Talk:Numerically controlled oscillator

Untitled

After the DAC there has to be a filter, this is correct. The main reasons for the filter are (a) to remove the spectrum above the nyquist frequency and (b) interpolate between samples. It is therefore called INTERPOLATION filter. i think aliasing refers here to (a). Aliasing generally occurs at the input of an ADC due to spectrum folding, so a antialiasing filter has to be used here, but not at DAC output! It is of course also useful for removing glitches faster than the Nyquist Frequency.

---

—Preceding unsigned comment added by 131.188.134.245 (talk) 14:53, 14 July 2009 (UTC)

A NCO and a DDS in my opinion are not equivalent. A DDS digitally synthesises a waveform (The Direct digital synthesis article even goes as far to suggest the waveforms final destination is a DAC). A NCO creates a periodic waveform of a given frequency and sample rate.

I think it is accepted that NCO's are an integral part of a DDS. However its not implied that the output of an NCO will ever see a DAC, or be part of a DDS system. e.g. The output may be used for further digital modulation, or as a phase reference for an another signal. The latter example I wouldn't consider to have any relationship to a DDS. —Preceding unsigned comment added by 202.37.96.11 (talk • contribs)

I agree. The two are closely related, with one used in the other, but they are not the same thing. - mako 20:11, 10 February 2006 (UTC)

I don't think they should be merged.

Ditto, no merge --Hooperbloob 03:51, 30 March 2006 (UTC)

don't both methods use a phase-accumulator and lookup table for the waveform?

may i suggest that someone craft words that explicitly differentiate between NCO (or DCO, the same thing, right?) and DDS? and also remove the merge tag? Rbj 02:07, 20 April 2006 (UTC)

Another vote for not merging the two articles. On the contrary, I'd like to see a clearer separation between some things that this article currently confuses, such as the various things that have come under the broad banner 'Digital Oscillator' or 'Digitally Controlled Oscillator'. These go from the programmable-counter-plus-waveshaper DCO's of the Juno series synths, through to the full DDS phase-accumulator-plus-wavetable-lookup. Currently 'DCO' in the synth context refers us to this article on NCOs, which is completely inappropriate in many cases. Perhaps I should just write a 'DCO' article for the synth pages..,Electricdruid 22:33, 28 May 2006 (UTC)

One further distinction that has recently come to my attention is mentioned by Chris Meyer (ex-Sequential Circuits) in his article "Birth of the Prophet VS" (http://www.vintagesynth.com/sci/birth.txt). As he says there, you can make a digital oscillator one of two ways: either use a short waveform and vary the playback rate (as is done in the PPG and Prophet VS) or use a long waveform and skip some samples in it to alter how long it takes to playback (The DDS/NCO scheme, used in the Korg DW series). Electricdruid 08:35, 16 June 2006 (UTC)

Moved text

The following isn't written in an encyclopediac tone and is at best confusing (and I suspect is at worst actually incorrect). Shop manuals for gear you have around do not make readily accessible references. --Wtshymanski 17:48, 16 August 2006 (UTC)

---

Confusion: The description above is about a pure Digital Oscillator (DO) - not a DCO. However, the term "digitally controlled oscillator" is often used incorrectly to describe an analogue oscillator being digitally controlled. While A DCO uses no DAC, the generated waveform is not analog either.

Analog oscillators have analog cores and produce either a triangle or sawtooth waveform originally. A standard DCO IC (82C53- The C for CMOS), as found in say, a Roland Juno or JX model synth, is a *digital* or programmable counter IC. Completely synchronous and without phase independence or control. Digital bits go in and one digital bit comes out, completely locked in phase to the main CPU clock. If such an oscillator were indeed a digitally controlled analog oscillator, the oscillator section itself could drift, change phase and contain other qualities. This is not possible because the heart or core of these mis-named oscillators is, indeed, digital. In fact, in a Juno-106, this signal is still called a clock (in the Roland schematics themselves) when it reaches the waveshapers and filters. It should also be noted that few, if any, true analog oscillators produce a square wave (as DCOs do) as their original, core waveform. All true analog oscillators will drift if heated or if directly exposed to high humidity. None of these conditions are true for DCOs because they have no analog core to begin with. There's nothing to drift. In fact, a Juno doesn't need or use an autotune circuit or pitch correction circuit whatsoever, because the pitches are certain and definite to begin with and cannot vary. This is something only a digital source can provide. Proof can further be found in the fact that when placed in unison mode, no phasing can be observed at all. There is no possibility of six non-synced analog oscillators, even when controlled by a digital source, of achieving such a feat. A DCO stops at the digital counter. Analog waveshaping is not part of a DCO. For this reason, no noise, grunge or other artifacts (including ringing and overshoot) are present at the output of the DCO. It's a purely digital output equal to any other digital logic line as found in a computer. An 8253 counter contains only digital circuitry. The main confusion regarding what an oscillator is or isn't must be reduced to its core, since waveshaping comes afterwards in both cases. The phase, whether shaped by analog conversion or not, still remains perfect between DCOs. This is the main difference between an analog core and a DCO. There is no drift. Drift is always caused by analog sources, something a DCO doesn't have. Since all sound is analog, a line must be drawn between the digital componentry and final output. If such a convention did not exist, all digital audio devices such as CD players could be considered analog. Additionally, it would suggest that a suboctave output of a VCO, such as found on a Roland SH-101 synthesizer, obtained by a digital flip/flop, would be digital and thus a DCO. This is not so. A DCO is not a digitally controlled, nor digitally modified VCO. Inversely so and well accepted, a DCO is also not an analog oscillator because it starts digitally and ends digitally. It only has analog waveshaping circuitry afterwards. Both the inputs and outputs remain digital, in phase and at standard logic levels. Treating digital logic outputs as analog does not make a DCO an analog component.

Verifiable information: To avoid further re-edits and misinformation it is suggested that future authors first consult the Roland Juno-106 Service manual on page 5. Clearly shown there are block diagrams labeling the 8253 counters as DCOs. This shows that Roland Corporation themselves regard the 8253 digital counters as DCOs and did not include any analog waveshaping circuitry coming afterwards as part of their DCO designations. This proof of terminology can be additionally found on page 8 of the same service manual where it is stated "...DCOs IC33 and IC34 which are 16-bit Programmable Interval Timers."

---

The following information was extracted from various debates in the Harmony Central synth discussion groups that involved a co-designer of the DSI Evolver series. It contradicts other definitions of the Digitally-Controlled Oscillator (DCO).

In a conventional DCO such as those found on the Roland Juno and JX series synthesizers, the actual "oscillation" (variable pitch generation) is generated in the digital domain (square wave cycles emitted via a programmable counter IC). However, these are only clocked pulse signals that have absolutely no audio signal content, and therefore cannot be considered the end of a DCO construct by any reasonable measure. There is analogue waveshaping going on beyond the IC counter portion of a DCO, producing an analog waveform. This is all part of the DCO construct. The output of a DCO is a time- and amplitude-continuous waveform, which is achieved entirely without the use of DACs (digital/analog converters). It's an audio signal. It's analog. DCOs can have noise, grunge, overhang and all of the artifacts associated with VCOs because they are free-running just like VCOs.

This combination of digital and analog circuitry makes the DCO a hybrid construct in itself; therefore, it's technically neither an analog nor a digital oscillator, but a hybrid oscillator.

Since this issue comes up every so often here's a case study for those around that are interested in checking the down and dirty details of DCOs for themselves:

Let's use a most typical DCO based hybrid synth for which the schematics are available, the Roland JX-3P/MKS-30. The MKS-30 has 4 x 8253 counter/timer chips. Each 8253 has 3 16 bit counters for a total of 12 counters (DCOs!). The counters are each configured to count through a programmed range. The 8253 counts at a fixed rate, so the range of each counter determines the frequency at wihch the counter completes and starts again (or wraps around). Each counter outputs a pulse or a square wave period every time it counts through its range - this signal is just a logical level on a pin of the counter chip.

The outputs from the counters are routed to the analogue waveshaping circuitry. This is still part of the DCO operation. Each counter also has an input pin which is configured to reset the counter, and this is used for oscillator sync. The counter range only need to be reprogrammed each time the oscillator changes frequency, so the DCO's are (or can be) free running in this case.

The JX-8P and JX-10 work exactly the same way, and most other DCO based synths work similarly to this, although the counter/timer chips/software will vary. In a modern hybrid synth like the Evolver, a single DSP chip can generate several DCO pulse trains, with PWM if wanted, using software and/or built in counter/timers.

There are even earlier implementations of the DCO (such as the ones found in the Crumar DS-1 and DS-2 synthesizers, for example) that only have the CV processor and expo converter in the "digital domain", and send a control current (via DAC) to the oscillator core. Everything could be analog from there. Hence another hybrid, but more "analog" than using a timer to generate a square or other stepped waveform, and then waveshaping.

GA Review

This review is transcluded from Talk:Numerically-controlled oscillator/GA1. The edit link for this section can be used to add comments to the review.

Reviewer: Shirik (Questions or Comments?) 02:08, 3 April 2010 (UTC)

Well written

(a) the prose is clear and the spelling and grammar are correct

Well written. There was an incredibly minor grammar issue that I went ahead and corrected myself.

(b) it complies with the manual of style guidelines for lead sections, layout, jargon, words to avoid, fiction, and list incorporation

Remember to wikilink technical terms when possible. Waveform and sinusoidal are not typical English vernacular so it is appropriate to wikilink those from the article. I have gone ahead and corrected that and found no other issues.

Factually written and verifiable

(a) it provides references to all sources of information in the section(s) dedicated to the attribution of these sources according to the guide to layout

It would be advisable to get additional sources of references, but the content is referenced. A wider source of references would only serve to provide a wider view on the topic and/or allow for cross-referencing.

(b) it provides in-line citations from reliable sources for direct quotations, statistics, published opinion, counter-intuitive or controversial statements that are challenged or likely to be challenged, and contentious material relating to living persons—science-based articles should follow the scientific citation guidelines

While in-line citations are sparse, they are where they need to be. This could be improved, but it certainly satisfies this criterion.

(c) it contains no original research

No evidence of original research. Content appears to be referenced appropriately.

Broad in its coverage

(a) it addresses the main aspects of the topic

OK

(b) it stays focused on the topic without going into unnecessary detail (see summary style).

Detailed where it needs to be, but does not digress or over-complicate things (beyond what would be expected from an engineering article).

Neutral

it represents viewpoints fairly and without bias.

The article appears to be neutral, but for further improvement, a wider set of references would not be a bad idea.

Stable

it does not change significantly from day-to-day because of an ongoing edit war or content dispute.

No content dispute

Illustrated, if possible

(a) images are tagged with their copyright status, and valid fair use rationales are provided for non-free content

Both images are free and appropriately tagged, however I strongly advise converting them both to SVG as both are definitely candidates for it.

(b) images are relevant to the topic, and have suitable captions

Surely relevant and captioned, however consider placing the first image at the very top of the page instead.

General comments

This article is well written for a technical article. There is still a bit of work to be done (some details can be fleshed out, etc.) but the article gives a good background knowledge to any reader. See the above comments on potential places for continued improvement.

Overall

Having satisfied the above criteria, I see no reason not to pass this article, however I advise that the recommendations above be followed for further improvement. --Shirik (Questions or Comments?) 02:21, 3 April 2010 (UTC)

Undid revision 365380425 by 194.128.29.93 (talk)

The editor was confused. The diagram is normalized to one cycle. So the text is correct, Delta F is the integer step size and the diagram is correct, when the phase is normalized to one cycle, the step-size is Delta F/2N. JPatterson (talk) 05:01, 8 June 2010 (UTC)

Claim about dither as mitigation for amplitude spurs

I think the following is only part of the story: "Amplitude truncation spurs can not be mitigated in this fashion. Introduction of noise into the static values held in the PAC ROMs would not eliminate the cyclicality of the trunction error terms and thus would not achieve the desired effect."

What the author doesn't mention is that WGN can be added after the ROM and would indeed eliminate the cyclicality of the truncation error terms.

I recommend revising the article. — Preceding unsigned comment added by Rhearty1 (talk • contribs) 20:17, 3 September 2013 (UTC)

"WGN can be added after the ROM and would indeed eliminate the cyclicality of the truncation error terms."

This is incorrect. Dither is a linearization technique. It does no good if applied after the non-linearity (in this case, amplitude quantization). The original author was correct. — Preceding unsigned comment added by 162.72.213.208 (talk) 04:46, 4 February 2014 (UTC)