ve7fim

ve7fim, 5 months ago

@sxpert

Indeed! At 20 sliders per module, they add up really quickly.

ve7fim, 5 months ago

@sxpert

I put up all the boards and BOMs for ordering from JLC, but this one isn't finished yet... https://github.com/dslik/nanomia

ve7fim, 5 months ago

Progress! All the alarms are cleared, and it's 50% through updating the almanac.

Time to make some nice custom-length colour-coded Ethernet cables to tidy everything up while it finishes.

ve7fim, 5 months ago

Ethernet cables are done, and everything looks neat now.

But the clock just wouldn't sync with the other PTP GM, with a near constant 38 ns offset. Turned out that I had an old configuration for my antenna cable length, which I needed to update to try to fix the problem.

It never ceases to amaze me how precise you can get things nowadays.

So, back into alarm-land...

ve7fim, 5 months ago

Back to no alarms, and everything syncs up between the two PTP systems.

Good to have that up and running again.

I can actually cross a project off the "to do" list!

ve7fim, 5 months ago

Looks like I spoke too soon — the system is still having some problems. Specifically, it seems to be occasionally loosing GNSS tracking, which then forces the oscillator loop to have to recover.

I have a good view of eight SVs right now, so I wonder if this is another one of the intermittent broadband interference issues I've been trying to track down.

ve7fim, 5 months ago

@sxpert

I'm running this through a SyncSwitch TC100, but the problem is upstream (GNSS tracking/sync).

ve7fim, 5 months ago (edited 5 months ago)

@sxpert Oh, that's just for the dedicated PTP / IEEE 1588 timing network.

My edge network switches are HPE 2910al and my core network switch is an HPE 5406Rzl2. I run 10GigE between the core and edge (and to some computers/devices).

It's all old equipment, and I need to upgrade soon because I've been running more SMPTE 2022 and 2110 streams over the network and those sure do suck up the bandwidth.

ve7fim, 5 months ago

@sxpert Yeah. I quite like those Blackmagic 2110 IP Converters.

Most of my stuff is Evertz and AJA, though.

ve7fim, 5 months ago

@sxpert

The used market can be weird that way. Often the BlackMagic stuff is more expensive!

ve7fim, 5 months ago

@sxpert

Sometimes you can get really good deals, like https://www.ebay.com/itm/165604312472

ve7fim, 5 months ago

@sxpert

I have one of UBNT's eight port 10G aggregation switches for NAS connectivity. Inexpensive and works fine for that use case. That new Hi-capacity aggregation switch looks really nice!

Wish I had the $$ to get one of BMD's Cloud Store Mini, but that's out of my budget too.

ve7fim, 5 months ago

@sxpert If you're just looking for 3G SDI over 10G, there are some good used prices for Evertz MMA10G-IPGM-6 modules. Things get a little more expensive when you jump up to 12G SDI.

The problem with the Evertz ecosystem is that you need to get all this other stuff (controllers, shelves, etc) to assemble a working system.

ve7fim, 5 months ago

@sxpert True, but then it won't be a step up fro my current NAS.

ve7fim, 5 months ago

@sxpert

I know that feeling!

ve7fim, 5 months ago

@sxpert

The nice thing about the MMA10G-IPGM-6 is that it's six SDI in and six SDI out per card, so you get reasonably high density. And it has built-in scalers.

I still mostly work with 1080p, with only a few things running in 4K so far, so lots more opportunities for used stuff since everyone is getting rid of their older kit.

ve7fim, 5 months ago

@adorfer

I believe it is based on a meme where people green-screen in different things in to the sinkhole. I just found the author's choice of what they put in the sinkhole to be hilarious because of how unexpected it was.

ve7fim, 5 months ago

@stephlahs I feel seen. (Looks over at desk full of almost-finished projects waiting for firmware work to be done...)

ve7fim, 5 months ago

Software is checked in.

Noticed some UART corruption when echoing back pasting blocks of text. E.g. "clear" becomes "slear". The data is received fine, but the echoed data is corrupted somewhere on the way out.

(01100011 -> 01110011)

This only seems to happen in certain configurations. In the full debug/dev env, I can't reproduce it at all. When I power it from the computer USB port, it is most noticeable, and when I power it off of a battery, it happens less frequently.