LSF0108: SPI level shifter

user4979899

Part Number: LSF0108
Other Parts Discussed in Thread: SN74LVC3G17, SN74LVC2G17, SN74AUC17

Hi Sir,

My customer is looking for SPI level shifter solution and I just suggest them to use SN74LVC2G17 or SN74LVC3G17 for SPI level-shifter.

But customer concerned about propagation delay. There is just <1ns delay in LSF family, but 5.4ns in SN74LVC.

Could you please help to suggest which family is the most suitable for SPI level shifter? LSF or LVC or TXB/TXS family?

Please help to advice.

Thanks

over 4 years ago

0 Emrys Maier over 4 years ago

TI__Guru* 96681 points

Hi Anne,

I would highly recommend watching the video series on the LSF family to understand how it works:

Understanding the LSF Family of Bidirectional, Multi-Voltage Level Translators

The SN74LVCxG17 devices are not voltage translators, but can produce a down-translation effect from the over-voltage tolerant inputs.

You haven't told me what voltages or frequencies the customer needs to level shift, or the characteristics of the connected devices (capacitive load, drive strength), so I can't really make any recommendation.

0 user4979899 over 4 years ago in reply to Emrys Maier

TI__Expert 3080 points

Hi Emrys,

Here is the requirement for the SPI level shifter.

Please help to suggest a suitable solution for it.

Thanks

1.The voltage of SPI host is 3.3V, and slave is 1.8V. Frequency is max. 100MHz.

2.There are two slave devices with independent CS. So total are required 5 signals.

3.Drive strength may be 5~10mA, and capacitance is max. 3.5pF.

0 Clemens Ladisch over 4 years ago in reply to user4979899

Guru 258710 points

For unidirectional translation, a device with real buffers (LVC/AUC/AVC/AVX etc.) is preferrable to a device with automatic direction sensing; the LSF has no drive strength at all.

The LSF propagation delay is so low because it uses passive switches; when the signal is low, there is essentially a wire betweeen the A/B pins. But the higher LVC propagation delay does not really matter because all channels have approximately the same delay (the variation is less than 1 ns).

Instead of multiple LVC buffers, I'd recommend a single hex buffer like the SN74AUC17.

0 Helios Yang over 4 years ago in reply to Clemens Ladisch

TI__Prodigy 530 points

Hi,

We followed up the discussion with customer and they're asking for bandwidth information in data sheet to confirm 100MHz is ok. Help to check if we have related info or we could calculate from the Switching Characteristics table in P.3? Thanks

0 Clemens Ladisch over 4 years ago in reply to Helios Yang

Guru 258710 points

At 100 MHz (200 Mbps), the length of a single bit is 5 ns, so this is OK.

How to Select Little Logic says:

Logic

Logic forum

LSF0108: SPI level shifter