<azonenberg>
Soooo the new AKL-AD3 prototype is much better from a mechanical standpoint
<azonenberg>
Still has some issues (minimum spacing of the jaws is too large)
bvernoux has joined #scopehal
someone-else has joined #scopehal
someone-else has quit [Quit: Connection closed]
someone-else has joined #scopehal
GenTooMan has quit [Quit: Leaving]
GenTooMan has joined #scopehal
Stephie has quit [Quit: Fuck this shit, I'm out!]
Stephie has joined #scopehal
<azonenberg>
And electrically, I'm an idiot
<azonenberg>
I forgot to put a ground plane on the board
<azonenberg>
like, the ground layer just has a few ratsnest traces and no zone fill
<azonenberg>
sooo all of my impedances are completely off
<azonenberg>
i spent so much time simulating this
<azonenberg>
then never actually drew the layout i simmed
* azonenberg
facepalms with a giant trout
<azonenberg>
So i guess i'm respinning the amplifier
<azonenberg>
I was gonna make some mechanical tweaks anyway, the current flex has two extra boards which i think are good
<electronic_eel>
ouch
<azonenberg>
I cant reuse the flex on this prototype because right now the AKL-AD3 design is kind of a one-way assembly process
<electronic_eel>
but this was just an oshpark board, right? not some fancy expensive board house
<azonenberg>
Yes, but it has a $40 diff amplifier on it
<azonenberg>
and two $6+ linear tech low noise LDOs
<azonenberg>
So it's not *cheap* to respin
<electronic_eel>
can't you desolder the parts?
<azonenberg>
Not nondestructively
<azonenberg>
The assembly is one-way
<azonenberg>
what happens is, first i put the amplifier board sandwiched between the two enclosure halves
<azonenberg>
then i mate the flex to the connector on it, through the opening
<azonenberg>
then i glue the flex down to the enclosure
<azonenberg>
on the tip side
<electronic_eel>
the amps are glued in too? i thought they were just plugged into the flex board
<azonenberg>
once that happens there's no way to lift the flex enough to unmate the connector
<electronic_eel>
and the flex was glued in
<azonenberg>
if you cut the flex, you can unscrew it and remove the amp
<azonenberg>
but there is no way to access the amp without damaging some part of the system once populated
<azonenberg>
In the production rev, the amp board will be accessible by removing the top half of the enclosure because there won't be a connector
<azonenberg>
it'll be flex-rigid
<azonenberg>
but the flex end will still be glued to the jaws
<azonenberg>
Anyway, i can scrap the first two revs in a pinch if i need parts
<azonenberg>
but right now i have one more amplifier in inventory and five more on order from digikey who said they were in stock when i clicked buy
<azonenberg>
and i have i think eight or nine boards worth of LDOs
<azonenberg>
and plenty of passives
<azonenberg>
So i can do several more iterations without being hit by the shortages, at least
<electronic_eel>
hehe, right now i just believe that something is in stock when they send me the shipping confirmation...
<azonenberg>
Production is a different story but the mechanical design will likely need a few iterations before i think about making these in volume anyway
<azonenberg>
I have one in inventory right now at home
<azonenberg>
So this respin is assured
<azonenberg>
everything i need to populate it is on hand physically
<azonenberg>
Anyway, off to get some sleep, it's 10am on saturday and i've been awake since some unknown hour friday afternoon
<electronic_eel>
have a good night
<azonenberg>
i already have the new ground plane layout drawn
<azonenberg>
but am holding off on ordering boards until i've run a few more simulations to triple check
<azonenberg>
and make sure this is the only problem :p
<electronic_eel>
do you have enough test points on the amp board? when you can't access it enough to desolder the amps, it might be difficult to properly measure all important poinst too
<azonenberg>
There isn't really a lot of things i could probe
<azonenberg>
the output signal is a SMA, the input diff signal is broken out to the connector
<azonenberg>
the only other thing on there is power and i have test points outside the shielded enclosure for 2V5_N and 2V5_P
<azonenberg>
The board is literally the diff amp, two LDOs, and some connectors and passives
<electronic_eel>
can you measure the whole input trace, like from the pins of the probe to the amp? you could measure that with a vna or tdr to find some weak spots or causes for unwanted resonances and so on
<electronic_eel>
"input diff signal broken out to the connector" sounds to me like you only can measure either the flex or the amp, but not both plugged together
<someone-else>
I wonder what people think about amplifier-less diff probes which use 2 scope channels..
<someone-else>
while using 2 channels might be inconvenient, there is no bandwidth limit imposed by the amp and CMRR/flatness can potentially be better since we can calibrate them out in digital domain
<electronic_eel>
someone-else: a diff amplifier in hardware will have better common mode noise rejection. you'll be able to use all the bits of your scope for the actual signal you want to measure
<someone-else>
electronic_eel: could be, but from what I've seen CMRR for diff probes @GHz frequencies is often worse than scope's SNR
<someone-else>
depends on relative common/diff mode magnitude, though
<someone-else>
if common mode is larger, then sure
<electronic_eel>
shouldn't you be able to cal out flatness issues of a diff amp in the digital domain too?
<someone-else>
electronic_eel: yes, but only if it's not limiting the overall system bandwidth
<someone-else>
e.g. if I have 5ghz diff amp and 10ghz scope, calibrating the amp into 10ghz bw might be problematic
<electronic_eel>
yeah, sure
<electronic_eel>
you won't be able to go much higher than what the diff amp can do. but that is an issue with single ended active probes too
<someone-else>
but not so much with the passive ones
<someone-else>
I was just wondering why there are no commercial products like that (i.e. cheap passive diff probe which uses two scope channels)
<someone-else>
could be because nobody wants such a thing or perhaps big scope vendors never bothered..
<azonenberg>
So actually
<azonenberg>
when i build my scopes
<azonenberg>
i'm starting to seriously consider having native differential inputs
<azonenberg>
with the option of terminating IN_N if you have a single ended probe
<azonenberg>
You would also be able to terminate IN_P and use IN_N only, thus giving you a 2:1 mux on your inputs but IN_N is inverted
<azonenberg>
but most importantly you would be able to hook a diff pair up to a single scope input without using a second channel
<azonenberg>
or an external diff probe
<azonenberg>
just two separate 50 ohm inputs
<azonenberg>
I'm amazed this isn't more popular
<GenTooMan>
azonenberg, that's how scopes USE to be set up. When digital scopes came into being the bureaucratic silliness of such businesses solved this by using SOFTWARE instead. So they subtracted channel and from channel B in software, and thus the resolution of the difference was limited by the input channel range selected.
<GenTooMan>
This methode works ... not at all, and gives useless results to very poor results.
<azonenberg>
GenTooMan: you miosunderstand
<azonenberg>
i don't mean a hardware subtractor between two separate channels
<azonenberg>
I mean twice as many input ports as ADCs
<azonenberg>
with partially shared frontends
<azonenberg>
I'm pretty sure I could do it in my front end by adding just a handful of components
<azonenberg>
you'd double up the input connector, protection circuit, and antialiasing filter
<azonenberg>
then in the offset stage, add a 2:1 mux where you either subtract IN_N or a DC offset from IN_P
<azonenberg>
(this approach would mean no vertical offset control for differential signals in hardware, but I think that would be OK?)
<azonenberg>
then you'd probably also add a second mux to the positive side of the amp, hard wired to one path, just to balance the paths better
<azonenberg>
I think when I get back to the scope project i'm going to include this feature on the prototype
<azonenberg>
also is it bad that i'm thinking of putting those $$$$ Rosenberger SMAs on the 500 MHz scope input?
<azonenberg>
they're expensive but also super durable with the wide footpritn and screw mount
<azonenberg>
for a front panel connector that will get a fair bit of use i think it makes sense
<GenTooMan>
do you plan on using a MUX or RF relays to route signals for the differential input?