This is the first time I'm implementing a CPLD and so I came here looking for some advice. The following schematics show how I've wired the JTAG connector and also the bypass caps. I intend to run all IO banks at 3.3V. http://i.imgur.com/z1ReB.png http://i.imgur.com/LEpF4.png And this is the PCB layout: http://imgur.com/VDyPd.png Things I'm looking for advice on: Bypass Cap Position and Values. Bypass Cap Position and Values. Bypass Cap Position and Value Power Distribution Network - Is the broken ground plane ok? JTAG Pull Up/Down Resistor Values A question regarding JTAG, specifically. The core voltage of the Max V is 1.8V - but I'm using the IO banks at 3.3V. Do I provide 3.3V to the JTAG header or 1.8V? Please note that CPLD is intended to implement a 100 bit Shift Register. The frequency will be controlled by a MCU when it interfaces via SPI. I intend to use the slowest frequency (62.5kHz), so the requirements are not of a high speed digital design. However, I'd still like my design to be robust have high signal integrity. I would appreciate any form of input on this matter.

Hi, JTAG pins are powered by the same VCCIO as the remaining pins on the I/O bank. So, in your case, 3.3V. I don't have the manuals at hand. The JTAG looks OK but... tripple check. At higher frequencies, return paths are important and having signals over plane cuts is a problem. But at 62.6 kHz, it's pretty much a matter of whatever.

Thanks, I'll take a look at JTAG again. I had another question: None of the implemented logic in the CPLD requires an external oscillator. It's just a 100-bit shift register that sends data out serially when it receives a clock signal from a microcontroller. So, my question is - does the CPLD still need an external oscillator or can I have a fully working CPLD for my application without one?

--- Quote Start --- So, my question is - does the CPLD still need an external oscillator or can I have a fully working CPLD for my application without one? --- Quote End --- You do not need an external oscillator. If you find you need a clock source later on, then you can use the on-chip ~4MHz oscillator (its part of the UFM block). Cheers, Dave

--- Quote Start --- I had another question: None of the implemented logic in the CPLD requires an external oscillator. It's just a 100-bit shift register that sends data out serially when it receives a clock signal from a microcontroller. So, my question is - does the CPLD still need an external oscillator or can I have a fully working CPLD for my application without one? --- Quote End --- The only clocks you need are the ones your logic uses. There's no "hidden" need for clocks in the CPLD. So, if I understood your design correctly, the only clock signal you need is the SPI clock provided by the MCU. As Dave mentioned, the MAX V has an internal oscillator. However, it's frequency will vary depending on voltage, temperature and from chip to chip.

Thanks guys. I was going over the section "Power Up Sequencing" in the Max V Handbook. While I know that the Max V supports any power-up sequence, the following caught my eye: --- Quote Start --- When VCCIO and VCCINT are supplied from different power sources to a MAX V device, a delay between VCCIO and VCCINT may occur. Normal operation does not occur until both power supplies are in their recommended operating range. When VCCINT is powered-up, the IEEE Std. 1149.1 JTAG circuitry is active. If TMS and TCK are connected to VCCIO and VCCIO is not powered-up, the JTAG signals are left floating. Thus, any transition on TCK can cause the state machine to transition to an unknown JTAG state, leading to incorrect operation when VCCIO is finally powered-up. To disable the JTAG state during the power-up sequence, pull TCK low to ensure that an inadvertent rising edge does not occur on TCK --- Quote End --- The last paragraph concerns me. I do have a pull down resistor (1K) on TCK. Does the paragraph imply that I should pull this line low ONLY during power-up or should is it implying that a pull down resistor is all thats needed to avoid this unknown JTAG state issue.

Altera_Forum

Honored Contributor

14 years ago

Max V Hardware Implementation

This is the first time I'm implementing a CPLD and so I came here looking for some advice. The following schematics show how I've wired the JTAG connector and also the bypass caps. I intend to run all IO banks at 3.3V.

http://i.imgur.com/z1ReB.png

http://i.imgur.com/LEpF4.png

And this is the PCB layout:

http://imgur.com/VDyPd.png

Things I'm looking for advice on:

Bypass Cap Position and Values.
Bypass Cap Position and Values.
Bypass Cap Position and Value
Power Distribution Network - Is the broken ground plane ok?
JTAG Pull Up/Down Resistor Values

A question regarding JTAG, specifically. The core voltage of the Max V is 1.8V - but I'm using the IO banks at 3.3V. Do I provide 3.3V to the JTAG header or 1.8V?

Please note that CPLD is intended to implement a 100 bit Shift Register. The frequency will be controlled by a MCU when it interfaces via SPI. I intend to use the slowest frequency (62.5kHz), so the requirements are not of a high speed digital design. However, I'd still like my design to be robust have high signal integrity.

I would appreciate any form of input on this matter.

18 Replies

Altera_Forum
Honored Contributor
14 years ago
The reason for short wide traces is that long thin traces have higher inductance and, at high frequencies, that inductance becomes high impedance.
If you're only working at low frequencies, there's no need to worry that much about it.

So, I'd still put the 47 µF capacitor. But I wouldn't go to much trouble about trace length and width. I'd just put it somewhere between the power connector and the 1 µF capacitors.
Altera_Forum
Honored Contributor
14 years ago
Well, after some hours this is what I came up wth. Caps 1000 and 1001 are large 1000uF caps. Caps 100,101,102 and 103 are 47uF Tantalums.

Note that all the power routed to the decoupling caps is via these caps. Two of the caps are for VCCINT and the other two are for VCCIO. However, each CPLD has two - one for the VCCIO and one for VCCINT. Altera recommend one cap per two IO banks and hence I have one cap.

I'm seeking feedback regarding:

1) Power Distribution
2) Placement of Bulk Caps.
3) Overall routing.

http://i.imgur.com/JtHHm.png

P1 is power for VCCIO and P2 is for VCCINT.
Altera_Forum
Honored Contributor
14 years ago
Actually it's not perfectly information.........
Altera_Forum
Honored Contributor
14 years ago
--- Quote Start ---
Actually it's not perfectly information.........
--- Quote End ---

I'm sorry, what?
Altera_Forum
Honored Contributor
14 years ago
I'm not sure it will make routing easier in your case, but the JTAG can be chained, so you only need 1 connector.

In the blue plane, sometimes the empty space between traces looks a bit thin. Making it wider will make manufacturing simpler and more reliable.

Nitckpicking (meaning, I wouldn't bother myself to go back and redo it)
There's a few places where you have sequences of thinner and thicker traces. Ie, between C102 and U1. It also happens when you change planes, ie, traces on blue plane being thicker than on red plane.

Instead of some thick traces, you could just use use fills to fill all the available areas in the red plane with 3.3 and 1.8V zones, since copper is free.
Altera_Forum
Honored Contributor
14 years ago
I don't worry about thin supply lines as long there's no explicite high current requirement for a power net. But I don't like the rutted ground plane. You should at least place a few "jumper" traces to make the ground more continuous. Bypass capacitors near to each supply pin are the next basic requirement. This is achieved in the design, ground escape traces should be generally shorter however.
Altera_Forum
Honored Contributor
14 years ago
Thanks a lot for the insight, guys.

FvM, I tend to be a bit clumsy and pull jumpers all the time. And I feel the main advantage of a PCB is that it can be completely jumper free. Jumpers will just make the entire thing more fragile.

I REALLY would prefer a continous ground plane but it can get quite hard with so many traces. I am going to investigate if its worthwhile getting the final PCB manufactured from abroad (this is just a prototype). If so, I'll go with a 4 layer board and then have continuos ground and power planes.

Do you, however, feel that the current layout will give me trouble? Please note, this is just a prototype and my only requirement is that it lets me hit the ground running - allows me to write the firmware for the CPU and get the CPLDs to talk to the MCU.

Also, could you elaborate what you meant by the term ground escape traces?

Looking at the layout, I think I can take the thick blue trace that passes near R2 and bring it over to the top layer. In that case, I'll need to route the signal around the J3 connector.
Altera_Forum
Honored Contributor
14 years ago
I told about "jumper" traces, not wires. As an example, see the ground plane below U1. It's effectively cut in two halves. It would be appropriate to either change the layer of the supply trace below or place jumper traces bridging the gap.

But most likely, the circuit doesn't involve any fast signals and thus will be fine as is.

Forum Discussion

Max V Hardware Implementation

18 Replies

Recent Discussions

Arria 10 GX RX max intra-differential pair skew

MAX10 Bitstreams Authentication

Cyclone 10 GX development board collaterals

Agilex 7 FPGA Availability on Cloud Platforms (AWS, Azure, GCP)?

AGRW027R28A2I2V Thermal Model