Pentax engineers are usually more open talking about technical problems than their collegues from other companies. OK1000 Pentax blog translated an interview about K10D DSLR design and decision process at Pentax.
A few cuts from the interview showing problems with camera board design:
"It doesn’t take long to produce an acceptable level of quality images at the base sensitivity (ISO 100), however, as the sensitivity increases step by step, at some point, terrible images will be produced. The real effort starts from this point, trying to pin point possible reasons, mainly by trial and error, often having to make new Cbs (circuit boards). At the same time, firmware is developing along and each group has to closely coordinate in order to slowly improve image quality. Fine tuning can only commence after this point.
The frequency of designing new circuit boards might be hardest hurdle in the whole K10D project.
Homogeneous noise across the whole image could often be permitted, but countermeasures against localized noise is difficult to achieve.
Of course, each case is closely analyzed and dealt with, but there are a number of cases of localized noise and it is really time consuming to solve these. Changing Cb’s is one measure, but that has to be repeated many times.
We were fully aware that we were going to face very critical eyes on the noise issue because of increased resolution. In addition, the base sensitivity of the sensor was lower, which made tuning to the acceptable level for production very difficult.
Recent Cbs might look at one piece, but there are actually layers to this and a slight change in wiring alone significantly changes how noise appears. In a digicam, DDR2 memory that functions at a very high frequency and many other parts all produce radio waves. A TV tuner portion of a videocard for PC, for example, is tightly shielded in order to avoid the interference between the digital circuit and analog circuit. If noise enters before signal’s digitization, the digital image output would almost be unusable. When increasing the gain for sensitivity, at ISO1600 for example (4 stops over base sensitivity), just a slight voltage variation in the order of a few milli-volts will impose a huge impact on image quality.
Separating grounding circuits is one way, but it is not simply a matter of separation. Sometimes, good results could be obtained by a larger grounding area. This is another area requiring the trial and error method."
Some other design trade-offs and considerations of DSLR image sensor customers make this interview worth reading.
A few cuts from the interview showing problems with camera board design:
"It doesn’t take long to produce an acceptable level of quality images at the base sensitivity (ISO 100), however, as the sensitivity increases step by step, at some point, terrible images will be produced. The real effort starts from this point, trying to pin point possible reasons, mainly by trial and error, often having to make new Cbs (circuit boards). At the same time, firmware is developing along and each group has to closely coordinate in order to slowly improve image quality. Fine tuning can only commence after this point.
The frequency of designing new circuit boards might be hardest hurdle in the whole K10D project.
Homogeneous noise across the whole image could often be permitted, but countermeasures against localized noise is difficult to achieve.
Of course, each case is closely analyzed and dealt with, but there are a number of cases of localized noise and it is really time consuming to solve these. Changing Cb’s is one measure, but that has to be repeated many times.
We were fully aware that we were going to face very critical eyes on the noise issue because of increased resolution. In addition, the base sensitivity of the sensor was lower, which made tuning to the acceptable level for production very difficult.
Recent Cbs might look at one piece, but there are actually layers to this and a slight change in wiring alone significantly changes how noise appears. In a digicam, DDR2 memory that functions at a very high frequency and many other parts all produce radio waves. A TV tuner portion of a videocard for PC, for example, is tightly shielded in order to avoid the interference between the digital circuit and analog circuit. If noise enters before signal’s digitization, the digital image output would almost be unusable. When increasing the gain for sensitivity, at ISO1600 for example (4 stops over base sensitivity), just a slight voltage variation in the order of a few milli-volts will impose a huge impact on image quality.
Separating grounding circuits is one way, but it is not simply a matter of separation. Sometimes, good results could be obtained by a larger grounding area. This is another area requiring the trial and error method."
Some other design trade-offs and considerations of DSLR image sensor customers make this interview worth reading.
