Jitter at the converter is an exclusive Specifications Pro file. The claim under test is jitter. In a converter lab in Tokyo the sheet offers picoseconds or a plot. On a working day in a mastering room in Portland that figure is only meaningful beside where it was measured: input or output. This desk does not sell the device and does not rank a brand. We read the footnote. A high-intent buyer already knows the headline number. What is missing is the sentence a successor can check without the salesperson in the room: Measure on the clock you will distribute. If that sentence cannot be written, the datasheet is a poster. Posters are allowed in a lobby. They are not allowed on a purchase order. The sections below stay specific to jitter work. They are not a generic luxury essay and not a reprint of a manual.
Start with the test condition, because the number without it is theater. For jitter, where it was measured: input or output. Write those words next to picoseconds or a plot before you compare two quotes. In a converter lab in Tokyo the condition is controlled. In a mastering room in Portland it is not, unless you bring the control with you. Exclusive reporting on this desk keeps meeting the same quiet failure: the converter was excellent on its internal clock and not on the house clock. That failure is not a scandal. It is what happens when a careful adult trusts a headline. Ask the vendor to point at the paragraph in the sheet that states the condition. If they point at a photograph, you do not yet have a specification. You have a campaign. Campaigns are not what Timing is for.
Units and tolerances are the product. Picoseconds or a plot must name the unit, the tolerance, and the sample. A single hero unit in a converter lab in Tokyo is not a production spread. Ask how many units were measured and what was excluded: warm-up outliers, pre-production firmware, a golden sample. Then put the same question to the unit that will ship to a mastering room in Portland. If the vendor cannot say whether your serial is inside the measured population, price the risk as real. Specifications Pro will not invent a tolerance the sheet does not have. We will insist the missing tolerance is a blank line, and blank lines do not ship.
The interface can void a true number. Here the interface is a long cable and a dirty source. A measurement that stops before that interface is a measurement of a different product. Draw the path from the claim to the thing a human hears, sees, or stores. Every box on that path can add noise, latency, color, heat, or a second clock. In a mastering room in Portland the path is rarely the path in a converter lab in Tokyo. Write the path you will actually build, then ask which part of picoseconds or a plot survives it. If the answer is 'the spec assumes a direct connection,' you have learned the limit of the spec. Limits are useful. Implied infinity is not.
A bench protocol, short enough to do. Day one: copy picoseconds or a plot and where it was measured: input or output into a one-page sheet titled with the legal model, not the nickname. Day two: measure or inspect a long cable and a dirty source and photograph the screen that states the mode, not the screen that says congratulations. Day three: run the failure on paper. The failure we keep seeing is this: the converter was excellent on its internal clock and not on the house clock. If your installation cannot prevent it, write the mitigation beside it. Day four: take the page to a second human who was not in the sales call. If they cannot reconstruct the claim, the page is not finished. Day five: only then compare price. Price without the page is how a mastering room in Portland buys a lab story that will not survive the first week.
The purchase order is where specifications become enforceable. The sentence we would not ship without is: Measure on the clock you will distribute. Add the firmware or revision, the included accessories that were present during the test, and the lead time for the spare that keeps the number true. Do not write 'or equivalent.' Equivalent is how a matched pair, a profile, a clock, or a lamp bin disappears between the quote and the crate. If the seller refuses the sentence, they are telling you the number is not a promise. Believe them. Buy a different promise or budget for a test you will run yourself on delivery.
A specification can pass and the job can still fail. Hold both ideas at once. The converter was excellent on its internal clock and not on the house clock The sheet in a converter lab in Tokyo did not lie. The job in a mastering room in Portland asked a different question. List three job questions the sheet is allowed to ignore: the room, the operator, and the deadline. Then decide which of those three you will control. If you control none of them, you are not buying jitter. You are renting a demo. Demos are legitimate. They are not a substitute for where it was measured: input or output. Write 'demo' on the file if that is what you attended, so a later reader does not treat a showroom as a lab.
Compare two live quotes and let the second one win. Put both through the same page: picoseconds or a plot, where it was measured: input or output, a long cable and a dirty source, and the order sentence. If the first vendor matches the second in an afternoon, the first quote had slack. If they cannot match and cannot explain the gap with a document, you have found the real product. Do not let a brand narrative replace the gap. A famous door in a mastering room in Portland can still ship a unit that was not the unit measured in a converter lab in Tokyo. The desk's refusal is specific: we would not accept an undated plot, a verbal tolerance, or a firmware 'you can update later' as if later were free of risk. Later is a second product.
Field sheet for jitter readers. One: legal model and revision. Two: picoseconds or a plot. Three: where it was measured: input or output. Four: a long cable and a dirty source. Five: the failure you are buying against, in ordinary words: the converter was excellent on its internal clock and not on the house clock. Six: the order sentence: Measure on the clock you will distribute.. Seven: spare lead time and match. Eight: who else can read this page if the buyer is on a plane. If a line is blank, the specification is not finished. Reopen the page when the firmware changes, the room changes, or the vendor is acquired. Those events rewrite jitter without asking you. The question that reopens it is the same question the file started with, asked of the unit you hold, not of the unit in the photograph.
What this desk would not do: we would not treat picoseconds or a plot as a personality. We would not hide a failed check because the case is beautiful. We would not let a long cable and a dirty source be 'included' without a drawing. We would not sign in a mastering room in Portland against a PDF that names a different revision than the crate. The test is dull on purpose. Can a careful adult, using only the page, predict the converter was excellent on its internal clock and not on the house clock and say whether this installation prevents it? If the answer is no, the missing line is the thing you have not bought. The object can wait a week. The line often cannot be reconstructed after the crate is unpacked and the salesperson has left.
Method. Specifications Pro commissioned this exclusive explainer for the Timing desk. We do not take payment to bury a warning. Display advertising is sold apart from the newsroom and does not choose these sentences. We follow Google publisher rules and Google Ad Exchange rules: original copy, no encouraged clicks, no scraped manuals, ads kept off buttons. Sulvo and the other parties named in ads.txt may measure and render ads under their own fraud rules. They do not edit the desk. This is general information, not an engineering stamp, a calibration certificate, or legal advice. Standards and firmware change. The current datasheet and your measurement win. Corrections: send the URL, the clause, and the date. We update a page when a material error is confirmed, not when a vendor dislikes a specific sentence. If you keep one move from Jitter at the converter, keep this: write Measure on the clock you will distribute. and do not accept picoseconds or a plot without where it was measured: input or output.
Applied note for a buyer acting this month. Open a single page. Title it with the legal model of jitter. Under it, paste picoseconds or a plot and, in the next line, where it was measured: input or output. Under that, draw a long cable and a dirty source as boxes. Under the boxes, write the failure in your own words: the converter was excellent on its internal clock and not on the house clock. Leave four lines for the answer you were given, the name, the date, and the document. Then copy the order sentence and do not soften it: Measure on the clock you will distribute. A printed copy stays with the quote. A second copy stays with the person who will operate the system in a mastering room in Portland when the buyer is gone. If you cannot finish the page in one sitting, you have found the hole while it is still a hole. The Timing desk publishes files like this one only on Specifications Pro. Use the section index for the next file. Do not trust a snippet that points at a look-alike domain. The standard on every desk is the same: a footnote a successor can read, a number with its condition, and a clear separation between the sentence and any advertisement beside it.
