LaPOD: Low cost Logic Analyzer probe for MSO5k, DHO900 and more!

[ebastler]

All traces, from the 50 pin connector, to the output pin, should all be exactly the same length for each signal I didn't length match pairs only, For obvious reasons, if I measure a 2 channel signal, across 2 pods, I don't to care which input is matched, but want to ensure the delay is identical everywhere. That's why some of the lines look a bit weird, while not being snakes, because they are length matched too.

I did not express my thought well, I'm afraid. I realize that you have properly length-matched everything, and that it must have taken a lot of tweaking!

What I meant to say is that you can potentially simplify the length-matching. It looks to me like, on the pod PCB, you have (a) length-matched the 4 signal traces from the 2*4 connector to the LMH7324, and (b) also length-matched the 8 signal traces from the LMH7324 to the HDMI connector. But the requirement is only that the sum "length of trace before + length of trace after" the LMH7324 need to be matched between the four channels. So two of the channels may have shorter traces from the input connector to the near side of the LMH7324, and compensate by having longer traces from the LMH7324 to the HDMI.

This thought doesn't matter if you do not need to touch the layout any further; it is clearly good as it is. But if you needed to move things around, get the capacitors closer or such, it might help to work with that relaxed constraint, because it can probably reduce the overall trace lengths and wiggles somewhat.

I got [the blocking caps] as close as I could, there's traces underneath for power already, so dropping them isn't an option really. The caps are 6mm away, which I admit is further then I'd prefer too. Also, I really wanted to avoid via's, as that introduces other problems of course. There's always the possibility to add extra caps at the bottom though, like what dren.dk did, but I didn't want to compromise on the heat dissipation ability, as the bottom plane is where the ground plane gives the heatsink, adding those capacitors reduce that by quite a bit.

OK, I have not seen the inner layers of course. If they are too busy, my suggestion may be impractical -- never mind!

I would also prefer to keep the bottom side free of components -- better heatsinking, more mounting options, also easier to home-build.

[oliv3r]

All traces, from the 50 pin connector, to the output pin, should all be exactly the same length for each signal I didn't length match pairs only, For obvious reasons, if I measure a 2 channel signal, across 2 pods, I don't to care which input is matched, but want to ensure the delay is identical everywhere. That's why some of the lines look a bit weird, while not being snakes, because they are length matched too.

I did not express my thought well, I'm afraid. I realize that you have properly length-matched everything, and that it must have taken a lot of tweaking!

What I meant to say is that you can potentially simplify the length-matching. It looks to me like, on the pod PCB, you have (a) length-matched the 4 signal traces from the 2*4 connector to the LMH7324, and (b) also length-matched the 8 signal traces from the LMH7324 to the HDMI connector. But the requirement is only that the sum "length of trace before + length of trace after" the LMH7324 need to be matched between the four channels.

Yeah, I thought of that too, but that's actually harder, because then you need to do math a lot kicad can't relate and sum traces between components. Also the 'win' there is minimally anyway, but a descent idea anyway.

I'll try to route the traces around the capacitors, to get them closer. I realized, the extra 1 - 2 cm are insignificant, as a) they are differential digital signals, and b) look at the wiggles on the breakout board, I'm not carying about length there either ...

Regarding the HDMI vs mini/micro HDMI discussion, before the raspberry pi, mini-hdmi is something I always thought of, as those cheap tablets used to have them, so there always was some availability. But with the PI, getting cables might actually be quite feasable, so micro HDMI might not be that weird to aim for. I'll have to look at the pinout, but routing may not be a problem either, as the signal traces may be at the top row, and the bottom row be the grounds.

So then, mini-hdmi or micro-hdmi I wonder, they cost about the same, so that's not the issue.

edit: The signals are not all on the top, so routing would be much more painfull, so for now, i'll try mini-hdmi

[ebastler]

So then, mini-hdmi or micro-hdmi I wonder, they cost about the same, so that's not the issue.

Technically, mini-HDMI is probably the better choice: A bit sturdier if one bumps against the plugs by accident, and small enough to fit all four connectors onto a PCB not wider than the scope's port.

Future-proofing might speak in favor of micro-HDMI. As devices keep getting smaller, micro-HDMI is probably taking over and mini-HDMI cables will become harder to get in the long run. But I would not be overly concerned about this: You don't need to plan for a 10-year production lifecycle, and if push comes to shove someone could always add a micro-HDMI breakout board later.

[oliv3r]

How do we feel about this? In terms of dimmensions, the PCB itself is now 68mm wide

[ebastler]

How do we feel about this?

 

[UK]

How do we feel about this?

It looks neat, but it also looks like a big lever and possible strain on the scope port... which will clearly affect its service life.

I even liked the vertical connector idea, but those special IDC connectors make me a bit cringe :p

Is there any real reason for that cringe?! it definitely will put less strain on the scope port.
Also, LEDs can be more visible from the front view, especially while using the DHO900 on the vesa mount.

[oliv3r]

Hmm, that mini-HDMI part is intended to be 'in-pcb' mounted, e.g. requiring a cut-out. If you can find another mini-hdmi port that's 'just flat' that'd be good

How do we feel about this?

It looks neat, but it also looks like a big lever and possible strain on the scope port... which will clearly affect its service life.

I even liked the vertical connector idea, but those special IDC connectors make me a bit cringe :p

Is there any real reason for that cringe?! it definitely will put less strain on the scope port.
Also, LEDs can be more visible from the front view, especially while using the DHO900 on the vesa mount.

The cringing part is that it's quite deep, but more importantly, it's a single-supplier part, it costs 10 times more then a regular connector (9 euro's a piece), will make routing probably quite hard, also because it's a through-hole part.

Yes, the led's visibility will be better, but then there's also 'side-entry' RGB leds. I simply used this one because I used it in a different project (and it's nice and big, side entry RGB's tend to be super small)

[UK]

I agree with you point. I've just recall that also already used 'side-entry' leds on several projects.

[ebastler]

Hmm, that mini-HDMI part is intended to be 'in-pcb' mounted, e.g. requiring a cut-out. If you can find another mini-hdmi port that's 'just flat' that'd be good

LCSC part number C2682170, you mean? Doesn't look to me like it would need a cutout. Were you looking at the same drawings?
https://datasheet.lcsc.com/lcsc/2110112030_XKB-Connectivity-A71-05H4-111N1_C2682170.pdf

[oliv3r]

Hmm, that mini-HDMI part is intended to be 'in-pcb' mounted, e.g. requiring a cut-out. If you can find another mini-hdmi port that's 'just flat' that'd be good

LCSC part number C2682170, you mean? Doesn't look to me like it would need a cutout. Were you looking at the same drawings?
https://datasheet.lcsc.com/lcsc/2110112030_XKB-Connectivity-A71-05H4-111N1_C2682170.pdf

No, I used the wrong part, well a different part, which did have the cutout I saw the name in the pdf, which made me trigger that it was the wrong one. I have too many HDMI parts in my lcsc library now :p

[oliv3r]

Anyway, lets break down our options for a moment.

'top full-szie 2x2' entry:
cons:
* requires expensive extra deep/long IDC connector, hard to source, single supplier
* vertical HDMI connectors sticking out quite a bit to the front
* 'base' quite 'high' (top-bottom)
* vertical HDMI connectors have many connection pins, but put a lot of strain on the connector
pros:
* easily fits 4 connectors
* standard leds easily face forward
* strain on the connectors is more equally distributed?

vertical HDMI connectors
cons:
* increases height (makes the breakout fat) and sticks out further due to their longer length
* leds sit next to the connector, making it harder to identify 'is it the one to the left or right'
pros:
* makes all connectors fit nicely

mini HDMI connectors
cons:
* cables are harder to source
* connections are less strong due to the more fragile connector

micro HDMI connectors
cons:
* Much harder/more ugly to route
* much more fragile
* cables not the easiest to source (but certainly possible)

full HDMI connectors
pros:
* easier to route
* much more robust
cons:
may not fit on non-mso5k

(I'm sure we can increase and work on this list more :p) but for now, I think just doing two layouts, mini and full boards. We can always do a 3rd or a 4th too :p the breakout boards are after all the easiest to do. If for example you have both an MSO5k and a DSO900, you may get one each or just use 2x2 one

[ebastler]

for now, I think just doing two layouts, mini and full boards. We can always do a 3rd or a 4th too :p the breakout boards are after all the easiest to do. If for example you have both an MSO5k and a DSO900, you may get one each or just use 2x2 one

Thanks! I agree that having two options, mini and full-size HDMI, should cover all bases.

For the full-size version, I quite like UK's 2*2 proposal. Maybe the exotic, tall 2*25 connector can be avoided by simply stacking two regular connectors and glueing them together? For the HDMI connector, C168715 looks like the only LCSC part which is meant to be mounted that way. (They also have C2962411 and C711355, which are sold for SMD mounting on a PCB edge but could probably be "abused" as vertical through-hole parts.)

Those vertical connectors should get some mechanical support from a 3D-printed enclosure. But I trust UK will find a stylish solution there! 

[oliv3r]

for now, I think just doing two layouts, mini and full boards. We can always do a 3rd or a 4th too :p the breakout boards are after all the easiest to do. If for example you have both an MSO5k and a DSO900, you may get one each or just use 2x2 one

Thanks! I agree that having two options, mini and full-size HDMI, should cover all bases.

For the full-size version, I quite like UK's 2*2 proposal. Maybe the exotic, tall 2*25 connector can be avoided by simply stacking two regular connectors and glueing them together? For the HDMI connector, C168715 looks like the only LCSC part which is meant to be mounted that way. (They also have C2962411 and C711355, which are sold for SMD mounting on a PCB edge but could probably be "abused" as vertical through-hole parts.)

Those vertical connectors should get some mechanical support from a 3D-printed enclosure. But I trust UK will find a stylish solution there! 

I wouldn't dare putting all those cable stresses on an SMD footprint, but without further ado, the mini HDMI version

One thing I just realized, is that I didn't match the PCB size, but then it also shouldn't matter so much, as the width is different anyway, so having the same depth doesn't add any value.

I did swap the LED pins on the HDMI connector, so I'll have to go back to the other bards and fix that, but still, I think this came out well.

[ebastler]

That looks very nice and tidy! 

When I mentioned the "board edge" HDMI connectors, that may have read a bit confusing. If you look at the drawings for the three part numbers I mentioned, you will find that they all look very much the same, but the manufacturers give different footprints for them: One with vertical mounting, and through-holes for the mechanical mounting tabs as well as all the pins. The other two for mounting horizontally on the board edge, with the PCB between the two rows of pins, and SMD mounting only. But the parts seem to be more or less the same, so all three could probably be mounted vertically (with through holes).

Anyway, as long as the C168715 part is in stock, that's probably the one to get, because one can be reasonably sure that the footprint is correct.

[UK]

Those vertical connectors should get some mechanical support from a 3D-printed enclosure. But I trust UK will find a stylish solution there! 

For sure!

C168715 looks like the only LCSC part which is meant to be mounted that way.

Here you go, C5204145 , C2682172 and short one C711353

I think this came out well.

That's true! May be I missed that but why you didn't route channels D0 and D16?

[ebastler]

C168715 looks like the only LCSC part which is meant to be mounted that way.

Here you go, C5204145 , C2682172 and short one C711353

Oh, nice! It is reassuring to see that there is a larger selection. I had scanned through the thumbnail pictures too, not trusting LCSC's keywords, but did not see (or recognize) those.

I wonder what the "brick nogging" tag means? Is that a standard industry term or the unexpected result of a Chinese-to-English translation?


Edit: Ah....

As a native Chinese speaker, "brick nogging" is gibberish and completely incomprehensible to me.   I was curious and just looked it up, apparently it was a mistranslation of "立贴".
"立" means "standing" or "vertical", and "贴" is the short-hand for "贴片", which means "pick-and-place" or "SMD", so it just means "vertical SMD".

[oliv3r]

So here's the new LM7324 POD with the capacitors hugging the chip as closely as possible.

One thing I just came to realize though, is that the pin-order on the POD's are all 3, 2, 1, 0. This was due to the breakout board, swapping that order makes routing really really hard, as all lines cross each other. I don't know how the original pod solves this, or if the order is also backwards? But the solution is of course simple, flip the pod around but that puts the led on the bottom, and since I don't want to do dual-sided components ... maybe I'll do the leds on both sides ....

Flipping the connector btw won't help that much, because it's a side-entry connector, so flipping it just puts it up-side down, but pin 0 is still on the right ...

May be I missed that but why you didn't route channels D0 and D16?

I don't understand, they are routed aren't they?

[ebastler]

So here's the new LM7324 POD with the capacitors hugging the chip as closely as possible.

One thing I just came to realize though, is that the pin-order on the POD's are all 3, 2, 1, 0.

Cool, that layout looks really nice and tight now! 

I would not worry at all about the bit order. To the contrary: Isn't that how multi-digit binary numbers should be written, low bit on the right? 

[UK]

I don't understand, they are routed aren't they?

I've read the mask on the board with mini-HDMIs - 1-3,4-7,9-11,12-15... looks like 0 and 8 are missing

One thing I just came to realize though, is that the pin-order on the POD's are all 3, 2, 1, 0.

I forgot to tell you last time, that the length of the PCB should be extended for proper placement of the 8p socket, since it has to be lying on the PCB (at least half of it) and not floating in the air. Because anyway we have this empty space in the shell under the 8p socket. So then if the length will be extended and we get additional PCB space, why just not reroute the pins order on the PCB backside like this?

After all, we have 6 spring-connection points for each channel, so 2 more VIAs should not drastically affect on the signal.

[oliv3r]

I don't understand, they are routed aren't they?

I've read the mask on the board with mini-HDMIs - 1-3,4-7,9-11,12-15... looks like 0 and 8 are missing

ohh, fair point, the silkscreen is wrong

One thing I just came to realize though, is that the pin-order on the POD's are all 3, 2, 1, 0.

I forgot to tell you last time, that the length of the PCB should be extended for proper placement of the 8p socket, since it has to be lying on the PCB (at least half of it) and not floating in the air. Because anyway we have this empty space in the shell under the 8p socket. So then if the length will be extended and we get additional PCB space[/quote]
I wanted to keep the PCB length <5cm because that's where you get the cheaper rates with the board houses, or you can put 2 on a 10x10 board
I realize its floating in mid air, and would only be held in place by the solder and the housing, which is a bit weird, but just more economical I suppose. I can now put 3 pod designs on a 10x10 PCB order

, why just not reroute the pins order on the PCB backside like this?

To flip the order? Yeah I was thinking that too, though it would also mean length matching traces again etc. And, having a MSB - LSB layout isn't that bad :p

After all, we have 6 spring-connection points for each channel, so 2 more VIAs should not drastically affect on the signal.

Yeah, I'll sleep on it some, though those connection points etc are all on those differential channels. This is now on the 'more anloguey' input. fair nuff, it's a digital input in the end, so an additional via won't hurt of course.

So it would end up something (a little prettier) like this, length matching should be duable, I just don't like the fact that I have to use really thing traces now (0.25mm), because much more won't fit between those pins of course ...

[ebastler]

So it would end up something (a little prettier) like this, length matching should be duable, I just don't like the fact that I have to use really thing traces now (0.25mm), because much more won't fit between those pins of course ...

That should work. You can make the traces nearly twice as thick as they currently are and still keep the minimum distance from the pins, it seems.

And you should not need much extra room for length matching in that area: Channels 0 and 3 have extra wiggles above the voltage dividers anyway; if you straighten those, the lengths should be reasonably well-matched already?

[UK]

I wanted to keep the PCB length <5cm because that's where you get the cheaper rates with the board houses, or you can put 2 on a 10x10 board

I don't understand why you even need this since previously you told us you won't mass produce them... currently, every PCB manufacturing service offers a minimum 5pcs of PCB which is 4 pods + 1 extra... what a point to place 3 pcs on 1 PCB if you do not plan to get in total 15 pods

You can make the traces nearly twice as thick as they currently are and still keep the minimum distance from the pins, it seems.

Totally agree!
You can route them like this to get nearly perfect length matching.

Edit: really forgot we have here 4 layer PCB 

[mwb1100]

A little bit of an interjection:  I'm really a software guy so pretty much everything here is over my head, but I have been loving following this thread.  Seeing the issues involved in designing a layout AND seeing the amazing collaboration going on here is just fantastic!

[oliv3r]

So while fiddling with it, it wasn't actually that hard and not even that ugly. I've put the traces all in the middle, so that a) ground planes help, and b) I did really need that room.

I wanted to keep the PCB length <5cm because that's where you get the cheaper rates with the board houses, or you can put 2 on a 10x10 board

I don't understand why you even need this since previously you told us you won't mass produce them... currently, every PCB manufacturing service offers a minimum 5pcs of PCB which is 4 pods + 1 extra... what a point to place 3 pcs on 1 PCB if you do not plan to get in total 15 pods

Well I also want some spare/give away's :p and .. options/premature optimization :p My OCD also won't cope going just a few mm over 5cm board size

As for mass-producing, if there's tons of people wanting these boards, I might order a bigger quantity of course, but I won't be going into business selling these, that's true.

You can make the traces nearly twice as thick as they currently are and still keep the minimum distance from the pins, it seems.

Totally agree!
You can route them like this to get nearly perfect length matching.

Yep, I played with it a bit, and got to 0.4mm which isn't too horrible I thought.

The wiggles above the voltage dividers are needed, because the center two channels come from the top of the chip, so need to go all the way from the top, to the bottom.

A little bit of an interjection:  I'm really a software guy so pretty much everything here is over my head, but I have been loving following this thread.  Seeing the issues involved in designing a layout AND seeing the amazing collaboration going on here is just fantastic!

Same here, I couldn't properly design this, I don't know enough about EE, but we got the reverse engineered pods and earlier designs, that I'm hopping will prove to be true :p

It's mostly just me taking other peoples great idea's and re-spinning them, listening to ebastler and UK for feedback/wishes

P.S. I asked the admin/moderate to spin this thread into its own forum post, so lets see if that'll happen

[ebastler]

The wiggles above the voltage dividers are needed, because the center two channels come from the top of the chip, so need to go all the way from the top, to the bottom.

No, I don't think they are needed. They are in channels 0 and 3, right? And now you have created additional wiggles in channels 1 and 2 just above the 2*4 plug. They pretty much even out, so just do away with all four of them -- that's what I meant.

Otherwise, great work, and I am excited to see the pod getting its final touches!