BeMicro CV Questions

[weilawei]

I've recently received the BeMicro CV FPGA dev board, and it's a cute little bugger. I love it and I'm super pleased with my purchase, but it's got an 80-pin card-edge connector, and I can't seem to find the right one. In a post I wrote for my blog, I noted that the suggested mate is a Samtec MEC-140-02-L-D-RA1. The photos show a plastic divider, but no such corresponding cutout exists on the card edge! I've counted pins and it doesn't appear that it would even fit half the connector or something like that. Does anyone have any suggestions on how to make use of it (without soldering because I try to reuse all my parts)?

Also, I'm pretty new to serious electronics, but I'm a very experienced programmer (though new to Verilog). I've played with 100-in-one kits early on, but my enthusiasm was damped by spring terminals as a kid, so I invested heavily in the programming side of my skillset. Since my buddy gave me a breadboard, it's been a whole different experience.

I've got a little Hello World demo going on the board (also run the Propeller P8X32A on it!!) showing a few logic gates, but there's a quirk. In the sim, the code acts as you would expect (an AND gate acts like an AND, OR like OR, XOR like XOR), but on the board, an AND acts like an OR, an OR like an AND, and an XOR like an XNOR! The code in my post is written so that it'll light up the LEDs in the "intuitively correct" fashion (to my mind at least).

A good friend of mine who's an actual EE explained that I could use a chunk of the logic elements to swap things around, but I'd like to know: what's the standard practice for dealing with this sort of mirrored logic? (It's not really an inverse to my mind as a programmer, because you can't just negate the gate or the inputs as the truth tables for AND and OR aren't symmetrical.)

Ultimately, I plan on implementing my own small processor inside it (to tackle a specific crypto problem I have in mind), but I'd love any advice or corrections to the information contained in my post.

Thanks everyone! (And especially thank you Dave for teaching me so much about electronics!)

[miguelvp]

Too bad you didn't post before purchasing.

I got the BeScope Bundle for just $50 which includes the SDP Interposer with the mentioned connector.

My research shows me that the connector used by hitex on the original BeMicro is Edit(NOT) the one you mentioned.
Edit: correction, the one used is a MEC6 not MEC1 as you linked.

http://www.hitex.com/index.php?id=1804
Look at the schematic of the last product here that supposedly can be used with the BeMicro series including their new MAX10 offering.
http://www.hitex-download.de/lpc/lpc-stick/USR_SCM_LPC-Stick-Prototype-A1-2.pdf

But I haven't purchased such connector yet. At $10 it puts in in the expensive category for me:
http://www.newark.com/samtec/mec6-140-02-l-d-ra1/connector-card-edge-rcpt-80pos/dp/85T5569

And I don't have a need for it since I have the interposer and I can get the other connector cheaper (I'll have to look it up again what is on the other side of the interposer but as I recall the connector was only a couple of dollars)

[miguelvp]

Forgot to put the link to the bundle:

http://parts.arrow.com/item/detail/arrow-development-tools/bescopebundle

It comes with:

BeMicro CV Eval Kit
SDP Interposer
mini USB cable
oscilloscope probe
BeScope Daughter Card

I already had a BeMicro CV so now I have two

One problem is that the software distributed with the BeScope is missing some crucial file, but some one got it from the nCk-research team that Arrow hired to do the software. A batch file is still missing but it's not crucial.

Edit: link to where to get the missing file (BeServer.tcl) in this thread from parallax.com:
http://forums.parallax.com/showthread.php/157115-BeScopeBundle-System-Files

There is also a batch file missing but as mentioned before it's not crucial, but convenient to be able to run the scope code easier.

[tggzzz]

Looks like a good price for a reasonable board - except for the 40 pin connectors. One only has two GNDs, the other has 10 but they are somewhat clumped together.

I question whether there might be (unnecessary) signal integrity issues with tightly specified high speed signals.

[miguelvp]

Looks like a good price for a reasonable board - except for the 40 pin connectors. One only has two GNDs, the other has 10 but they are somewhat clumped together.

I question whether there might be (unnecessary) signal integrity issues with tightly specified high speed signals.

One of the 40 pin connectors is compatible with Terasic daughter boards (including all the DE series boards that have 40 pin connectors)
So you can find boards like this 3 channel full 8bit per color high speed (330Msps) video DAC board shown here with a DE0-Nano.


BTW is a good DAC:


The other 40 pin connector is not compatible with Terasic boards so you gotta be careful on using the right one.

[weilawei]

Fantastic! Thank you so much miguelvp! It's cool to see that other people have this board as well, since I've had some difficulty cobbling together documentation. I wish I had posted first, but hey--live and learn. I've gone and updated the blog post to reflect your advice.

I might eventually buy the kit and get a second board, but first, I think I'll get that connector and make myself a breakout board. Right now, I'm not really looking to use existing daughterboards, but more to make pins available for wiring up things to experiment with. I'm sure I'll get there eventually, but for now, it's all about the GPIO.

How's the quality of the BeScope?

[miguelvp]

How's the quality of the BeScope?

I only played with it briefly just to make sure I could make it work, and posted my thoughts here:
https://www.eevblog.com/forum/blog/eevblog-661-mailbag/msg513818/#msg513818

Excerpt:

Pictures of the device generating 2.5MHz and 5.0MHz ringing square waves as well what the DS2202 sees.

2.5MHz:


5.0MHz:


Rigol:

[miguelvp]

From that BeScope post I mentioned the connector on the other side in the interposer:

Not bad at all for just $50, the thing I'm more exited about is the SDP Interposer so I can connect to other Analog Devices boards.

http://wiki.analog.com/resources/alliances/altera

Also the connector Analog Devices uses is a Hirose FX8-120S-SV that mates with the FX8-120P-SV in the Interposer, they are fairly cheap.

So If you get the SDP Interposer you can use the Hirose FX8-120S-SV in your board. or if you get the connector you could also get the FX8-120P-SV and find out the schematic for the Interposer (I know it's somewhere on the net) and make your own interposer that is compatible with the Analog Devices boards

[tggzzz]

How's the quality of the BeScope?

Pictures of the device generating 2.5MHz and 5.0MHz ringing square waves as well what the DS2202 sees.

For us to understand the significance of that ringing, you have to define the probe used and how it was connected to the board.

And then, when considering the lack of GND pins on the 40-way connectors, see what happens w.r.t. ground bounce when many outputs simultaneousl switch in the same direction.

[miguelvp]

For us to understand the significance of that ringing, you have to define the probe used and how it was connected to the board.

The ringing is just be cause of the oscillator generating the 2.5MHz and 5.0MHz test signal.
This is the schematic that you can find in the link I gave above with more details:


I was just probing the in board test signal using both the ground and the signal pins on the test signal.

And then, when considering the lack of GND pins on the 40-way connectors, see what happens w.r.t. ground bounce when many outputs simultaneousl switch in the same direction.

As far the lack of GND pins I'm not sure what are you talking about, it has 2 GNDs one 3.3V and one 5V on the Terasic 40 pin connector.
The custom one has 10 GND pins out of the 40, and the 80 pin connector has 23 pins that I can't find assigned, I will have to look at the schematic if they are all connected to ground, pins [30-34] and [61-80]

[weilawei]

As far the lack of GND pins I'm not sure what are you talking about, it has 2 GNDs one 3.3V and one 5V on the Terasic 40 pin connector.
The custom one has 10 GND pins out of the 40, and the 80 pin connector has 23 pins that I can't find assigned, I will have to look at the schematic if they are all connected to ground, pins [30-34] and [61-80]

From schematics (not 100% sure they're correct, this is untested, will edit after testing).

On the card edge, using the format of: CEC Pin -> Label -> Pin Planner (IO Bank) (Notes)):

30 -> EG_P45 -> P14 (D)
31 -> EG_P13 -> AB18 (D)
32 -> GND
33 -> GND
34 -> EG_P46 -> R14 (D)

61 -> EG_P26 -> W21 (D)
62 -> EG_P58 -> V20 (D)
63 -> EG_P27 -> V21 (D)
64 -> EG_P59 -> U20 (D)
65 -> EG_P28 -> U22 (D)
66 -> EG_P60 -> P18 (E)
67 -> EG_P29 -> U21 (D)
68 -> NC
69 -> NC
70 -> NC
71 -> NC
72 -> NC
73 -> NC
74 -> NC
75 -> NC
76 -> GND
77 -> NC
78 -> EXP_PRESENT -> P19 (E) (R32, 10K to GND)
79 -> VCC5P0
80 -> VCC5P0

[tggzzz]

For us to understand the significance of that ringing, you have to define the probe used and how it was connected to the board.

The ringing is just be cause of the oscillator generating the 2.5MHz and 5.0MHz test signal.

Very unlikely, since the oscillation frequency is independent of the clock rate, probably around 100MHz.

I'll bet your probe has a 6" ground lead and croc clip, and a ~15pf tip capacitance.

See  http://www.ganssle.com/video/episode1.html http://www.ganssle.com/video/episode2_whats_in_a_scope_probe.html for an indication why, then read
Quick probe overview: http://www.williamson-labs.com/scope-probes.htm
Good intro to probes: http://circuitslab.case.edu/manuals/Probe_Fundamentals-_Tektronix.pdf
Quirky but informative about probes: http://www.dfad.com.au/links/THE%20SECRET%20WORLD%20OF%20PROBES%20OCt09.pdf

I was just probing the in board test signal using both the ground and the signal pins on the test signal.

Details matter where signal integrity is concerned.

And then, when considering the lack of GND pins on the 40-way connectors, see what happens w.r.t. ground bounce when many outputs simultaneousl switch in the same direction.

As far the lack of GND pins I'm not sure what are you talking about, it has 2 GNDs one 3.3V and one 5V on the Terasic 40 pin connector.
The custom one has 10 GND pins out of the 40, and the 80 pin connector has 23 pins that I can't find assigned, I will have to look at the schematic if they are all connected to ground, pins [30-34] and [61-80]

Precisely.

I prefer to see a 1:1 signal:ground for high speed signals, but 4:1 is usually not a problem particularly with differential signals, provided the grounds are near the signals, which implies evenly distributed. "Professional" high-speed board connectors meet those conditions.

You need to get a basic understanding of signal integrity. There's no such thing as a digital signal (except when considering individual electrons); everything is analogue and/or RF.

To give you an inkling of the issues, consider a single output that drives a 50ohm line through 1V in 1ns. That's a current change of 20mA in 1ns, or 20MA/s (and I mean MA/s not mA/s). Now a piece of wire (e.g. in a connector or IC lead) has an inductance of roughly 1nH/mm. Given a 3mm lead in a connector, the 20MA/s will induce a voltage across the wire/inductor of 60mV, i.e 6% of the signal voltage. Now have 10 outputs switching simultaneously, and you have an induced voltage of 600mV, or 60% of the signal.

[Dongulus]

Very unlikely, since the oscillation frequency is independent of the clock rate, probably around 100MHz.

I'll bet your probe has a 6" ground lead and croc clip, and a ~15pf tip capacitance.

Yeah, this is probably the case. You can make a quick short ground probe using a piece of insulated wire. Start by taking the "witch's hat" off of the oscilloscope probe then strip both ends of the insulated wire and wrap one end around the metal ground shield of the probe. Thicker gauge wire is a bit better because you can bend it and it will keep its shape. Pick a ground reference point close to where you are probing.

It takes some effort to master holding the probe tip and the ground probe in place with one hand. We have some real Tektronix short ground probes at work, but most of the other engineers tend to prefer the wire wound probes better. This will likely remove some of that edge oscillation if you keep the wire short.

[miguelvp]

There are plenty of high speed transceivers and lvds pins.

you not always need ground.

And it's just a test signal. it reads my 1khz test signal from the scope without ringing.

Also probing without ground since the test signal ground is the same as the scope module produces the ringing, so I'm not worrying too much about it and it's kind of off topic. I was just saying that for $50 he can get the bundle instead of paying $49 for just the dev kit. The interposer alone goes for another $50 for example. And the BeScope by itself is $45, the probe is a cheapy $10 100MHz, so you are getting $150 worth of things for just $50

[miguelvp]

Back to the OP message, I did look at your article and I want to confirm that both my boards have the 5CEFA2F23I7N higher specked FPGA, didn't even bother to look before so thank you 

As for the memory, you can use the NIOSII lower level soft core so you don't need to be tethered (I haven't tried it yet). Also you gotta check on hamster_nz's website he made his own SDRAM controller. There might be other controllers that don't need Avalon, but you got to initialize them with a State Machine.

http://hamsterworks.co.nz/mediawiki/index.php/SDRAM_Memory_Controller
http://hamsterworks.co.nz/mediawiki/index.php/Simple_SDRAM_Controller

Also the BeScope uses the memory via the Avalon interface without the NIOS soft core, so you can use the DRAM controller untethered.

There are two versions of the documentation for the BeMicro CV I will try to hunt them down where I got them from (nevermind you address that on your blog).

As for the inverted logic, it's not really inverted, a 1 is active High. Now your inputs might be inverted, didn't look at the details but if your input ports are set to 3.3V and you get one high and the other one low (0V) an AND will produce 0V as expected. Or maybe your output is inverted and the LED is pulled high, didn't look at the details but the board is not running on inverted logic levels.

[miguelvp]

More detail on the inverted logic problem you are experiencing.

Here are the schematics of the switches, buttons and LEDs on the BeMicro CV.


The switches and buttons are pulled high when off (putting VCC on the pin) and grounded when on or pushed (putting 0V on the pin)

On top of that the LEDs are pulled high, so when floating or off on the pin they light up. And they turn off when you output a 1 on the pin.



So the logic is valid within the FPGA but your inputs are inverted and so it's your output, so you'll have to add not gates to both to make it work properly.

You can do that by adding a schematic top level to define the pins and adding the appropriate Not gates.

Here are the DE0-Nano schematics, you'll notice that they are what you expect, and on top of that the buttons are hooked via a hex Schmitt-trigger buffer providing debouncing for the switch.

[miguelvp]

Updated:
https://www.eevblog.com/forum/microcontrollers/bemicro-cv-questions/msg523257/#msg523257

with link to where to get the BeServer.tcl file that is missing in the BeScope.

Also, do get the source for the BeScope to learn how to make your own Avalon Master so you don't need the NIOS II soft core.

[tggzzz]

And it's just a test signal. it reads my 1khz test signal from the scope without ringing.

So in one toy example there aren't any problems. I can believe that.

Now try to use the 40-way connectors "for real", and there may be subtle intermittent failures that can be difficult to diagnose and impossible to cure.

Hint: what's important is the signal's rise/falltime and the setup/hold times on inputs, and edge ringing causing false clocking. And all of those are as relevant for a 1kHz signal as they are for a 50MHz signal.

Also probing without ground since the test signal ground is the same as the scope module produces the ringing, so I'm not worrying too much about it and it's kind of off topic. I was just saying that for $50 he can get the bundle instead of paying $49 for just the dev kit. The interposer alone goes for another $50 for example. And the BeScope by itself is $45, the probe is a cheapy $10 100MHz, so you are getting $150 worth of things for just $50

I was addressing your point and pictures about "device generating 2.5MHz and 5.0MHz ringing square waves". I made no comment about anything else you said.

[miguelvp]

I've driven that DAC up to 1600x1200 at 60Hz, so 162 MHz pixel clock. Edit (using that 40 pin connector)

In this capture it shows the time for 256 values. So multiply the 1 over delta X to get the approximate pixel clock.



Edit: 1600x1200 @60Hz output (phone camera picture)


Pushing the pixel clock to 330MHz beyond what the FPGA should be capable off I was getting some noise, you can clearly see 16 bigger spikes over the 256 value range and noisier.


On the previous page it was a 25.6MHz pixel clock

All of them 24 pins for the RGB plus HSync and VSync


Edit: And if you are compelled to know how fast the edge connector is, feel free to send me any one of these boards:

http://wiki.analog.com/resources/alliances/altera

[tggzzz]

Very unlikely, since the oscillation frequency is independent of the clock rate, probably around 100MHz.

I'll bet your probe has a 6" ground lead and croc clip, and a ~15pf tip capacitance.

Yeah, this is probably the case. You can make a quick short ground probe using a piece of insulated wire. Start by taking the "witch's hat" off of the oscilloscope probe then strip both ends of the insulated wire and wrap one end around the metal ground shield of the probe. Thicker gauge wire is a bit better because you can bend it and it will keep its shape. Pick a ground reference point close to where you are probing.

It takes some effort to master holding the probe tip and the ground probe in place with one hand. We have some real Tektronix short ground probes at work, but most of the other engineers tend to prefer the wire wound probes better. This will likely remove some of that edge oscillation if you keep the wire short.

Personally I like the bayonet/spear type ground connection on a low-impedance Z0 probe, which is demonstrably good to >1GHz (and potentially 6GHz if you have the right Keysight probe!).

I suspect your "real Tektronix short ground probes" are bayonet/spear grounds. They are surprisingly easy to crete at home with a needle, phosphor bronze sheet and Fimo modelling clay! They are also a prime candidate for a small project on a 3D printer.

There are good descriptions of how to make your own GHz Z0 probes on the web.

Putting all that together, I'm surprised that people are prepared to publish diagrams showing ringing digital signals; it looks as if they don't know what they are doing.

[tggzzz]

I've driven that DAC up to 1600x1200 at 60Hz, so 162 MHz pixel clock. Edit (using that 40 pin connector)
In this capture it shows the time for 256 values. So multiply the 1 over delta X to get the approximate pixel clock.

Very pretty pictures, and I am glad they work for your application.

But I don't think the pictures can be taken as indicating too much about the signal integrity. The waveforms shown are too far removed from the potential problem, so there's no indication of how close to the edge of the envelope (or eye diagram in this case) the waveforms are.

Is there any possibility you could show diagrams when using a low impedance Z0 probe plus very short ground lead/bayonet/spear? That probing technique is good to >1GHz, and actually the "low impedance" Z0 probe probably has a higher impedance than the probe I suspect you are using.

In addition, you state the signals in question are (very sensibly) differential, which will of course avoid degrading signal integrity - as I noted in an earlier post. If someone else in a different design uses single-ended signals, the results might be quite different.

[miguelvp]

I said the pins can be configured to do differential signals. Those are not the case in there.

Also my 200MHz scope with 300MHz probes at 2Gsps falls short even for the 330MHz measurement.

In any event, the BeScope measuring its own test signal with the ground clip removed showed the ringing even with a shared ground with the test signal.

[tggzzz]

I said the pins can be configured to do differential signals. Those are not the case in there.

Also my 200MHz scope with 300MHz probes at 2Gsps falls short even for the 330MHz measurement.

Of course it does. What else would you expect with a 200MHz scope looking at a 330MHz signal? Even the fundamental will be significantly attenuated, and the harmonics will be grossly attenuated. The 2GS/s spec is effectively meaningless in this context.

(If you mean a 330Mb/s signal then the effects won't be quite as pronounced, but they will still be significant. The rule of thumb is that to look at the signal integrity of an X MHz clock, you need a scope with a bandwidth >5X MHz)

In any event, the BeScope measuring its own test signal with the ground clip removed showed the ringing even with a shared ground with the test signal.

What else would you expect to see if you don't have a properly grounded probe?! At these frequencies there's no such thing as a "shared ground".

Seriously consider reading those references I gave in an earlier post - you'll learn a lot and will get better results.

[miguelvp]

The ground on the test signal is the same ground on the scope barrel, are you clamming that's not properly grounded if I use the probe with the ground clip removed?
(the signal is labelled as CLK Out).

Edit: the board is designed to do 250MSps topping at 50MHz. At least my Rigol uses a x10 ratio 200MHz with 2GSps.



Schematic here:
http://download.siliconexpert.com/pdfs/2014/8/12/6/18/35/157/arrowd_/manual/bescope-revg6schematics.pdf

[tggzzz]

The ground on the test signal is the same ground on the scope barrel, are you clamming that's not properly grounded if I use the probe with the ground clip removed?
(the signal is labelled as CLK Out).

Oh good grief!

Look at the references I previously posted in this thread.

They will explain the theory and show experiments demonstrating that the theory matches practice.