DS160PR410: use Sigcon Architect

抓取的时候连接上设备后，在high level page勾选all devices channel0，然后refresh后，点击save to file。之后弹出如下界面invalid address or device not active，选择continue，然后回到configuration后重新detect地址（地址一直是正确的），之后重复上面抓取file动作一直跳出提示

现在抓出来的cfg文件一直是空的，选择继续在演示模式下，channel0/1/2/3均有数据且数据一样，应该在哪种模式下抓取

Hi,

The error message seems to indicate that there is a communication issue between the USB2ANY and DS160PR410.

Can you please share your schematic in .pdf format and the SigCon setup?

When you set test with SigCon Architect, on the Configuration Page, are you able to see the LED of the USB2ANY toggle when you click "Toggle LED?" This will help verifying that you are properly controlling the USB2ANY.

On the DS160PR410, are you setting the EN_SMB pin to L3? 

For the hard disk bandwidth issue, what DS160PR410 GAIN and EQ setting are you using right now?

Thanks

David

HI David:

    First,The schematic diagram has been uploaded as an attachment.

    And I can see the lights turning on and off when I click "Toggle LED".

    On the DS160PR410, I have set the EN_SMB pin to L3.

    For the hard disk bandwidth issue, we use DS160PR410RNQR and there is no firmware burned into the chip.

9MY62EX0NR-Y2-V20B.pdf

Hi,

Looking at the schematic, the address is 8-bit address. But the USB2ANY uses 7-bit address. Are you using 7-bit or 8-bit addressing with the SigCon GUI? Below table shows the 7-bit address.

Also, in I2C slave mode, 

• Leave ALL_DONE_N floating
• READ_EN_N needs to be pulled up to 3.3V through a 4.7k resistor

Thanks

David

HI David:

I use 8-bit address.

And the four redrivers currently used in our schematic diagram, the READ_EN_N of the previous chip is the ALL_DOWN_N of the next chip. It can't  rework all four chips as you suggested.

1、If I read each chip one by one, for instance, by first pull up the READ_EN_N signal of the first chip (now, the ALL_DOWN_N signal of the second chip will be at a high level), would it be necessary to connect the tool only to the first chip?    Can the other three chips still function properly?

2、Check the chip manual. It indicates that ALL_DOWN_N is an output signal，and requires an external 4.7K pull-up. Is it possible to directly apply a 4.7K pull-up to both ALL_DOWN_N and READ_EN_N signals?

Hi,

For ALL_DONE_N, the external 4.7k pullup is required when in I2C master mode. In I2C slave mode, it is High-Z and can be left floating. 

So the READ_EN_N signal is the only signal that requires an external 4.7k pullup to 3.3V so the device I2C state machine is not hold in reset.

Thanks

David

好的了解，我先尝试R3009上件4.7K，R3005去掉，并用7-bit地址看下能不能读取第一个芯片。

HI David：

    我尝试R3009上件4.7K，R3005去掉，主板无法进系统了。无法确定是不是因为后面串起来的re-driver芯片READ_EN_N也悬空导致。

Hi, 

But are you able to power up the first DS160PR410 and able to access it using the SigCon GUI? The log shows USB device enumeration, I am not sure how this is related to the issue of not able to enter OS?

Thanks

David

HI，

   1、第一颗redriver芯片是可以使用工具读到的，软件上使用的地址得用8bit而不是7bit，触发USB2ANY的LED也有反应。在low level page寄存器是有东西的，但是在high level page保存出来的cfg文件内容仍然是空的，具体请看截图。

2、然后量测第一颗芯片的ALL_DOWN_N信号为低电平，我想请问下你所说的动作需要把上一级redriver的all_down_n和下一级的READ_EN_N断开么？我现在是没割线仍然串在一起的状态。

3、这个不仅系统的log在这里枚举USB设备之后，就会进系统了，现在卡在这里。之前我尝试过去修改芯片GAIN和VOD两个pin上下拉，有的搭配也会出现卡在进系统之前，有的搭配不会但是仍然掉带宽。

4、我们掉带宽的设备一直是三星的3.0硬盘，三星4.0和Intel3.0/4.0均未出现问题。

Thanks

Christine

Hi,

On the Low level Page, are you able to write and then read the DS160PR410 registers? If you are, then you are in full control of DS160PR410. 

The High Level Page provides a user friendly GUI interface to control the DS160PR410. If you go the "Block Diagram" tab, and change the EQ index. Are you able to see the register value change correspondingly in the low level register? 

What current EQ and DC gain are you setting the DS160PR410? I am still not clear on how the USB enumeration issue is related to DS160PR410, do you have a block diagram to show your overall setup?

Thanks

David

你好：

目前第一个redriver可以读取和写寄存器了，也可以修改EQ。请帮忙准确回答以下问题：

1、用软件修改EQ参数后是不是得重启服务器才可以生效？

2、整个链路是CPU  x8lane----->re-driverx4----->X8 MCIO------>背板两个硬盘，出现的问题是开机进系统后，在OS下查看硬盘发现前4lane对应的硬盘掉带宽。我是否需要收集这4颗redriver芯片在low level page下的寄存器信息给你们分析即可？

3、如果要按照第二点收集，那我要如何rework才可以读到所有的redriver，现在只能读TX[0:3]对应的redriver，后面的设置都串起来了，除了割线我不知道要怎么去rework。请帮忙给出意见

请按照上述问题分点回复，让我可以更加高效的debug

关于rework动作导致无法开机进OS问题，我会用一块新的主板去验证是否是由于rework动作导致，请暂时忽略这点。

谢谢！

Hi,

You can change the EQ on the fly without power cycling the DS160PR410. 

I just verified on the EVM that with even ALL_DONE being connected to READ_EN_N, you can still use SigCon GUI to access both devices. Please make sure each device using an unique I2C address.

Thanks

David 

你好：

请问如果想看掉带宽的时候redriver的ltssm（我们希望可以找到原因并解决），是否如上次我回复的第二点那样？

Hi,

There is no way to look at the DS160PR410 re-driver LTSSM. If you change the DS160PR410 EQ and DC gain, are you seeing an improvement in the SSD BW?

Thanks

David

您好：

我重新拿了一块板子连接，我在四颗redriver的ALL_DOWN_N和READ_EN_N均上拉的情况下能够读到芯片并修改寄存器。有新的问题想请教下：

1、在configuration页，EVM model我选择SA160PR410EVM-RSC可以同时读到上游和下游共8个地址，但是在high level page会出现invalid address or device not active。

EVM model我选择SA160PR410EVM-SMA可以单独读一颗redriver，并且也可以成功修改寄存器也没有出现问题提示。请问这两种EVM model有什么区别？

2、由于掉的盘是X8lane的前4lane对应的盘，并且跑压测较容易复现到。所以我先修改了TX[3:0]和RX[3:0]的EQ值为7，DC GAIN没做修改。目前跑压测验证中。

针对上述操作，想请问下

①EQ值有16种，DC GAIN有两种，有这么多的搭配，有什么优先推荐修改的参数值么？

②X8lane从CPU的同一个Port出，我只修改前4lane，对后4lane的会有影响么？

Hi,

The DS160PR410EVM-RSC has four DS160PR410 shared the same I2C bus while the DS160PR410EVM-SMA has only one DS160PR410 on the I2C bus. When using SigCon GUI with DS160PR410EVM-RSC EVM model, the GUI is expecting to receive data from all four DS160PR410 devices, otherwise it will give you the invalid address warning. 

For the DC gain, I would keep it at 0dB. The EQ would depends on the insertion loss between the CPU and the DS160PR410 with EQ trying to compensate for the insertion loss as close as possible.  

For the CPU x 8 lanes, does it go to one endpoint or two end points with two DS160PR410?

Thanks

David

您好：

那我是否可以直接使用DS160PR410EVM-SMA模式，只针对前面两个re-driver分别设置EQ来调试呢？

我们的拓补和相关部分原理图请查收。

1030.SCH.pdf

Hi,

You can go to "Device" -> "Manage Devices", and create a list of loaded devices as shown below. I created the first one with address 0x30, and second one with address 0x40.

Once you created the list, you should see two devices under selection and then control them individually.

Thanks

David

您好：

好的了解。另外还有个问题想请教下，我们复现问题的时候是需要做不掉电重启+压测的测试的，这个时候修改EQ是可以一直生效的。但是这个板子上redriver并没有外挂eeprom，假设我后面验证了preset设置为P9的时候正常，那redriver要怎么去固定这个设置呢？是否可以出厂的时候你们可以直接设置芯片内部寄存器？还是说只能每次开机都用软件去设置？

Hi,

You can put DS160PR410 into pin strap mode, and use EQ0_ADDR0 and EQ1_ADDR1 pin to set the EQ and the Gain pin to set the DC gain.

Thanks

David

您好：

想请问下有bypass EQ或者设置RX_DET_BY的外部硬件设置的方法么？Bypass EQ是只能针对16GT/s的设备么？

我们目前掉带宽出现的条件是做冷重启（此时主板从S5状态重新启动，pcie会重新初始化），跑硬盘压测或者做热重启目前未复现到问题。之前我处理retimer的情况有出现DC后降速掉带宽问题，是通过设置RX_DET_BY取消RX_DET过程解决，想看下我们这个问题能不能类似解决。

Hi,

The DS160PR410 has EQ Boost 1 and Boost 2, you can bypass EQ Boost 1 but not EQ Boost 2. So in both pin strap and SMBUS mode, there is no way to completely bypass the DS160PR410 EQ.

Thanks

David