diff --git a/apps/casette/encode.py b/apps/casette/encode.py
index a0d8fa7..322f81c 100644
--- a/apps/casette/encode.py
+++ b/apps/casette/encode.py
@@ -20,6 +20,11 @@ class FSKWriter:
         self.phase = 0.0
         self.samples = bytearray()
 
+        self.samples.extend(b'\0'*int(1.0*self.sample_rate))
+
+    def finalize(self):
+        self.samples.extend(b'\0'*int(1.0*self.sample_rate))
+
     def _tone(self, frequency: float, sample_count: int) -> None:
         phase_step = 2.0 * math.pi * frequency / self.sample_rate
         scale = int(32767 * self.amplitude)
@@ -41,7 +46,7 @@ class FSKWriter:
         for _ in range(8):
             self.bit(1)
             self.bit(0)
-        self.byte(0x0f)
+        self.byte(0xf0)
 
     def bit(self, value: int) -> None:
         if value:
@@ -66,6 +71,7 @@ class FSKWriter:
         self.byte(l&0xff)
         self.bytes(data)
         self.byte(self.chsum(data))
+        self.samples.extend(b'\0'*int(0.3*self.sample_rate))
 
     def chsum(self, data: bytes) -> int:
         cs = 0
@@ -93,6 +99,7 @@ def main() -> None:
     tones.block(0, args.input.name.encode())
 
 
+    tones.finalize()
     with wave.open(str(output_path), "wb") as wav:
         wav.setnchannels(1)
         wav.setsampwidth(2)
diff --git a/apps/casette/tape_loader.asm b/apps/casette/tape_loader.asm
index f89ee11..8d8728f 100644
--- a/apps/casette/tape_loader.asm
+++ b/apps/casette/tape_loader.asm
@@ -4,6 +4,8 @@
 .org    TAPE_LOADER_ROM_ADDRESS
 
 TAPE_PORT       = 0x18
+MAX_CYCLE_COUNT = 20
+MAX_BLOCK_SIZE  = 0x100
 
 STACK_TOP       = 0xBFFF
 WORK_BASE       = 0xB000
@@ -15,6 +17,10 @@ LONG_LEN	= WORK_BASE + 0x04	; byte
 TRESHOLD	= WORK_BASE + 0x05	; byte
 SHORT_COUNT	= WORK_BASE + 0x06	; byte
 LONG_COUNT	= WORK_BASE + 0x07	; byte
+BLOCK_LENGTH	= WORK_BASE + 0x08	; word
+BLOCK_NUMBER	= WORK_BASE + 0x0a	; byte
+
+BLOCK_BUFFER	= WORK_BASE + 0x100	; MAX_BLOCK_SIZE
 
 ; #################################################3
 ; Entry
@@ -38,42 +44,23 @@ save_iff:
         ld      hl,#msg_start
         call    print_with_irqs
 
-	; Calibrate receiver
-	call	calibrate
-
-	; Print parameters
+	; Read the first block
+	call	read_block
+	jr	z,1$
+	ld	hl,#msg_error_checksum
 	ei
-	ld	hl,#msg_short_len
 	call	print_string
-	ld	a,(SHORT_LEN)
-	call	print_hex8
 	call	print_crlf
-
-	ld	hl,#msg_long_len
+	di
+	jp	exit
+1$:
+	; Filename stored in buffer
+	ei
+	ld	hl,#msg_filename
 	call	print_string
-	ld	a,(LONG_LEN)
-	call	print_hex8
-	call	print_crlf
-
-	ld	hl,#msg_treshold
+	ld	hl,#BLOCK_BUFFER
 	call	print_string
-	ld	a,(TRESHOLD)
-	call	print_hex8
 	call	print_crlf
-
-	ld	hl,#msg_short_count
-	call	print_string
-	ld	a,(SHORT_COUNT)
-	call	print_hex8
-	call	print_crlf
-
-	ld	hl,#msg_long_count
-	call	print_string
-	ld	a,(LONG_COUNT)
-	call	print_hex8
-	call	print_crlf
-
-	call	BIOS_FLUSH
 	di
 
 ; #################################################3
@@ -132,6 +119,19 @@ wait_cycle:
 	ld	d,a
 	ret
 
+; #################################################3
+; Eat A cycles
+; #################################################3
+eat_cycles:
+	or	a,a
+	ret	z
+	ld	c,a
+1$:
+	call	wait_cycle
+	dec	c
+	jr	nz,1$
+	ret
+
 ; #################################################3
 ; Average 16 cycle intervals
 ; Returns A = average loop count
@@ -157,6 +157,10 @@ average16:
 	ret
 
 ; #################################################3
+; Calibrate
+; Gets short and long pulse lenghts and bit pulse counts
+; from training sequence
+; carry = succeeded
 ; #################################################3
 calibrate:
 	; establish comparator state
@@ -164,6 +168,9 @@ calibrate:
 	ld	d,a
 	; Wait for the first edge
 	call	wait_edge
+	; Wait for the next edge to be back in phase with the
+	; audio stream
+	call	wait_edge
 
 	; Measure shoft pulses
 	call	average16
@@ -210,6 +217,9 @@ calibrate:
 	cp	a,c
 	jr	nc,5$
 	inc	e
+	ld	a,#MAX_CYCLE_COUNT
+	cp	a,e
+	jp	c,error_cycle_count
 	jr	4$
 5$:
 	ld	a,e
@@ -223,13 +233,181 @@ calibrate:
 	cp	a,c
 	jr	c,7$
 	inc	e
+	ld	a,#MAX_CYCLE_COUNT
+	cp	a,e
+	jp	c,error_cycle_count
 	jr	6$
 7$:
 	ld	a,e
 	ld	(LONG_COUNT),a
-	
+
+	; First cycle of short was already consymed
+	ld	a,(SHORT_COUNT)
+	dec	a
+	call	eat_cycles
+
+	; Consume rest of training bits
+	; 6 L/S pairs then last L
+	ld	h,#6
+8$:
+	ld	a,(LONG_COUNT)
+	call	eat_cycles
+	ld	a,(SHORT_COUNT)
+	call	eat_cycles
+	dec	h
+	jr	nz,8$
+	ld	a,(LONG_COUNT)
+	call	eat_cycles
+
+	; Test calibration with sync byte
+	call	check_sync_byte
 	ret
 
+error_cycle_count:
+	ei
+	ld	hl,#msg_error_cycle_count
+	call	print_string
+	call	print_crlf
+	di
+	jp	exit
+
+
+; #################################################3
+; Read one bit
+; carry = value
+; #################################################3
+read_bit:
+	call	wait_cycle
+	ld	a,(TRESHOLD)
+	ld	c,a
+	ld	a,b
+	cp	a,c
+	jr	c,2$
+
+	; zero bit
+1$:
+	ld	a,(LONG_COUNT)
+	dec	a
+	call	eat_cycles
+	or	a,a 		; clear carry
+	ret
+	; one bit
+2$:
+	ld	a,(SHORT_COUNT)
+	dec	a
+	call	eat_cycles
+	scf			; set carry
+	ret
+
+; #################################################3
+; Read byte
+; A = byte
+; #################################################3
+read_byte:
+	ld	e,#0
+	ld	h,#8
+1$:
+	call	read_bit
+	rl	e
+	dec	h
+	jr	nz,1$
+	ld	a,e
+	ret
+
+; #################################################3
+; Check sync byte
+; carry = succeeded
+; #################################################3
+check_sync_byte:
+	call	read_byte
+	ld	e,a
+	ld	a,#0xf0
+	cp	a,e
+	jr	z,1$
+	or	a,a
+	ret
+1$:
+	scf
+	ret
+
+error_wrong_sync_byte:
+	ei
+	ld	hl,#msg_error_sync_not_found
+	call	print_string
+	call	print_crlf
+	di
+	jp	exit
+
+; #################################################3
+; Read a block
+; checksum in a
+; #################################################3
+read_block:
+	call	calibrate
+	jp	nc,error_wrong_sync_byte
+
+	; Get block number
+	call	read_byte
+	ld	(BLOCK_NUMBER),a
+
+	; Get block length
+	call	read_byte
+	ld	b,a
+	push	bc
+	call	read_byte
+	pop	bc
+	ld	c,a
+	ld	(BLOCK_LENGTH),bc
+	
+	exx
+
+	ld	bc,(BLOCK_LENGTH)
+	ld	hl,#BLOCK_BUFFER
+
+1$:
+	exx
+	call	read_byte
+	exx
+
+	ld	(hl),a
+	inc	hl
+
+	dec	bc	
+	; Check for zero
+	ld	a,b
+	or	a,c
+	jr	nz,1$
+
+	; Read checksum
+	exx
+	call	read_byte
+
+	; Calculate checksum of block
+	ld	hl,#BLOCK_BUFFER
+	ld	bc,(BLOCK_LENGTH)
+	push	af
+	xor	a,a
+	ld	e,a
+2$:
+	ld	a,e
+	add	a,(hl)
+	ld	e,a
+	inc	hl
+	dec	bc
+	; Check for zero
+	ld	a,b
+	or	a,c
+	jr	nz,2$
+
+	pop	af
+	; e is calculated checksum
+	; a is received checksum
+	ld	h,a
+	ld	l,e
+	cp      a,e	
+	ret
+	
+
 ; #################################################3
 ; IO Helper functions
 ; #################################################3
@@ -254,6 +432,15 @@ print_crlf:
         ld      a,#10
         jp      BIOS_PUTC
 
+print_hex16:
+	push	hl
+	ld	a,h
+	call	print_hex8
+	pop	hl
+	ld	a,l
+	call	print_hex8
+	ret
+
 print_hex8:
         push    af
         rrca
@@ -302,3 +489,20 @@ msg_short_count:
 msg_long_count:
 	.ascii "Long count: "
 	.db	0
+
+msg_error_cycle_count:
+	.ascii "Cycle count out of bounds"
+	.db	0
+
+msg_error_sync_not_found:
+	.ascii "Sync not found"
+	.db	0
+
+msg_error_checksum:
+	.ascii "Checksum error"
+	.db	0
+
+msg_filename:
+	.ascii "Filename: "
+	.db	0
+
diff --git a/apps/casette/test2.wav b/apps/casette/test2.wav
index 9037cd5..cb94fed 100644
Binary files a/apps/casette/test2.wav and b/apps/casette/test2.wav differ