'Dimmable' G9 lamps: experiments in performance, by Claire Tomara & Kevan Shaw


Revision A, 22.02.2019

Specifying LED retrofit lamps is part of the daily life of lighting designers, especially on hospitality projects. There is a great range of LED filament lamps on the market, especially E27 base, that claim to be dimmable. On many projects, decorative fittings use G9 rather than E27 lamps. The cost of decorative fittings using G9 or E27 lamps is less than fittings with integral LED modules. Many companies are not producing decorative fittings with integrated LED. While LED retrofit G9 lamps might sound like a lighting designer’s nightmare, there are a few companies available on the market that offer dimmable G9 LED lamps. Is it a dream come true or a continual nightmare?

As part of one of our recent hospitality projects, we decided it was necessary to test a few of these products for their dimming ability and control. 



The equipment used for the testing of G9 lamps is listed below:
- A Pendant fitting with three G9 bases
- Lighting control system with Trailing edge dimming and DALI outputs
- DALI Electronic trailing edge dimmer for incandescent lamps, electronic transformers and line voltage LED lamps
- Control software running on a web interface
- LED G9 dimmable lamps by two different manufacturers
- Halogen G9 to serve as a comparison
- Integration sphere, made in-house
- Conventional light meter
- Assenstek Essence Spectrophotometer
- MiniDSO 211 storage oscilloscope


In order to begin the test, we set different levels of intensity through the control software;  the levels rise in 5% steps starting from 0% to 100%, with 3 second fade time.

At the beginning, we dim the lamp up and down from 0 to 100 with 7 seconds fade time to see the general dimming. Then, we start dimming up from 0 and every 5% until 100% to see and record any perceivable difference and flicker behaviour. 

After the general dimming test, we repeat the test using the integration sphere. This helped us to take relative light measurements at each dimming level. The measurements taken create the dimming curves that we will illustrate below.

We repeat the test using the trailing edge dimmer and DALI electronic trailing edge dimmer.  The minimum rated load for the trailing edge dimmer is not indicated. The minimum rated load for the DALI electronic trailing edge dimmer is 3 VA.


Some of the difficulties we encountered while engaging with the test were the relative stability of light meter readings. The light meter was giving varying readings for a particular control setting. An average was taken to draw the dimming curve. In some cases, the light output was visibly unsteady, therefore we assume the variation in meter reading was in response to variations in light output of the lamp being tested through the duration of the reading.

The pendant we used for testing has three G9 bases. In some cases, we encountered the issue of LED lamps being on, even when the level was on 0%. We believe this to be a result of insufficient current to keep the dimmer triac latched off. This can be made worse through the very poor (capacitive) power factor of these lamps. In order to increase the load current, we used the halogen lamp in addition to the tested G9.  This allowed us to creating the curves for lamps that could not be individually controlled. In these situations, we are concerned that the actual dimming curve will vary when a resistive load is not applied to the dimming circuits in addition to the LED. However, the dimming curves of the LED still diverge from what we would expect.


Lamp 1: Halogen G9 40W

Halogen lamps used to be the dominant light source in residential applications. They are known for their quality of light and specific dimming behaviour.  Halogen lamps are being taken off the market by EU regulations. We tested the halogen lamp in order to use it as point of reference for LED retrofit lamps.

When using the trailing edge dimmer, the lamp started visibly illuminating at 25%. The perceived fading up lasted until 90%-100%. While dimming down, the lamp stayed on until 20% and then faded off completely. The change in colour temperature of the lamp was not considered during this test.   

G1: This graph shows the dimming curve of the halogen lamp on the trailing edge dimmer

G1: This graph shows the dimming curve of the halogen lamp on the trailing edge dimmer

The graph shows a logarithmic curve that reaches a plateau at 90%. While on the trailing edge dimmer, the fading up and fading down curves are quite distinctive. While fading down, the light readings resulted in higher values than those during fading up. This must be caused by the internal control circuitry in this dimmer. 

G2: The graph shows the dimming curve of the halogen lamp on the DALI trailing edge dimmer

G2: The graph shows the dimming curve of the halogen lamp on the DALI trailing edge dimmer

The behaviour of the lamp stayed similar while fading both up and down, following a logarithmic curve. One of the main differences was that the lamp visibly illuminated from 1% gradually until 100% without reaching plateau.

G3: Graph of halogen lamp on mains dimming at 60%

G3: Graph of halogen lamp on mains dimming at 60%

G4: Graph of halogen lamp on DALI trailing edge dimmer at 60%

G4: Graph of halogen lamp on DALI trailing edge dimmer at 60%

What is the reason for the changing behaviour? A look at the graphs from the oscilloscope might give us a better understanding.

We can see some differences in the traces. Firstly, while they are taken from the same level, the curve is not cut at the same point. It seems that when the lamp is on, the mains dimming the curve is almost cut at its peak point, while on the DALI dimmer, it is cut at a lower point. In addition, the flat line looks smoother on the DALI dimmer than on the mains dimming. Both these devices are exhibiting different dimming curves and different trace patterns indicating different electronics to achieve the same anticipated results.

Lamp 2: LED G9 Dimmable 3W

After testing the halogen lamp, we continued the process testing LED dimmable G9 lamps from different manufacturers.

We have just one product type of Lamp 2. When we placed it to the pendant and on the trailing edge dimmer, the lamp was on at 0%.  At 10%, it flashed once and started fading up from 15%. It seemed to reach a plateau at approximately 50-60% as there was no significant change in intensity after 60%. The perceived fading down of the lamp started at 20% until 5% while it stayed on at 0%.  

G5: The graph shows the dimming curve of the single Lamp 2 on the trailing edge dimmer

G5: The graph shows the dimming curve of the single Lamp 2 on the trailing edge dimmer

The curve seems to be a square root curve which reaches a plateau at 90%. There are slight differences in the curves while fading up or down, while the lamp stayed on at 0%.

In order to stabilise the dimmer, we used the halogen lamp to increase the load, and repeated the test.

When Lamp 2 was placed with the halogen lamp, its dimming behaviour changed radically. Firstly, Lamp 2 was off at 0%. It turned on at 10% with a little delay and started dimming up gradually until 80%. At 85% there was a big step change in the intensity and then a gradual fade up at 90% when it reached plateau. 

Lamp 2 started fading down at 85%. At 80% there was a big step change in intensity and then it gradually faded down until 5%. At 5% it stayed on for a while but then turned off after few seconds. It  stayed off at 0%.

G6: The graph shows the dimming curve of Lamp 2 with the halogen lamp on the trailing edge dimmer

G6: The graph shows the dimming curve of Lamp 2 with the halogen lamp on the trailing edge dimmer

The graph shows a slightly off-linear dimming behaviour that reaches plateau at 90%. The fading up and fading down curves show slight differences, as the lamp stayed on, even for just a few seconds at 5% while fading down, although it only turns on at 10% with a little delay. It seems that the change in the load smoothes the square root curve of Lamp 2.

We repeated the test using the single Lamp 2 on the DALI electronic trailing edge dimmer.  The lamp was on at 0%. Its intensity was significantly less than when it was on the mains dimming. It stepped up at 1% and then gradually faded up until 100%. While dimming down, the lower levels seemed to be brighter than when the lamp was fading up. 

G7: The graph shows the dimming curve of the single Lamp 2 on the DALI electronic trailing edge dimmer

G7: The graph shows the dimming curve of the single Lamp 2 on the DALI electronic trailing edge dimmer

The dimming behaviour resulted in a linear curve with two peak points in the beginning at 1%, and at 95% at the end. While the first peak step was quite obvious to the naked eye, the second peak was not as perceivable as shown in the graph.

We repeated the test again using the halogen lamp to increase the load.  

G8: The graph shows the dimming curve of Lamp 2 with the halogen lamp on the DALI electronic trailing edge dimmer

G8: The graph shows the dimming curve of Lamp 2 with the halogen lamp on the DALI electronic trailing edge dimmer

As was the case on the trailing edge dimmer, the presence of the halogen lamp smoothed the linear curve of Lamp 2.  At the first peak at 1%, the lamp was obviously flickering. It then seemed to stabilise at 15%. The high peak was clearly perceivable from 90 to 95%.

A look at the traces from the oscilloscope demonstrates the differences.

G9: Single Lamp 2 on mains dimming at 10%

G9: Single Lamp 2 on mains dimming at 10%

G10: Lamp 2 with halogen on mains dimming at 10%

G10: Lamp 2 with halogen on mains dimming at 10%

G11: Single Lamp 2 on DALI electronic trailing edge dimmer at 10%

G11: Single Lamp 2 on DALI electronic trailing edge dimmer at 10%

G12: Lamp 2 with halogen on DALI electronic trailing edge dimmer at 10%

G12: Lamp 2 with halogen on DALI electronic trailing edge dimmer at 10%

When Lamp 2 is placed along with the halogen lamp, the diagrams look more as expected. Even here, we can see significant difference at the cut point of the curve, as when the lamp is on the DALI dimmer, the cut point is higher than when on mains.

The diagrams relating to Lamp 2 when the single lamp is in the pendant, look quite different to what was expected.  When on mains dimming, it seems that the wave never quite reaches a flat linear value, while on DALI dimmer, it seems to get closer to linear values, although we still cannot see the same linear curve as when Lamp 2 is placed along with the halogen lamp.

There are some very significant interactions between the electronics of the dimmers and the electronics in the lamps, particularly when there is a low load on the dimmer. This accounts for the unpredictable dimming and flickering. 

Lamp 3: LED G9 Dimmable 5W

We continued the experiment with another LED lamp, from a different manufacturer.  We had three items of the same product so that we could use them all at once on our pendant. This provided a more realistic situation, and at 15W, a minimum load.

The lamps were off at 0%, and started slightly turning on at 30%. Then, they gradually faded up until 95%, when they seemed to reach plateau. The lamps started gradually fading down after 90% until 40%. They turned completely off at 25%.  

G13: The graph shows the dimming behaviour of 3 x Lamp 3 on trailing edge dimmer

G13: The graph shows the dimming behaviour of 3 x Lamp 3 on trailing edge dimmer

The dimming curve tends to match the logarithmic curve of the halogen lamp, although it has a smaller curve radius that looks almost linear, with a starting point of 25% as 0. 

We repeated the test, dimming the lamps with the DALI electronic trailing edge dimmer.

Lamp 3 started to be very slightly on at 15%, and gradually faded up until 100%. While dimming down, it gradually faded off until 10% when it turned completely off. 

One of the main differences in behaviour of Lamp 3 while on the two different dimming systems was that on the DALI electronic trailing edge dimmer, the lamp flickered and looked unstable below 45%. This behaviour was not visible when the lamp was on the trailing edge dimmer. 

G14: The graph illustrates the dimming behaviour of 3 x Lamp 3 on the DALI electronic trailing edge dimmer

G14: The graph illustrates the dimming behaviour of 3 x Lamp 3 on the DALI electronic trailing edge dimmer

The dimming curve smoothed into a logarithmic curve, although slightly different from the curve   of the halogen lamp.

A look at the traces from the oscilloscope provides us with an interesting comparison.

G15: 3 x Lamp 3 on mains dimming at 30%

G15: 3 x Lamp 3 on mains dimming at 30%

G16: 3 x Lamp 3 on DALI electronic trailing edge dimmer at 10%

G16: 3 x Lamp 3 on DALI electronic trailing edge dimmer at 10%

The two traces are quite different. The second trace looks more like what we would expect     than the first trace. The cut point of the wave seems to be the same. The flat line of the second graph looks a little shaky, while it is completely distorted on the first trace, creating more curves.


Our experiments showed that each lamp displayed different dimming behaviour on the different dimming systems. The dimming curves seem to be smoother when the lamps were on the DALI electronic trailing edge dimmer. In addition, they demonstrated a greater dimming range. 

How much of this range we can actually use or perceive? As discussed in the Lamp 3 results, the lamp demonstrated visible slow flicker, and was unstable for almost half of the dimming range.  This behaviour was not visible on the mains dimming. However, the dimming range of Lamp 3 on mains dimming was much reduced, compared with when the lamp was on the DALI dimmer. 

Lamp 2 demonstrated a greater dimming range on both dimmers. However, in both situations it needed the presence of the halogen lamp in order to stabilise the load and to ensure the lamp is turned off at 0%. In addition, it seemed that the DALI dimmer smoothed the dimming behaviour, even when there was a single Lamp 2 in the pendant.


On the basis of the testing undertaken so far, we do not believe that there is a safe specification for dimmable G9 lamps. We expect to see different performance depending on lamp type, dimmer type and total load per circuit. Dim to extinction is also unreliable and will vary with the load (number of lamps) per circuit, different wattages of lamps and different makes of lamp.

As different lamps of similar specification and performance show distinctly different curves, lamp replacement also becomes a potential problem, requiring the attendance of a controls programmer to re-set scenes according to the performance of new lamps.

Many of the lamps we received were in generic boxes, and most were not properly labelled to comply with EU EcoDesign regulations. The majority of lamps were purchased through online retailers, both specialist lamp sellers and Amazon. One manufacturer sent us a sample lamp, and within a week had sent another sample to the same specification. The only differences were the size of the box, the printing to the base of one lamp, the envelope of the other, one visibly different electronic component, and different codes printed on each lamp. 

It is not safe to assume the performance of, or to specify any of these products without a significant disclaimer as to their performance, and putting the onus on the installing contractor to undertake the necessary testing on site before bulk ordering lamps for any project.

Be sure to follow our posts, as we will continue experimenting and publishing our results. Until then, enjoy a small video demonstration of the varied dimming behaviour of three different lamps on mains dimming. 

Claire Tomara and Kevan Shaw, February 2019

Revision A, 22.02.2019
We recognised that the breaks appearing in the graphs relating to Lamp 2, in the earlier version of this report, were due to the auto-range error of the light meter. We repeated the readings using a light meter with controlled error range. We updated the graphs for Lamp 2 accordingly.

KSLD BBQ 2017 - We had a blast!

KSLD held their annual BBQ on 23 June 2017 at Cramond Beach in North Edinburgh. Staff, both former and current, and their families joined together to celebrate a year well done.

The weather was the best it's been in years, with blue skies and a not insane wind (even for the Forth). Warm and pleasant, the breeze knotted the balloons around the table, but didn't cause any big problems.

This year, the activity was shooting off water-bottle-rockets. You can buy the kit yourself from Rocket. It's simple, but effective: take a an empty plastic bottle, decorate it, fill it 1/3 with water, add the adapter, then attach to a bike pump and inflate until lift off!

We had a variety of bottles available: 1.5L, 2L and even some 3L. We found that the 1.5L and 2L bottles worked best. They flew with a height and distance that would have tired out even the most marshmallow-filled child. 

The 3L bottle, while it did eventually fly, tended to be more unpredictable in its flight as the water would shift the rocket in different directions, and it would have packed quite a punch if it had connected with anyone. Perhaps more experimentation is needed to see about add-ons to increase stability.

It was a smaller get-together than usual, but the weather was nice, the food delicious, and the rockets dynamic.

More Blogs

VR at Kelvin Lighting's Party Recap Video

At Kelvin Lighting's 10 year birthday party, KSLD’s Eric Berntsson got the opportunity to demo his VR models for architects and lighting designers. The project demonstrated was one Eric was working on. He took the lighting design KSLD made and turned it into a virtual reality space to illustrate how a potential project could appear in VR. The project is in construction now, so this is a sneak peek at what the finished project will look like. 

Watch below to see what users have seen when they saw things in VR and how the VR compares to reality.

Scott Kelly's reaction alone is worth a look!

More Blogs

KSLD is a Scottish Living Wage Accredited Employer

KSLD has today been accredited as a Living Wage Employer by the The Living Wage Foundation.

The Living Wage commitment will see everyone working at KSLD, regardless of whether they are permanent employees or third-party contractors; receive a minimum hourly wage of £8.45 - significantly higher than the national minimum wage of £6.95 and the new minimum wage premium for over 25s of £7.20 per hour introduced this April.

The Living Wage is an hourly rate set independently and updated annually. The Living Wage is calculated according to the basic cost of living using the ‘Minimum Income Standard’ for the UK. Decisions about what to include in this standard are set by the public; it is a social consensus about what people need to make ends meet

"KSLD has always been committed to supporting our employees. What's great about the Living Wage Accreditation is that is also applies to our regular third-party subcontractors. It's great to know that even the cleaners who come in and do a fantastic job are now receiving the Scottish National Living Wage," says Design Director Kevan Shaw.

Employers choose to pay the Living Wage on a voluntary basis. The Living Wage enjoys cross party support, with public backing from the Prime Minister and the Leader of the Opposition.

Living Wage Foundation Director, Katherine Chapman said: “We are delighted to welcome KSLD to the Living Wage movement as an accredited employer.

“The best employers are voluntarily signing up to pay the Living Wage now. The Living Wage is a robust calculation that reflects the real cost of living, rewarding a hard day’s work with a fair day’s pay.

“We have accredited nearly 3,000 leading employers, ranging from independent printers, bookshops and breweries, to well-known companies such as Nationwide, Aviva and SSE. These businesses recognise that clinging to the National Living Wage is not good for business. Customers expect better than that."

More Blogs

Virtual Reality Meets Lighting Design & Architecture.

Eric Berntsson on VR and the challenges and opportunities it presents for lighting designers - take it from one who knows!


Virtual reality (or VR) has been around for a much longer time than many would think. You can find info about VR headsets dating back to the early 90s.
For most people it’s just a stupid mimic of “The Future”. The reason for this is that the majority of VR headsets that people try on are in general quite bad. You get dizzy, light headed or even nauseous.

However, if you ever have the opportunity to try one of the newer, higher quality beasts - one that is not an app on your smart phone; one that requires a monster of a computer to bring you all that eye-candy graphics and physics - stay in the queue.

The HTC-Vive is one of these headsets, and it’s the one I decided to buy for myself. The first time you try one of these high end headsets you realize how immersive they really are. You can walk around in the virtual space, pick up objects and look at them in detail.
You want to stand on the Empire State Building? No problem, “Google Earth VR” is out there and you can visit (or revisit) any part of the world.
You want to dive deep under water? Go ahead and try “TheBlu” and freak out when a whale swims next to you.

So that is all cool and fun stuff. But what about lighting design and architecture? Well... there is none. There are a lot of educational demos and experiences out there for medical, historic and biological purposes, but no architecture. So I decided to try to model up one of our projects at KSLD myself using Unity - a free development program. Add some textures, load it up in the headset and suddenly you stand in the middle of the project - in real life scale.

The lighting is a huge challenge by itself. linear lighting is not a standard thing you can just throw in a model and move around in real time when using Unity. A powerful computer like the one I own struggles when calculating these things and it took me a good while to get the lighting to look somewhat realistic.

With the problem of good looking lighting solved, I could create more models, try different environments like lobbies, bars and offices. I soon realized that I had quite a few models ready. I brought these in a demo to show the office along with a few invited designers, engineers, and decision-makers.
Putting the “First time in Virtual Reality” shock away, I got some really good feedback on the models and ideas of what to change and add.

One of the joys of owning the headset is to look at other peoples faces when they try out VR for the first time! The future seems to have found its way here after all.

More Blogs

5 Ways to Spruce up Your Lights for Winter

Ah, the nights’ a fir drawin’ in! With the increasing darkness, people are spending more time indoors, and those that are outside need more light. How are the lights on your property? Keep reading to identify the ways you can make your property light beautiful.

Stobo Castle

Stobo Castle

The first thing you need to do is walk all the way around your property in the dark, with all your current lighting on in its normal setting. As you walk, you need to look for the following things:

1) Are all the lights working?

You’d be amazed how many properties have rubbish lighting simply because a lamp or two has gone out and not been replaced. If you notice a light out, or a location that is too dark, the first thing you need to do is call your electrician to either replace the lamp or repair the fixture. This does not need to be an expensive exercise, and keeps your property clean and attractive. If your fixtures are so old that you cannot replace the lamps, contact a lighting designer or manufacturer to help you find out which fixtures are appropriate for your property.

2) Are all the lights the same colour temperature?

It’s a little appreciated fact that different lights have slightly different colours of light. In LEDs, the distinction is usually between “cool” white and “warm” white. Having a row of cool white lights with one warm white light in the middle can really frustrate people. Again, you should be able to fix this easily by simply buying the correct lamp for your fixtures.

3) Do any lights shine directly in people’s eyes?

When you look at the fantastic oil painting on the wall, or when you sit in the chair in the sitting room, is there a light shining right into your eyes? Why is it doing that? Is it trying to light something else? Did it slip and get crooked? Try seeing if you can point the fixture in a different direction. If you’re really stuck, a quick consultation with a lighting designer should be able to resolve any minor problems without a whole redesign.

4) Are the correct things and places lit?

The Edinburgh Festival Hub

The Edinburgh Festival Hub

If you have art that you want to be seen in dark winters, it needs to be lit. If you have dark stairwells that could be dangerous, they need lit. If you sit in your favourite chair to read, is there enough light to read by? Think about safety and the things that are important to you, and make sure you have enough light. However, if you regularly open your home to the public, you may need to meet specific health and safety lighting regulations; please ask your lighting designer which regulations can apply.

5) Is there… something missing?

Light can be functional, but light can also be beautiful. It can provide an atmosphere, and turn a simple bulb into a feature. It can hide faults and accentuate perfection. Maybe you need a feature light, or maybe it needs a bit more sensuality and cosiness. Is it too modern, too dated? Or do you just want something that is beautiful? Light can do all those things, and it can be achieved with as little as adding a feature light, or changing the location of a couple of fixtures. Historic buildings need tender loving care, but respond best to good lighting. Get in contact with KSLD if you think your lights need an overhaul, or if you just feel over your head. Your property is important, and deserves the best lighting.


More Blogs

Cleaning the Old KSLD Neon Sign

Back when Kevan first started KSLD as Kevan Shaw Lighting Design some 20-odd years ago, he had a custom neon sign created. Visitors to our office may have noticed its absence over the last few months as it was away getting restored and repaired. The sign is now back, but needs a good clean.

The sign has returned to us, but unfortunately our restorer, Kemps Architectural Lighting, was hit by the terrible floods that plagued North England and the Scottish Borders this past winter. We were actually quite fortunate, as this light is attached to a wooden frame, and so he said he just walked in to find it floating in the workshop. While I’m sure he cleaned it up significantly and had to do additional repairs because of the flood damage, it still came to us a touch grimy.

You can see the levels of dust and water spots on the frame.

The 20-odd year old neon sign needed cleaned.

Cleaning a neon sign can be rather delicate work. You don’t want to damage the delicate bulbs, and in the case of this sign, damage any connecting wires. Nor do you want to use harsh chemicals that might damage the paint on the frame, the paint on the tubes, or the tubes themselves.

The tools used for this clean were an air duster, a soft cloth, dish soap, warm water, and some paper wipes.

First as much loose dust as possible was removed by using the air duster.

Then came the soft cloth (no loose fibres to catch on wires sticking out) with warm water and dish soap. Cleaning had to be careful and gentle.

To get underneath the tubes bolted to the frame, the cloth was folded and dragged under each section, with a motion similar to flossing your teeth.

Cleaning the tubes themselves also provided some challenges. To minimize risk of damaging the tubes, as little physical contact as possible was needed. Rather than holding the cloth and rubbing the tubes, instead the cloth was folded and held some distance away from the tubes, like dusting.

Having the cloth folded also created the opportunity for cleaning the underside of the tubes by placing the tubes in the folds of the cloth and again employing a flossing style technique. Folding the cloth into smaller sections allowed access to tight corners and spaces.

It is important to rinse the cloth often, as if you don’t rinse the cloth then you’re just pushing dirt around.

As you can see by the colour of the water, the cleaning was indeed successful.

After using the cloth, smaller spaces and missed spots were tidied up with the paper wipes. These wipes are the ones you use for cleaning electronic things, so have little lint. These were quite useful for sections very close to the frame.

The clean neon sign.

After cleaning and before hanging you need to test the light to make sure it wasn’t damaged in the cleaning process. Apparently some neon signs which are really dirty can be soaked and washed in water. That type of cleaning would require appropriate drying time to reduce risk of electrical problems. As this was dry-wiped, there was little moisture on the light and frame, so it was ready to test very quickly.

The light works! And is clean!

Get your qualified individual to install the light in its location. Here, that’s Kevan installing the light back where it was before.

IMG_20160511_165146257 copy.jpg

And now the light is completely functional! The word “design” was re-blown (topped up), and is brighter than the others, but you can clearly see that each letter is lit, albeit in different brightness. Because the light hasn’t been on for some month, this should improve with more constant use.

Thanks to Becky for cleaning the light, Kevan for hanging it, and Kemps Architectural Lighting for restoring and repairing it. KSLD always recommends that you have a qualified professional service, restore and repair your neon signs.

More Blogs