really slow PLL

At minimum it likely won't scale very well. Why implicitly discourage one's customers from buying only a limited number of units

I wonder if there's an advantage to using the closure phase for an array that large. With 17 oscillators you've got 136 independent phase differences, so maybe there's a way to get 22 dB instead of 12 dB improvement.

Cheers

Phil Hobbs

but they are anot free runnign are they? they are all reacting to midi

Given that it's so simple to do it right, why not do that?

Cheers

Phil Hobbs

There's a system clock in each one surely but they don't try to sync their system clocks, they receive an instruction "do X for Y ms" and their processor figures out how long Y ms is, and does it for that long.

It is literally good enough for rock & roll, but whether it's good enough for power supply sequencing IDK, there is gonna be some latency.

HP used to have GPIB on their power supplies, I've never used it but I expect it wasn't really useful for tight synchronization.

My power supply box has 8 plugin modules of various types. Some don't need much air but some really do. The two big fans are howlers at 48 volts.

Each module can present one bit to the motherboard software: I want more air, or I don't. If any board wants more, ratchet the fans up a bit. If none want more, jog the fans down.

We cut our EE labs and went sailing or something, and faked all the reports one night at the end of the semister. Of course we got As.

Time synchronization of programmable power supplies and loads is precisely one selling feature that my customers want but nobody else seems to make. It works fine in one box but I want to extend the function to multiple boxes in a rack.

The controller in each box is a MicroZed and doesn't support the PTP thing, and my customers may not be able top provide it anyhow. The 1 PPS thing works with just a BNC cable.

Besides, what I do is design things.

Yup. We do the Class H supplies for our TEC driver boards like that--if the linear amp rails, immediately jack up the supply by 0.5 V or so, then gradually ramp it down again. We could use two ADC channels, of course, but one comparator is simpler and works very well.

Cheers

Phil Hobbs

If there's a way to get by letting each box synchronize to GPS on its own and then using some simple network protocol to sequence them that's what I'd do, yeah.

Trying to get their system clocks to sync up over a cable makes me uncomfortable, why do they need to share info about their inner lives.

The Group Execute Trigger command does allow quite tight synchronisation between different GPIB devices.

John

So if it works fine in one box why can't you just send some simple packet data that says like OK box 4, run program 2 now for 1,084 ms.

If they're all already locked individually to GPS or whatever other standard they know how long a ms is. There will be some overhead latency but syncing a bunch of boxes within a ms doesn't seem unreasonable.

It's at least easier to ballpark how well a digital scheme like that would scale on paper. The BNC scheme is analog, how do you know.

And to think you went to art school.

Cheers

Phil Hobbs

A TEC driver doesn't mind a little lag on its supply rail being ratcheted up. My boards, as they heat up, also don't mind a little lag in the air flow being modulated. The fan control slew time can be in the ballpark of the thermal time constants.

Each board will have one or more temp sensors, and a resident cal table in eeprom that has the temperature targets. Simple.

Mathematicians often like music. In my experience, music fandom is negatively correlated to engineering design skill. Different brain structure or something.

One other thing I see a lot is undue respect for standards. As in "you can't do that because it violates SCPI standards." Where are the SCPI Police when you need them?

Am 21.07.22 um 16:15 schrieb Phil Hobbs:

-v ?

what do you mean with closure phase? Where do the 22 dB come from?

The idea was simply to have all 16 regulated to the be synchronous and then feed them into a 16-to--1 Wilkinson combiner. The phase noise should average out among the

16 units. Just as proof of concept. The MTI-260 are quite ok, but with bleeding edge oscillators that could be interesting. In the region where you just cannot improve an oscillator.

Gerhard

The BNC would be 1 PPS pulses. They could come from GPS if it's available, or one of the boxes could output the 1 PPS and the others would sync to that.

When we designed the box, we added two BNCs, open drain drivers with weak pullups and schmitt receivers, connected into our FPGA. We didn't know what they were for, but they turn out to be handy.

One is START RUNNING YOUR SEQUENCES NOW and the other is the 1 PPS lock. That should work.

Each module has a sequencer table in its local fpga RAM, sort of a primitive program with 5 opcodes. A table entry can write any other register in the FPGA, ie do anything, and one command is WAIT UNTIL, against the 1 MHz start counter.

Customers can write fairly simple SCPI script files to program events in each module, load them up, and fire off a shot and keep the whole experiment time synchronized forever.

All we want to do is revolutionize the power supply business.

It's actually useful to me to type and discuss this sort of thing. Ideas evolve at their own rates.

I see.

Gotcha

Sounds good

I'm just sayin there's also a standard (uh oh!) dating back to the 50s for synchronizing equipment using time-code:

up to 10,000 pulses per second. I understand now that there's no facility to program or sequence one box from the others they each run their own "script" from memory.

It would seem simpler to flip a switch to designate one box as the master and have the others watch a timecode it sends out at a higher resolution than the desired sync error. If the master then needs to itself be synced to real-world time via a GPSDO then ah...well I guess there should've been a 3rd BNC :-(

Sure. Thing is, that wastes a lot of information that you could maybe use to get 10*log(136) = 21.3 dB improvement instead of 10*log(17) =

12.3 dB. [136 = N(N-1)/2 when N = 17.]

Closure is a really cute idea, which I first came across in the context of very long baseline interferometry (VLBI) radio telescopes. See the discussion from BEOS 3e here:

. It's on P. 341 of BEOS 3e and P. 385 of BEOS 2e for those who have copies.

I don't have a scheme right handy, but it works at DC for measuring noise by the correlation method, where N meters get you 20 log N - 6 dB improvement.

It seems like it might be worth a bit of chasing, especially (as you say) in the regime where any improvement becomes very difficult to get. It would make an interesting paper if nobody's done it already.

Cheers

Phil Hobbs