GraphWorks User Guide

GraphWorks is Headphones.com's website for exploring headphone and in-ear monitor measurements, comparing frequency responses, and sending supported measurements to GraphWorks EQ.

Use this guide in two ways: follow Getting Started for your first graph, or use the contents to look up a particular control. You do not need to understand every measurement term before starting. The most useful first skill is learning to identify what you are looking at before drawing a conclusion from it.

Choose Your Starting Point#

The basic workflow: find a product, choose a measurement, check the reference and alignment, inspect the response, then share or export the result. An item is the product; a measurement is a dataset for that product; a reading is an individual capture within that dataset. Keeping these three levels separate prevents many common mistakes.

Access Levels#

GraphWorks has three access levels.

Free Access requires an account created with an email address and password; it does not require a paid subscription. Paid Access is GraphWorks Access, by subscription or included with a linked membership: unlimited comparisons, MIRE measurements and GraphWorks EQ. A control can still be unavailable when the selected data does not support it, regardless of access level.

A measurement is public, unlisted or private:

• Public measurements are listed and viewable for everyone.
• Unlisted measurements do not appear in other users' item pages, Visualizer lists, search results, or catalog measurement counts. The uploader can still find their own entries. Anyone with a shared Visualizer link or embed that names one can view it, signed out included, subject to their access level: MIRE data still needs Paid Access, and the comparison limit still applies.
• Private measurements are available only to their uploader and authorized GraphWorks staff. Sharing a link does not make them available to other users. A shared link or embed that names a private measurement draws the rest of its comparison without it for viewers who cannot access it.

Items (the products) have the same three levels. Public items appear in the catalog and search. Unlisted items are omitted from other users' catalog, brand, and search results, including Search items and the Visualizer's Add item search, but anyone with a link to the item page or a visualization naming it can open it. Private items are available only to their creator and authorized GraphWorks staff. A paid subscription does not override an item's or measurement's privacy setting.

Getting Started#

View Your First Measurement#

1. Open GraphWorks.
2. Use Search items, or open Items to browse the catalog.
3. Select a headphone or in-ear monitor to open its item page.
4. Find the measurement you want. Check the measurement name, measurement rig, and uploader rather than assuming every measurement of the same model is interchangeable.
5. Click the measurement entry to open the Visualizer.
6. Start with Single line for a straightforward overview, or Left & Right lines to examine channel differences when your access allows it.
7. Read the legend and check whether compensation and preference bounds are enabled before interpreting the curve.

The number of measurements you can display and the types of data available depend on your access level. For more than the anonymous single-measurement view, sign in first.

A First-Session Walkthrough#

For your first visit, choose a product you already know. Familiarity gives you a listening reference, but do not expect every feature of the graph to match your experience exactly.

1. Open one measurement from its item page. Confirm that its name, rig, and uploader describe the dataset you intended to open.
2. Select Single line. This reduces visual clutter while you learn the axes.
3. Leave Automatic (by rig) selected unless you specifically want to reproduce an older or newer comparison. Keep compensation on for this exercise.
4. Set Normalization Center (Hz) to 630, Octave Range to 2 octaves, and Y-Scale (±dB) to 20. These are a reproducible starting point, not a requirement for every analysis.
5. Check that the measurement's manual offset is 0. Turn on preference bounds if you want the reference region visible.
6. Read from left to right: look at bass, midrange, then treble. Describe broad regions first, rather than treating every small ripple as significant.
7. If available, switch to Left & Right Lines to see what the average concealed. Then try Percentile to inspect variation among the available readings.
8. Return to Single line before adding a second measurement. This makes it easier to tell whether a new line is another product or another channel.

At this point you should be able to answer: which product and dataset am I viewing, which reference am I using, and am I seeing an average, channels, or variation? If any answer is unclear, check the sidebar, settings, and legend before interpreting the response.

Compare Two Measurements#

1. Open the first measurement in the Visualizer.
2. Click Add item, type the model or brand, and select the matching item.
3. Use the measurement's Show measurement eye button to add its trace.
4. Hide any additional measurements you do not want in the comparison.
5. Choose Single line initially and give the two measurements different colors.
6. Use the same normalization settings for the comparison. Prefer comparable rigs where possible; see Preference Bounds and Compensation when comparing different rigs.

Free Access supports up to two selected measurements. Paid Access has no product-imposed limit on the number of selected measurements, although very large comparisons may take longer to render and are harder to read.

Adding an item to the sidebar is not the same as selecting all of its measurements. The eye controls determine which measurements are drawn.

A Starting Configuration You Can Return To#

When a comparison becomes confusing, simplify it before adding more data:

Apply these settings individually; this is a recommended setup, not a dedicated reset command. A shared URL may restore different settings, so do not assume opening a link starts from this configuration.

Accounts and Access#

Create a Free Account#

1. Choose Login, then New here? Create an account., or open the registration page.
2. Enter a username, email address, and password.
3. Click Create account.
4. Open the verification email and follow its activation link.
5. Return to GraphWorks and sign in.

Your username is your display name; enter it without an @ prefix. Your email address is used to sign in. Free Access starts with the verified account and does not require entering payment details.

Sign In and Sign Out#

Enter your email and password, then click Login with email. Remember me requests a persistent login; avoid using it on a shared computer. The show/hide password button only changes whether the password is visible while typing.

After signing in, the site may return you to the page you came from or direct you to your account's landing page. Use the account menu to open Settings or Log out.

Recover a Password#

Choose Forgot password, enter your GraphWorks email, and click Reset my password. Open the email link and complete the reset page. Then sign in using the new password.

On the reset page, enter New password and choose Submit. The success message confirms that the password was reset; you can then return to Login.

Check spam or junk folders if the email does not arrive. Use the newest reset link and request a fresh one if it is no longer valid. Never send a password or password-reset link to support.

Get Paid Access#

GraphWorks Access is the website's paid subscription. Open its page from the header or from Settings. When subscriptions are available, the page displays the current plan, price, billing interval, and payment choices.

Sign in before subscribing so the purchase is attached to the correct GraphWorks account. Choose one of the payment methods offered on the page, complete the provider's checkout, and return to GraphWorks. Read the displayed currency, tax, and recurring-payment terms before confirming. The checkout page is the authoritative source for the amount you will pay.

If checkout returns to a confirmation or waiting page, allow payment confirmation to finish. Check Settings > Membership before starting another checkout; a pending purchase does not always mean you need to pay again.

Link an Existing Sonic+ Membership#

An active linked Sonic+ membership includes GraphWorks Access. You do not need to buy a separate GraphWorks Access subscription for the same benefits.

1. Sign in to GraphWorks.
2. Open Settings and find Sonic+ Membership, or use Link an existing Sonic+ membership on the subscription page.
3. Enter the email address used for your Headphones.com shop membership. It may differ from your GraphWorks login email.
4. Click Link Sonic+ membership.
5. Retrieve the code sent to that email address.
6. Enter it in Code and click Verify.
7. Check that the membership shows as linked and your GraphWorks Access status has updated.

For an already-linked membership, Re-check membership asks the site to refresh its status. Linking verifies an existing membership; it does not purchase one. Active linked Sonic+ tiers receive the same GraphWorks Access benefits.

Finding Headphones and Measurements#

Site Navigation#

On compact layouts, Menu opens the navigation drawer and Close dismisses it. Clicking outside the drawer also closes it. The account/person button opens Settings and Log out for signed-in users; Login is shown when signed out. The sun/moon button changes the site theme.

Footer destinations are external resources, not additional graph-analysis modes. Menus and navigation choices can differ on the home page and with account access.

Search and Browse#

Search items provides a short list of matching products as you type. Select a result to open the item page. If there are more matches, Show ... more... opens the larger Items results list. The clear button removes the search text.

Use Items for the catalog and Brands to browse manufacturers. On catalog or measurement lists, use the visible search field and sorting selector to narrow the results. The sorting options shown on that page describe its available ordering; searching within an item's measurements is separate from searching the entire product catalog.

Product lists use Best match while searching and Latest when no search is entered, unless you explicitly choose an order. Prev and Next move through catalog/brand result pages; disabled buttons indicate the first or last page. Changing the search or sort returns you to the first page.

An item's measurement list offers Latest, Oldest, A to Z, and Z to A, and loads more entries as you scroll. Measurements are also grouped/ordered by rig in the display, so do not assume the selector removes that organization.

An item can have several measurements: different rigs, different uploaders, samples, or measurement conditions. A displayed measurement count does not mean all of those entries are active in your graph, or that every entry is available to your access level.

Choose Measurements for Your Question#

Choose the question before choosing the dataset. Two measurements of the same model can be useful for different purposes.

Read any supplied description or condition information. A recent upload is not automatically a better measurement, and several entries are not automatically several independent product samples. When the dataset does not establish a condition, keep that uncertainty in your conclusion.

Item Details and Favorites#

The item page shows the product image, brand, model, available category tags, release year when supplied, description, and measurements. Read more... expands a shortened description; Read less... collapses it. A placeholder photo or No description provided means that content has not been supplied, not that measurements are necessarily missing.

For signed-in users, the heart beside each measurement's EQ button saves that specific measurement to favorites. A filled heart indicates it is favorited; click again to remove it. Wait for the success message before assuming the account update completed. Public measurements of public items appear on your public profile and link to that exact measurement in the Visualizer. On your own profile, use the up and down arrows to order favorites; Open Visualizer opens the first five in that order. Use Import CSV on your own Ratings page to load a file with a Headphone or IEM column and a Score column. Scores are rounded to the nearest tenth, and the CSV file itself is not retained. On another member's Ratings page, sort by score or select Compare my scores to show your rating beside theirs; an em dash means you have not rated that item. Favoriting does not save graph settings or an EQ preset, and does not unlock data. To return to a particular comparison, save its visualization URL.

Read a Measurement Entry#

Paid Access provides additional viewing and EQ features; it does not grant permission to upload, edit, or delete catalog entries or measurements.

Understanding the Graph#

Frequency and Level#

The horizontal axis shows frequency from 20 Hz to 20 kHz on a logarithmic scale. Equal distances represent equal frequency ratios, not equal numbers of hertz. Bass is on the left; treble is on the right.

The vertical axis shows amplitude in decibels (dB). A higher point means more measured level at that frequency relative to the graph's chosen reference. A response curve is not a direct volume-meter reading, and the graph's normalization can shift an entire trace up or down.

Use the axes and legend rather than judging only by the curve's apparent steepness: a different vertical scale changes how large differences look.

On a logarithmic axis, 100 to 200 Hz takes the same horizontal distance as 1 to 2 kHz: both are one octave. A broad feature spans a range of frequencies; a narrow peak or notch is concentrated in a small region. This distinction matters more than how dramatic a feature looks at one zoom or scale.

Read a Comparison in Four Passes#

1. Identify the view. Read the measurement names, rigs, render mode, bounds selection, and compensation state. Confirm whether multiple lines mean products, channels, or individual readings.
2. Establish the baseline. Check normalization, Y-scale, and manual offsets. A higher trace is not proof of a more sensitive headphone or a louder listening level.
3. Compare broad regions. Follow each trace across bass, midrange, and treble. Ask where it is higher or lower than the other trace, not just whether it is close to a shaded area.
4. Investigate the detail. Use readouts, channel views, and percentile modes to determine whether a small feature is stable across the available readings. Return to the overview before making a general claim.

For example, if trace A is above trace B across much of 40-150 Hz after you have checked the settings, you can describe more measured bass relative to the chosen alignment. That does not establish a precise perceived-loudness difference, and it does not by itself prove which product you will prefer.

Inspect a Frequency With Readouts#

Enable the cursor/tooltips control in the Visualizer toolbar, then move the pointer across the graph. Use the displayed frequency and response values instead of estimating only from the grid. If the graph is crowded, hide unrelated measurements first.

When recording a numerical comparison, state the frequency, measurements, and relevant settings. For example: "At approximately 100 Hz, A is about 3 dB above B in this compensated, normalized view." This is a description of that view, not a claim that A is 3 dB louder overall. Do not report more precision than the readings and repeatability support.

Traces, Bands, and Reference Lines#

• A measurement trace represents measured response data or a derived average.
• A measurement percentile band describes spread among the included readings, such as repeated seatings.
• The preference bounds are a separate reference region used to contextualize a response; they are not that measurement's uncertainty band.
• The bounds center line is drawn only with the Old preference bounds and is not another headphone measurement. In the compensated view it is the 2023 reference slope (about -1 dB per octave, about 10 dB across the band); uncompensated, it is the middle of the rig's bounds. Measurements are aligned to the middle of the bounds (halfway between their edges), not to the slope, so a trace can sit slightly off the line by design. The New preference bounds and MIRE listener bounds have no center line.
• The legend identifies the rendered measurements, channel modes, and bounds reference.
• The GraphWorks watermark identifies the source of the visualization; it is not data.

The center line is part of the Old bounds display. Hiding preference bounds hides their shaded region and, with the Old bounds, their reference line together; there is no separate center-line visibility control.

What a Graph Can and Cannot Tell You#

Frequency response can help describe tonal balance, channel differences, and variation among measurements. It is not a complete assessment of sound quality, comfort, isolation, distortion, dynamics, or personal preference.

Measurements on an artificial fixture do not exactly reproduce every listener's head and ears. Seal, pads or tips, positioning, individual anatomy, and sample variation can change the result. Be particularly cautious about reading narrow high-frequency peaks and dips as exact predictions of what you will hear.

A curve within the preference bounds is not a guarantee that you will prefer that product; a curve outside them is not automatically a defective product. Use measurements alongside listening and the context supplied with the dataset.

Using the Visualizer#

The Visualizer brings together an item/measurement list, graph settings, the graph itself, and a header toolbar. On smaller windows the arrangement can change and sections may need scrolling.

Find Your Way Around#

On a phone or narrow window: scroll to locate the measurement list and graph settings if they are not alongside the plot. Select the desired dataset before looking for its line. Some settings remain disabled until a measurement is selected. Use the device's comfortable orientation and keep the comparison simple; a smaller display does not change the measurement values or export resolution.

Before closing a comparison you intend to revisit, copy or bookmark its current URL. Favorites are not a substitute for saving the visualization link.

Add and Remove Items#

Add item opens a product picker. Type a search term and choose an item. Its measurements appear in the sidebar. The same item is not added twice.

Use the remove-item button beside an item heading to remove that item from the comparison when the button is available. Removing an item from the Visualizer does not delete it from the database. To stop drawing an individual measurement while keeping its item in the list, use Hide measurement instead.

Select Measurements#

Click a measurement's name area or eye button to show or hide it. Dimmed entries are not drawn. A selected entry's color corresponds to its trace in modes that use per-measurement colors.

The measurement limit counts selected datasets, not visible channel lines. One dataset drawn as left and right still represents one selected measurement.

Manage Overlapping Traces#

First hide measurements you do not need. If the remaining traces overlap, use distinct colors in Single line, then adjust line thickness if necessary.

You can drag legend entries to change the measurement drawing order. In Difference, legend entries represent headphones, so dragging changes the order of their fields instead. This changes which response or field is drawn in front; it does not change the values or rearrange the measurement sidebar. Try it when a filled region obscures another dataset, but do not interpret the foreground trace as having higher priority or quality.

The top of the legend order draws in front. You can also focus a legend entry and use the arrow keys to move it. Collapsing an item's measurement group only makes the list more compact; use its measurements' eye controls when you want to change what is plotted.

Keep Headphones and IEMs Separate#

An active headphone comparison cannot add an IEM measurement alongside it, and an active IEM comparison cannot add a headphone measurement alongside it. Those categories use different reference data. Deselect the existing category before selecting the other category, or start a separate visualization.

If a measurement is already active, it can still be deselected. A category-lock message is not a paid-access limit.

Name the Visualization#

Signed-in users can edit the title and description at the top of the Visualizer. The initial title is generated from the selected products and rigs; once you edit it, your custom title is used.

These fields label the visualization and its image export. They do not rename the product or uploaded measurement. Finish editing the description by moving focus out of the field before exporting.

Visualizer Button Reference#

Hover an icon to read its tooltip. Some icons change to indicate the action available in the current state.

There is no need to match a particular browser zoom setting to export an image. The graph layout scales consistently with the display, and export resolution is selected separately.

Trace Render Modes#

Trace Render Mode determines how readings within each selected measurement are drawn. Switching mode does not change the uploaded data. A measurement may contain one or more readings per channel, including repeated positions or seatings.

Choose a Mode#

The word percentile here describes the distribution of measurement readings, not a product's preference score. The word Difference describes an envelope, not a subtraction curve. Neither mode ranks products.

Single Line#

Draws one average response per selected measurement by combining the available left and right readings. It is a good starting point for comparing overall tonal balance.

Response averages are calculated in the pressure-amplitude domain and then converted back to dB; they are not a simple arithmetic average of dB values. A single line can conceal channel mismatch and positional variation, so use the other modes when those are important.

Left & Right Lines#

Draws the left-channel average in blue and the right-channel average in red, when right-channel data exists. Each channel's repeated readings are averaged separately.

Use this mode to inspect channel matching. With many selected measurements, repeated blue/red pairs can be difficult to distinguish; use the legend and hide unnecessary datasets. A missing right line may mean the dataset contains only left-channel data.

Percentile#

Displays a measurement's seating/positional-variation bands:

• The lighter outer region spans the 5th to 95th percentiles, or approximately 90% inclusion.
• The darker inner region spans the 25th to 75th percentiles, or approximately 50% inclusion.

The calculation derives variation around each channel's own average, then combines the channel bounds. It is designed to show seating variation rather than treating a systematic left/right mismatch as additional positional spread.

These are descriptive percentiles of the available readings, not confidence intervals for all headphones of that model or for all listeners. With very few readings, the bands offer limited statistical information. A dataset with a single entry may render as a line rather than a visibly wide region.

Left & Right Percentile#

Shows percentile regions separately for each channel: blue for left, red for right. It uses the same outer 5th-95th and inner 25th-75th percentile structure.

Use it when you want to see each ear's positional variation without losing the separation between channel averages. Channel data must be present to produce a meaningful channel band.

Difference#

Draws an envelope across each headphone's selected measurements: the highest upper variation bound and the lowest lower variation bound at each frequency enclose that headphone's field.

Difference is not a conventional A-minus-B subtraction trace centered at zero. A field describes the combined range of one headphone's selected responses and their variation. Each headphone is rendered as its own field, using the color of its first selected measurement, with one legend entry, so measurements from two headphones show as two fields; the legend describes the derived result.

Use this mode to examine the extent of disagreement across a set of responses, not to read a signed numerical difference between two products. For a simple comparison, use two colored Single line traces and their readouts.

No Averaging#

Draws the individual left readings in blue and the individual right readings in red rather than collapsing them into averages.

This is useful for seeing the source spread and identifying unusual seatings. It can become busy quickly. No averaging is not the same as uncompensated: compensation and normalization remain separate settings.

Why a Mode May Be Unavailable#

The number of entries does not make a mode unavailable: Percentile and Difference take measurements with one entry and with more than one entry together. In Percentile, each measurement's bands come from its own readings only, so a single-entry measurement is drawn as its own line beside the others' bands and never changes their percentiles. Its legend chip ends in "Avg." rather than "50th & 90th % Inclusion".

Access level, missing channels, and dataset structure are separate considerations. Without an account, Left & Right Lines and Left & Right Percentile are listed with "(sign up)": choosing one offers a free account and keeps the current mode, and a shared link or embed that uses one is drawn as Single line or Percentile instead. The available menu is the practical guide to which modes you can select for the current combination.

Preference Bounds and Compensation#

Preference Bounds#

The Preference Bounds selector chooses the reference set shown behind the measurements and used by their target-relative alignment. Automatic (by rig) is the default: it uses Old bounds when any B&K 4128, GRAS KB006x or HMS II.3 measurement is visible; a single 5128 or RA0402 uses its rig-specific New bounds; and a 5128 + RA0402 comparison uses shared New bounds. IEM and MIRE data keep their existing bounds. You can manually select Old or New at any time.

The options, at every access level (a shared link keeps its Automatic, Old or New selection):

Changing bounds can change both the reference region and the normalization of a trace. Compare like with like: do not treat a curve aligned against one reference as directly identical to a curve aligned against another.

Use Bounds as Context, Not a Score#

Start with Automatic (by rig) for a new analysis. Choose Old Preference Bounds or New Preference Bounds when you want to override the rig-based selection. Keep the same choice when making a sequence of comparison images.

Read the bounds frequency by frequency: they provide reference context for the response, not a single pass/fail box for the product. Do not count the number of visible crossings as a quality rating. The bounds do not replace checking the actual measurements, variation, or listening conditions.

If shading makes a trace hard to follow, hide preference bounds with the header eye button and inspect the lines alone. This hides the reference display; it does not turn compensation off or remove normalization. Turn the bounds back on when discussing a response relative to that reference.

Compensated View#

Compensation removes the applicable measurement-rig reference response from the normalized measurement. This makes the trace easier to interpret relative to the compensated preference bounds and is especially useful when comparing supported rigs.

The compensation reference is not the same thing as the bounds center. Therefore 0 dB in compensated view does not necessarily mean the selected preference-bounds center at every frequency.

Raw / Uncompensated View#

Turning compensation off shows the response without that rig-reference subtraction. Normalization still applies, so this is not a promise to preserve the original recording's absolute SPL calibration.

Expect a substantial change in overall shape and level between raw and compensated views. Use the appropriate rig-specific reference and avoid interpreting a raw fixture response as a direct prediction of response at your own eardrum.

Mixed Rigs and Custom Compensation#

When selected measurements use different rigs, the Visualizer forces compensated display and disables the compensation toggle. This prevents a raw cross-rig comparison from being presented as if the rigs had a common baseline.

Some older measurements were uploaded already compensated to a custom HRTF. They are drawn as uploaded and also disable the compensation toggle. An unavailable toggle does not mean the eye/bounds display control is unavailable too.

Separate the Three Operations#

Think of compensation as changing the reference shape, and normalization and manual offset as changing vertical placement. Y-scale changes only how the result is framed. These controls answer different questions and should not be used interchangeably to make curves look similar.

MIRE Data#

MIRE is a distinct measurement type with its own reference and flat compensation treatment: GraphWorks does not subtract a conventional rig DF-HRTF from a MIRE trace. Selecting MIRE data chooses the listener's preference bounds automatically.

MIRE data is part of Paid Access: without it, MIRE measurements are listed but cannot be shown. MIRE measurements can be sent to GraphWorks EQ like rig measurements. The transferred curve keeps the same flat compensation and is read against the MIRE listener's Preference Bounds, whichever bounds version is selected for other rigs.

Normalization and Vertical Scale#

Normalization Center (Hz)#

This field selects the center frequency of the region used to vertically align a measurement to its target reference. Enter a frequency between 20 and 20,000 Hz, then press Enter or leave the field to apply it.

The interface rounds newly entered center values to the nearest 10 Hz. Invalid values are rejected and the last valid value is restored.

Normalization changes a constant level offset; it does not reshape the response, apply an EQ filter, or change playback volume. Both channels of a measurement share the computed offset, preserving their relative channel mismatch.

Octave Range#

Choose 0 through 5 octaves. The selected number describes the spread to either side of the center, rather than the total width across both sides.

For a center frequency f and a setting n, the nominal window extends from f / 2^n to f * 2^n, with the usable range constrained to the graph's frequency limits.

Examples at 630 Hz:

0 octaves aligns at the nearest available reading; the data does not need to contain an exact point at the chosen frequency. A shared or older graph can show Custom when its saved normalization range is not one of the standard choices.

A broad window can be more stable than aligning on one narrow peak or dip. Changing the window changes the comparison baseline, so keep it consistent when making before/after images.

Choose an Alignment Deliberately#

For an initial comparison, 630 Hz / 2 octaves gives a broad alignment region. It is not a universally correct reference for every product or question. If you change it, ask what part of the response you are treating as the baseline and explain that choice when sharing the result.

For example, matching two traces at a narrow bass frequency makes their differences elsewhere relative to that bass point. Matching over a broad midrange region can leave the bass difference more apparent. Neither operation changes the measured response shape; both can change the impression created by the overlay.

Try this exercise with two selected measurements:

1. Use Single line, fixed bounds, compensation on, and manual offsets of 0.
2. Set the center to 630 Hz and the range to 2 octaves. Note the relative bass and treble placement.
3. Change only Octave Range to 0 octaves. Observe whether the relative positions move.
4. Restore 2 octaves. The original alignment should return if the other settings are unchanged.

If a conclusion depends entirely on one narrow alignment point, describe that dependence rather than presenting it as a setting-independent result. Avoid placing a single-point alignment on an unusually sharp peak or dip without a specific reason.

Y-Scale (+/-dB)#

The displayed control is Y-Scale (±dB). It sets the vertical range above and below zero: 20 means -20 to +20 dB, not a total span of 20 dB. The interface offers a range from 15 to 60, in steps of 5.

Increasing the scale fits larger level differences but makes them look flatter. Decreasing it magnifies the apparent separation. It changes the display only; it does not change the measurements or normalization.

Browser zoom affects the size of the page, not this numeric dB range.

When comparing screenshots, check both graphs' scale labels. The same 3 dB separation occupies more vertical space on a ±15 dB plot than on a ±30 dB plot. Use a consistent scale for a series of images so readers do not mistake different framing for different measurement behavior.

Measurement Appearance#

Color#

Open the measurement row's Colour, offset and thickness control and use its color picker to choose the trace color. Pick distinct colors for a comparison. In Left & Right Lines, Left & Right Percentile, and No averaging, the fixed blue/red channel colors take precedence over the chosen per-measurement color.

Vertical Offset#

Open the measurement's appearance/settings panel. The up/down-arrow field sets a manual vertical offset in dB. A positive offset raises the measurement; a negative offset lowers it. Set it back to 0 to remove the manual adjustment.

The offset is applied with the automatic normalization baseline. It affects the selected measurement's channels together and does not rewrite its uploaded data. Always disclose intentional offsets in a comparison image if they affect its interpretation.

Line Thickness#

The pencil-marked field in Measurement settings changes trace thickness. Use a value of 1 or greater. Thickness is a styling parameter, not a measured level, and it is not necessarily a literal CSS-pixel width at every display/export scale.

Some derived band/envelope modes use their own edge widths and fills, so their appearance may not respond like a simple line. A thick line can hide small local differences; choose a readable value rather than the largest one.

Style Without Changing the Claim#

Use color, thickness, theme, and drawing order to improve legibility. Check the result in the theme you intend to export, because a color that is clear on a dark background may be difficult to see on a light one.

Keep manual offsets at 0 for an ordinary normalized overlay. Offsets can be useful for deliberately separating crowded responses, but the resulting vertical distance is no longer solely a measurement difference. State the applied offsets prominently when sharing such a graph. If the goal is simply to reveal an obscured trace, try hiding other measurements or changing legend order before shifting the data vertically.

Frequency Bands#

Use the swatch button on the graph to show or hide a strip labeling these broad regions:

These labels help orient you on the frequency axis. They are not filters, selections, or a diagnostic score. Ear gain refers to the frequency region associated with ear-related gain; it does not instruct you to flatten all measured energy in that region.

The strip is on by default. Its visibility is remembered by the browser and carried into image exports, but is not saved in the shared visualization URL. A recipient can therefore see a different strip state. If browser storage is blocked, the choice may last only for the current page. Toggling it does not change the measured response or the bounds values.

Downloading and Sharing Graphs#

Download an Image#

Image downloads are available to signed-in accounts, including Free Access, for the measurements that account can view.

1. Select the measurements you want to show.
2. Choose trace mode, bounds, compensation, normalization, scale, theme, colors, and frequency-strip visibility.
3. Set a useful title and description, and leave the description field to commit the edit.
4. Click Share visualization in the toolbar.
5. Choose Download Image (4K) or Download Image (1080p).
6. Wait for rendering to finish and retrieve the PNG from your browser's downloads.

The menu names refer to export resolution. The complete exported image is not 16:9: it includes a title/metadata area and trimmed side bands. The graph within the image uses a 16:9 layout.

The image includes the visualization title, measurement rigs, uploader attribution, graph, legend, and GraphWorks watermark. The description is included only when the title fits on one line; a two-line title occupies the description's space. Titles are limited to two displayed lines, and long description/metadata lines can be shortened with an ellipsis. Use concise labels and inspect the PNG before publishing it.

The output uses the current theme and frequency-band visibility. Hover tooltips are excluded. Export dimensions do not depend on browser zoom or matching your window to a particular aspect ratio.

The filename is GraphWorks - [title] (1080p).png or GraphWorks - [title] (4K).png, with unsafe filename characters removed and very long titles shortened. A blank title falls back to Untitled. Exporting makes an image; it does not change the database or create an EQ profile.

Check an Export Before Publishing#

Open the downloaded PNG and check the artifact itself, not just the on-screen graph:

• The intended measurements and only those measurements are represented in the legend.
• The title identifies the comparison without hiding an important description through wrapping.
• Colors, line widths, shading, and draw order remain readable at the size people will actually see.
• Bounds, compensation, normalization, scale, and intentional offsets match the claim in your accompanying text.
• Uploader/source attribution and axis labels remain visible and legible.
• No private information appears in the title, description, or accompanying screenshot.

Use the higher-resolution export when the image needs more detail or will be resized. A larger PNG does not increase measurement resolution or make the data more accurate. Prefer the built-in export over a screenshot when you want a clean graph without browser or account controls.

Share the Visualization URL#

Copy the browser's address while the visualization is open to share a link to its configuration. The URL can contain selected item/measurement identifiers, rendering choices, colors, and offsets, so check the resulting view before distributing it.

A link does not grant an access level: the recipient must have the access level needed for the selected data. It does let them view an unlisted measurement it names (see Access Levels); a private measurement remains restricted to its uploader and authorized GraphWorks staff. Login, access restrictions, unavailable datasets, or later measurement changes can affect what they see. A PNG preserves the rendered view at the time it was downloaded; a link is not an immutable archival copy.

Treat titles and descriptions included in a shared URL as shareable content. Do not put private information in them.

Save and Reopen a Comparison#

1. Finish selecting measurements and changing settings. If you edited a text field, leave it so the edit applies.
2. Copy the complete current browser URL, not just the product-page address or /visualizer without its query string.
3. Save it as a browser bookmark or in your own notes. Give it a name that identifies the products and purpose.
4. Reopen the saved link and check the selected measurements, render mode, bounds, alignment, and offsets.
5. When sending it to someone else, explain whether they need an account or Paid Access to reproduce the full view.

Use a bookmark to return to a comparison. Favoriting an item preserves only that item association, not the comparison configuration. Browser Back is navigation, not a reliable undo tool for individual graph-setting changes.

What a Link, Image, or Favorite Preserves#

Graph settings such as selection, mode, reference choice, alignment, colors, offsets, thickness, and drawing order are represented in the visualization configuration. Site theme and frequency-strip visibility should not be assumed to match for a recipient. If exact presentation matters, send a PNG alongside the interactive link.

Viewing Embedded Graphs#

You may encounter a GraphWorks graph embedded on another page. Embedded graphs follow the same access rules as the Visualizer: one measurement without an account, up to two with Free Access, and unlimited comparisons with Paid Access. MIRE data requires Paid Access, and private measurements remain restricted.

If an embedded comparison contains more data than your access allows, its header offers the relevant account or access option in a new tab. An embed does not grant access to the rest of a comparison simply because it appears on another website.

GraphWorks EQ#

This section covers the website's EQ download, activation, and measurement-transfer workflow. It is not a manual for every control inside the separate GraphWorks EQ desktop application.

Get the App#

Open GraphWorks EQ from the site's navigation or a locked EQ button. The download page explains the app, access requirements, supported platforms, and license agreement.

The download page lists:

• Windows: Windows 10 or later, 64-bit x64 PC.
• Mac: Apple silicon, M1 or newer, macOS 14 Sonoma or later. Intel Macs are not supported by the listed Mac build.

With qualifying access, choose Download for Windows or Download for Mac (Apple silicon). Read the linked GraphWorks EQ license agreement. Always check the download page for the latest platform requirements and version. Updates are delivered inside the app.

Install and open the application, then use its sign-in/activation process with the GraphWorks account that holds access. If that process opens the browser, complete the requested account authorization and return to the app.

Send a Measurement to the App#

1. Install, activate, and open GraphWorks EQ on the same computer as your browser.
2. Sign in to the website with your eligible account.
3. Find the measurement on its item page or in the Visualizer sidebar.
4. Click that measurement's EQ button.
5. Allow the browser to open the desktop app if prompted.
6. If the browser asks permission to connect to the app on your computer/local network, choose Allow if you intend to perform this transfer.
7. Wait for Sending curve..., then Sent to GraphWorks EQ.
8. Inspect the imported measurement and reference in the application before making playback adjustments.

The button transfers that measurement, not an image, not all selected measurements, and not the Visualizer's combined Difference envelope. It prepares a curve from supported channel data and includes measurement/source information. The Visualizer placement uses the applicable bounds version and normalization range; changing a line's color or thickness does not create an EQ filter.

The website handoff is not automatic optimization of your headphone. The app is where you inspect the imported curve and make processing adjustments. Begin listening at a conservative volume, especially when testing tones or gain increases.

Choose the Right Curve Before Sending#

The EQ control belongs to an individual measurement row. If two products are on the graph, use the button on the dataset you actually intend to import. Hiding other datasets can make that choice easier to verify, but a two-product comparison is not itself an EQ target or a filter preset.

Before transferring, check the product name, measurement conditions, rig, and channel availability. In the Visualizer, also check the bounds version and normalization window: those settings are relevant to the handoff. Color, line thickness, hidden preference shading, and the displayed Y-scale are presentation choices, not instructions for audio processing.

After Sent to GraphWorks EQ., verify the imported source and reference in the desktop app. Do not assume that a successful transfer means audio processing is already enabled, that filters have been chosen, or that a particular output device is configured. Those are desktop-app tasks outside this website guide.

EQ Button States#

Additional errors can report too few readings, no readings in the audible or normalization range, out-of-range values, or curve data the app cannot accept. Those are data/transfer problems, not instructions to buy another subscription.

Device Limit#

GraphWorks EQ can be active on up to three computers at a time. Settings > GraphWorks EQ devices lets you identify, rename, and sign out devices to free slots.

The device cap is separate from the website's simultaneous-measurement limit. Renaming a device helps identify it; signing it out revokes that activation when the app next renews. It does not uninstall the app or cancel the account's subscription.

Settings and Account Management#

Open Settings from the account menu. You must be signed in.

Account Details#

Use the editable Email and Username fields, then click Update account. Usernames are entered without @. An email change sends a verification message to your outgoing/current email inbox; follow the requested verification before assuming the new address is active. Do not discard access to the old inbox before completing the change.

Password#

Enter Current password and New password, then click Change password. Log out of all devices requests revocation of other website sessions when the password changes. Use Forgot password if you cannot provide the current password.

Website sessions and EQ device activations are managed through different controls; use the EQ device list for an EQ activation slot.

Membership and Billing#

The GraphWorks Access card shows your status and the actions appropriate to it. It can reflect paid access, linked Sonic+, pending payments, cancellations, or unavailable billing services.

• Get GraphWorks Access opens the subscription page when a purchase is available.
• Manage billing opens the card provider's billing portal, where the offered payment and subscription actions can be managed.
• Cancel subscription, if shown for a PayPal subscription, opens a confirmation dialog. Keep dismisses it; the confirmation cancels that subscription's future renewal.
• Manage on PayPal, if shown, opens PayPal's automatic-payments settings.
• Re-check membership refreshes an already-linked Sonic+ membership.

For a normal paid-period cancellation, access continues until the end of the period already paid for, as shown by the account's status. Cancellation is not an automatic refund. Read the displayed confirmation and applicable purchase terms.

If you have more than one access source, canceling one does not necessarily remove the other. For example, a linked active Sonic+ membership can continue granting access after canceling a redundant GraphWorks subscription. If the account reports two subscriptions, cancel the one you do not need rather than assuming the site automatically cancels it.

GraphWorks EQ Devices#

The device list shows name, status, platform, and last-used time.

A signed-out device learns it must authenticate again at its next renewal; this is not an immediate remote shutdown of the application.

Delete Your Account#

Delete account is permanent and is not the same as Log out or Cancel subscription.

1. Resolve subscriptions or unfinished checkouts that could still charge the account, following the site's instructions.
2. Choose Delete account in Account settings.
3. Enter your password and submit the deletion form.
4. Read and accept the final confirmation only if you intend permanent deletion.

A wrong password or unresolved billing state can prevent deletion. If the site refuses, read the explanation and resolve it before retrying. Successful deletion ends the account's session and opens the account-deleted page. Contact support first if you are unsure about billing or data consequences.

Practical Workflows#

Compare Tonal Balance#

Use two comparable measurements in Single line, with the same compensation/bounds selection and normalization window. Assign distinct colors. Look for broad differences across bass, midrange, and treble before focusing on narrow peaks. Re-check the result in Left & Right Lines if channel matching matters.

For a repeatable comparison:

1. State the question: for example, "How does the bass balance of A compare with B?"
2. Select one relevant measurement for each product, preferably on the same rig. Hide the rest.
3. Apply the starting configuration, checking for nonzero manual offsets inherited from a shared link.
4. Identify both colors in the legend and follow each response across the region of interest.
5. Inspect channels and variation before treating a small separation as a stable difference.
6. Save the URL and, if useful, export a PNG. Include the rig, bounds, and alignment when describing the result.

A careful conclusion is "A has more measured bass relative to this alignment in these datasets," not "A is better" or "every listener will hear exactly this difference."

Check Channel Matching#

1. Select only the dataset you want to investigate.
2. Choose Left & Right Lines. Confirm that blue is left and red is right.
3. Look for regions where the channel averages separate. Check both broad differences and localized features.
4. If repeated readings are available, switch to Left & Right Percentile, then No averaging. Ask whether the difference persists across readings or is dominated by one placement.
5. Repeat with another suitable dataset if available, keeping the analysis settings consistent.

Both channels share the automatic normalization offset. Their visible separation is not caused by independently aligning the two channel averages. However, positioning, seal, and the measurement process can still influence channel differences; a graph alone does not diagnose a defective unit.

Examine Fit or Seating Variation#

Start with No averaging to see individual readings, then use Left & Right Percentile where available to distinguish each channel's spread. Wider variation can indicate sensitivity to position or seal, but the dataset alone does not establish its cause.

1. Select a dataset with repeated readings and hide unrelated measurements.
2. Use No averaging to see whether the captures closely overlap or diverge in particular regions.
3. Switch to Percentile for a summary, or Left & Right Percentile to keep channels separate.
4. Distinguish the measurement's colored variation bands from the preference-bounds background. Temporarily hide preference bounds if necessary.
5. Note the regions where the spread widens. Avoid inferring a cause that is not documented in the measurement conditions.
6. Compare a second dataset only after you understand the first. More captures or different capture conditions can change the apparent spread.

For example, consistent low-frequency variation across seatings may justify investigating seal or fit, but the graph alone cannot establish which placement matches your own head or ears. A very narrow band is evidence about those available captures, not a guarantee of identical performance for every listener.

Compare Measurement Rigs#

Use compensated display and the applicable references. GraphWorks forces compensation for mixed-rig selections. Check the metadata and do not assume compensation removes every difference in fixture geometry, sample, measurement method, or placement.

1. Find measurements of the same product on the rigs you want to compare.
2. Select Single line, use a single bounds version, set a common normalization window, and remove manual offsets.
3. Confirm that compensated display is active. A disabled compensation toggle is expected for a mixed-rig graph.
4. Start with broad similarities and differences, then inspect variation within each dataset before attributing a narrow feature to the rig.
5. Label the comparison with the rig names and any known sample or condition differences.

Agreement between compensated responses is useful context, not proof that raw outputs are interchangeable. If you want to study an uncompensated response, inspect the rig separately where the compensation control allows it.

Compare Conditions on the Same Product#

Use this workflow for pad, tip, or other condition comparisons only when the available measurement labels actually identify those conditions.

1. Open the product and select the two relevant condition datasets.
2. Prefer the same rig, and check whether the dataset describes the same sample and measurement procedure.
3. Use Single line, distinct colors, identical bounds/alignment settings, and zero manual offsets.
4. Describe the measured change across the affected frequency region.
5. Inspect channel and seating variation to see whether the difference is large relative to the available spread.

Do not attribute every difference to the named condition if other factors also changed. If only one condition has been measured, the site cannot supply the missing comparison by changing render mode or reference bounds.

Prepare an Image for a Review or Discussion#

Choose only the measurements needed, keep colors readable, use a consistent vertical scale, and write a title identifying the products or question. Put important conditions or intentional offsets in the description. Export a PNG with the legend and source attribution intact. Mention the trace mode and normalization if they affect the claim you are making.

For an accompanying caption, use a compact structure such as: [Products/conditions], [rigs], Single line, compensated, New Preference Bounds, 630 Hz / 2-octave normalization, ±20 dB scale, manual offsets 0 dB. Add the dataset/uploader names when they are not already unambiguous in the image. This records the view without turning the caption into a claim about every listener's experience.

Move From a Graph to EQ#

Choose a supported measurement representative of the product and conditions you intend to use. Inspect its channel data and variation first. Send it with EQ, then verify the import in the desktop application and make conservative listening-based adjustments. A measurement is a starting point, not a guarantee of the response on your own head.

Reproduce a Graph Someone Else Shared#

1. Open the full visualization link rather than searching for a similar-looking measurement by name.
2. Sign in if the comparison needs more than the anonymous view. Check whether any selected data requires Paid Access.
3. Confirm the measurement identities, render mode, bounds version, compensation, alignment, Y-scale, and offsets.
4. If matching an image, also match the theme and frequency-strip state; those may differ between browsers.
5. If a measurement is missing, consider access or visibility restrictions before substituting another dataset.

If you have only an image, you may not have enough information to reproduce the original graph exactly. Ask for the visualization URL and the relevant settings instead of inferring them from the line shape.

Recover From an Overcrowded or Confusing Graph#

1. Save the current URL if you may want to return to it.
2. Hide all but one or two relevant measurements.
3. Return to Single line and check the legend.
4. Restore a known bounds/alignment setup and set manual offsets to 0.
5. Choose readable colors and a common Y-scale.
6. Add complexity back one step at a time: channels, variation, then additional datasets if your access allows them.

Change one setting at a time when investigating unexpected behavior. If you change the mode, bounds, alignment, and offsets together, it becomes difficult to identify which change explains the result.

Troubleshooting#

Work Through Problems in This Order#

1. Selection: is the intended measurement actually active, or have you only added its item?
2. Eligibility: does your account allow the selected count and data type, and are the measurement categories compatible?
3. Presentation: are mode, bounds visibility, compensation, scale, and offsets appropriate for what you expect to see?
4. Reproduction: can you reproduce the issue with one measurement and a known configuration? Save the affected URL before refreshing or simplifying it.
5. Support: if the problem remains, report the exact steps and settings rather than only saying the graph looks wrong.

Do not use another purchase, account deletion, or repeated checkout as a general troubleshooting step. Check the account and the specific error first.

Glossary#

Support#

Use Have feedback? (on a phone, Feedback in the footer) or email info@headphones.com.

For a useful report, include the product and measurement names, rig, page/visualization URL, selected trace mode and bounds, compensation state, normalization/offset settings, browser and operating system, and the steps that caused the problem. If appearance is the issue, include browser zoom and window size. An image can help, provided it contains no private account or payment information.

You can structure a report like this:

Page / visualization URL:
Product, measurement name, rig, and uploader:
Account access level (no password or payment details):
Render mode, bounds, and compensation:
Normalization center / octave range, Y-scale, and manual offsets:
Steps to reproduce:
Expected result:
Actual result / exact error message:
Browser, operating system, and window size / zoom if relevant:

Remove private information from links, text, and images before sending. For billing or account issues, use the support channel rather than posting account details in a public discussion.

Never send passwords, verification codes, reset links, payment-card details, or authentication tokens.