Table detection

[fl4p]

table2matrix

Datasheets contain merged cells if a unit or condition applies to multiple rows. headers might also be merged. when iterating the data row wise, we need to first resolve the merged cells and copy the value across all rows within the span.

• resolve row-span and col-span

  
  +-----------+---+---+---+
  | A         | B | C | D |
  |           +---+---+---+
  |           | E         |
  |           +---+---+---+
  |           | E | C | C |
  +---+---+---+---+---+---+
  | E | C | C | C | C | C |
  +---+---+---+---+---+---+
  
            V

[['A', 'A', 'A', 'B', 'C', 'D'], ['A', 'A', 'A', 'E', 'E', 'E'], ['A', 'A', 'A', 'E', 'C', 'C'], ['E', 'C', 'C', 'C', 'C', 'C']]

# Tabula
* the web version has an auto-detect function, which performs much better than tabula-java (CLI)
* it does not auto-detect nested tables
* selecting the whole page leads to poor results

<details open>
  <summary>EPC2306</summary>

![EPC2306](https://github.com/user-attachments/assets/29800f31-cbb2-49ea-bd08-43d5cf7d40f0)

</details>

![image](https://github.com/user-attachments/assets/a0298fc5-33cf-4b75-a389-8832337a4c7a)

## Stream
(FIELD* headers are actually empty, the CSV->MD converter put them)

<details>
  <summary>Table</summary>

|FIELD1   |FIELD2                                               |Dynamic Characteristics# (TJ = 25°C unless otherwise stated)|FIELD4|FIELD5|FIELD6|FIELD7|
|---------|-----------------------------------------------------|------------------------------------------------------------|------|------|------|------|
|         |PARAMETER                                            |TEST CONDITIONS                                             |MIN   |TYP   |MAX   |UNIT  |
|CISS     |Input Capacitance                                    |                                                            |      |1777  |2369  |      |
|CRSS     |Reverse Transfer Capacitance                         |VDS = 50 V, VGS = 0 V                                       |      |5.8   |      |      |
|COSS     |Output Capacitance                                   |                                                            |      |616   |803   |pF    |
|COSS(ER) |Effective Output Capacitance, Energy Related (Note 1)|                                                            |      |730   |      |      |
|VDS = 0 to 50 V, VGS = 0 VCOSS(TR)|Effective Output Capacitance, Time Related (Note 2)  |                                                            |      |882   |      |      |
|RG       |Gate Resistance                                      |                                                            |      |0.4   |      |Ω     |
|QG       |Total Gate Charge                                    |VDS = 50 V, VGS = 5 V, ID = 25 A                            |      |12.3  |16.2  |      |
|QGS      |Gate to Source Charge                                |                                                            |      |4.3   |      |      |
|QGD      |Gate-to-Drain Charge                                 |VDS = 50 V, ID = 25 A                                       |      |1.1   |      |      |
|         |                                                     |                                                            |      |      |      |nC    |
|QG(TH)   |Gate Charge at Threshold                             |                                                            |      |3.1   |      |      |
|QOSS     |Output Charge                                        |VDS = 50 V, VGS = 0 V                                       |      |44    |57    |      |
|QRR      |Source-Drain Recovery Charge                         |                                                            |      |0     |      |      |

</details>

## Latice
* headers (min) are off

<details>
  <summary>Table</summary>

|Dynamic Characteristics# (TJ|25°C unless otherwise stated)                        |FIELD3                          |FIELD4|FIELD5|FIELD6|FIELD7|
|----------------------------|-----------------------------------------------------|--------------------------------|------|------|------|------|
|PARAMETER                   |TEST CONDITIONS                                      |MIN                             |TYP   |MAX   |UNIT  |      |
|CISS                        |Input Capacitance                                    |VDS = 50 V, VGS = 0 V           |      |1777  |2369  |pF    |
|CRSS                        |Reverse Transfer Capacitance                         |                                |5.8   |      |      |      |
|COSS                        |Output Capacitance                                   |                                |616   |803   |      |      |
|COSS(ER)                    |Effective Output Capacitance, Energy Related (Note 1)|VDS = 0 to 50 V, VGS = 0 V      |      |730   |      |      |
|COSS(TR)                    |Effective Output Capacitance, Time Related (Note 2)  |                                |882   |      |      |      |
|RG                          |Gate Resistance                                      |                                |      |0.4   |      |Ω     |
|QG                          |Total Gate Charge                                    |VDS = 50 V, VGS = 5 V, ID = 25 A|      |12.3  |16.2  |nC    |
|QGS                         |Gate to Source Charge                                |VDS = 50 V, ID = 25 A           |      |4.3   |      |      |
|QGD                         |Gate-to-Drain Charge                                 |                                |1.1   |      |      |      |
|QG(TH)                      |Gate Charge at Threshold                             |                                |3.1   |      |      |      |
|QOSS                        |Output Charge                                        |VDS = 50 V, VGS = 0 V           |      |44    |57    |      |
|QRR                         |Source-Drain Recovery Charge                         |                                |      |0     |      |      |

</details>

## Findings
* stream extraction method appears to be more usable here
* it doesn't provide an output that supports merged cells?
* the JSON output contains raw cell coordinates. cells are already grouped in rows. a cell with a rowspan (merged vertically) occurs in the first row. all subsequent rows within the span (and the same column) will have empty content. we can easily fill these rows with the same value
* auto-detect does not find nested tables and output quality suffers when selecting the whole page
* when we know the table bbox, the output can be good

#  BSB028N06NN3GXUMA2.pdf

![image](https://github.com/user-attachments/assets/d6e665fa-d736-4cbb-91ec-976124f8062a)

## tabula stream
| Parameter                    | Symbol   | Conditions            |      | Values |       | Unit |
|------------------------------|----------|-----------------------|------|--------|-------|------|
|                              |          |                       | min. | typ.   | max.  |      |
| Dynamic characteristics      |          |                       |      |        |       |      |
| Input capacitance            | C iss    |                       | -    | 8800   | 12000 | pF   |
|                              |          | V GS=0 V, V DS=30 V,  |      |        |       |      |
| Output capacitance           | C oss    |                       | -    | 2100   | 2800  |      |
|                              |          | f =1 MHz              |      |        |       |      |
| Reverse transfer capacitance | Crss     |                       | -    | 64     | -     |      |
| Turn-on delay time           | t d(on)  |                       | -    | 21     | -     | ns   |
| Rise time                    | t r      | V DD=30 V, V GS=10 V, | -    | 9      | -     |      |
|                              |          | I =30 A, R            |      |        |       |      |
| Turn-off delay time          | t d(off) | D G,ext=1.6 W         | -    | 38     | -     |      |
| Fall time                    | t f      |                       | -    | 6      | -     |      |

* results are good. it even uses the column headers from the "previous" table (min/typ/max)

## pix2image
![image](https://github.com/user-attachments/assets/98837843-7830-47e2-b2f3-0fb94306c080)

* for the first 3 rows it detects a single rowspan=3 , across all columns. this could be simplified. still it its not a usable result.

[fl4p]

pix2text

p2t predict -l en --resized-shape 2048 --file-type pdf -i datasheets/epc/EPC2306.pdf -o epc2306.md \
    --save-debug-res output-debug-p2t 

• it supports merged cells
• it appears to ignore lines mostly and put a new table grid from text boundaries

MD table (no merged cells)

| | PARAMETER | TEST CONDITIONS | MIN | TYP | MAX | UNIT | | --- | --- | --- | --- | --- | --- | --- | | CIss | Input Capacitance | Vos=50V,Vcs=0V | | 1777 | 2369 | pF | | Cass | Reverse Transfer Capacitance | | 5.8 | | | Coss | Output Capacitance | | | 616 | 803 | | CoSSER | Effective Output Capacitance, Energy Related (Note 1) | Vos=0to 50V,VGs=0V | | 730 | | | CossTR | Effective Output Capacitance, Time Related (Note 2) | | 882 | | | Re | Gate Resistance | | | 0.4 | | Q | | QG | Total Gate Charge | Vps=50V,Vcs=5V,lb=25A | | 12.3 | 16.2 | nC | | QGs | Gate to Source Charge | Vps=50V,lp=25 A | | 4.3 | | | QG | Gate-to-Drain Charge | | 1.1 | | | QGirn) | Gate Charge at Threshold | | 3.1 | | | Qoss | Output Charge | Vps=50V,Vcs=0V | | 44 | 57 | | QRR | Source-Drain Recovery Charge | | | 0 | |

HTML

p2t = Pix2Text.from_config()
doc = p2t.recognize_pdf('../datasheets/EPC/EPC2306.pdf', page_numbers=[1], resized_shape=2048)
table = doc.pages[0].elements[9]
print(table.meta['html'][0])

	PARAMETER	TEST CONDITIONS	MIN	TYP	MAX	UNIT
CIss	Input Capacitance	Vos=50V,Vcs=0V		1777	2369	pF
Cass	Reverse Transfer Capacitance			5.8
Coss	Output Capacitance			616	803
CoSSER	Effective Output Capacitance, Energy Related (Note 1)	Vos=0to 50V,VGs=0V		730
CossTR	Effective Output Capacitance, Time Related (Note 2)			882
Re	Gate Resistance			0.4		Q
QG	Total Gate Charge	Vps=50V,Vcs=5V,lb=25A		12.3	16.2	nC
QGs	Gate to Source Charge	Vps=50V,lp=25 A		4.3
QG	Gate-to-Drain Charge			1.1
QGirn)	Gate Charge at Threshold			3.1
Qoss	Output Charge	Vps=50V,Vcs=0V		44	57
QRR	Source-Drain Recovery Charge			0

Another example

HTML and MD tables

Parameter	symbol	Values			Unte	Note I Test Condition
		Min.	Typ.	Max.
Drain-source breakdown voltage	V(BR)DSS	100	-	-	V	Ves=0V, Io=1 mA
Gate threshold voltage	Vesth	2.2	3.0	3.8	V	Vos=Ves, /D=72 uA
Zero gate voltage drain current	ls	:	0.1 10	1 100	uA	Vos=100V, Ves=0 V, T=25°0 Vps=100 V, Ves=0 V, Tj=125°0
Gate-source leakage current	less	-	10	100	nA	Ves=20 V,Vos=0\V
Drain-source on-state resistance	Rosom	:	4.3 5.3	5.0 7.1	m2	Ves=10 V,D=50A Ves=6V,D=25 A
Gate resistance)	Re	-	1.2	1.8	Q	-
Transconductance	Ofs	50	100	-	S	|Vos|>2|/p|Ros(on)max,|b=50 A

| Parameter | symbol | Values | Unte | Note I Test Condition | Min. | Typ. | Max. | | --- | --- | --- | --- | --- | --- | --- | --- | | Drain-source breakdown voltage | V(BR)DSS | 100 | - | - | V | Ves=0V, Io=1 mA | | Gate threshold voltage | Vesth | 2.2 | 3.0 | 3.8 | V | Vos=Ves, /D=72 uA | | Zero gate voltage drain current | ls | : | 0.1 10 | 1 100 | uA | Vos=100V, Ves=0 V, T=25°0 Vps=100 V, Ves=0 V, Tj=125°0 | | Gate-source leakage current | less | - | 10 | 100 | nA | Ves=20 V,Vos=0\V | | Drain-source on-state resistance | Rosom | : | 4.3 5.3 | 5.0 7.1 | m2 | Ves=10 V,D=50A Ves=6V,D=25 A | | Gate resistance) | Re | - | 1.2 | 1.8 | Q | - | | Transconductance | Ofs | 50 | 100 | - | S | |Vos|>2|/p|Ros(on)max,|b=50 A |

[fl4p]

img2table

• looks promising
• pure OCR solution