EEVblog #1127 - Mailbag Monday

[tombi]

Python is a garbage. There is absolutely no excuse to write and ship software on Python.
It is a scripting language. Yes, it is okay to write some quick and dirty script to process some data on Python, run it, and then throw the script away.
And no, it is not okay redistribute those scripts for users who are not "experts" in Python scripting.

Actually I disagree in part - there are some excellent scientific tools for python including systems and control libraries, plotting (they clearly used matplotlib) and symbolic maths libraries. It might be good to use Python if you wanted to use this thing for some sort of signal processing type application. Providing python bindings would definitely be useful.

I don't think the language matters much. I think that they should have hidden the fact that it is Python more during installation. Also - that scope screen seems to be straight Matplotlib plot - that's why most of the menus don't make sense. The software is barely there - looks to be a small bit of code wrapped around the hardware. I think if they actually implemented some oscilloscope software for it things might be different.

Maybe in some contexts that would be ok - have you ever used a HackRF? Not easy to get started but you expect it to be that way.

The problem is they say it is an oscilloscope. I think Dave is right - it's a DAQ - or a DIY oscilloscope-ish kit (with no front-end worth a damn).

If you want to see what you can do with Python in EE checkout APMonitor.com on youtube. Basically FOSS Matlab.

[NivagSwerdna]

I agree that Github and Python are mainstream tools nowadays but the requirement to have to tinker with them to get a trace out of a scope is due for criticism.

My requirement for a scope is that if I turn it on and connect the probe I get a trace after a bit of knob twiddling or pressing AUTO 

w.r.t. the screwdriver.... I do think this is garbage and looks like a paid for promo (that didn't go quite according to plan thanks to a bit of Dave objectivity)... A nice set of manual Wika electricians screwdrivers is around £25 here in the UK.... and if you really have 1kV would you really want an electric screwdriver running amok... and what happens when the screw is fully unscrewed and drops out.... Vorsprung Durch Bulls**t

[SparkyFX]

As far as the WIHA screwdriver, I was excited to see it work...then fail on every count. Ease of operation 0, Functionality 0, Necessity 0, Battery 0, A clean yellow parts surface after 3 months or real use 0, price -4

There is actually a whole product category just for torque limited setting of screws/threaded inserts/pins, there might even be a limit on contacts. But most of it is in the area of industrial connectors and pogo pins, i.e. not philips or flat head.

Most of these obviously don´t make much sense for the homegamers, as their typical applications are due to the lack of documentation and based on expectation more likely abuse resistant, than having a low defined torque. So i see this Wiha thing as kind of a bridge solution to get into this kind of market, and trying to satisfy some of the electronic screwdriver "hype" thingy. Plus there probably is an ergonomic benefit if you do e.g. disassemble whole compartments of contactors and breakers or rewire with the correct rated wire and do not want to break out the (lowest torque setting) power drill.

There might also be areas with a requirement to work with VDE rated equipment only, which quickly inhibits the use of other alternatives. And not to forget the thing looks like a conventional screwdriver, so is less likely to be stolen.

[GeorgeOfTheJungle]

But please, tell me, what good stuff would you say about it?

4 channels. 60 MHz. An fpga! It works! Looks good! Promising! But the software is unfinished... I don't know. How would I know? Poor guy. I feel sorry for him.

[jnissen]

Sorry but will not sit through 20 minutes of just getting to 2 packages. Worthless at that point. 

Brutalized the scope card. Ive never seen a USB scope ever lead me to believe they are a replacement for the real thing. Why get upset that it does not meet your expectations! Like complaining that your apple looks like an orange.

[haastyle]

Don't worry, I'm doing fine!
I made an attempt at cleaning up the schematic:
https://github.com/drandyhaas/Haasoscope/blob/master/eagle/max10adc_v9.0_schematic.pdf

The software does work well, but takes some practice using. I of course agree it could be made easier to use. And I also agree python has its major drawbacks for this sort of application. But don't give up on the Haasoscope!

[GeorgeOfTheJungle]

Don't worry, I'm doing fine!

 

[maginnovision]

Sorry but will not sit through 20 minutes of just getting to 2 packages. Worthless at that point. 

Brutalized the scope card. Ive never seen a USB scope ever lead me to believe they are a replacement for the real thing. Why get upset that it does not meet your expectations! Like complaining that your apple looks like an orange.

Because the letter stated they hoped it'd change his mind on USB scopes. When the design didn't even properly support oscilloscope probes.

[thm_w]

If there are no more than two packages per week sent, then we don't even need this discussion. - New format is OK If there are more than few and some packages are going to never be opened in front of the camera - then there's problem.

Or that's just life and your package does not get shown on air. If its a "promotional" item then you could send it to mickmake or someone else instead.

Don't worry, I'm doing fine!
I made an attempt at cleaning up the schematic:
https://github.com/drandyhaas/Haasoscope/blob/master/eagle/max10adc_v9.0_schematic.pdf

The software does work well, but takes some practice using. I of course agree it could be made easier to use. And I also agree python has its major drawbacks for this sort of application. But don't give up on the Haasoscope!

Looks much better.
If you are still looking for ideas: could still get rid of those big round circles, and the "pas 1", "sup 0" texts. Also move text if its vertical or overlapping with other text.
BNC circuit can be horizontally laid out instead of vertical.

[ogden]

If there are more than few and some packages are going to never be opened in front of the camera - then there's problem.

Or that's just life and your package does not get shown on air.

I was hoping someone will say exactly that. - I will not mail anything if I am not sure that my letter will be ever read by Dave. Others may think same way. In result you can rename Mailbox to just "review of product X and Y" video.

If its a "promotional" item then you could send it to mickmake or someone else instead.

I don't care about business here. They most likely will send their wireless ESD wristbands or something like that no matter what.

[edit]  Screwdriver seemed like very promotional item. Far from package from some kid who want his first PCB to be checked by Dave.

[maginnovision]

If there are more than few and some packages are going to never be opened in front of the camera - then there's problem.

Or that's just life and your package does not get shown on air.

I was hoping someone will say exactly that. - I will not mail anything if I am not sure that my letter will be ever read by Dave. Others may think same way. In result you can rename Mailbox to just "review of product X and Y" video.

If its a "promotional" item then you could send it to mickmake or someone else instead.

I don't care about business here. They most likely will send their wireless ESD wristbands or something like that no matter what.

[edit]  Screwdriver seemed like very promotional item. Far from package from some kid who want his first PCB to be checked by Dave.

Except mailbag monday could allow him to actually get to the more time sensitive stuff in a timely manner. I don't think his plan is to throw half in the trash just stick to a couple items. That way he spends less time editing and less time recording while still checking things out. 8 items a month is plenty. Just because this mailbag seemed like two reviews doesn't mean they all will. That's just the nature of these items he picked.

[Bud]

Because the letter stated they hoped it'd change his mind on USB scopes.

Do not know about Dave but it did change My mind on USB scopes, from 'junk' to 'total junk'

[ogden]

Do not know about Dave but it did change My mind on USB scopes, from 'junk' to 'total junk'

Maybe narrow your junk category to "Open source USB scopes"? I would not call PicoScope, CleverScope or Keysight U2702A as junk

[station240]

I, too, wondered about the 1KV rating for the screwdriver.  I would like to see an example of the use case they had in mind - but even so, what if you are unscrewing something that did need that rating ... and then dropped a screw from the tip.   

You just need a 1KV rated Magnetic Pick Up Tool 

I have to admit, despite what Dave said I like everything about the WIHA Electric Screwdriver, except the price.
Dave would never part with the thing, so no point even asking for one.

As for the HaasoScope, I'd expect that level of mistakes in a prototype. Making something open source, is not a replacement for actually fixing the problems with it.

[FrankBuss]

Don't worry, I'm doing fine!
I made an attempt at cleaning up the schematic:
https://github.com/drandyhaas/Haasoscope/blob/master/eagle/max10adc_v9.0_schematic.pdf

This is slightly better. Some tips:

• even in Eagle you can use multiple pages. Maybe a good idea to group things like power supply and decoupling to one page, front end to another page, and the digital stuff on a 3rd page
  
• the front ends part still doesn't look nice. In many schematics you have all inputs on the left side, and all outputs (or outputs to other parts of the schematic) at the right side. You could organize the 4 blocks vertically, and then the coaxial connector at the left side, then right to it the range selection and relay.
• even better: use the hierarchy feature of Eagle (which it has finally in newer versions) and draw the front end only once, then just instantiate 4 blocks. Many professional schematics use this feature as well for one main page with blocks for all logical parts and the connections between them, and then the blocks itself are listed on other pages
• the "front end and gain offset" block looks very crowded. Leave some more space between the components and use the Eagle "smash" feature to move the labels, and then maybe use sub-circuits for the MAX4416, so that each channel looks the same (see e.g. the 7400 for an example how it is done, or read the manual). Again, it is better if all inputs are at the left side and all outputs are at the right side.
• in other postings it looks like you used the auto-router and no ground plane. That's not a good idea for the FPGA. And looks like you connected the FPGA just in sequence with the peripherals. How I do such things: I place the parts on the board and look at the ratsnest (that's how unrouted wires are called in Eagle). Then I use the "pin-swap" function a lot, to make it look less like a rats nest, with nice parallel lines and not many intersections. Then I try to route it by hand, using pin-swap again while routing. Newer Eagle version can do push-and-shove routing, like KiCad which I use nowadays, so it is not that difficult anymore. You can use the auto-router to give you hints, where components might be not placed optimally or more pin-swaps are necessary, but then rip-off the routed traces of the auto-router again (or don't save the auto-routed board and reload it). Auto-routers are nice for low-speed designs.
  
• for such a high speed design, you really want a nice solid ground plane, with no or only very short traces on it. And when routing, start with the supply voltage, with very direct connections, then route the signals. If you have 4 layers, you can reserve one dedicated layer for the supply voltages as well.
• some parts don't need to be blocks, you should connect them directly to the FPGA, e.g. the crystal oscillator etc., because then it is easier to see which parts belong to which other parts

The software does work well, but takes some practice using. I of course agree it could be made easier to use. And I also agree python has its major drawbacks for this sort of application. But don't give up on the Haasoscope!

Nowadays software should explain itself. And users, especially on Windows, expect a nice "setup.exe" to install it. Try to imagine the user experience of an engineer, who wants a nice USB scope, but has never heard of Python. He/she wouldn't recommend your scope, that's for sure.

[NiHaoMike]

Python is basically used as a free alternative to Labview at many companies I have worked at. I like it not only because it's hobbyist friendly but also because Labview (when I last used it) quickly becomes a mess when dealing with complex programs.

[Muttley Snickers]

I have to admit, despite what Dave said I like everything about the WIHA Electric Screwdriver, except the price.
Dave would never part with the thing, so no point even asking for one.

As a prolific power tool junkie I thought the screwdriver was an overpriced joke, the last thing any technician wants to risk doing is breaking or stripping out the termination screws on an RCD, MCB, Relay, Contactor, Bus Bar or other device inside an electrical fuse or distribution panel, not only could it result in a hazardous condition but it would also ruin your entire day chasing a replacement part.   

Furthermore, from my experience most cable and terminal connections inside these cabinets really only require a couple of turns from a regular screwdriver to properly and securely terminate a connection, I somehow doubt that people would bother dicking around trying to find the correct or appropriate torque setting for a particular termination. The bullshit packaging lost me from the get go but if anybody wants one of these tools then the fellow below said he was giving a few away. 

[Brumby]

... But don't give up on the Haasoscope!

We won't give up on it - but I hope your next iteration takes into consideration all the issues that have been raised.  It may sound brutal - but having clear criticism is much kinder to the development process.  Some of the design decisions that may appear to have equally valid alternatives in theory, often turn out to have one very clear winner in practice.

Your bold invitation for Dave to reassess his opinion on USB scopes attracted big interest - and the fact that it failed to deliver on some of the fundamental issues with these didn't help.

Certainly, some software improvements are necessary - but, IMHO, you will need to address some improvements in the hardware side - such as available vertical ranges and ease of changing these - and sweep ranges.


There is potential here - but it will need more effort to start closing in on the USB scope that will change Dave's mind about them.

[f4eru]

Python is basically used as a free alternative to Labview at many companies I have worked at.

Very good news !

Labview quickly becomes a mess when dealing with complex programs.

Yes. it's a real mess.

[Noon]

Asking the important questions here, but does anyone know about the red t-shirt Dave wears in this video? I'm really digging the design...

[haastyle]

• for such a high speed design, you really want a nice solid ground plane, with no or only very short traces on it. And when routing, start with the supply voltage, with very direct connections, then route the signals. If you have 4 layers, you can reserve one dedicated layer for the supply voltages as well.

Thanks for the tips. There is a gnd plane though (with no traces on it). The exact way I've done the routing/layers is explained in the manual:
https://docs.google.com/document/d/1euyzCGzQo0u9APwEeior9fpyhcYWbkNxYQaV7SuFfoE

Best, Andy.

[tszaboo]

Don't worry, I'm doing fine!
I made an attempt at cleaning up the schematic:
https://github.com/drandyhaas/Haasoscope/blob/master/eagle/max10adc_v9.0_schematic.pdf

The software does work well, but takes some practice using. I of course agree it could be made easier to use. And I also agree python has its major drawbacks for this sort of application. But don't give up on the Haasoscope!

Step in the right direction. Not perfect, but it is a start.
But you also need to clean up your layout, and that is even more effort. There are functional reasons, why this is important. Basically, you need to read high speed PCB design techniques, and apply them to your board.
Read this: http://www.analog.com/en/analog-dialogue/articles/high-speed-printed-circuit-board-layout.html
and this: http://www.ti.com/lit/ml/slyp173/slyp173.pdf
Have someone review your board before next release. Pay him if you need to. You received money from other people to make this, use it. As I said, the layout has many issues, dont ship it as it is, karma is going to bite back.

[FrankBuss]

Thanks for the tips. There is a gnd plane though (with no traces on it). The exact way I've done the routing/layers is explained in the manual:
https://docs.google.com/document/d/1euyzCGzQo0u9APwEeior9fpyhcYWbkNxYQaV7SuFfoE

Ok, a good gnd plane is good. But I think you need more decoupling capacitors for the FPGA and they should be closer to the FPGA. But your layout still looks like Dave's "don't trust the auto router" T-shirt  I don't know if this is good for such a high speed design. With pin-swap you could use much less vias and more direct connections, there is lots of space on the board, and when you increase the board in order to use standard scope probes, you would have even more space.

Compare your layout:



To my hand routed layout of my Kerberos cartridge, where I used lots of pin-swaps, because you can assign them all in the FPGA (CPLD in my case) as you like, if the power bank is right:



It is a 4 layer board, with one dedicated GND layer and one (mostly) dedicated power supply layer, and I managed to route the rest on 2 layers. There are also some other tricks for easier routing, like you don't need to connect D0-D7 of the external bus to D0-D7 of the RAM IC, because it doesn't matter, if the RAM data bit 7 is accessed by the bus data bit 3.

[Quarlo Klobrigney]

I'm just migrating from the other forum thread.

This is how a properly drawn Eagle Schematic (everything from center to left of center) is drawn. No grounds on top, VCC on bottom. Confusing when there is real estate to do it right. The caps US symbol types in the power supply's are upside-down. They are also renamed instead of their reference C#'s. Same with the LED resistors. Your CH340 part grabber is way off center and on the wrong layer (it's all black). I didn't do all of it and I know I missed some items, but I have other things to do. Multiple stray leads. Do you even do an error check for both sch and brd? I say that because you would see what I see.
Also, no fill planes on board. I'm sure there are other no-no's.

Even though it's a different layer, it's bad form as it looks like a short. Like this on the board:
And the strays errors in the schematic:

[McBryce]

you don't need to connect D0-D7 of the external bus to D0-D7 of the RAM IC, because it doesn't matter, if the RAM data bit 7 is accessed by the bus data bit 3.

I hate when people do this, because it makes debugging a nightmare when the only accessable data bus pins are in the wrong order.

McBryce.