D300 built-in flash, Part 1: Blinding speed

Started Dec 25, 2008 | Discussions
nikonsrme Senior Member • Posts: 1,461
Thanks again Marianne

Thats a pretty clever idea to monitor the mirror movement. The 6ms delay in manual mode is interesting. When you measured that was the exposure mode in manual also?

I use FV lock quite often as I do find kids escpecially to be fast enough to be just starting to close their eyes when the shutter opens after the preflashes. With these traces you can see exacly how fast those kids have to be. This is one of the most interesting threads I have read here in a long time.

BTW, not sure which HP calculator that is but it looks similar to mine, a 32S II RPN.

I hope you have a "Happy New Year" and keep up the good work.

-- hide signature --

'You can observe a lot just by watching.' YOGI BERRA

Jerry Faircloth
Pbase Supporter
http://www.pbase.com/nikonsrme

 nikonsrme's gear list:nikonsrme's gear list
Nikon D300 Nikon D810 Nikon AF-S Teleconverter TC-14E II Nikon AF-S Teleconverter TC-17E II Nikon AF-S Nikkor 24-120mm F4G ED VR +1 more
OP Marianne Oelund Veteran Member • Posts: 7,777
Camera used

OleThorsen wrote:

What kind of camera did you use for the shutter sequence pictures
giving you down to 0.37ms between shots (2700fps)?

D3, of course - it has plenty of speed to do this.

David Gay Regular Member • Posts: 178
Re: Thanks again Marianne

Seconding the very nice thread comment, and...

nikonsrme wrote:

BTW, not sure which HP calculator that is but it looks similar to
mine, a 32S II RPN.

It's an HP48S(X?)...

David Gay

clarknj_1234 Regular Member • Posts: 185
Re: Thanks again Marianne

Thanx Marianne
--
Nigel

OleThorsen Veteran Member • Posts: 3,111
A wild guess

Marianne Oelund wrote:

OleThorsen wrote:

What kind of camera did you use for the shutter sequence pictures
giving you down to 0.37ms between shots (2700fps)?

D3, of course - it has plenty of speed to do this.

Hehe, of course?!?

A wild guess would then be several rounds of 11fps output from the D3, so you would have enough frames to cover the whole shutter sequence. That would also explain the random gaps between the captures. But how did you then get the timings of the single frames?

Regards
Ole Thorsen
http://www.pbase.com/ole_thorsen

  • OMNISCIENCE

Knowing what
thou knowest not
is in a sense
omniscience.
(Grook by Piet Hein)

 OleThorsen's gear list:OleThorsen's gear list
Nikon D70 Nikon D200 Nikon D300 Nikon 1 V1 Nikon 1 V3 +32 more
veejay chatterjee Regular Member • Posts: 157
Re: D300 built-in flash, Part 1: Blinding speed

A most interesting series, I am appreciating it.
--
Veejay Chatterjee

pwilly
pwilly Senior Member • Posts: 1,171
Timing by Tektronix?

Marianne uses oscilloscopes to time her experiments.
--
Paul
Just an old dos guy

OP Marianne Oelund Veteran Member • Posts: 7,777
Part 4: Commander Mode Timings

Basic Pulse and Data Encoding

Commands and data sent by the D300's built-in flash in Commander Mode use a low-power pulse of duration 12-15us, which is 1/320 of full power, or 1.3 stops lower in energy than a Manual mode 1/128 power pulse. The only exceptions to this are the camera's own monitor flashes if TTL mode is used (which are 1/16 and 1/64 power), and the flash used during the actual exposure. In the latter case, the camera will revert to the 1/320 power flash pulse if the built-in flash is set to Off, in order to synchronize the remote flash units.

This brings up a question of CLS reliability if the built-in flash is set to TTL mode and fires at very low power during the exposure. As shown in earlier postings, this pulse can be as short as 4us and have a power of only 1/3300 of full power. In my own experiments, CLS operation is reliable at distances up to at least 20 feet, even when the remote flash sensors are not facing the D300, but I would still recommend setting the built-in flash to Off instead of TTL in cases where flash power will be very low.

Since the data sent by the D300 flash for Commander Mode uses pulses of uniform amplitude, the information is encoded by pulse interval alone. Intervals range from 150us to 600us, in steps of 75us.

Maximum-Length Command Sequence

Let's jump in with both feet, and start with a look at the most complex command sequence, which corresponds to all groups (including the D300 itself) set to TTL mode. Note that EV and power settings do not affect the length of the command sequence.

Here is a trace of the sequence, with marks and labels denoting the segments as discussed below. As before, you can make timing measurements from these traces, since 100 pixels horizontally corresponds to 20ms:

Here is the content and range of timings for each of the command segments:
1. Shutter Release Delay

This is quite variable, and can be anywhere from 2ms to 17ms. There is no correlation between the length of this delay, and the details of the command sequence itself. The variation is random, and you will see a number of different values in the scope traces here.
2. Channel Number and Flash Mode Data

This cluster of pulses contains the channel number, and the user's selected modes for the external flash groups. It consists of about 7 to 12 individual pulses, and takes 2.2ms to send. Following the cluster, there is a processing delay of 5-6ms.
3. D300 Flash Monitor Pulses

If TTL mode is selected for the built-in flash, the monitor pulses will appear here. The first is at 1/16 power, and a second pulse at 1/64 power may also be used if the subject returns a high level of light from the first pulse. The total time for this segment is 18-19ms if two pulses are used, or 8-9ms for a single pulse. This segment will be skipped if the built-in flash is set to Manual, or Off.
4. Group A Monitor Pulses

This segment begins with a short (5-6 pulse) cluster which identifies the group being addressed (group A). This is followed by a pair of pulses which trigger the remote flash(es) to produce a 20us monitor pulse. If in TTL mode and this monitor pulse is insufficient, a second one (which will be 120us) is requested 10ms later. Total time for this segment is
25-26ms for TTL mode if two monitor pulses are used,
15-16ms for TTL mode if one monitor pulse is used, or
21-22ms for AA mode (always single monitor pulse).
This segment is skipped if Group A mode is Manual or Off.

5. Group B Monitor Pulses
This segment is as described for Group A above.

6. Power Level Data

If a TTL exposure mode is used for Group A or Group B, this segment begins with a 6ms processing delay. There is no processing delay if Manual mode only is used for external groups. Following this is a cluster of pulses which communicate the calculated power levels to the remote flashes. The cluster duration is 2.2ms for a single external flash group, or 3.3ms for two groups, thus the total time for this segment is 2.2ms, 3.3ms, 8.2ms or 9.3ms.
7. Shutter Cycle

This is the mechanical shutter cycle. For 14-bit recording, time from the start of this segment to exposure is 90ms; for 12-bit RAW or JPEG, it is 47ms.

Summary

Using the above data, you can now calculate the maximum and minimum shutter delays for any combination of Commander Mode settings - just add together the max or min timings for each of the 7 segments. For example, maximum shutter delay for the all-TTL Commander Mode sequence shown above is (17 + 8 + 19 + 26 + 26 + 9 + 90)ms = 195ms for 14-bit RAW, or 152ms for 12-bit RAW or JPEG.

Example Variations

Following are four examples of shorter Commander sequences. Each of these are the same as the maximum-time all-TTL sequence described above, except for one setting parameter, which is listed on the tracing. The segment which is affected by the changed setting is between the a and b cursors. All of these traces were taken with 14-bit recording.

Speeding Things Up

With some subjects, the long Commander Mode delays and pre-flash sequences will be problematic. One solution is to use Manual mode only for all groups, which eliminates most of the communication delay, as shown here:

FV Lock with Commander Mode

Use of this feature provides best speed, and retains the mode flexibility of CLS. When FV lock is entered, the D300 executes the full Commander sequence, but ends with a special power-level data cluster (there is no 6ms processing delay associated with this). When the shutter release is subsequently pressed, the D300 only needs to send the actual power-level data cluster (2.2ms or 3.3ms), then immediately starts the mechanical shutter cycle.

David314 Veteran Member • Posts: 4,790
SB800 and bands

nice shutter sequence pictures

Marianne Oelund wrote:

Although
they pertain to use of the SB-800 flash, the effect on D300 built-in
flash use is very similar. See
http://forums.dpreview.com/forums/read.asp?forum=1021&message=30453868
If you will be using flash with your D300 at the highest 1/320 sec
sync speed and high flash power settings, I recommend that you
perform some experiments to determine if the curtain shadow problem
will be significant to you.

as far as i can tell with the speed lights turn the flash off and on quick enough to avoid the dark band

in fact i would expect to see a bright bands on the edges if the flash was slow to turn off or quick to turn on

but perhaps i am missing something

for instance SB800 at full power but 1/250 shutter

OP Marianne Oelund Veteran Member • Posts: 7,777
Close, but not quite there
1

David314 wrote:

as far as i can tell with the speed lights turn the flash off and on
quick enough to avoid the dark band

Not quite. At 1/250 shutter, since the curtain transit time is 2.4ms (and because the camera does not initiate the flash pulse before the shutter is fully open), there is at the very most 4.0ms - 2.4ms = 1.6ms for a flash pulse without shadowing. As shown by both your trace and my measurements, the flash continues producing light until 1.8ms after it's triggered. Thus the rear curtain has started closing over the image sensor for at least 0.2ms by the time the flash turns off, and this results in a narrow band at the top edge of the image that is slightly darker.

You can demonstrate this for yourself. Take two photos of a uniform surface, one with the SB800 at full power, and another at half power (both at 1/250 shutter). Then display the images side-by-side on your computer, and you will see the darkening at the top in the full-power image.

for instance SB800 at full power but 1/250 shutter

Note that although the flash has been set for full power, it is not actually producing full power, as it has been limited in duration by the camera. A full power discharge continues to decay exponentially for many milliseconds and exhibits no cutoff edge.

If you want to see the true peak characteristic of the flash output, you need to place your detector a bit farther away from the flash, as it's clipped the signal.

Edward Amarante Regular Member • Posts: 248
VERY impressive posts, Thanks n/t
-- hide signature --

n/t

David314 Veteran Member • Posts: 4,790
Re: Close, but not quite there

Marianne Oelund wrote:

Then display the images side-by-side on
your computer, and you will see the darkening at the top in the
full-power image.

ah yes, the curtain closing makes it darker yes, i see that now and yes you can see it in side by side pictures - note black on bottom is a black border of a screen

it is interesting to note that you can turn the tables on where the darkening occurs by using FP flash with a radio trigger or shooting full power on the flash through SU-4 mode

A full power discharge continues to decay exponentially

for many milliseconds and exhibits no cutoff edge.
If you want to see the true peak characteristic of the flash output,
you need to place your detector a bit farther away from the flash, as
it's clipped the signal.

no it wasn't clipped by the measurement - it was turned off by the flash unit - full power same setup just 1/60 instead of 1/250

Kluso Senior Member • Posts: 1,904
It was clipped - you just cant see it

David314 wrote:

Marianne Oelund wrote:

If you want to see the true peak characteristic of the flash output,
you need to place your detector a bit farther away from the flash, as
it's clipped the signal.

no it wasn't clipped by the measurement - it was turned off by the
flash unit - full power same setup just 1/60 instead of 1/250

I tried in vain to tell you it was saturated 4 months ago, see here

http://forums.dpreview.com/forums/read.asp?forum=1034&message=29422415

The point you are missing is that you are using a device with an analog output with gain over which you have no control.

You dont even know what the schematic of the internal circuit is - you are NOT seeing the true peak characteristic of the flash output -

See how the OP's traces are not flat ever at the top.

-- hide signature --

Inspector Kluso

OP Marianne Oelund Veteran Member • Posts: 7,777
Detector Range

David:

Thanks for sharing your results, which are a nice addition to the information on this thread.

David314 wrote:

no it wasn't clipped by the measurement - it was turned off by the
flash unit

Clipping refers to the vertical axis, i.e., the signal amplitude. Any real-world device has an amplitude limit, and clipping results when a signal attempts to exceed that limit. The amplitude limit of your detector is 1.6V, and if a signal approaches that level, it won't be represented accurately.

In order to obtain waveform fidelity, all you need to do is move your detector back from the flash a bit, and compensate by increasing the scope's vertical sensitivity. Here is an example of an over-driven detector, and restoring waveform fidelity by reducing the light input:

David314 Veteran Member • Posts: 4,790
you are back?

Kluso wrote:

I tried in vain to tell you it was saturated 4 months ago, see here

http://forums.dpreview.com/forums/read.asp?forum=1034&message=29422415

See how the OP's traces are not flat ever at the top.

yes and i tried to explain how it made no difference in measuring the duration of the pulse and how you don't understand the circuit i was using either

notice the OP's measurements agree with mine for full duration flashes and showed you plenty of evidence that full duration flashes are longer that 1/1000 of a second and tried in VAIN to explain this to you

now you are back home you can show your proof that the full power flash on a Sb800 is only 1/1000 of a second

David314 Veteran Member • Posts: 4,790
Re: Detector Range

Marianne Oelund wrote:
David:

Thanks for sharing your results, which are a nice addition to the
information on this thread.

David314 wrote:

no it wasn't clipped by the measurement - it was turned off by the
flash unit

Clipping refers to the vertical axis, i.e., the signal amplitude.
Any real-world device has an amplitude limit, and clipping results
when a signal attempts to exceed that limit. The amplitude limit of
your detector is 1.6V, and if a signal approaches that level, it
won't be represented accurately.

regardless of the input and distance, this light sensor tends to show a flatter top than i have been seeing from other scope traces i have seen on the web

in this particular trace i was intent on capturing the duration of the pulse and this device seemed to have a minimum threshold affect so i had to pick my poison so to speak because of a limited dynamic range i suppose

there also maybe an adjustment in the light response curve hard to tell

i do know when it truly saturates, it flat lines and makes a perfectly smooth line

again, this is not a photo diode, but a light sensor with an analog output

at any rate, i think what is of most interest is the duration of the flash and you have shown once again very nicely in your original post the various flash lengths

my 'friend' Kluso seems to think that full power flash on the SB800 is only 1/1000 in duration

I am glad to see yet another scope trace confirming that the pulse is much longer in duration and behaves as a decaying exponential

good shooting

David

Kluso Senior Member • Posts: 1,904
Re: you are back?

David314 wrote:

Kluso wrote:

I tried in vain to tell you it was saturated 4 months ago, see here

http://forums.dpreview.com/forums/read.asp?forum=1034&message=29422415

See how the OP's traces are not flat ever at the top.

yes and i tried to explain how it made no difference in measuring the
duration of the pulse and how you don't understand the circuit i was
using either

You did write that however just because you say it is so or belive it is so does not make it so.

notice the OP's measurements agree with mine for full duration

No, they do not, the OP's traces are as expected, yours are clipped.

Here are my traces, 2 versions clipped and not clipped

http://forums.dpreview.com/forums/read.asp?forum=1034&message=30550321

flashes and showed you plenty of evidence that full duration flashes
are longer that 1/1000 of a second and tried in VAIN to explain this
to you

now you are back home you can show your proof that the full power
flash on a Sb800 is only 1/1000 of a second

-- hide signature --

Inspector Kluso

David314 Veteran Member • Posts: 4,790
ok, lets see your 1/2 power trace

Kluso wrote:

notice the OP's measurements agree with mine for full duration

No, they do not, the OP's traces are as expected, yours are clipped.

the OP's traces clearly show that for full power there is a decaying exponential function that exceeds 1/1000

here is the OP's view of a clipped trace and not

notice - in spite of the clipping the length of the pulse is the same

btw this is a 1/2 power pusle -

Here are my traces, 2 versions clipped and not clipped

and here is your full power curve

you attribute the slow decay to lack of active circuitry to pull down the signal

that maybe true - to verify that we would have to see a 1/2 power or less pulse

but note the OP's circuit appears to be an active circuit - you can see this from the 1/2 power pulse

so just forget my scope traces - notice in the OP's traces the difference between 1/2 power and full power - if the full power is affected by the capacitance of the circuit and is wrongly showing a decaying exponential, how is it that the 1/2 power and less pulses decay so fast?

i suggest you go back into the lab and capture 1/2 power and less pulses along with a full power pulse

but please read this first from the OP

"Triggering and Power Control

Flashes work by discharging a storage capacitor into the flash tube. When full power is used, the discharge is allowed to run its full course. To obtain lower powers, the pulse is shut off mid-way. Thus whatever "power level" is used, the pulse profile follows the same curve, with lower-power pulses simply being shorter.

Once the storage capacitor is charged and ready, the process begins with the application of a high-voltage trigger pulse to a trigger electrode on the flash tube. After this, it takes about 10-12usec for the flash tube to start producing light. The initial build-up is exponential until about 15usec after trigger (light level is still very low here) and the light output then ramps up roughly linearly until about 30us, where the light level is at half peak brightness. After that, it rounds off slowly to its peak value at 120-140us, then begins an exponential decay as the storage capacitor loses voltage. "

or this from the OP in regards to full power flash at 1/250 shutter speedr

"Note that although the flash has been set for full power, it is not actually producing full power, as it has been limited in duration by the camera. A full power discharge continues to decay exponentially for many milliseconds and exhibits no cutoff edge."

so now you better work hard on discrediting Marianne Oelund's results

David314 Veteran Member • Posts: 4,790
Another comment

Marianne Oelund wrote:

D300 Shutter Timing
First, let's have a close look at the D300 shutter dynamics.

i just read how you used a D3 to capture these images using flashes to measure the relative timing?

Did you try using the sync output? just put a pull up on it and you would see the start stop pulse

also, it would be nice to see the flash in the shutter image - i assume that was cropped out?

you did an immense amount of work on this - well done

makes one want to buy one of those casio high speed cameras and shoot the shutter with light from the flash visible

good shooting

David

Kluso Senior Member • Posts: 1,904
Sure - in the other thread

http://forums.dpreview.com/forums/read.asp?forum=1034&message=30551677
I do not wish to hijack the OP's excellent thread with which I find no fault.
--
Inspector Kluso

Keyboard shortcuts:
FForum MMy threads