John Curl's Blowtorch preamplifier part II

Status
Not open for further replies.
I am interested in a diy way of estimating jitter “quantity” in audio equipment (calibrating in standard terms is not of my concern)
DIY way.. Vertical axis is relative level of unwanted "artifacts" in Dunn (J-test) spectra..
 

Attachments

  • jitter.png
    jitter.png
    11.5 KB · Views: 206
The Meitner Interview is 20 years old! It's nice to know DIY has caught up.

OK I didnt look at the date, I presumed something being linked here for conversation was recent. OK well 20years ago it was obviously quite forward thinking, but no way was it equal to what is being done today; the technology just wasnt there.

I notice he never did what he said was just around the corner 20years ago...

I have however seen and used one of his more recent olive workstations (if hes still involved with them). it was a huge, expensive, clunky, closed system that was bettered by the laptop based system next to it.

I have been told the recent meitner dac is quite good though, but its just a plain vanilla dac.
 
Last edited:
the presence and height/spread of harmonics/sidebands next to the base of the fundamental is the indication of jitter. 1audio/Demian has access to higher end kit and more experience measuring jitter than just about anyone on this forum...

LOL! You guys don't even recognize when someone's making fun of you.

That spectragram that 1audio showed is meaningless! What about reducing the frequency span for a starters? i.e. zooming in on the center frequency..

For comparision - I just fed 9978 Hz sine to the analog input of an oldish "plain vanilla" Digidesign sound card residing in this same computer I'm typing this text on. The card itself is nothing fancy - 74HCU04 oscillator with off the shelf 48MHz crystal, residing in "harsh" enviroment and fed from the switched PSU.
Details on the settings are seen in the picture. I made the span as small as possible and activated 64 Bit analysis (see how "deep down" the analyzer can "look", if fed with the appropriate source + sufficiently low noise floor).

Do we see any sidebands? None! Nada! Zilch! It didn't matter in my case whether I used Hanning. Hamming, Triangular or Rectangular window.

What I am trying to say with all this - one has to have very low noise floor and/or very jittery clock feeding the A to D converter chip to see any sidebands by testing it with the signal fed to its input. If you see any that means the A to D converter is no good.

I'm affraid, it takes the specialized equipment ($$) to do any serious jitter measurements. I have a 43MS/s 25ps resolution analyzer here and even that is sometimes not "good enough" (right picture, for illustration only).

Best,
 

Attachments

  • no_sidebands.jpg
    no_sidebands.jpg
    258 KB · Views: 206
  • 11MHz_hist.jpg
    11MHz_hist.jpg
    105.2 KB · Views: 192
Last edited:
Member
Joined 2004
Paid Member
Demian,

I got another denizen of this loony bin to solve my jitter issue. But where did you learn the trick of using an FM detector? :) (I need to have real numbers to present rather than I tweak it a bit.)

ES

Someone here mentioned it and I took the bait. It turns out to be more involved than it might seem. I posted my incomplete notes here on using a tuner: http://www.diyaudio.com/forums/blogs/1audio/983-fm-tuner-jitter-analysis.html

What kind of numbers did you get for jitter?
 
LOL! You guys don't even recognize when someone's making fun of you.

That spectragram that 1audio showed is meaningless! snip...

umm, maybe thats because your post was pretty obtuse and you have gone on to mention a system you have that is orders of magnitude worse than the system I mentioned in the post previous to yours... so good on you!! you scored a point...less

the system I linked was 8Gb/s 2ps.. its not cheap, but can be had for less than the $10K mentioned.

I gather Demian could probably zoom in... I gather its not the system he has at work... i'll let him combat you though, since that appears to be what you want...:rolleyes:


@ simon:

it seems the latest meitner top of the line EMM labs dac uses basically exactly the same technology I was talking about thats available here on the forum to eliminate jitter, except the EMM version (called MFAST) does up to 24/192 vs 32/384 for the $250 i2s fifo buffer/DSP available here. no digital crossover systems to be seen anywhere in the pages

clearly the man knows what hes doing and has advanced the technology in the marketplace, but any company that has to come up with a marketing acronym/trademark/buzzword for every... single... function... turns me off a bit and he has simply continued to build on what I was talking about.

• MFAST™ asynchronous technology for instant signal acquisition, jitter-free
performance
• MDAT™ signal processing technology:
• Provides 2x DSD upsampling for PCM playback
• Preserves phase, frequency and dynamic integrity of the waveform
• MDAC™ discrete dual differential D-to-A conversion circuit
• MCLK™ high-purity master clock that establishes new benchmarks in jitter
performance
he even has to make a real abbreviation into a trademark acronym

at least on spec, the dac i've built, mostly from assorted PCBs and modules bought here and some tweaks of my own, equals or betters every spec on paper and sounds great to me. bear in mind its taken 2 years and cost over $4K for a quad mono ES9018 dac that actually IS built to function as the front end for a 2 way digital crossover that can be controlled via ipad

edit: actually not quad mono only 2 x ESS dacs, each DAC is used for a 'mono' 2 way balanced output for the L or R channel, instead of a stereo output, so I suppose its kinda dual stereo balanced, but the dac power supply, analogue stages and the amps are quad mono balanced.

that someone like me can put together such a system, with a lot of learning along the way, speaks of where the DIY scene is now.

maybe all this annoying marketing fluff speaks more of the consumers, but at any rate its pretty comedic.
 
Last edited:
Member
Joined 2004
Paid Member
LOL! You guys don't even recognize when someone's making fun of you.

That spectragram that 1audio showed is meaningless! What about reducing the frequency span for a starters? i.e. zooming in on the center frequency..

For comparision - I just fed 9978 Hz sine to the analog input of an oldish "plain vanilla" Digidesign sound card residing in this same computer I'm typing this text on. The card itself is nothing fancy - 74HCU04 oscillator with off the shelf 48MHz crystal, residing in "harsh" enviroment and fed from the switched PSU.
Details on the settings are seen in the picture. I made the span as small as possible and activated 64 Bit analysis (see how "deep down" the analyzer can "look", if fed with the appropriate source + sufficiently low noise floor).

Do we see any sidebands? None! Nada! Zilch! It didn't matter in my case whether I used Hanning. Hamming, Triangular or Rectangular window.

What I am trying to say with all this - one has to have very low noise floor and/or very jittery clock feeding the A to D converter chip to see any sidebands by testing it with the signal fed to its input. If you see any that means the A to D converter is no good.

I'm affraid, it takes the specialized equipment ($$) to do any serious jitter measurements. I have a 43MS/s 25ps resolution analyzer here and even that is sometimes not "good enough".

Best,

First, the point was to see if the ADC was free enough of artifacts to see problems with the signal from the 48 KHz word clock. It is. And the system can show problems (see the Apogee plot I posted).

Your not looking in the right place using the analyzer you mentioned. You are looking at cycle to cycle jitter and won't see the audio frequency variations. Your measurement of a 9978 Hz signal did not indicate what the source is. If its a loopback no jitter will be visible. I checked several very high performance sources that I did not post and things like power supply modulations show just fine. Its the same process used by Stereophile which has several benefits. First its the audio output that matters. Second Stereophile has published a number of measurements to compare to.

Being able to look at the spectrum of a word clock directly is useful if only to see if there are real problems. Also, while having a clean word clock is important it doesn't tell you anything about the pll used in the DAC/ADC and what that adds. Eventually you must look at the audio outputs.

I have made really high resolution measurements and they do show issues occasionally. The QA400 is limited to 65K point resolution. My Praxis setup which is more costly (around $1200) can make up to 16 million point fft. It takes a while and you don't see much more. I have access to much more exotic equipment (e.g.: DSA8300 Sampling Oscilloscope | Tektronix ) but it still won't show these audio frequency problems and is a monumental PITA to use. I stopped looking at clock jitter directly since it was becoming fruitless.

If you are really obsessed with analyzing the clocks this is the go to instrument: http://www.symmetricom.com/products...ation-test-sets/3120A-Phase-Noise-Test-Probe/ It was $5K when it was introduced, I don't know the current price.

.
 
Sounds like Carver :) BTW I was glancing at the FIFO thread, good stuff!

;) indeed it is not a unique methodology.

yep, Ians fifo is one of the most complete solutions ive come across here. its a pure brute force solution that does what its designed to do. it does turn a dac into something resembling mission control, but it gets results and with a refreshing highly engineered approach. it is missing something that goes bing, otherwise its audio geek heaven ;)
 
Sorry Jacco, if I appeared to 'diss' you, by using your input to respond to something that even I hate in the hi end audio industry.
One of my 'disappointments' in high end audio is NOT that something costs a lot, because some assembly techniques and materials really DO cost a lot to use and process.
What 'bugs' me are the designs that make a really fancy case and then just put IC's or something even more 'crude' inside, and they jack the price way up high.
This makes for embarrassing examples that the rest of us have to contend with, from the comments of our online critics and others who are skeptical of the audio industry.
The first time that I saw this was in the late 1970's, when a French audio company put forth a preamp with a fancy name, a control stick, like on a plane, for a volume control, AND when you looked inside, you found only very dated IC's and cheap parts. YET, the cost of the preamp was significantly more than previous examples in the marketplace.
A few years later, I just about laughed, out loud, when Matti Otala's Citation power amp built by HK, came out at $5,000 dollars and the only fancy add on was a gold plated circuit board. This was another extreme mark-up example.
My last example, but not the least, was another HK product that took another company's internal circuit board and re-packaged it at several times the price of the original unit.
Hi End design doesn't necessarily come cheap. There is the inefficiency of small product runs, the extra cost of labor to get the 'fit and finish' right, etc. BUT that does not excuse where the actual circuitry is relatively cheap, YET the final product price is enormous. I agree with everyone here that there are examples of this in hi end audio, BUT not usually, just sometimes.
Other times, products like Nelson Pass, Charles Hansen, and I make are highly regarded, and IF EXPENSIVE, have the obvious reason for the added expense obvious when you look inside. It is NOT that everything has to be expensive, but IF it is expensive, then let it be because we went 'all out' to produce the best product possible, and our approach took us in a particular direction.
For example, do you realize the Nelson, Charles and I use the SAME jfets in our better designs? Why, when they are now so hard to obtain? It is because they work better than anything else, and if this increases the price, so be it.
It may well be that IC's if properly applied, can 'fool' most of the people all the time, and be perfectly adequate for 90% of the listening public in virtually every case, but there will be exceptions and they tend to be our customers, if they can afford it.
 
Member
Joined 2002
Paid Member
He is playing up the difficulties and expense to design things properly to avoid the jitter etc. when it can be eliminated completely with a module bought here on the forum for $250 + 2 clocks and works with i2s or spdif sources alike, with any dac with i2s input, producing the same very high quality regardless of the source.

he speaks of cutting edge things that may be possible in the future that are pretty much already being done here by hobbiests without the polish and aircraft grade alloy chassis.

qusp
When I am defending or bushing a person I do it in an obvious way. In that article, it was only the technical issues and ideas that I noted. Only!
And that article was on a commercial (promotion) magazine. Meitner has published on jitter also in AES (I don’t have access there): "Time Distortions Within Digital Audio Equipment Due to Integrated Circuit Logic Induced Modulation Products", Ed Meitner and Robert Gendron, presented at the 91st AES Convention, New York, October 1991, Preprint 3105
Me, just as you, haven’t paid attention to the date. The article writes: Published: Mar 1, 1993. Ed saw it.
That a diy module can solve the issue is very encouraging. :)

George
 
Last edited:
Member
Joined 2002
Paid Member
DIY way.. Vertical axis is relative level of unwanted "artifacts" in Dunn (J-test) spectra..

BV thanks for the chart.

George, take a look at this approach:
DAC Overview
Best,

Elektroj
Piece of cake. But what is the physical quantity that this methodology show?

Someone here mentioned it and I took the bait. It turns out to be more involved than it might seem. I posted my incomplete notes here on using a tuner: http://www.diyaudio.com/forums/blogs/1audio/983-fm-tuner-jitter-analysis.html

Thank you Demian. I’ll look into it.

George
 
Status
Not open for further replies.