DS160PR410: Can this redriver operate with 4 Gbps Fibre Channel

Hi Arley,

DS160PR410 is typically used in interfaces which use 100 Ohm differential impedance, where as Fibre Channel is 150 Ohm differential impedance. Therefore, this device has not been characterized in a 150 Ohm environment.

I have looped in one of my colleagues to see if he has any suggestions. I believe one of our Ethernet redrivers may be more suitable for this application: https://www.ti.com/interface/ethernet/retimers-redrivers-mux-buffers/products.html#1241=Mux%20Buffer%3BRedriver&1326max=4.25%3B56&

Best,
David

Hi David,

Our PCB traces are 100 ohm differential so that wouldn't be a problem. The controller we are using (Marvell EP2722) also specifies 100-ohm. With that in mind, would this be compatible? 

Hi Arley,

From an impedance perspective, if the controller supports 100 Ohm, this portion of the compatibility should be fine.

Care should be taken to ensure that the Fibre Channel spec is electrically compatible with the DS160PR410's electrical specifications. 

I may also recommend a device which is more suitable for your Nyquist frequency, such as DS64BR111. Do you have any information about your channel's insertion loss at 2GHz (4Gbps)?

Best,
David

Hi David,

We are designing a COTS mezzanine circuit card that plugs into a customer’s host card that passes the Fibre Channel signals down a backplane to another circuit card. 

 There will be many different customers using our board each with different signal paths in the channel so therefore we do not know the insertion loss.

Since the channel loss is unknown we want to use redrivers that give us the most flexibility on signal conditioning to provide the best performance reasonably possible. 

We looked at DS160PR410 along with DS100BR210. These seem like they would work. The one issue we had with these parts is they use 4-level strapping pin inputs. These pin input levels are set using resistor dividers. We would prefer to interface a FPGA to control the strapping inputs so with a 4-level setup, the FPGA can’t be involved in any changes we might want to make to the default settings. 

Hi,

These devices have I2C registers that can be used to set EQ.  Can the customer use their FPGA to set device registers via I2C?

Please keep in mind that redriver devices like DS160PR410 and DS100BR210 will have less boost at 2 GHz (4 Gbps) than at their target data rate.  We were inquiring about insertion loss in order to understand how much compensation the channel would require.  Is there an upper bound on insertion loss for this application?

Thanks,

Drew

Hi David,

 Yes, we certainly will use I2C to reconfigure the redriver and we plan to provide access to our customers with driver application so they can reconfigure. I brought up the 4-level strapping because we just prefer to set the default factory configuration with the strapping pins. We could alternatively use an EEPROM to set the default. We were just hoping for a part that would allow our FPGA to set that up instead of using EEPROM and the 4-level configuration prevents FPGA interface.

As far as insertion loss upper bound,  we don’t know the upper bound or budget. We really don’t have control over what our customers’ setup will be (type of connectors, backplane PCB, cable splices, etc.) The upper bound insertion loss for fibre channel will depend on their setup.

 We would like to pick a part that would give us the best chance it will work for our customers at the 2 GHz (4 Gbps) data rate.

Are DS64BR111or DS100BR210 redrivers the best options with TI? Any better choices?

Would it be better to choose a part that has better rate margin?  Would  DS100BR210 (10.3 Gbps) be a better choice than DS64BR111 (6.4Gbs)? Or the other way?

Hi,

Arley Lozada said:

We were just hoping for a part that would allow our FPGA to set that up instead of using EEPROM and the 4-level configuration prevents FPGA interface.

I am a bit confused in regards to this.  Does your FPGA not support I2C?

Alternatively, if your FPGA GPIO uses the same logic level as the redriver and supports setting output impedance as HIGH-Z (either disabling, or set GPIO to input), 4-level control can be achieved with two pins using the following mapping.

For this application, I think it's important to understand the bound on insertion loss and consider this when selecting a device.  Please keep in mind that redrivers will generally have a maximum boost setting of around 30 dB at their intended data rate, but much less boost at lower rates.

For example, DS100KR800 applies up to 34 dB of boost at 5 GHz, but somewhere between 16 - 30 dB of boost at 2 GHz.

DS64BR111 and DS100BR210 are reasonable choices for this data rate.  I'll review our portfolio and get back to you early next week if I see any other suitable devices for consideration.

Thanks,

Drew

Hi Drew,

Thanks for your response. Our FPGA does support I2C. We wanted the initial default configuration to be setup with direct connections from the FPGA to the pin strapping so we thought the 4-level was not going to work. However, your recommendation for using two GPIOs to setup the 4-level pin strapping seems like it would be a good option for us in addition to having an EEPROM and I2C in the circuit as options for configuration. 

Let us know if you find other options to consider. 

-Arley

Hi Drew,

We need a industrial part (-40 to +85). The  DS50PCI401/402 parts doesn't meet that.

The DS50PCI401/402 are designed for PCIe just like the DS60PR410. DS60PR410 does support industrial temp. Is DS60PR410 not a suggested part? 

Or do we focus on DS64BR111or DS100BR210?

Thanks.

Arley

Hi Dave,

It seems like that for a 2 Ghz application like ours a part that can boost up higher frequencies in the Gen 4 rates will be boosting frequencies higher than we want.

Does TI have a part that is max at Gen 3 PCIe rates that you can recommend for us? Gen 2 parts are not as available and possibly have a shorter life.

Thanks

Arley