What does an accelerator chip do?

stevebrot · 07-24-2020, 12:59 PM

Originally posted by luftfluss

I've had several shots with my K-70 where there was oddness in the rendering of fine detail. They were shot late in the evening at ISO 6400 (RAW), where bits of fox fur and grass were rendered as small smudgy blocks. None of the shots were keepers, anyway, so no harm no foul.

I remember your post of those.

Steve

photoptimist · 07-24-2020, 01:08 PM - **5 Likes**

The effects of the accelerator on the power spectrum of the noise (Baffling Pentax KP Read Noise: Photographic Science and Technology Forum: Digital Photography Review) and it's strong suppression of chromatic noise suggest the chip analyzing the neighborhoods around every pixel (probably over about a 5 pixel diameter) to statistically adjust "impossible" or "unlikely" RGB combinations seen in the RAW sensor data.

The color filters on the sensor are not perfect. Every wavelength of light induces some signal in the sensor in every measured color band. Thus a RAW RGB value of [0,0,20] from the sensor, for example, is impossible -- there's not enough red and green for the measured level of blue. The chip would probably assume that the red and green channels should have been a little higher and maybe the blue channel might be a little lower so that [0,0,20] becomes something like [1,2,17]. This kind of processing would not eat stars, would have little or no impact on most details.

The accelerator chip is almost certainly a hard-coded, specialized device, not a "fast" general purpose processor. The chip probably only does one task -- take the stream of RAW data from the main sensor and crank out an adjusted RAW signal that goes to the CPU. In fact, it probably does this task BEFORE the RAW data even goes into RAM (storing the RAW data, retrieving it from RAM, processing it, and storing the results would add time and slow down the camera). Instead, it looks like the digital data goes directly from the ADC into the chip, is processed in a stream, and comes out. The CPU and RAM get what they think is RAW sensor data.

**FozzFoster** · 07-24-2020, 01:26 PM - **2 Likes**

Originally posted by photoptimist

The CPU and RAM get what they think is RAW sensor data.

Damnit. They can think for themselves now? I knew this day would come!

*cough* I, for one, welcome our new robotic overlords... *cough*

Not a Number · 07-24-2020, 02:31 PM

Just going from memory of the various videos, interviews and most significantly an interview in one Japanese eMagazine the AC is doing more than just noise reduction. Claims of being able to render blue colors better and that processing was tweaked for each individual ISO setting. Ricoh has made the same claim as they do with the Limited lens series of factors that are not easily measured numerically. Graphs that show the AC doesn't "kick" in until ISO 640 bear out this claim since Ricoh says processing is going on at every ISO setting.

**clickclick** · 07-24-2020, 06:26 PM

Based on the K-1 to K-1 II upgrade, i believe the accelerator may have impact on focus speed, as this was one of the changes in performance with the upgrade. I can't say if this is because processing was moved off other chips to the accelerator allowing them to commit more resources to the focusing, or if the accelerator is providing the boost, but there definitely was a change. A test here on PF made some real measurements of the differences in speed.

Pentax K-1 II vs K-1 Autofocus Test - Hands-On Tests | PentaxForums.com

So, does anyone know for sure if the AF change came from the accelerator, or was it a side effect? My recollection of the circuit boards we saw etc. was the accelerator chip addition was the core of the upgrade. Well, that and the little "II" logo where it used to say "SR".

**StiffLegged** · 07-25-2020, 12:29 AM - **1 Like**

Originally posted by aslyfox

I wonder

is it physically located to the right of the brake chip

Stuff and nonsense, it’s the twist grip under your right hand as you hold it correctly!

**mikesbike** · 07-25-2020, 07:19 PM

The description offered by photoptimist appears to make logical sense. It might also show good reason for the lack of actual specifics coming out, due to being protective of design secrets. My impression has been, whatever it is they've come up with- it works! I've been very happy with images out of my K-5 IIs, but the KP keeps giving me outstanding results! I have found gains in fine detail instead of losing any.

Last edited by mikesbike; 07-25-2020 at 07:45 PM.

stevebrot · 07-25-2020, 08:09 PM

Originally posted by mikesbike

the KP keeps giving me outstanding results! I have found gains in fine detail instead of losing any.

The significant increase in pixel resolution might also be a consideration.

Steve

**Kerrowdown** · 07-26-2020, 01:50 AM - **2 Likes**

Originally posted by Cerebum

What does an accelerator chip do?

Keeps you on the ground around fast corners... but that might not be the correct answer.

**mikesbike** · 07-27-2020, 01:52 PM

Originally posted by stevebrot

The significant increase in pixel resolution might also be a consideration.

Steve

Probably so. But at least this gain is then a visible gain, not ruined or compromised by the accelerator chip!

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Is this describing what the accelerator unit does?	JPT	Pentax DSLR Discussion	14	10-01-2019 05:24 AM
Pentax K-1 Mark II accelerator unit vs. denoise filters in Darktable	tax	Pentax K-1 & K-1 II	20	09-03-2018 06:09 AM
Accelerator	reh321	Pentax KP	25	05-02-2018 02:51 PM
Pentax K-70 with "accelerator unit"	Tau-Ceti	Pentax News and Rumors	89	08-27-2016 05:03 PM
Machinery Particle accelerator	DeadJohn	Post Your Photos!	10	08-05-2015 04:08 PM

07-24-2020, 12:59 PM	#16
stevebrot Otis Memorial Pentaxian Join Date: Mar 2007 Location: Vancouver (USA) Photos: Gallery \| Albums Posts: 42,007	Originally posted by luftfluss I've had several shots with my K-70 where there was oddness in the rendering of fine detail. They were shot late in the evening at ISO 6400 (RAW), where bits of fox fur and grass were rendered as small smudgy blocks. None of the shots were keepers, anyway, so no harm no foul. I remember your post of those. Steve

07-24-2020, 01:08 PM - 5 Likes	#17
photoptimist Pentaxian Join Date: Jul 2016 Photos: Albums Posts: 5,122	The effects of the accelerator on the power spectrum of the noise (Baffling Pentax KP Read Noise: Photographic Science and Technology Forum: Digital Photography Review) and it's strong suppression of chromatic noise suggest the chip analyzing the neighborhoods around every pixel (probably over about a 5 pixel diameter) to statistically adjust "impossible" or "unlikely" RGB combinations seen in the RAW sensor data. The color filters on the sensor are not perfect. Every wavelength of light induces some signal in the sensor in every measured color band. Thus a RAW RGB value of [0,0,20] from the sensor, for example, is impossible -- there's not enough red and green for the measured level of blue. The chip would probably assume that the red and green channels should have been a little higher and maybe the blue channel might be a little lower so that [0,0,20] becomes something like [1,2,17]. This kind of processing would not eat stars, would have little or no impact on most details. The accelerator chip is almost certainly a hard-coded, specialized device, not a "fast" general purpose processor. The chip probably only does one task -- take the stream of RAW data from the main sensor and crank out an adjusted RAW signal that goes to the CPU. In fact, it probably does this task BEFORE the RAW data even goes into RAM (storing the RAW data, retrieving it from RAM, processing it, and storing the results would add time and slow down the camera). Instead, it looks like the digital data goes directly from the ADC into the chip, is processed in a stream, and comes out. The CPU and RAM get what they think is RAW sensor data.

07-24-2020, 01:26 PM - 2 Likes	#18
FozzFoster Custom User Title Join Date: Jan 2016 Location: Alberta Photos: Albums Posts: 6,806	Originally posted by photoptimist The CPU and RAM get what they think is RAW sensor data. Damnit. They can think for themselves now? I knew this day would come! cough I, for one, welcome our new robotic overlords... cough

07-24-2020, 02:31 PM	#19
Not a Number Moderator Join Date: Mar 2012 Location: Venice, CA Posts: 10,526	Just going from memory of the various videos, interviews and most significantly an interview in one Japanese eMagazine the AC is doing more than just noise reduction. Claims of being able to render blue colors better and that processing was tweaked for each individual ISO setting. Ricoh has made the same claim as they do with the Limited lens series of factors that are not easily measured numerically. Graphs that show the AC doesn't "kick" in until ISO 640 bear out this claim since Ricoh says processing is going on at every ISO setting.

07-24-2020, 06:26 PM	#20
clickclick Loyal Site Supporter Join Date: May 2016 Location: East Coast Posts: 2,903	Based on the K-1 to K-1 II upgrade, i believe the accelerator may have impact on focus speed, as this was one of the changes in performance with the upgrade. I can't say if this is because processing was moved off other chips to the accelerator allowing them to commit more resources to the focusing, or if the accelerator is providing the boost, but there definitely was a change. A test here on PF made some real measurements of the differences in speed. Pentax K-1 II vs K-1 Autofocus Test - Hands-On Tests \| PentaxForums.com So, does anyone know for sure if the AF change came from the accelerator, or was it a side effect? My recollection of the circuit boards we saw etc. was the accelerator chip addition was the core of the upgrade. Well, that and the little "II" logo where it used to say "SR".

07-25-2020, 12:29 AM - 1 Like	#21
StiffLegged Site Supporter Join Date: Jan 2018 Photos: Gallery Posts: 4,616	Originally posted by aslyfox I wonder is it physically located to the right of the brake chip Stuff and nonsense, it’s the twist grip under your right hand as you hold it correctly!

07-25-2020, 07:19 PM	#22
mikesbike Site Supporter Join Date: Dec 2016 Location: Southeastern Michigan Posts: 4,549	The description offered by photoptimist appears to make logical sense. It might also show good reason for the lack of actual specifics coming out, due to being protective of design secrets. My impression has been, whatever it is they've come up with- it works! I've been very happy with images out of my K-5 IIs, but the KP keeps giving me outstanding results! I have found gains in fine detail instead of losing any. Last edited by mikesbike; 07-25-2020 at 07:45 PM.

07-25-2020, 08:09 PM	#23
stevebrot Otis Memorial Pentaxian Join Date: Mar 2007 Location: Vancouver (USA) Photos: Gallery \| Albums Posts: 42,007	Originally posted by mikesbike the KP keeps giving me outstanding results! I have found gains in fine detail instead of losing any. The significant increase in pixel resolution might also be a consideration. Steve

07-26-2020, 01:50 AM - 2 Likes	#24
Kerrowdown PEG Moderator Join Date: Aug 2008 Location: Highlands of Scotland... "Hold Infinity in the palm of your hand" - William Blake Photos: Gallery Posts: 57,816	Originally posted by Cerebum What does an accelerator chip do? Keeps you on the ground around fast corners... but that might not be the correct answer.

07-27-2020, 01:52 PM	#25
mikesbike Site Supporter Join Date: Dec 2016 Location: Southeastern Michigan Posts: 4,549	Originally posted by stevebrot The significant increase in pixel resolution might also be a consideration. Steve Probably so. But at least this gain is then a visible gain, not ruined or compromised by the accelerator chip!