Preliminary Research Report on Canon 580EX II Flash Failures By Jim Clark & Tim Ambrose, Research and Development LPA Design Background: Since the launch of its PocketWizard MiniTT1 and FlexTT5 models in early 2009, LPA Design (A.K.A. PocketWizard) has been receiving periodic calls and reports from customers who are using their Canon 580EX II flashes with the PocketWizard ControlTL radios and expressing concern that perhaps their flash may have been damaged by using a PocketWizard radio trigger with it. As of the time of this report (Jan, 2011), there are about 120‐140 PocketWizard customers who have reported having one or more 580EX II flashes damaged with common symptoms over the past 18+ months. If the failure mode was related to our product, this would work out to less than 0.5% of our MiniTT1/FlexTT5 radio units which were connected to a flash that failed and less than 1% of the 580EX II flashes that were connected were damaged. LPA Design has conducted a serious in‐depth investigation for several months as we continue to collect data and evidence of the specifics of the failure mode. In addition we have just started the outside research assistance of analytical chemistry professor Giuseppe Petrucci at the University of Vermont. Research: The symptom of the failure is that the 580EX II will no longer output controlled amounts of light. It will only output full power dumps. This full power dump will happen even on a pre‐flash, therefore leaving no stored energy within the capacitor left for a main flash exposure. The testing quickly revealed that an IGBT (transistor) within the flash is damaged. This part is made by Renesas (part # RJP4301). The IGBT is used to turn off the flow of current through the flash tube when the desired amount of light has been reached. This IGBT is also used to control (modulate) the brightness of the light when in HSS (High Speed Sync) with the help of a fiber optic cable feeding a light sensor within the main body. Although LPA Design has been able to identify the symptom of the problem, LPA Design has not been able to trace the cause of the problem to our PocketWizard products and we have no knowledge of the incidence rates of this problem arising independently without PocketWizard, i.e. some number of 580EX II flashes likely fail without PocketWizard ControlTL radios interacting with them. We received reports of this IGBT failure in all spectrums of use. It spans multiple different Canon camera models, below X‐ Sync and above in HSS (more often in HSS), TTL or manual modes and on any version of firmware that the PocketWizard MiniTT1 and FlexTT5 have had in public release. Even brand new flashes could fail. We have ruled out heat buildup as a cause since several have failed within the first dozen shots after being pulled out of the camera bag. This is all based on verbal descriptions from customers. It also has failed as many times with a PocketWizard AC5 RF Soft Shield as it has failed without using the AC5. So there appears to be no change in occurrence when using the AC5 or AC7.
Transcript
Preliminary Research Report on Canon 580EX II Flash Failures By Jim Clark & Tim Ambrose, Research and Development
LPA Design
Background:
Since the launch of its PocketWizard MiniTT1 and FlexTT5 models in early 2009, LPA Design (A.K.A. PocketWizard) has been receiving periodic calls and reports from customers who are using their Canon 580EX II flashes with the PocketWizard ControlTL radios and expressing concern that perhaps their flash may have been damaged by using a PocketWizard radio trigger with it.
As of the time of this report (Jan, 2011), there are about 120‐140 PocketWizard customers who have reported having one or more 580EX II flashes damaged with common symptoms over the past 18+ months.
If the failure mode was related to our product, this would work out to less than 0.5% of our MiniTT1/FlexTT5 radio units which were connected to a flash that failed and less than 1% of the 580EX II flashes that were connected were damaged.
LPA Design has conducted a serious in‐depth investigation for several months as we continue to collect data and evidence of the specifics of the failure mode. In addition we have just started the outside research assistance of analytical chemistry professor Giuseppe Petrucci at the University of Vermont.
Research:
The symptom of the failure is that the 580EX II will no longer output controlled amounts of light. It will only output full power dumps. This full power dump will happen even on a pre‐flash, therefore leaving no stored energy within the capacitor left for a main flash exposure.
The testing quickly revealed that an IGBT (transistor) within the flash is damaged. This part is made by Renesas (part # RJP4301).
The IGBT is used to turn off the flow of current through the flash tube when the desired amount of light has been reached. This IGBT is also used to control (modulate) the brightness of the light when in HSS (High Speed Sync) with the help of a fiber optic cable feeding a light sensor within the main body.
Although LPA Design has been able to identify the symptom of the problem, LPA Design has not been able to trace the cause of the problem to our PocketWizard products and we have no knowledge of the incidence rates of this problem arising independently without PocketWizard, i.e. some number of 580EX II flashes likely fail without PocketWizard ControlTL radios interacting with them. We received reports of this IGBT failure in all spectrums of use. It spans multiple different Canon camera models, below X‐Sync and above in HSS (more often in HSS), TTL or manual modes and on any version of firmware that the PocketWizard MiniTT1 and FlexTT5 have had in public release. Even brand new flashes could fail. We have ruled out heat buildup as a cause since several have failed within the first dozen shots after being pulled out of the camera bag. This is all based on verbal descriptions from customers.
It also has failed as many times with a PocketWizard AC5 RF Soft Shield as it has failed without using the AC5. So there appears to be no change in occurrence when using the AC5 or AC7.
We purchother com
Most sign
One commreplacem
The secondischarge
Repaired periods w
So it beca
Eventuallyand soot bplace.
Replacemfail with o
Below are
hased many ommon elemen
nificant obser
mon theme went.
nd common t from one or
units that havwhen used dir
me clear that
y we found thbuildup aroun
ment of the flaor without a P
e examples of
f the failed flnts between t
rvations:
was that ALL f
heme was thmore ends o
ve only the IGectly on a Ca
t the IGBT wa
hat the flash hnd the ends o
ash heads solvPocketWizard
f the 580EX II
ashes from ethe units that
failed flash un
at ALL failed f the flash tu
GBT replaced non camera w
asn’t the only
head (tube/reof the flash tu
ved the failurd radio.
flash heads t
nd users so wt failed.
nits have had
units have visbe to the met
will frequentwithout a Poc
y failure comp
eflector) was ube. Sparks a
re problems a
that were cau
we could take
the IGBT dam
sible signs of tal reflector.
tly fail again wcketWizard ra
ponent within
also bad andand electrical
and the flashe
using IGBT fai
e them apart a
maged and ne
electrical arc
within relativeadio involved
n the system.
d showed visibarcing were a
es could no lo
ilures to repe
and study the
eeded
cing / electrica
ely short time.
ble signs of asapparently ta
onger be mad
eat –
e
al
e
sh aking
de to
Figure 1: oxidizatio
Figure 2: possible o
Figure 3: housing is
Flash tube enon of metal re
Flash tube enoxidization of
In more seves near the tub
nd showing beflector near r
nd with triggemetal reflect
ere cases the be.
lack soot behrubber end se
er ring showintor near rubb
flash tube gla
hind tube on real.
ng black soot ber end seal.
ass is charred
rubber moun
behind tube
d in a ring aro
nt. Also show
on rubber m
und where th
ws possible
mount. Also sh
he metal refle
hows
ector
Figure 4:
Figure 5: E
Another exam
Example of d
mple of soot
amage from r
and electrica
reflector to tr
l damage at e
rigger ring aft
end of flash t
ter electrical
ube.
arcing
Electrical
We captu
For referethe IGBT e
Below is a
Figure 6:
Analysis of F
red failure ev
ence, IGBT cuemitter lead.
an example of
In the middle
Failure Mode
vents as well
rrent is sense
f a failure we
e of a High Sp
:
as good work
ed by measur
e have capture
peed Sync pul
king units to v
ring voltage a
ed –
se train, the
verify the pro
cross a 0.01 o
IGBT fails and
oblem using a
ohm 1% resis
d shorts Colle
n oscilloscop
tor in line wit
ector to Emitt
pe.
th
ter.
Figure 7: spark jum
Close up viewmped to the re
w of IGBT eleceflector and t
ctrical signalshen into the
s. Note that collector of t
the IGBT gatehe IGBT.
e was turningg off when the
e
Making th
In many cfront edgein feedba
We are sethe negat
Figure 8:
Note: Top
he flash more
cases (but note of the housck to the sens
eeing that thistive flash tube
Short curren
p trace is the c
e prone to fai
t all) we are sing or it has asor.
s fiber optic ce cathode ove
t spikes on IG
control line g
ilure:
eeing that tha kink / pinch
cable can maker to the trigg
GBT emitter c
going to the tr
e fiber optic c in the fiber l
ke the flash mger ring of the
aused by arci
rigger coil driv
cable has eithine which mig
more sensitivee tube.
ing to reflecto
ver circuit.
her pulled baght be causin
e to having a
or or trigger r
ck some fromng some redu
spark arc fro
ring.
m the ction
m
Figure 9:
Note thatcircuits. (H
In theory fire at thethat cause
To furthewith the tas elimina
Same as Figu
t the arc pulseHappens afte
this figure 9 se moment thaed the arcing
r confirm thatrigger ring anation of the c
ure 8 (zoomed
e to the trigger arc not befo
shows that that is circled ab.
t this is in facnd saw a signircled pulse o
d in) Short cu
er ring in this ore which ind
he flashes elebove. This als
ct arcing to thificant reducton the driver c
rrent spikes o
case propagadicates that th
ctrical controso indicates t
he trigger ringtion in currencircuit line (to
on IGBT emitt
ates backwarhe control sig
ol circuits did that it’s not a
g, we placed ant ringing on top trace).
ter caused by
rds through thgnal did not ca
not cause thCPU firmwar
a 47K ohm rethe circled ev
y arcing.
he trigger coiause the arcin
e trigger coil re driven even
sistor in serievent above as
l ng.)
to nt
es s well
Figure 10the 580 EX
Note that
This flash
: Typical tracX II on a good
t the current p
has a good p
ce showing a pd working flas
profile does n
positioning of
pre‐flash (whsh that does n
not have any
the fiber opt
hile using a Ponot damage IG
spikes in the
tic and no sign
ocketWizard MGBTs.
middle of the
ns of kinks or
MiniTT1 / Fle
e IGBT ON pe
r bends in the
xTT5) to trigg
eriods.
e fiber.
ger
Flash Tub
Another sextending
One theometal refl
Manually shortened
We have f
Figure 11the flash t
be Alignment:
significant obsg into the refl
ry is that this lector the sho
moving the fd did increase
found that w
: NOTE: the stube.
:
servation is thector are mo
tube trigger ortest thereby
flash tube shoe the rate at w
e cannot get
sealed glass c
hat flashes thre prone to f
ring positiony enabling arc
own on bottowhich arcing o
the IGBT to f
covers are rem
hat have the fail again.
ing (top imagcing to happe
om image to toccurred. (H
ail if the fron
moved from t
flash tube alig
ge) makes theen much easie
the right suchowever it did
t glass cover
the head asse
gned with the
e spark gap dier.
h that the spad not cause th
is not in plac
emblies for ea
e trigger ring
istances to th
rk gap was he IGBT to fai
e.
asier viewing
he
l)
of
Ozone fro
Ozone is g
(source W
In a sealeelectrical shorter di
Ozone cangeneratedseparated
High Speedischarge
In theory
The head that hold
LPA Desigdamage t
To furthedid cause
So in theo
With thes
(Model #
Pictures o
om UV and El
generated by
Wikipedia)
d container (sbreakdown vistances.
n be generated in the presed by a glass tu
ed Sync mode in a head tha
this generate
assembly on the flash tub
gn theorized the IGBT.
r test this theIGBTs to fail.
ory the best w
se theories in
106‐M) The
of the test set
ectrical fields
using UV ligh
such as sealevoltage of air
ed in the presence of rapid ube is how co
e (HSS) of the at has a misal
es higher and
the 580EXII he in place. Th
that ozone bu
eory, we remo. With the gla
way to help pr
mind, LPA De
model 106‐M
tup are on the
s:
ht, high voltag
d flash head)such that arc
sence of UV frepeated elemmercial / in
580EX II genligned tube a
higher levels
has a glass cohe ozone has
uild up in the
oved the glasass cover rem
revent ozone
esign rented
M reads Ozone
e following pa
ges (10‐20KV
ozone can bucing (electrica
rom flash tubctrical dischandustrial ozon
erates a lot olong with imp
s of ozone tha
over over the nowhere to g
flash head is
s front cover moved, we can
build up is to
an Ozone Gas
e as low as 10
age (Figs 12,
V) or an electr
uild up to danal discharge) c
bes. Howevearge. Pulsed hne generators
of repeating pproper feedb
at can’t escap
front and a rgo.
causing the f
from a flash n’t get the IG
o ventilate th
s Analyzer fro
0ppb (10 part
13).
ric spark (coro
ngerous levelcan happen m
r it’s far morehigh voltage es work.
pulses of electack from the
pe the head a
ubberized se
failure events
head assembBT to fail.
e head assem
om Ozone So
ts per billion)
ona).
ls and reducemuch easier a
e likely to be electrical cha
trical arc / fiber optic ca
ssembly.
al on both en
s that ultimat
bly that typica
mbly.
lutions.
.
e the t
rge
able.
nds
ely
ally
Figure 12Ozone gas
Figure 13
: This 580EX Is could be sam
: Ozone Solu
I had the glasmpled via the
tions Gas ana
ss cover remoe front of the
alyzer model
oved from thehousing thro
106‐M
e internal flasough the plast
sh head assemtic Fresnel len
mbly so that ns.
Ozone A
Our first tseals the fthe flash v
We confir700ml pehole in th
Next we mthe flash t(HSS mod
The Ozonup of ozoreading af
Figure 14
nalysis of th
test was to coflash tube hevia an OC‐E3
rmed that Ozor minute. In oe far back of
moved the 58to fire specifie). A Canon
e analyzer hane vs. time. Tfter averaging
:
he 580EX II F
onfirm we couad assembly cable to fire a
one was startorder to enabthe flash hea
80EX II flash tocally 80 popsCP‐E4 battery
as a voltage oThe Ozone ang for each 10
Flash:
uld detect Ozinside the flaat 1 second in
ting to be detble air to be dd near the pi
o be off the c at 1.5 secondy pack had to
utput port thnalyzer takes second perio
one. To do thsh head. Wentervals for a
tected. The Odrawn from thvot/hinge.
camera runnind intervals ato be used to e
hat was connesamples of aiod.
his we removee then set thebout 40 seco
Ozone analyzehe flash head
ng via its ownt maximum poenable firing t
ected to an osir every 10 se
ed the glass ce camera thatonds.
er samples aid, we had to d
n optical triggower at 1/80this fast.
scilloscope toeconds and ou
cover plate tht was connect
r at a rate of drill a tiny air
ger input. We00th shutter s
o show the buutputs a new
hat ted to
vent
e set speed
uild‐
As can betriggeringis used asefficient H
Shown be
Figure 15control sypossible w
Since the energy thflash head
Our theorII is what
To furthe(see Figur
seen in the sg has a much f the trigger. HSS timing co
elow (fig 15) a
: Showing fuystem. The rewhen using th
light output wat was delived was half as
retical conclucauses most
r confirm thisre 16 on follow
scope traces sfaster buildupThis lower rantrol system
are the two lig
ll power HSS ed trace showhe PocketWiz
when using thered by the flamuch when u
sion is that thof the ozone
s, we also firewing page)
shown in Figup of Ozone inte of ozone bused by the P
ght pulse out
light output fws the shorterard FlexTT5 a
he PocketWizash tube was using the Poc
he rapid firingbuildup and
ed a long serie
ure 14, the Caside the flashbuildup is likePocketWizard
tputs as meas
from 580EX IIr firing time bas the triggeri
zard radio wathe same butketWizard Fle
g of the 10KVnot the brigh
es of non‐HSS
anon native syh head than wely related to d ControlTL sy
sured by a Th
I when fired vbut brighter ming system.
as brighter but the resultinexTT5.
V trigger circuhtness of the
S full power li
ystem using twhen the Pocthe much moystem.
or Labs DET3
via Canon’s nmaximum out
ut half the durg ozone build
it repeatedly light output.
ight pulses fo
their own optketWizard Fleore energy
6A light sens
ative optical tput that is
ration, the todup inside the
inside the 58
or comparison
tical exTT5
or.
otal e
80EX
n.
Figure 16buildup wbrightnes
Test modeCanon’s nthe IGBT fsystem m
However ring arounlight emitfull poweHSS pulse
We are prmuch thetrigger rinresearche
: When firingwas nearly zers of Canon’s
e in FlexTT5 fnative pulse dfail. We had ight have inc
the tests did nd the flash tted by the flar non‐HSS flae has only a si
reparing to ru ozone will chng to the refleer in the field
g the flash usiro. However native optica
firmware 5.21uration and bthought that reased the ch
not have thaube were likeash since a lonsh pulse is farngle firing of
un tests at thehange the eleector metal hof ozone and
ing only a fullwhen we trielly triggered H
12 (never a pubrightness of perhaps the hances of a fa
t outcome. Wely the more snger durationr brighter andthe trigger ri
e University oectrical breakdousing. Chemd will be assist
We found thasignificant can HSS pulse hd yet generatng on the flas
of Vermont (Udown of air smistry profesting us to furt
p in non‐HSS te the exact sah) the ozone b
) was to test oETTL system t output capa
at the 10KV eluse of the ozas many mortes nearly zersh tube at the
UVM) in the cuch that an essor Giuseppether our rese
(below X‐Syname pulse dubuildup was m
our theory thmight be lessability of the P
lectrical pulsezone formatiore firings of tho ozone. Thie beginning o
coming weekselectrical spare Petrucci of Uearch and the
nc) the ozone uration and much greater
at duplicatings prone to haPocketWizard
es from the tron and not thhe trigger rings full power nof the light ou
s to find out hrk will jump frUVM is a ories.
r.
g ving d
rigger e UV g. A non‐utput.
how rom
Preflash Boost:
PocketWizard introduced back in April of 2009 a firmware feature that enables a brighter preflash to be selected by the user. The default was to have the brighter preflash enabled.
In September of 2010 as part of our on‐going research into this IGBT failure problem in 580EX II flashes, we tried a test where we changed the default of the preflash to be disabled.
After 5 months of collecting data since defaulting to preflash boost OFF, we have found there was no change in reported failure rate since. In fact, the preflash boost was confirmed to be disabled in most of our recent reports.
Our current conclusion so far is that the Preflash Boost feature is not related to the 580EXII’s IGBT failure.
An important thing to note is that Canon’s preflash is a short form of HSS. It is just a 400us long HSS
pulse. So a preflash can generate a quantity of the high voltage pulses similar to the HSS main flash
mode. So even if a user is shooting at or below X‐Sync with non‐HSS exposures, the preflash may still
be generating ozone.
Further tests will be needed to determine if preflashes alone could generate enough ozone to cause the
failure even if we disable any main exposure light output.
Other testing:
We conducted many experiments around heat, static, and other factors.
Using an ESD generator, we discharged 11,000 – 20,000 volts into many places on a 580EX II both alone
and mounted on a FlexTT5 with and without the AC5. No issues were noted. The 580EX II and FlexTT5
are both CE marked products which means they must be able to handle a certain level of ESD to qualify
for sale in European countries.
Heat tests over long periods, “torture tests” of hundreds of triggers in various configurations, and
electrical short tests on the shoe pins were conducted.
None of these tests yielded any repeatable failure modes or conclusive data around IGBT failure.
As part of
designed.
people. W
the failure
occurred
The Cano
Figure 17
reflector t
theoretica
The rubbe
f the research
The newest
While we have
e was actually
right after ini
n 430EX II Fla
: The 430EX I
than used in t
ally for extra
er end seal &
T
h efforts we a
430EX II flash
e had a single
y related to a
itial purchase
ash Head Ass
I has 3 times
the 580EX II.
venting and c
mount used
he 430EXII do
also wanted to
h from Canon
e failure repo
n IGBT or if it
e, it may have
sembly:
more air gap
Additionally,
clearance from
on the 580EX
oes not dama
o review a few
n is one that i
rt, the flash w
t had any rela
e been a manu
space betwe
, a cutout wa
m electrical a
X II is eliminat
age IGBTs.
w other flash
is working ext
was not sent t
ation to ozone
ufacturing de
een the trigge
s added direc
arcing.
ted on the 43
hes and look a
tremely well
to us and we
e. Since this
efect.
er ring wire an
ctly below the
30EX II.
at how they w
and reliable f
could not co
one failure
nd the metal
e flash tube
were
for
nfirm
Figure 18
flash uses
The 430EX
It has bec
improvem
models.
The older
that the f
to get acc
optical fee
problems
: The 430EX
s.
X II was desig
ome LPA Des
ment of the de
Canon 580EX
lash tube is h
cidentally alig
edback senso
.
II does not us
gned and start
sign’s theory t
esign on thes
X flash does h
eld in positio
ned in the wr
or in the 580E
se a sealed en
ted shipping
that design p
e elements to
have a head d
on directly by
rong position
EX is very diffe
nd cap to hold
before Pocke
roblems with
o prevent fail
design that is
a printed circ
by hand asse
erent and ma
d the flash tu
etWizard Cont
h the 580EX II
lures of a sim
somewhat si
cuit board an
embly worke
ay be less susc
be in position
trolTL produc
were realize
milar nature on
imilar to the 5
d has no way
rs at a factory
ceptible to ha
n like the 580
cts existed.
ed and needed
n future flash
580EX II, exce
y for the flash
y. Also, the
and assembly
0EX II
d
h
ept
h tube
y
Conclusions: (The Perfect Storm)
It appears that some combination of elements comes together to create the risk and increased
probability of an IGBT failure within the 580EX II.
