Symmetrical double sided welds
When the weld is symmetrical about the joint with identical symbols and dimensions on both sides, System A deletes the dashed line. The two symbols sit opposite each other on a single solid line.
Fabrication reference
How to read a welding symbol under ISO 2553 and AWS A2.4, written by a working steel fabricator rather than assembled from other charts. It starts with the side convention, because that is the one that puts welds on the wrong face of the plate and costs real money.
The one that catches people
In AWS A2.4, below the line means arrow side. In ISO 2553 System A, position means nothing: the solid line is arrow side and the dashed line is the other side, and the dashed line is preferably drawn below. Same position, opposite meaning.
Both standards answer the same question, which side of the joint gets the weld, and they answer it in incompatible ways. Get this wrong and the fabrication is wrong, not just the paperwork.
| What you see | AWS A2.4 | ISO 2553 System A |
|---|---|---|
| Symbol below the reference line | Arrow side | Depends entirely on which line it sits on |
| Symbol above the reference line | Other side | Depends entirely on which line it sits on |
| Symbol on the solid continuous line | Not a distinction AWS makes | Arrow side |
| Symbol on the dashed line | Not a distinction AWS makes | Other side |
| Symbols on both | Weld both sides | Weld both sides |
Now the part that does the damage. In System A the dashed line may be drawn above or below the continuous line, and the standard states a preference for drawing it below. That is precisely where AWS A2.4 says arrow side. So a fabricator trained on AWS, looking at a System A drawing, seeing a fillet symbol on the lower line and welding the arrow side, has welded the wrong face. This is not an obscure edge case. It is the default presentation of both standards colliding head on.
One more distinction worth being exact about, because it is a separate mistake: arrow side and other side are not near side and far side. They are defined relative to the joint the arrow points at, not to the person reading the drawing. On something like a flare bevel in an angle heel, the other side can be the inaccessible interior rather than the opposite leg.
The dashed line is not always there, and its absence does not mean the drawing is AWS. ISO 2553 drops it in three situations.
When the weld is symmetrical about the joint with identical symbols and dimensions on both sides, System A deletes the dashed line. The two symbols sit opposite each other on a single solid line.
Arrow, reference line and tail only, with no weld detail. No dual reference line is needed because no weld information is being carried. The standard notes this form is often used to show where tack welds go.
Welds made at the interface between two components. This one is corroborated from secondary sources rather than read directly in the standard, so treat it as indicative and check a licensed copy.
The trap that follows from the first case: a symmetrical double fillet on a System A drawing has no dashed line and looks identical to an AWS symbol. You cannot identify the system from that one symbol. Identify it from the title block first, then from the other symbols on the same sheet.
Systems A and B were introduced in ISO 2553:2013, the fourth edition, which replaced ISO 2553:1992. The current edition is ISO 2553:2019, the fifth, which kept both systems and made mostly editorial and clarification changes. You will see it written that the 2019 edition introduced the two systems. It did not.
Dual reference line: one continuous, one dashed. The side is shown by which line carries the symbol. Descended from ISO 2553:1992, and the system you will meet on UK and European drawings.
Single reference line, side shown by position, the same mechanism AWS uses. But ISO describes System B as based on standards used by Pacific Rim countries, not on AWS A2.4, so calling it the AWS system is a near miss a working engineer will spot.
ISO 2553 clause 4.3 requires that the systems are not mixed and that a drawing clearly indicates which one it uses, along with its units. If a drawing does not say, it is non compliant, and the right response is a technical query rather than a guess.
The difference reaches further than the reference line, which is worth knowing before you assume a symbol means the same thing in both. In the elementary symbol table of ISO 2553, the entries covering plug, resistance spot, projection, fusion spot, resistance seam, fusion seam and stud welds are split into separate System A and System B columns. And one supplementary symbol, specified root reinforcement on butt welds, carries a second meaning in System B. Which individual glyphs differ needs a licensed copy to state precisely, and this page will not guess at them.
A cause almost nobody writes about: older CAD releases followed ISO 2553:1992, which has no Systems A and B, so they could only produce System A symbols. A drawing can therefore be in System A because the software could not do anything else, with nothing in the title block saying so.
Title block first, always. If that does not settle it, these are the practical tells. Treat them as what to look for rather than as rules from the standards.
| What you see | What it points to |
|---|---|
| A dashed line parallel to the reference line | ISO 2553 System A. The dashed line exists in no other system. |
| Fillet size written with an a or z prefix, such as a6 or z8 | ISO 2553, either system. AWS has no such prefix. |
| Fillet size as a bare number, or two numbers for unequal legs | AWS A2.4. |
| A numeric process reference in the tail, such as 135 or 111 | ISO, per ISO 4063. |
| Letter process codes in the tail, such as GMAW or SMAW | AWS, per its own process designations. |
| Intermittent weld written as n x l (e) | ISO. |
| Intermittent weld written as length then pitch, such as 2-6 | AWS. |
| Inch fractions and decimal inches throughout | AWS, almost certainly. |
| Title block citing BS EN ISO 2553, EN 1090 or ISO 5817 | ISO System A, almost certainly, on UK and European work. |
UK note worth knowing on older work: BS 499, the British standard that came before the ISO and EN versions, placed symbols above or below a solid line the way AWS does. The dashed line convention arrived with the ISO and EN standards. Genuinely old UK drawings still in circulation may therefore use the position convention. Confirm against a copy before relying on it.
a is the nominal throat thickness, the height of the largest isosceles triangle that fits inside the weld section. z is the leg length, measured across the fusion face. On an equal leg fillet they are related by the square root of two.
| Measure | a6 as specified | z6 as built |
|---|---|---|
| Leg length | 8.49 mm | 6.00 mm |
| Throat thickness | 6.00 mm | 4.24 mm, about 29 percent short |
| Weld cross section | 36.0 mm squared | 18.0 mm squared, exactly half |
Most sources describe this as roughly 30 percent undersized, which is true of the throat and badly understates the loss. The section is exactly half, because area goes as the leg squared over two and the square root of two squared is two. Half the weld metal is the number an estimator and a welder will both remember.
It costs money in the other direction too. Reading z6 as a6 and building an 8.5 mm leg is a 100 percent overweld: double the metal, double the arc time, double the distortion, on a job priced at half that.
The UK angle that makes this worse rather than better: shop practice has traditionally sized fillets by leg length, while European practice and BS EN 1993-1-8 work in throat. So on a UK job the shop floor thinks in legs, the Eurocode designs in throats, and a single letter on the drawing is the only thing arbitrating between them. The letter is mandatory on an ISO drawing for exactly that reason, and a bare number in front of a fillet symbol is an incomplete callout, not a leg length by default.
Worth knowing that AWS took this seriously enough to include an informative annex on ISO 2553 in AWS A2.4:2020, containing a dedicated figure comparing a fillet weld with the a and z modifiers. If a standards body publishes a figure specifically to stop this confusion, it is not a minor point.
A missing dimension is not always a missing instruction. Sometimes it is a defined default, and the defaults are not the same on both sides.
| Situation | ISO 2553 | AWS A2.4 |
|---|---|---|
| Butt weld with no size given | Full penetration, unless dimensions or a referenced procedure say otherwise | Complete joint penetration for single groove welds and for double groove welds with symmetrical geometry |
| Fillet weld with no size given | No default size. The drawing is incomplete | No default size. The drawing is incomplete |
| No length given after the symbol | Weld continuous over the whole length of the joint | Continuous unless the callout says otherwise |
| Intermittent weld spacing | n x l (e), and see the note below on what the bracket means | Length then pitch, and AWS pitch is center to center |
On intermittent welds, sources genuinely disagree about whether the ISO bracketed figure is the clear gap between elements or the center to center pitch. This page is not going to pick one and sound confident about it. What works either way is the arithmetic: if the bracketed number is smaller than the element length, it has to be the clear gap, because otherwise the welds would overlap. So 4 x 50 (30) means four 50 mm welds with 30 mm gaps, an overall run of 290 mm and an effective pitch of 80 mm. If the bracketed number is larger than the element length, the callout is ambiguous on its face and is worth an RFI rather than a guess.
A worked example doing the rounds online claims a 50 mm weld at 70 mm centers leaves 10 mm clear between welds. It leaves 20 mm. That arithmetic error has been copied between sites, which is a fair warning about how much of this material is transcribed rather than checked.
You will find pages claiming the American Welding Society attributes 80 or 85 percent of design errors to misread drawing symbols, or that symbol errors cost firms tens of thousands a year. We went looking for the source of those numbers and could not find one. They contradict each other, they trace to no AWS publication we could locate, and they are not repeated here.
We also could not find a publicly documented incident where a misread symbol caused a named failure, so there is not one on this page either. What does exist is consistent trade press and practitioner reporting on the patterns that recur.
Because groove welds default to full penetration, designers assume fillets have a default too. They do not. It shows up most at joints with no dimensional constraint, like angle heels and pipe column base plates.
Flare bevel and flare V groove welds are necessarily partial penetration, so the effective throat has to be stated in parentheses. The radius on its own is not the weld capacity.
Used at complex three dimensional joints, full depth stiffeners and HSS connections, it can leave insufficient throat at the transitions and it hides fitup from inspection.
Geometry reduces the effective throat, so these need the throat specified rather than the leg. Sizing by leg on a skewed joint quietly undersizes the weld.
All around, field weld, contour and finish symbols are visually small and get read past. Each one changes what you have to do or where you have to do it.
The symbol is one input. Drawing notes, joint detail, material spec and tolerances are the rest of it. Most people were never formally taught to read symbols and learned from old drawings, which is how habits spread.
Welding symbols FAQ
Each answer is short enough to use at the bench, with the reasoning above if you need to defend it.
AWS A2.4 encodes the side by position: a symbol below the reference line means arrow side, above means other side. ISO 2553 System A encodes it by line type instead: the symbol sits on the solid line for arrow side and on the dashed line for the other side, and the dashed line can be drawn above or below. So position tells you nothing on a System A drawing, and reading one by AWS habits can put the weld on the wrong face.
Yes. In AWS A2.4 the arrow side is the side of the joint the arrow points to, and arrow side information goes below the reference line while other side information goes above it. That stays true no matter which way the arrow comes into the line. Symbols both above and below mean weld both sides.
It identifies the other side of the joint, the side the arrow is not pointing at. It only exists in ISO 2553 System A, always alongside a solid continuous line. If you see a dashed line parallel to the reference line anywhere on a drawing, you are looking at System A.
Yes, it may be drawn either above or below the continuous line, and the standard states a preference for below. That is the crux of the whole problem: the preferred ISO layout puts the other side symbol in exactly the position where AWS puts arrow side. Never read a System A drawing by position.
System A uses a dual reference line, one continuous and one dashed, and the side is shown by which line the symbol sits on. System B uses a single reference line and shows the side by position, the way AWS A2.4 does. ISO describes System B as based on standards used by Pacific Rim countries, so it is not accurate to call it the AWS system. The two systems must not be mixed on one drawing.
The drawing is required to tell you: ISO 2553 says the system in use must be clearly indicated, so start at the title block. After that, the tells are a dashed reference line for System A, an a or z prefix on a fillet size for ISO, numeric process codes in the tail for ISO against letter codes like GMAW for AWS, and inch fractions for AWS. If nothing on the drawing says, that is a technical query for the designer, not a guess.
a is the nominal throat thickness, the height of the largest triangle that fits inside the weld section. z is the leg length, measured along the fusion face. On an equal leg fillet, a equals z multiplied by 0.7071, and z equals a multiplied by 1.4142. The letter is mandatory on an ISO drawing, so a bare number in front of a fillet symbol is an incomplete callout.
You lose exactly half the weld metal. a6 requires an 8.49 mm leg. Building a 6 mm leg instead delivers a 4.24 mm throat against the 6 mm required, so the throat is about 29 percent short, and because fillet area goes as the leg squared over two, the deposited section is exactly half. Read it the other way, building z6 as a6, and you have doubled the weld metal, arc time and distortion on a job you priced at half.
Under ISO 2553, yes: a butt weld is full penetration unless dimensions or a referenced procedure say otherwise. Under AWS the default applies to single groove welds and to double groove welds with symmetrical geometry, so the qualifier matters. Fillet welds have no default size under either standard, so an undimensioned fillet symbol is an incomplete drawing rather than a standard size.
In ISO notation an intermittent weld is written as number of elements, element length and a bracketed spacing figure. Sources genuinely disagree on whether that bracketed figure is the clear gap or the center-to-center pitch. The reliable check: if the bracketed number is smaller than the element length it has to be the clear gap, because otherwise the welds would overlap. If it is larger, the callout is ambiguous and worth an RFI. In AWS the notation is length then pitch, and AWS pitch is center-to-center.
Written from the published standards and from trade practice, by a working UK steel fabricator. ISO 2553 and AWS A2.4 are copyrighted documents, so nothing here reproduces a table or a figure from either. What this page does instead is explain the system in original words and point you at the clause to look up in a licensed copy.
Where the sources disagree, or where something could only be corroborated from secondary material, that is said in the text rather than smoothed over. For anything you are going to weld to, work from the controlled drawing and the applicable standard, not from a web page.