954-A (TRANS PNP 80V 0.3A TO92). In a way, then, the 4558D is sort of "preamp" for the power delivery (AVR ckt).
Interesting options that Kenwood chose for topology in this CDP: sophisticated power/regulation. But the analog output opamp (after DAC) is as low cost as it gets! Maybe Kenwood used a basic design (topology; PCB) for several models. So higher models may have better opamps.


Above: The schema screams: STUDY ME!!!!!!!!
 
Last edited:
Measuring clocks is fairly subtle; the behaviour is much more complex than can be summed up in the headline ps jitter number. For a typical oscillator, the spectrum falls as the offset from the central, nominal frequency increases, until it levels out in a flat white noise far from carrier background. Bad systems will have some spikes on the flat background. In many applications, the close to carrier noise is the big worry, as it so much higher in level; one big contributor to it in many circuits is the 1/f noise in the active device intermodulating the carrier.

When you make a measurement, you end up with a spectrum that is the convolution of the measurement clock with the signal under test. This makes testing high performance oscillators VERY difficult; you need a measurement device that is a factor of ten better than the thing being tested. At work, I worked on a project building a production test rig for some electronics that were assembled at a third party . We ended up having to hire in some fancy Rhode & Schwartz spectrum analysers to repeatably do the measurements. Even then we could see close in shoulders either side of the carrier on most samples at offsets of a few hundred hertz, before it fell back to the white noise floor; from memory these were maybe 90dB down on the carrier, and we thought they were probably real, but never had confidence.
 
Yes, Martin. Your predictable obligatory response.
PigletsDad: but this thread isn’t about clocks and oscillators.
 
You must log in or register to reply here.


advertisement


Back
Top