app notes and competence

[rob d]

About 20 years ago I had three interviews at Altera and came so close to designing development boards which the rest of the world would have looked at for inspiration and guidance. I was 31 and trust me, the world has designed better Altera based PCBs as a result of not using my designs as reference material.
The May 2019 app note (ug 583) for Xilinx ultrascale plus decoupling caps was terrible and made me wonder yet again who writes this stuff. It has been replaced in September 2019 with the way that Howard Johnson recommended many, many years ago, ie plonk down the biggest capacitance in the smallest package.


As I said, who writes this stuff, graduates who don't know what they don't know probably.

Caveat Emptor

Rob

 

[Rick C]

On Tuesday, October 1, 2019 at 4:02:14 AM UTC-4, rob d wrote:

About 20 years ago I had three interviews at Altera and came so close to designing development boards which the rest of the world would have looked at for inspiration and guidance. I was 31 and trust me, the world has designed better Altera based PCBs as a result of not using my designs as reference material.
The May 2019 app note (ug 583) for Xilinx ultrascale plus decoupling caps was terrible and made me wonder yet again who writes this stuff. It has been replaced in September 2019 with the way that Howard Johnson recommended many, many years ago, ie plonk down the biggest capacitance in the smallest package.


As I said, who writes this stuff, graduates who don't know what they don't know probably.

First you complain about your own work being crap, then you appear to be complaining about Xilinx work. I'm not sure if you are saying Howard Johnson's work is good or bad.

What are you trying to say with all this?

--

Rick C.

- Get 2,000 miles of free Supercharging
- Tesla referral code - https://ts.la/richard11209

 

[rob d]

My work wasn't crap, but at the time I had only ever put one FPGA onto a PCB (lots of DSPs and Processors though) and as an industry we look to the designs of the manufacturers as the best way to do things. Clearly if I had passed at the third interview the world would have used the output of an inexperienced FPGA guy as the way to design with them.

Howard Johnson was hired (in some manner) by Xilinx to help them as Signal and Power integrity became an issue and I have met him and have enormous respect for him. Xilinx UG583 was first issued in 2013 and it has taken them six years to change their capacitor recommendations from something which I thought was a bit silly and which I slightly ignored to exactly what Dr Johnson worked out with them many years ago.

I'm in the middle of having a very expensive board laid out and I just wanted to rant about the quality of the advice that we all have to work with.

Rob

 

On Tuesday, October 1, 2019 at 4:02:14 AM UTC-4, rob d wrote:

About 20 years ago I had three interviews at Altera and came so close to designing development boards which the rest of the world would have looked at for inspiration and guidance. I was 31 and trust me, the world has designed better Altera based PCBs as a result of not using my designs as reference material.
The May 2019 app note (ug 583) for Xilinx ultrascale plus decoupling caps was terrible and made me wonder yet again who writes this stuff. It has been replaced in September 2019 with the way that Howard Johnson recommended many, many years ago, ie plonk down the biggest capacitance in the smallest package.


As I said, who writes this stuff, graduates who don't know what they don't know probably.

Caveat Emptor

Rob

I am using a very nice part at the moment from Linear Tech. I am finding that they have a feature that they go to great lengths to explain. What they do not explain is how to disable/not use this feature. I have had to spend days convincing myself that I can disable this feature (go against some of the wording in the data sheet).

 

On Tue, 1 Oct 2019 01:02:09 -0700 (PDT), rob d <rjd4567@gmail.com>
wrote:

About 20 years ago I had three interviews at Altera and came so close to designing development boards which the rest of the world would have looked at for inspiration and guidance. I was 31 and trust me, the world has designed better Altera based PCBs as a result of not using my designs as reference material.
The May 2019 app note (ug 583) for Xilinx ultrascale plus decoupling caps was terrible and made me wonder yet again who writes this stuff. It has been replaced in September 2019 with the way that Howard Johnson recommended many, many years ago, ie plonk down the biggest capacitance in the smallest package.


As I said, who writes this stuff, graduates who don't know what they don't know probably.

Caveat Emptor

Rob

HoJos books are half OK and half nonsense. If you can tell which part
is which, you don't need the book. The current UG583 is a similar mix.

HoJo has brain-damaged a generation of engineers. One of my customers
recently reviewed one of our PC boards and started spouting about
"return currents" and insane bypassing. It was embarassing.

Most big chips now have a lot of on-silicon bypassing [1]. Nobody
specifies or mentions the on-chip stuff, so I measure it. I must
average a tenth of the caps than appnotes recommend. And I've never
used too few caps.

[1] FPGAs for sure. The ST ARM chips don't.





--

John Larkin Highland Technology, Inc

lunatic fringe electronics

 

[Bill Sloman]

On Tuesday, October 1, 2019 at 7:38:39 PM UTC+10, rob d wrote:

My work wasn't crap, but at the time I had only ever put one FPGA onto a PCB (lots of DSPs and Processors though) and as an industry we look to the designs of the manufacturers as the best way to do things. Clearly if I had passed at the third interview the world would have used the output of an inexperienced FPGA guy as the way to design with them.

Howard Johnson was hired (in some manner) by Xilinx to help them as Signal and Power integrity became an issue and I have met him and have enormous respect for him. Xilinx UG583 was first issued in 2013 and it has taken them six years to change their capacitor recommendations from something which I thought was a bit silly and which I slightly ignored to exactly what Dr Johnson worked out with them many years ago.

I'm in the middle of having a very expensive board laid out and I just wanted to rant about the quality of the advice that we all have to work with.

Howard Johnson seems to know the basic facts of high speed circuit design, but he's a rotten teacher. I bought his "Black Magic" Book on Win Hill's recommendation here, and found it unhelpful - I was reminded of E.C. Snelling's book on soft ferrites, which eventually presents pretty much all the information you'd ever need, but in a way that makes it very difficult to extract the specific bits of information you need to solve any specific problem - another pedagogic disaster.

Even John Larkin has been rude about Howard Johnson - if you think his advice is good, you must have rather low standards.

--
Bill Sloman, Sydney

 

On Tue, 1 Oct 2019 02:38:34 -0700 (PDT), rob d <rjd4567@gmail.com>
wrote:

My work wasn't crap, but at the time I had only ever put one FPGA onto a PCB (lots of DSPs and Processors though) and as an industry we look to the designs of the manufacturers as the best way to do things. Clearly if I had passed at the third interview the world would have used the output of an inexperienced FPGA guy as the way to design with them.

Howard Johnson was hired (in some manner) by Xilinx to help them as Signal and Power integrity became an issue and I have met him and have enormous respect for him. Xilinx UG583 was first issued in 2013 and it has taken them six years to change their capacitor recommendations from something which I thought was a bit silly and which I slightly ignored to exactly what Dr Johnson worked out with them many years ago.

I'm in the middle of having a very expensive board laid out and I just wanted to rant about the quality of the advice that we all have to work with.

Rob

The PCB planes are the best capacitors around, and they are free. The
discrete caps don't matter much.

See, you didn't have to buy a book.

We don't generally put caps on the bottom of the board, and we keep
topside caps a distance away from the BGAs so we can slide our BGA
inspection microscope head between. Always works.



--

John Larkin Highland Technology, Inc

lunatic fringe electronics

 

[John Larkin]

On Tue, 1 Oct 2019 03:18:42 -0700 (PDT), blocher@columbus.rr.com
wrote:

On Tuesday, October 1, 2019 at 4:02:14 AM UTC-4, rob d wrote:
About 20 years ago I had three interviews at Altera and came so close to designing development boards which the rest of the world would have looked at for inspiration and guidance. I was 31 and trust me, the world has designed better Altera based PCBs as a result of not using my designs as reference material.
The May 2019 app note (ug 583) for Xilinx ultrascale plus decoupling caps was terrible and made me wonder yet again who writes this stuff. It has been replaced in September 2019 with the way that Howard Johnson recommended many, many years ago, ie plonk down the biggest capacitance in the smallest package.


As I said, who writes this stuff, graduates who don't know what they don't know probably.

Caveat Emptor

Rob

I am using a very nice part at the moment from Linear Tech. I am finding that they have a feature that they go to great lengths to explain. What they do not explain is how to disable/not use this feature. I have had to spend days convincing myself that I can disable this feature (go against some of the wording in the data sheet).

IC data sheets are, in general, a disgrace. They often skip obvious
issues and sometimes outright hide hazards.

Support requests now mostly go to a forum, where users can share their
discoveries. I think it's hard to get decent support people (the good
ones get hired away fast) and they want to insulate the chip designers
from rabble like us.

I was thinking of starting a web site icbugs.com or something. That
would be fun.

 

[Jon Elson]

On Tue, 01 Oct 2019 10:18:57 -0700, John Larkin wrote:

I was thinking of starting a web site icbugs.com or something. That
would be fun.

Wow, I got a few of them! But, not sure any of these chips are really
relevant anymore.

Jon

 

[bitrex]

On 10/1/19 9:51 AM, Bill Sloman wrote:

On Tuesday, October 1, 2019 at 7:38:39 PM UTC+10, rob d wrote:
My work wasn't crap, but at the time I had only ever put one FPGA onto a PCB (lots of DSPs and Processors though) and as an industry we look to the designs of the manufacturers as the best way to do things. Clearly if I had passed at the third interview the world would have used the output of an inexperienced FPGA guy as the way to design with them.

Howard Johnson was hired (in some manner) by Xilinx to help them as Signal and Power integrity became an issue and I have met him and have enormous respect for him. Xilinx UG583 was first issued in 2013 and it has taken them six years to change their capacitor recommendations from something which I thought was a bit silly and which I slightly ignored to exactly what Dr Johnson worked out with them many years ago.

I'm in the middle of having a very expensive board laid out and I just wanted to rant about the quality of the advice that we all have to work with.

Howard Johnson seems to know the basic facts of high speed circuit design, but he's a rotten teacher. I bought his "Black Magic" Book on Win Hill's recommendation here, and found it unhelpful - I was reminded of E.C. Snelling's book on soft ferrites, which eventually presents pretty much all the information you'd ever need, but in a way that makes it very difficult to extract the specific bits of information you need to solve any specific problem - another pedagogic disaster.

Even John Larkin has been rude about Howard Johnson - if you think his advice is good, you must have rather low standards.

He should have stuck to making grilled cheese sandwiches!

 

[bitrex]

On 10/1/19 10:44 AM, jlarkin@highlandsniptechnology.com wrote:

On Tue, 1 Oct 2019 01:02:09 -0700 (PDT), rob d <rjd4567@gmail.com
wrote:

About 20 years ago I had three interviews at Altera and came so close to designing development boards which the rest of the world would have looked at for inspiration and guidance. I was 31 and trust me, the world has designed better Altera based PCBs as a result of not using my designs as reference material.
The May 2019 app note (ug 583) for Xilinx ultrascale plus decoupling caps was terrible and made me wonder yet again who writes this stuff. It has been replaced in September 2019 with the way that Howard Johnson recommended many, many years ago, ie plonk down the biggest capacitance in the smallest package.


As I said, who writes this stuff, graduates who don't know what they don't know probably.

Caveat Emptor

Rob

HoJos books are half OK and half nonsense. If you can tell which part
is which, you don't need the book. The current UG583 is a similar mix.

HoJo has brain-damaged a generation of engineers. One of my customers
recently reviewed one of our PC boards and started spouting about
"return currents" and insane bypassing. It was embarassing.

Most big chips now have a lot of on-silicon bypassing [1]. Nobody
specifies or mentions the on-chip stuff, so I measure it. I must
average a tenth of the caps than appnotes recommend. And I've never
used too few caps.

[1] FPGAs for sure. The ST ARM chips don't.

Howard D. Johnson's design prototype Canton, MA restaurant location
sadly closed sometime in the early 2000s, there's a Dunkin' Donuts
coffee shop and some "upscale" bar & grill in the building, now.

<https://youtu.be/kyBWBShpWdA>

 

[John Larkin]

On Tue, 1 Oct 2019 20:41:50 -0400, bitrex <user@example.net> wrote:

On 10/1/19 10:44 AM, jlarkin@highlandsniptechnology.com wrote:
On Tue, 1 Oct 2019 01:02:09 -0700 (PDT), rob d <rjd4567@gmail.com
wrote:

About 20 years ago I had three interviews at Altera and came so close to designing development boards which the rest of the world would have looked at for inspiration and guidance. I was 31 and trust me, the world has designed better Altera based PCBs as a result of not using my designs as reference material.
The May 2019 app note (ug 583) for Xilinx ultrascale plus decoupling caps was terrible and made me wonder yet again who writes this stuff. It has been replaced in September 2019 with the way that Howard Johnson recommended many, many years ago, ie plonk down the biggest capacitance in the smallest package.


As I said, who writes this stuff, graduates who don't know what they don't know probably.

Caveat Emptor

Rob

HoJos books are half OK and half nonsense. If you can tell which part
is which, you don't need the book. The current UG583 is a similar mix.

HoJo has brain-damaged a generation of engineers. One of my customers
recently reviewed one of our PC boards and started spouting about
"return currents" and insane bypassing. It was embarassing.

Most big chips now have a lot of on-silicon bypassing [1]. Nobody
specifies or mentions the on-chip stuff, so I measure it. I must
average a tenth of the caps than appnotes recommend. And I've never
used too few caps.

[1] FPGAs for sure. The ST ARM chips don't.



Howard D. Johnson's design prototype Canton, MA restaurant location
sadly closed sometime in the early 2000s, there's a Dunkin' Donuts
coffee shop and some "upscale" bar & grill in the building, now.

https://youtu.be/kyBWBShpWdA

I wish we had DD out here in the West. Their Boston Cremes are the
best.

And the nearest Chick Fil A is 40 miles away.

Engineering makes a boy hungry.

 

[bitrex]

On 10/1/19 9:57 PM, bitrex wrote:

On 10/1/19 8:52 PM, John Larkin wrote:

Howard D. Johnson's design prototype Canton, MA restaurant location
sadly closed sometime in the early 2000s, there's a Dunkin' Donuts
coffee shop and some "upscale" bar & grill in the building, now.

https://youtu.be/kyBWBShpWdA

I wish we had DD out here in the West. Their Boston Cremes are the
best.

DD is one of those coffee shop chains where most things on the menu are
good, except for the coffee. It seems traditional in NE to drink it with
a lot of cream and sugar which makes it palatable but might as well have
a milkshake at that point.

And the nearest Chick Fil A is 40 miles away.

there's one right down the street I don't buy that company's stuff,
though, never had it and guess I never will.

Engineering makes a boy hungry.


looks like they interrupted the President's lunch to have a press
conference lol:

https://cdn.vox-cdn.com/thumbor/qnoxbf7jMSIn9qcPWaApuUMRPK0=/0x0:5184x3456/920x/filters:focal(0x0:5184x3456):format(webp):no_upscale()/cdn.vox-cdn.com/uploads/chorus_asset/file/16306255/GettyImages_1128640000.jpg

the point here is that many semi-pro athletes grew up poor and this is
their one shot in life to rise above their former station, people who
grew up poor and then "make it" in my experience tend to not have a lot
of nostalgia for the only cheap fast foods they could afford when poor

Sort of like how a proper first date with a woman who grew up working
class is a fine restaurant and the art museum while a proper first date
with a woman who is from a very wealthy family is McDonalds and a
monster truck rally.

Serving fast food to Congress regularly and making them take the bus
everywhere would be a policy I could get behind, though.

 

[bitrex]

On 10/1/19 9:57 PM, bitrex wrote:

On 10/1/19 8:52 PM, John Larkin wrote:

Howard D. Johnson's design prototype Canton, MA restaurant location
sadly closed sometime in the early 2000s, there's a Dunkin' Donuts
coffee shop and some "upscale" bar & grill in the building, now.

https://youtu.be/kyBWBShpWdA

I wish we had DD out here in the West. Their Boston Cremes are the
best.

DD is one of those coffee shop chains where most things on the menu are
good, except for the coffee. It seems traditional in NE to drink it with
a lot of cream and sugar which makes it palatable but might as well have
a milkshake at that point.

And the nearest Chick Fil A is 40 miles away.

there's one right down the street I don't buy that company's stuff,
though, never had it and guess I never will.

Engineering makes a boy hungry.


looks like they interrupted the President's lunch to have a press
conference lol:

https://cdn.vox-cdn.com/thumbor/qnoxbf7jMSIn9qcPWaApuUMRPK0=/0x0:5184x3456/920x/filters:focal(0x0:5184x3456):format(webp):no_upscale()/cdn.vox-cdn.com/uploads/chorus_asset/file/16306255/GettyImages_1128640000.jpg

<https://www.wcpo.com/news/national/more-fast-food-at-the-white-house-trump-serves-chic-fil-a-to-north-dakota-state-football-team>

there were bowls of extra sauce

 

[bitrex]

On 10/1/19 8:52 PM, John Larkin wrote:

Howard D. Johnson's design prototype Canton, MA restaurant location
sadly closed sometime in the early 2000s, there's a Dunkin' Donuts
coffee shop and some "upscale" bar & grill in the building, now.

https://youtu.be/kyBWBShpWdA

I wish we had DD out here in the West. Their Boston Cremes are the
best.

DD is one of those coffee shop chains where most things on the menu are
good, except for the coffee. It seems traditional in NE to drink it with
a lot of cream and sugar which makes it palatable but might as well have
a milkshake at that point.

> And the nearest Chick Fil A is 40 miles away.

there's one right down the street I don't buy that company's stuff,
though, never had it and guess I never will.

Engineering makes a boy hungry.

looks like they interrupted the President's lunch to have a press
conference lol:

<https://cdn.vox-cdn.com/thumbor/qnoxbf7jMSIn9qcPWaApuUMRPK0=/0x0:5184x3456/920x/filters:focal(0x0:5184x3456):format(webp):no_upscale()/cdn.vox-cdn.com/uploads/chorus_asset/file/16306255/GettyImages_1128640000.jpg>