Lower & Longer # 2 : October 29, 2022

Reduced air Volume - Repeated

To place the two ‘Low Air’ experiments in overall context, the typical smelt :
- uses 22 - 25 kg ore analog
- uses air input in the range of 800 + LpM (in 25 cm ID furnace)
- charcoal consumption rates in the range of +/- 8 minutes per kg
- expected bloom weight roughly 5 - 6 kg
- production yields roughly 20 - 25 %

    Recently, a higher quality Omega HHF 1001 'hot wire' style air volume meter had been purchased by research fellow Neil Peterson. To determine the actual air volumes produced by the previously constructed ‘smelter’ bellows unit during the October 2021 test, both the normal electric blower and the bellows were paired to the air inlet piping. The rough average volumes measured using those bellows was just over 500 Litres per Minute.
    This smelter bellows unit had been built in 2008, as part of the ‘Vinland’ series. It has been used for a number of past smelts, where either remote locations or a desire for more historic equipment has required human powered air.


Air system set up, blower on left, bellows to rear, as used October 2021.

    Part of the intent for the June 2022 (# 91 - LAM, Lower & Longer) smelt was to reduce the air volumes down to be in line with the measurements made during the October 2021 test (# 90 - Wind and Weathering).


15:00 Schweitzer 6 541 50%
15:06 Peterson 5 491 42%
15:11 Markewitz 1 477 8%



17:06 Schweitzer 4 549 57%
12:10 Peterson 2 559 29%
17:12 Markewitz 1 542 14%



19:44 Schweitzer 2 532 50%
19:46 Peterson 1 487 25%
19:47 Markewitz 1 501 25%



Overall  average


# 90 - Table 2-1 : Bellows Air Delivery, October 2021
    The June bloom yield was significantly higher than was expected - at 28 %. To that end, the October 2022 smelt was a repeat of June, using the same furnace (with repairs), duplicating set up, ore, and process (as closely as possible).
    The furnace built for the June experiment was basically in good condition, with only two areas needing some repairs. The material gathered from Norstead at L’Anse aux Meadows was found to be baked to a loose sandy consistency, so that rectangular section was replaced with the same clay mix that made up the rest of the furnace. Some cracking was expected at the boundary lines, as this fresh addition was exposed to typical smelting temperatures.
    There had been the usual damage to the front extraction area of the furnace when the slag mass was pulled clear previously. A number of smaller solid clay construction bricks were placed to duplicate the existing plinth structure, with one brick used to fill the space of the original extraction arch. The result was an arch 11 cm tall by 22 cm wide. There was no tapping arch made, as it was felt the various gaps between the random selection of bricks would serve for this purpose.

    The copper tuyere was placed back into the same location as previous, the only significant difference was the insert distance was increased to 6 cm (from 5 cm proud used last time), and the base depth (distance below tuyere) increased to 20 cm (from 16 + cm). Fresh clay was added as required to seal the tuyere in place.

after 10-21
10-22 build
Damage after October 2021
 As rebuilt for October 2022 (early sunlight)

The air inlet system, other than the tuyere and associated fittings, was slightly different than that used in June. The connections this time were made via corrugated plastic sump pump hose with a minimum Internal Diameter of 3.2 cm ( 1 1/4”), replacing the smooth surfaced flexible metal pipe that had an ID of 5 cm. Although the new arrangement used two shorter pieces, the total length of the hosing was about the same as in June.

gauge set up
Air delivery with gauges in place (pipe diameters given)

    In practice, the air flow from the electric blower was set using the same sliding plate blast gate that has been employed for the blower source experiments since (at least) 2008. There is a roughly calibrated scale on the plate, marked to increments of 100 LpM, originally scribed using measurements via a pocket wind speed anemometer. Given the scale of the marks, the best accuracy that could be expected here would be roughly 15 %, which has not been considered critical in past experiments (primarily since this variation is the same in all cases).
    Also included in the instrumentation was a vane type anemometer (Mastech MS 6252B), also provided by Neil. This unit is really intended for measuring air flow inside heating / cooling ducts, and when confined to a small diameter fitting it's readings have been found not to conform to the results from other equipment. Also included was a fitting connected to a (much older) dial style direct pressure gauge, but this unit is intended for much higher pressure automotive applications, so barely registers in this system.

The Smelt :

    The ore used here was basically the same standard DD1 analog mix that was used in June. (1) A measurement of of water content was made for the October batch, with a 100 gm sample baked for 5 minutes on the top of the furnace to drive out any remaining water. The ‘bone dry’ weight was 88 gms, a water content of 12 %. This compares with past measurements, were most typically the water content has ranged between 8 - 10 % with the same test method. Although not tested for, it is most likely that the ore batch used for the June smelt may have had an even lower water content, as it had been sun dried over at least a full week, with daily conditions bright and sunny with temperatures into the range of 30 C. (See the discussion below of moisture content against the top of the furnace.)

Smelt Team:    
Darrell Markewitz - smelt master, extraction
Neil Peterson - lead hand, recording, consolidation
Ryan Kuntze - charcoal, consolidation
Ethan Khan Hafner - ore, consolidation
Max Hafner - materials processing

ore braking
first charcoal
(L-R) Ryan, Ethan, Max - breaking sun dried ore into suitable sized pieces.
First addition of rough charcoal, showing the working area. Instrumentation (not connected yet) in the foreground, blower to the extreme right. (image by Lezli Kuntze)

    The start time through preheat phase was a similar duration, although with a second use the furnace walls would have been baked dry to partially sintered. Once again the primary charcoal used was oak. As ore additions were made, the same sequence of amounts was used as done with the previous experiment (groups of three, at 1 / 1.5 / 2 / 3 kg each charcoal measure).
    It was obvious from the beginning however that the individual burn rates (time for standard measure) were longer than the June smelt. (Go to the sequence data) As calculated over the period of identical ore additions / total amount of 24.5 kg :
        total time        average burn - kg        ore average - kg

June       295 min.        12 min.                11.8 min.

October    391 min.        18.1 min.               16.0 min.

    In comparison, the ‘ideal’ burn rate is considered to be roughly 6 - 8 minutes per kg.

    Although there were two new assistants for this smelt (Ryan and Ethan) their activity was with breaking charcoal and ore, then during the smelt itself making the required ore and charcoal additions, under experienced guidance.
    Two initial ‘seed’ charges, each one kg, of an uncertain quality Lexington Limonite ore were made as was also done in June. These charges were in place of the usual amount of iron rich slag that are added to accelerate the creation of a working slag bowl system, a standard method for this team. (2)

ore add charcoal add
Ryan adding ore with Max observing, near the end of the sequence. (image by Lezli Kuntze)
Dramatic image of charcoal addition by Ethan (image by Lezli Kuntze)

    As was the case in the last test, and as expected with lower volume air, the slag bowl was initially observed to form higher than ideal in relation to the tuyere, potentially creating a problem with rising slag obscuring the air blast. Rather than tapping off accumulating slag, a steel rod was driven through the rear portion of the slag bowl from the top of the furnace, allowing excess slag to drain through into the charcoal fines beneath. This process was done twice, at 1430 (2 kg seed + 3 kg ore added) and at 1606 (2 kg seed + 7.5 kg ore added). Later in the smelt, the furnace would self tap (‘incontinent’) several times. Generally this smelt ran pretty much as normal, beyond the significantly increased (and expected) burn rates.
    It was decided to leave the extraction process in experienced hands. As was done in June, the extraction arch and lower plinth below was opened, then the remaining charcoal fines scrapped clear to expose the bottom of the slag bowl. Again, the bowl was firmly attached to the lower furnace walls. To break this clear, the log ‘thumper’ was used to pound the bloom down with the encasing slag bowl to break this all clear, The loose mass was then pulled free out the arch, which did result in more slag bowl remaining attached around the bloom as it was lifted to the compaction stump.
    Because of the greatly increased time of the main smelt sequence ( over 6 1/2 hours other than the more typical 3 - 4) and short November daylight, it was fully dark by the time extraction was undertaken. With first time workers Ryan and Ethan using unfamiliar sledges, breaking off encrusting slag and initial bloom compaction was not as efficient as might be desired. Working in darkness can be difficult for even experienced workers, and certainly any estimation of temperature via heat colour was distorted.

Compaction - short video by Max Hafner

    After initial clearing of slag, the bloom mass was rushed back to the workshop, and further compression was attempted using the hydraulic press. In truth by this point the metal was far too cold, certainly for true compacting, even through simple mechanical adhesion. The bloom had a very ‘crumbly’ texture as well, so the net result was considerable fracturing. It was decided to break up and cut the mass into several pieces, primarily so both Ryan and Ethan could have smaller sections to keep as a reward for their efforts.
    With the resulting pieces and any fragments larger than ‘half walnut’ (collected magnetically), the total bloom weight was 5 kg, compared to the analog ore input of 24.5 kg, a final yield of 20 %. Using a quick spark test, the metal showed some slight carbon content, estimated to be about half way between carbon free iron and mild steel (at 0.2% C), so potentially about 0.1 % C. (3)

Air Flow Measurements :

    Measurements of air flow were taken at a number of points over the main smelting sequence. This unlike experiment # 90, when continuous readings were inputted directly into a lap top computer. One continuing problem is converting the gauge readings (with instruments intended for other purposes) into our desired units of Litres per Minute :



1 kg ore in stack


about 3 kg ore in stack

about 4 kg ore in stack

    The previously measured plate setting was for 500 LpM (considered to be at best within 15 %), the set of measurements from the Omega unit are considered to be most likely. It would be expected that air flow would be reduced as more and more ore is added into the same stack height of burning charcoal.

Discussion :

    The past work of this team has been focused on working systems based primarily on Viking Age bloomery iron smelting. How close to these original processes it is possible to match is effected by the limits of archaeological prototypes, and significantly by differences in the materials that are available due to the physical location in Southern Ontario, Canada.

    Over the years there have been a number of smelts using hand powered bellows based on possible Norse design : an early oversized ‘Ubber-Bellows’, the more manageable ‘smelter’ bellows, and configurations using ‘blacksmith’ sized units :

bellows use

table A - bellows air : as html

    In so much as many of these smelts were carried out at various ‘remote’ locations (with less control over environment and general support), it might be expected that the overall results might also be somewhat random :
- The furnaces themselves are fairly consistent in size and overall layout, all of the ‘Norse Short Shaft’ type, almost all at 25 cm interior diameter and typically 40 + cm of working stack height.
- There is some variation in ore iron concentration, but the bulk of tests are using variations on the red iron oxide (Fe2O3) analog developed and the primary material used after 2008, with roughly 53 % Fe most typical.
- The total ore amounts used show more variation, but most smelts added between 20 - 28 kg, with the average (discounting the first three low volume tests and the single limited test) at 23 kg.
- When considering air delivery via a pumped bellows, produced volume will vary considerably both by individual operators (differences in physical abilities), and also over the long duration of a smelt (through simple fatigue).

    It is important to note the estimated (E) value for so many of the LpM volumes indicated. This is an amount estimated from the figures measured during the October 2021 test, placed in retrospect (there were no actual measurements made at the indicated experiments). This limits the amount of the indicated litres per square centimetre (at tuyere cross section) figure. With bellows use, this could be expected to be somewhat lower than the ‘ideal’ figure of 1.2 - 1.5 L/cm2 suggested by Sauder & Williams. (4) normally sought in other smelts.

    With so many of the other variables (furnace size and layout, ore content and amounts) being fairly consistent, it is hard not to attribute the large variation in ‘addition rate’ (time in minutes for one kilogram of ore, as an average over the entire smelt sequence) to differences in air delivery *method*, rather than just volume, over the duration of an individual smelts listed above.

For experiment # 92, air volume measurements were made at periodic points over the smelt, with readings from the two instruments.

Weather Impact ?

- On June 25, it was bright and sunny, and extremely hot at 20 C at the start, plus 30 C by mid day, falling to about 23 C by the end of the smelt
- On October 29, it was bright and sunny, and ‘seasonal’ at 0 C at the start, 14 C by mid day, falling to about 6 C by the end of the smelt.

    It has been seen however, that other smelts undertaken out of doors without any cover have not illustrated any particular impact of environmental conditions on to the quality of input air :
- Smelt # 14 (February 2006) was at Traverse City Michigan when day time temperatures only reached -15 C, with 29 kg of taconite ore into a 13.6 kg bloom (47% yield). A number of smelts have been undertaken with air temperatures into the 35 C range (June / July), down to the 5 C range (March / November) , with no observable impact on burn rates or resulting yields. It should be noted that compared to the internal operating temperature of these furnaces at plus 1350 C, variations in input air temperature such a small fraction as to be insignificant.
- Smelt # 47 (April 2011) was at Bristol, Rhode Island with almost constant rain or drizzle. Ore used was a combination of primarily low iron content Limonite (mainly producing considerable slag) with 11 kg higher iron ores (black oxide and Hematite), producing a 2.2 kg bloom (20% yield against those better ores). This was a teaching event however, and a number of major problems had occurred over the sequence. Other smelts have been undertaken at ocean shores ( # 11 / # 44 / # 74) but as these smelts depended on bellows driven air, the possible effect of damp air can not be determined. Smelts are almost always set to fixed schedules, either annual dates or set months in advance, and are conducted ‘rain or shine’. Even so, on reflection, few have been undertaken with rainy weather for comparisons, also noting that environmental conditions (beyond the exceptional) are rarely recorded.

    There exists the potential for charcoal to absorb moisture from the environment over time :
- The charcoal used in June was mainly recently purchased, where it had been stored inside, with a weight per standard measure of 1.8 kg
- The charcoal used in October had been purchased in mid June, and stored outside for four months, with a weight per standard measure of 1.8 kg
    In truth however, the considerable amount of waste heat at the top of a correctly designed furnace, will serve to bake out any moisture contained in freshly added charcoal, well before it descends far enough to react with the available oxygen in the stack. Temperatures at the top of a furnace (so here + 40 cm above the tuyere) have been regularly measured in the range of 600 C. Because of this, moisture content of the charcoal via atmospheric absorption is not considered to have any impact on the smelting process.

Conclusions (?)

    Comparing the June smelt (# 91) against October (# 92), especially with references to other experiments utilizing lower air volumes (admittedly all using bellows variations), it would appear the ‘odd’ result is in fact June, with that yield of 28 %. The results of the October smelt at 20 % more closely approach the values produced from those other experiments. At the limits of casual observation, there are only three differences between the two ‘Low Air’ smelts :
- Source air temperature was roughly 30 C in June, roughly 10 C in October.
- Charcoal used in June was ‘bone dry’, while than in October had absorbed a slight amount of moisture.
- Air from the blower delivered in a smooth interior pipe at 5 cm ID in June, a corrugated minimum 3.3 cm ID in October.
    This last element, at least on first glance, appears the only accountable difference. The measurements of air volumes as recorded are made within the sealed air input system however, so should still reflect the actual flow of air into the furnace.
    The last possibility is some difference in physical method by the two groups of workers (charcoal and ore additions) between the two smelts. In June, these tasks were mainly carried out by individuals who had long (decades) experience. In October, these were first time workers, but under the supervision and guidance of those same senior experimenters.

    After considerable discussion and reflection, it would appear that the results of the October experiment are the more dependable, in terms of judging the effects of low volume air into this standard furnace layout and materials. It may be that a combination of factors, individually so small to barely register, added up to enough effect to produce the exceptional results in June.
    The difference between the constant blast produced by blowers, and the pulsing flow produced by Norse styled twin bellows still remains to be further explored. (5)


1) Simple geography of South Central Ontario precludes the availability of any nature iron ores, including primary bog iron ore. The Dark Dirt series of analogs uses commercially available iron oxide powders as the base, with a donation of a large quantity of red oxide / Fe2O3 as the DD1 the main type for these experiments. A batch is prepared for each smelt, a 22.5 kg bag is mixed with 2.5 kg of whole wheat flour, plus enough water to create a paste about the consistency of mayonnaise. This is then spread on to trays to about 1.5 cm thick, which are placed outside in the sun to dry. In off season / foul weather, the drying process may be assisted by placing the trays under infra-red heat lamps. Individual batches can be expected to vary slightly in water content (as discussed), with total amounts normally varying primarily through spilling) between 23 to 25 kg.

2) In June there was no tap slag from previous smelts available, and some of the stockpiled (poor quality) Lexington Limonite had been substituted. The method of quickly establishing a working slag bowl by making initial charges (2 - 3 kg) of black iron rich slag was developed by Michael Nissen of Denmark, and communicated during the 2008 ‘Iron in Thy’ symposium.

3) Judging carbon content by observing the sparks through grinding is at best only a rough approximation, as much experience as anything. Compounding this is the known variation of carbon through a raw bloom, where the original top surface commonly has a higher carbon amount than the bottom side. Testing broken pieces will further result in random results. The process of welding up a bloom into a consistent bar will also both blend and modify the starting carbon contents.

4) Sauder, L., & Williams, S., 2002, A Practical Treatise on the Smelting and Smithing of Bloomery Iron, Historical Metallurgy 36 (2), (available as pdf) https://s3.amazonaws.com/images.icompendium.com/sites/eliz2406/sup/3694971-A-practical-treatise-on-the-smelting-and-smithing-of-bloomery-iron.pdf

5) Peterson, N., 2021, A lot of Hot Air: Exploring Air Production in Viking Era Bellows, presented at the 12th Experimental Archaeology Conference - EAC12 World Tour (publication pending)
    An important consideration is separating the specific operation of Norse twin bellows from other human powered air delivery systems. Norse bellows will produce a pulsing air supply, which would also be the case for still earlier drum or bag bellows systems, or the roughly contemporary Chinese box bellows.
The historically later ‘Great Bellows’ (mid 1300’s) has a ‘double action’ mechanism, where a moving first chamber supplies air to a second delivery chamber, thus actually creating a constant air blast. The same result will be found with any system that uses a collecting air bladder as a second stage, regardless of the initial air input. Unfortunately, too many living history based teams do not clearly describe exactly what equipment they are using when stating ‘used a bellows’.

Unless otherwise indicated :
All text and photographs © Darrell Markewitz, the Wareham Forge.