• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

Best spec ADC Chip currently.... ??

DonH56

Master Contributor
Technical Expert
Forum Donor
Joined
Mar 15, 2016
Messages
7,835
Likes
16,498
Location
Monument, CO
Re. programs, why not Python?

More on topic, is there a link to these soon-to-be products? @IVX ?
 

IVX

Major Contributor
Joined
Mar 3, 2019
Messages
1,409
Likes
2,773
Location
South of China, SHZ area, - Слава Україні
The link is here https://e1dashz.wixsite.com/index but I recommend joining our discord to stay tuned.
I had the plan to start SMT production tomorrow after tomorrow but sample #6 which shows so nice specs, was assembled by my hands to confirm that all prepared parts are fine. Unfortunately, that sample also did help me find that the Epcos crystals 1000pcs don't work with Comtrue CT7601. No idea why, the same Epcos crystals work in thousands of 9038S/D but these 1000pcs don't, at least they look identical. So, the production run will be a bit delayed.
Epcos_12MHz.jpg
 
  • Like
Reactions: trl

DonH56

Master Contributor
Technical Expert
Forum Donor
Joined
Mar 15, 2016
Messages
7,835
Likes
16,498
Location
Monument, CO
Layout changes around the crystal could muck them up... Load capacitance and coupling could change.
 

DonH56

Master Contributor
Technical Expert
Forum Donor
Joined
Mar 15, 2016
Messages
7,835
Likes
16,498
Location
Monument, CO
Crystal oscillators have an optimal load impedance typically set by a couple of shunt capacitors and a feed back resistor across the amplifier. Some require a series resistor at the amplifier's output to limit the power into the crystal. Improper loading will change the oscillation frequency and can over- or under-power the crystal itself. I assume you know all this, but it is strange the crystal manufacturer does not provide basic specifications for the crystal, like load capacitance, effective series resistance (ESR), drive level, and shunt capacitance. Those are needed to design a working oscillator, along with the amplifier's characteristics like gain and bandwidth, and differences in those parameters could explain why some crystals work and others do not. Debugging crystal circuits can be tricky -- high impedances so it takes special probes (I use active FET probes with <1 pF load and 1+ GHz BW), and any extra load or coupling can change the results.

Not too long ago a customer had a run of boards that were causing start-up problems in an oscillator. They copied our reference design, but the board layout used traces to the crystal about four times longer than what we did, and routed them very close together, so load and shunt C were much higher than spec. We ended up tweaking the bias of the on-chip amplifier to compensate, and the customer had to tweak their load capacitor values a little, to solve the problem. The ultimate solution was a board spin to fix the layout.
 

mansr

Major Contributor
Joined
Oct 5, 2018
Messages
4,685
Likes
10,700
Location
Hampshire
Crystal oscillators have an optimal load impedance typically set by a couple of shunt capacitors and a feed back resistor across the amplifier. Some require a series resistor at the amplifier's output to limit the power into the crystal. Improper loading will change the oscillation frequency and can over- or under-power the crystal itself. I assume you know all this, but it is strange the crystal manufacturer does not provide basic specifications for the crystal, like load capacitance, effective series resistance (ESR), drive level, and shunt capacitance. Those are needed to design a working oscillator, along with the amplifier's characteristics like gain and bandwidth, and differences in those parameters could explain why some crystals work and others do not. Debugging crystal circuits can be tricky -- high impedances so it takes special probes (I use active FET probes with <1 pF load and 1+ GHz BW), and any extra load or coupling can change the results.
What's your take on oscillator modules vs discrete crystals?
 

DonH56

Master Contributor
Technical Expert
Forum Donor
Joined
Mar 15, 2016
Messages
7,835
Likes
16,498
Location
Monument, CO
What's your take on oscillator modules vs discrete crystals?

If you're asking about audio, I tend to think the packaged oscillators sold to improve jitter performance are overkill the vast majority of the time. I have seen $100 DIY modules that are basically rebadged off-the-shelf oscillators up to $10k+ clock boxes with fs jitter performance. A recording studio may use a master clock box to synchronize everything, but I really doubt they are paying audiophile prices for it. Usually one of the components has buffered clock output they can use.

Bearing in mind my work is not audio, for us and our customers it is basically a "make or buy" decision. A crystal circuit is cheaper, assuming the amplifier is on-chip, but usually has lower performance. As long as the performance is good enough to meet spec and the layout (and chip) design are good then a crystal works fine. A pre-packaged module has all the work done for you, has greater slew rate and lower output impedance into the chip's clock input, and so may offer better performance -- at a substantially higher cost. It also usually requires a filtered power supply rail, and a few support components for enables and such, so component count and board area might not be much different than the crystal circuit. Beyond that, you can get much higher-performance packaged oscillators with lower noise and features like temperature control (built-in ovens), programmable frequency and output voltage, and so forth. For a consumer audio device, I'd expect a crystal into a DAC (or whatever) chip would be good enough and cheaper, always assuming the board designer does due diligence on the layout requirements, both for parasitic elements and things like thermal considerations.

Things that can go wrong with a crystal circuit, besides the long traces mentioned above, include things like placing the crystal too close to a SMPS regulator that couples noise into the circuit, not shielding the crystal traces from adjacent signals or noisy power planes, and placing it close to a heat source that may be the device it is connected too -- sometimes putting the crystal on the back side helps if the device consumes a lot of power. And so forth. I have just enough experience designing oscillators to know there are a lot of tricks and sneaky "gotcha's" that takes an expert with lots of experience to catch. Without that, or access to it, buying a prepackaged oscillator may be the best solution.

Many crystal and device manufacturers provide design guidance and sample layouts that need to be followed closely to prevent such problems.

HTH - Don
 

mansr

Major Contributor
Joined
Oct 5, 2018
Messages
4,685
Likes
10,700
Location
Hampshire
If you're asking about audio, I tend to think the packaged oscillators sold to improve jitter performance are overkill the vast majority of the time.
I'm talking about the little 4-pin modules available for $1 or so from most of the same vendors that sell plain crystals. Depending on the total BOM, they seem like the extra cost could be worthwhile to avoid the potential problems you've mentioned with load capacitance etc.
 

IVX

Major Contributor
Joined
Mar 3, 2019
Messages
1,409
Likes
2,773
Location
South of China, SHZ area, - Слава Україні
mansr, I saw some appnote a few years ago, I believe it was Vectron which is today a part of Microchip. Unfortunately, I didn't save the pdf and now I can't find that again(Vectron's papers are now dumped together with Atmegas without any structure) but it was about low phase-noise crystals compared vs standard crystals. The oscillator used in the test was a lab-grade standalone device, of course, low-noise crystals showed phenomenally low noise but I was surprised how cool there performed standard crystals! In fact, low-noise had lower noise only <100Hz or so, 1k, 10k, 100k etc noise was the same for both types of quartz crystals and it was near to -180db floor or something crazy like that numbers. After reading that I did understand that a good generator-oscillator circuit surely dominates in the task of low phase noise source of clock. Any $1/10 Epcos crystal + the right oscillator circuit outperforms lots of famous clocks jitter.
 

DonH56

Master Contributor
Technical Expert
Forum Donor
Joined
Mar 15, 2016
Messages
7,835
Likes
16,498
Location
Monument, CO
I'm talking about the little 4-pin modules available for $1 or so from most of the same vendors that sell plain crystals. Depending on the total BOM, they seem like the extra cost could be worthwhile to avoid the potential problems you've mentioned with load capacitance etc.

OK, got it. In bulk I think we are getting crystals for maybe 10~25 cents, and if a customer implements the circuit properly they save money. For customers without the expertise, or without access to design help (which my company will supply, as will the crystal companies), then the packaged modules are better albeit more expensive. When you are shipping millions of units cost is a big factor. But yah, I agree with you, especially since if a customer gets it wrong I am usually the one spending time fixing it. :)

For precision applications requiring wide operating temperature range, high vibration tolerance, and extremely low noise, packaged by experts is usually the best choice.
 

DonH56

Master Contributor
Technical Expert
Forum Donor
Joined
Mar 15, 2016
Messages
7,835
Likes
16,498
Location
Monument, CO
mansr, I saw some appnote a few years ago, I believe it was Vectron which is today a part of Microchip. Unfortunately, I didn't save the pdf and now I can't find that again(Vectron's papers are now dumped together with Atmegas without any structure) but it was about low phase-noise crystals compared vs standard crystals. The oscillator used in the test was a lab-grade standalone device, of course, low-noise crystals showed phenomenally low noise but I was surprised how cool there performed standard crystals! In fact, low-noise had lower noise only <100Hz or so, 1k, 10k, 100k etc noise was the same for both types of quartz crystals and it was near to -180db floor or something crazy like that numbers. After reading that I did understand that a good generator-oscillator circuit surely dominates in the task of low phase noise source of clock. Any $1/10 Epcos crystal + the right oscillator circuit outperforms lots of famous clocks jitter.

Agreed, for most designs the amplifier dominates the noise. I helped design a fs source many years ago and it was basically all about the amplifier. And the housing in that case; extreme vibration resistance was required. And mil-spec temp range, radiation hardness, etc. Not an audio application...

Most crystal circuits, or oscillator outputs, go into a PLL or something that further suppresses the LF noise.
 

Grooved

Addicted to Fun and Learning
Joined
Feb 26, 2021
Messages
679
Likes
441
mansr, I saw some appnote a few years ago, I believe it was Vectron which is today a part of Microchip. Unfortunately, I didn't save the pdf and now I can't find that again(Vectron's papers are now dumped together with Atmegas without any structure) but it was about low phase-noise crystals compared vs standard crystals. The oscillator used in the test was a lab-grade standalone device, of course, low-noise crystals showed phenomenally low noise but I was surprised how cool there performed standard crystals! In fact, low-noise had lower noise only <100Hz or so, 1k, 10k, 100k etc noise was the same for both types of quartz crystals and it was near to -180db floor or something crazy like that numbers. After reading that I did understand that a good generator-oscillator circuit surely dominates in the task of low phase noise source of clock. Any $1/10 Epcos crystal + the right oscillator circuit outperforms lots of famous clocks jitter.

Hi @IVX
not sure at all but is it something like this pdf? : https://www.vectron.com/products/literature_library/phase_noise.pdf
 

HpW

Member
Technical Expert
Joined
May 13, 2018
Messages
38
Likes
9
Location
Guacimara
As it goes with some 24.576MHz DIP-14 Chinese implementations :D and other custom frequencies...

Inside is a crappy TXCO at 19.2MHz... after some IC's we get finally 24.576MHz at the output !!??

Just go figure until you open them :D :D
 
Top Bottom