AR-Cluster and DXSpider / A side-by-side study
SKIMVALID and Qualitee: what is comparable?
Both use shared receiver evidence, but they do not deliver spots in the same way. SKIMVALID accepts a classified incoming report. DXS collects reports, chooses a frequency, and selects one retained report for each eligible user.
The source and live evidence here compare ARC6.6.159 CT1BOH on NC7J-3 with released DXS1.57 build686 on NG7M-1. Both studies used W3BES and W3BES-1 with a K receiver-country restriction. They were separate real-traffic observations, not a simultaneous replay of identical inputs; spot totals are not an accuracy or performance benchmark.
Similarly configured users
For the CW subset studied, these are the relevant configurations:
| AR-Cluster | DXS on NG7M-1 |
|---|---|
| |
AR-Cluster’s second account used SKIMMER AND SPOTTERCTY=K, giving the broader comparison feed. Both DXS accounts used the same curated configuration; a third, raw RBN subscription supplied the input comparison. These choices align receiver scope and CW examples, but do not equate the algorithms.
How the decisions differ
| Question | AR-Cluster SKIMVALID | DXS Qualitee, observed build686 |
|---|---|---|
| What reaches the filter? | Each incoming CW / RTTY skimmer report with a CT1BOH classification. | One processed collection of same-call reports; one retained report may represent it for this user. |
| How does the user enable it? | Use SKIMVALID as a clause in SET DX FILTER. SKIMMER alone admits the broader skimmer feed. | Use SET/SKIMMER CW, then ACCEPT/RBN / REJECT/RBN as needed. RBN output is curated before user selection. |
| Comparable receiver restriction | SPOTTERCTY=K tests the incoming report’s receiver. | BY_DXCC K tests retained receiver reports and can affect which representative is chosen. |
| What evidence confirms a new call? | The normal path finds two earlier same-call reports, then classifies this incoming report Valid. Stored prior quality need not be Valid. | NG7M-1’s current minimum is two collected reports including the first report. Collection and receiver deduplication are separate steps. |
| Must all evidence pass the user filter? | No. The shared CT1BOH history is built before user filtering. | No. The global group can contain other receiver countries; the Q marker is not restricted to K witnesses. |
| Does the filter change shared decisions? | No. It determines whether this classified incoming report is delivered. | It does not retune global grouping, Q or suppression. It does select a retained representative within the processed group. |
| History and timing | Nearby report history is scanned newest first, with a 27-minute cutoff. No new-call 30-minute waiting period. | Normally collect for 10 seconds; insufficient evidence can wait 5 minutes. A published key retains time and Q for an hour, rather than a full hour of raw reports. |
| Repeat reports | SKIMVALID itself tests quality each time; UNIQUE and other normal delivery conditions remain separate. | The observed node suppresses a published call / nearby-frequency key for 3 minutes globally, then builds a respot collection. |
| Which receiver is delivered? | The receiver on the incoming Valid report. Cached witnesses are not delivered retroactively. | Without a user filter, the weakest retained signal. With a filter in build 686, the last passing retained report. |
| Which frequency is delivered? | SKIMVALID does not replace the incoming frequency by a receiver vote. | A frequency chosen by weighted receiver-agreement voting; differing input frequencies produce the star marker. |
| What about a similar busted call? | A nearby competing Valid call can trigger the CT1BOH similarity / Busted or BustedOverride path. | The inspected RBN processor checks format/prefix and same-call corroboration; it does not implement that CT1BOH competing-call classification. |
| What is the reported result? | Valid, Unknown, QSY or Busted filter fields; SKIMVALID accepts Valid. | Q is capped at 9, can retain an older higher value on a respot, and is not a correctness probability. |
| Can the second user receive raw input? | SKIMMER without SKIMVALID supplies the broader locally received skimmer comparison feed. | An ordinary user rule cannot bypass the node’s curation. Use a separate read-only raw RBN connection for comparison. |
A practical way to read the two outputs
With SKIMVALID: “This new report was classified Valid using the node’s prior evidence, and its receiver passes my filter.” Read the linked cache example to see which earlier reports established that result.
With DXS: “This group passed the node’s collection and timing rules; at least one retained report passed my filter, and this receiver represents it for me.” Read Q as stored corroboration under those rules, rather than as a fresh count of matching receivers. Read the frequency as the node’s vote result.
A DXS q 3+ rule does not reproduce SKIMVALID. It filters a different result, and an earlier higher Q can survive into a later two-report respot. Changing a sysop’s minimum or respot interval can make the visible traffic volume more alike without adding CT1BOH’s classification path.
Use the complete examples
Inside SKIMVALID shows two prior cached witnesses establishing a new incoming Valid report, older history, scan order and a competing-call rejection.
Inside DXS Qualitee shows saved current collections, representative selection, frequency voting, and a higher remembered Q retained on a smaller respot collection.
Neither result guarantees a correct decode. This comparison explains observed behavior and evidence; it does not rank one implementation’s overall accuracy over the other.
Credits and references
Investigation and documentation collaboration: Dave N3RD and Max George NG7M. CT1BOH algorithm: José Nunes CT1BOH; original AR-Cluster implementation: Terry Gerdes AB5K (SK). DXS RBN processor and Qualitee: Dirk Koopman G1TLH.
The companion pages contain version-pinned source references and the individual spot-only evidence records. The DXS settings are values queried from NG7M-1, not assumed universal defaults.