USB to SPI for .NET

Overview

I started investigating the hardware architecture for Nusbio 2. With the new device I would like to achieve a transfer of at least 100 K byte /s and target 1 M byte/s.
(Update: 2017.04.02 with the FTDI FT4222 I achieved 100 Kbyte/S in I2C using a 32k byte EEPROM and 2.4 MebaByte/S using a NOR FLASH at 100Mhz)

I most likely will use a USB 2.0 to SPI chip and add a micro-controller like the ATiny84 to provide

• More GPIOs
• PWM
• ADC
• EEPROM storage

I may select a Full speed or High speed. High speed is faster, but more expensive and most likely does not exist in package SSOP-20, but rather LQFP-48 or QFN-48.

In this post, I will summarize my findings regarding the chips I tested. Obviously my end goal is to create a .NET compatible device.

USB to SPI performance

See blog: USB to SPI for .NET (performance)

Test Scenarios

1. Test1 - Transfert 32 K byte of data from EEPROM 25AA256 (support SPI 10MHz) to computer in C#.
2. Test2 - Transfert 128 K byte of data from EEPROM 25AA1024 (support SPI 10MHz) to computer in C#.

USB to SPI Chip

• MicroChip - MCP2210 Datasheet
  
• Evaluation board ADM00419 $15
• USB 2.0 Full speed
• SPI Bitrate 3 Mbps
• 128 byte buffer
• GPIO for select: 8
• Packages: 20-lead QFN, 20-lead SOIC, 20-lead SSOP
• Price: $1.98 for one
• .NET samples: Yes
• SPI Low Level Optimization: Most likely No

• Cypress - CY7C65211 Datasheet
  
• Evaluation board  CY8CKIT-049-42XX $4
• USB 2.0 Full speed
• Bitrate 3 Mhz
• 256 byte buffer
• GPIO for select: 5 
• Packages: 24-lead QFN
• Price: $2.61 for one
• .NET samples: No. C code.
• SPI Low Level Optimization: Most likely No
• Remark: Can be configured as UART, SPI or I2C

• FTDI - FT232H Datasheet
  
• Evaluation board  UM232H $20
• USB 2.0 High speed
• Bitrate 30 Mhz for SPI
• 64k byte buffer
• GPIO for select: 5
• Packages: 48-lead QFN
• Price: $4.25 for one
• .NET samples: Partially, Require C DLL.
• SPI Low Level Optimization: Yes
• Remark: Can be configured as UART, SPI, I2C, JTAG, FIFO, Bit banging.


• SILICON LABS - CP2130 Datasheet
  
• Evaluation board  CP2130EK $20
• USB 2.0 High speed
• 6.6 Mb/s for SPI (844Kbyte/S, Video)
• xx byte buffer
• GPIO for select: 11
• Packages: 24-lead QFN
• Price: $2.16 for one
• .NET samples:?
• SPI Low Level Optimization:?

Test Result

• MicroChip - MCP2210
• At the SPI Bitrate 3 Mhz (which cannot be increased)
• Downloading the 32k of  data in chunk of 32 pages of 64 bytes (2k)
• 4.559s 7.02 K byte/S
• Conclusion: Really bad performance.
• Cypress - CY7C65211 
• At the SPI Bitrate 3 Mhz  
• Downloading the 32k of  data in chunk of 16 pages of 64 bytes (1k)
• 12 K byte/S
• I asked on the Cypress forum for help - Page. So far no real help from the Indian support.
  
• FTDI - FT232H 
• At the SPI Bitrate 10 Mhz (Max supported by the EEPROM)
• Downloading the 32k of  data in chunk of 256 pages of 64 bytes (16k)
•  0.03s  1066.67 K byte/S
• Conclusion: Really very nice performance. and the FT232H can goes up to 30 Mhz.
• SILICON LABS - CP2130

More About The MCP2210

•  Kerry Wong, blogged a lot about the MCP2210 at http://www.kerrywong.com.
• In his post "Finicky MCP2210 Oscillator", John Davis added the following comment, which explain a lot.
  
  
  WARNING: According to our trouble ticket interactions with Microchip technical support, the MCP2210 is “throttled” internally with a mininum SPI inter-byte spacing of about 45uS REGARDLESS of SPI clock frequency (we got 12 MHz working on our board.) That limits the byte data rate to about 23KBytes/second max. However, the performance is still limited further by the fact that it is an HID device with a HID-limit 64KByte/second interface. Two USB 128 byte transactions of 3mS each with a wait in between are required for a transfer, so the effective data rate using full 60 byte packets works out to slightly less than 8KBytes/second (Verified with ‘scope.) THIS IS NOT A HIGH THROUGHPUT PART!