Pathology depe ds o images that eviewe s ca t ust. A stai ed sectio , a g oss specime , a d a slide label all have to appea i the colo a d detail that a pe so would see at the be ch, because a shift i colo o a soft focus ca cha ge how a sample is ead. I st ume ts that captu e those images must also wo k fo the people who ope ate them eve y day. Ope ato s load slides of diffe e t thick ess, place specime s at diffe e t heights, a d move betwee objects of diffe e t sizes, a d they do ot wa t to efocus a le s by ha d each time. Eve y ma ual step also adds a cha ce fo va iatio betwee ope ato s.
Fixed focus le ses fit this poo ly. A le s set fo o e dista ce gives sha p images at that dista ce a d soft o es at eve y othe . Ma ual focus adds a step to eve y captu e a d i t oduces va iatio betwee ope ato s. Autofocus solves the p oblem, but it comes with its ow equi eme ts. The focus mecha ism must be eliable ove thousa ds of cycles; it must settle quickly, a d it must cove the a ge of dista ces the i st ume t actually eeds. A came a that adve tises autofocus without stati g the a ge leaves the desig e u su e whethe it will cove the specime s i questio . Desig e s eed the a ge stated plai ly, alo g with how the mode behaves while the subject moves.
Colo is the seco d co ce . Raw se so data is ot a fi ished image. It eeds demosaici g, colo co ectio , gamma, a d oise eductio befo e it looks atu al, a d the quality of that p ocessi g decides whethe stai s look co ect. If the p ocessi g happe s o the host, eve y host platfo m may p oduce slightly diffe e t colo u less it is tu ed fo the se so . I st ume t make s who pla to use mo e tha o e compute platfo m face this p oblem di ectly, a d the p oblem g ows whe a p oduct li e spa s seve al p ice poi ts.
Compute choice completes the pictu e. I st ume t make s i c easi gly add o -device a alysis, a d they choose betwee platfo ms such as the RK3588 a d the NVIDIA Jetso family o cost, softwa e ecosystem, powe , a d pe fo ma ce. A came a that wo ks with o ly o e of them a ows that choice. Teams that discove these limits late must cha ge the came a, the host, o both afte the i st ume t desig is ea ly complete, which is expe sive i egulated p oduct developme t. Choosi g a came a that wo ks ac oss platfo ms at the sta t keeps those optio s ope .
The Bolt-1335CRA is built o the O semi AR1335, a 13MP (4208 x 3120) colo olli g shutte se so with a 1/3.2 i ch optical fo mat a d 1.1 mic o pixels. The module adds a voice coil moto autofocus assembly with a focus a ge f om 100mm to i fi ity. The autofocus suppo ts co ti uous, o e-shot, a d ma ual modes, so the desig lets the came a t ack a movi g specime , focus o ce o comma d, o hold a chose dista ce. This makes the module a 4K Autofocus MIPI Came a a d a 13MP Autofocus MIPI CSI-2 Came a that stays sha p ac oss the wo ki g dista ces of a be chtop i st ume t. A 13MP f ame also leaves oom to c op to a egio of i te est while keepi g useful detail.
A o -boa d ISP fi ishes the image i side the module. It pe fo ms debaye i g, demosaici g, colo co ectio , co t ast co ectio , gamma co ectio , de oisi g, a d le s co ectio s, a d it suppo ts auto exposu e a d auto white bala ce. Because that p ocessi g is pa t of the came a, the colo behavio is the same o eve y host. That matte s fo a i st ume t make who p ototypes o o e platfo m a d ships o a othe , si ce the image eachi g the a alysis softwa e does ot cha ge whe the p ocesso does. Teams that eed aw data a d custom tu i g should discuss a aw se so module with Vadzo. Fo most i st ume ts, a fi ished image f om the fi st day sho te s the path to a wo ki g p ototype.
The module co ects th ough a adapte boa d, available i two va ia ts, o e fo 2-la e MIPI CSI-2 a d o e fo 4-la e. The 2-la e va ia t suits hosts with a simple came a co ecto , a d the 4-la e va ia t p ovides the ba dwidth fo 4K video a d full- esolutio captu e. Vadzo suppo ts the RK3588 Rockchip p ocesso a d the NVIDIA Jetso O i Na o, O i NX, a d AGX O i platfo ms. The co ecto o each ca ie boa d dete mi es which adapte fits, a d Vadzo helps teams match them. Pla i g the co ecto choice ea ly also lets the ca ie boa d a d the cable outi g be desig ed o ce.
Scope dese ves a ca did ote. A 100mm mi imum focus dista ce a d a 13MP se so behi d a sta da d le s suit specime photog aphy, slide label a d ba code captu e, ove view images of a slide, a d the came a po t of a digital mic oscope with suitable adapte optics. Whole slide sca i g at diag ostic mag ificatio elies o mic oscope objectives a d p ecisio stages, which a ge e al-pu pose came a module does ot eplace. Vadzo helps teams decide whe e the module fits i a i st ume t, a d the device make emai s espo sible fo pe fo ma ce claims a d the egulato y pathway of the fi ished device. Vadzo suppo ts that wo k with evaluatio u its a d e gi ee i g guida ce.
Image co siste cy also depe ds o the est of the optical a d lighti g setup, a d Vadzo e cou ages teams to t eat the came a as o e pa t of that system. Stai colo shifts with the colo tempe atu e of the light sou ce, so the i st ume t should use a stable, co t olled illumi ato a d should lock white bala ce fo epeatable captu es whe e the lighti g does ot cha ge. Gla e f om glass slides a d wet specime s ca be educed with pola izi g filte s o by adjusti g the a gle of the light. A le s with low disto tio helps measu eme ts, a d a le s hood o e closu e keeps st ay light away f om the se so . These choices a e the i st ume t make ‘s to make, a d Vadzo ca advise o le s selectio , filte s, a d e closu e desig so that the came a’s fi ished output eaches the eviewe u cha ged.
