1. Digiface USB is not a Midi-interface, you would have to use a separate Midi-interface.
2. Yes, fine.
3. Yes control of interface, no Midi-interface
4. The demand for 96kHz is what makes your budget sick. 32 channels at 48 kHz is no problem for 1320 Euro.

Hmmmm well either i have to up my budget then, or go for 48 kHz.

Does having a seperate midi interface, introduce jitter when tracking? Some people on fora say that it is important to use the audio interface's internal clock to send out midi, but since there's a lot of contradicting opinions out there, I can't really form my own opinion on the matter...

Seems like a good option! I'm looking into it. Thanks for the advice:)

The TB Sonnet expansions box is a box where you fit in the RayDat card and the box connect to your MacBookPro.

A very clever solution for your budget.

Otherwise I would overthink the 96kHz, for human ears 44,1kHz sampling rate was discovered to be perfect.

Seems like a very good solution yes! But then I have another question, in the case of 44,1 kHz vs. 96kHz (or 48 kHz for that matter), what would I gain using the Ferrofish instead of let's say an ADA8200 or OctroPre? I was contemplating the Ferrofish because of its ability to do 16 ins/outs in 96 kHz, but is it so much better running at 44,1/48 vs. the cheaper options?

Yeah, I agree. It seems like a very interesting set-up for my needs!

I wouldn't worry here.

Thunderbolt is an established solution for providing (not only Apple) computers with additional interfaces and PCIe cards that would otherwise have no place on or in the device. Be it graphics cards, network cards (e.g., 10 Gbit) or serial interfaces to put some examples. This only works so well because it is effectively an extension of the PCIe bus to the outside.

To put an example: Thunderbolt 3 comprises four data lanes. It is PCIe 3.0 with 4 lanes.

It's important to note that various protocols are multiplexed over these different data lanes, enabling a versatile and high-performance connection.

PCIe is a serial protocol. Whether this happens inside your computer or over an up to 2m long cable makes no difference.
Remember "external PCIe" ...

PCIe is the evolution of PCI and has several advantages:
- allows higher data rates and thus bandwidth
- better scalability (it is easier to increase the bandwidth of only one line compared to multiple, see the legacy PCI bus)
- cost-efficient (fewer physical lines and electronic components than parallel transmission)
- energy-efficient (requiring less power for signal transmission, which is important for battery-operated devices like Laptops)

To sum up:

I cannot determine which of the two solutions is better. The RME manual lacks information on MIDI jitter, and I am unsure if Expert Sleeper provides any values for it. However, both the MIDI implementation from RME and the drivers have an excellent reputation.

If you believe, have heard, or read that the MIDI implementation on the RayDAT meets the desired quality, you can assume that this quality remains consistent, regardless of whether you install the RayDAT card in a computer's PCIe socket or a PCIe expansion housing from Sonnet or other providers connected via Thunderbolt.

I would prefer Sonnet for an external case here; their product range gives me a very good and solid impression.

Regarding my proposal

I would still prefer the RayDAT card to the Digiface USB because it provides you with additional I/O ports that you could use later.
If you should need to connect a preamp it could offer you even WC (word clock) when getting the optional WC module additionally, to be able to clock synch a connected preamp.

In my opinion, the RayDAT card is a safe bet here. You simply have a bit of room for improvement, more flexibility.

The Digiface USB, as nice as it is, is quickly fully utilized in terms of ports and features. It doesn't offer many options apart from ADAT and a headphone output, which you could easily use in a studio or on stage.

A small Sonnet expansion housing, which ideally also offers a little space for the daughter card and an optional WC card (i.e., at least 3 PCIe socket widths), is also easy to transport.

Maybe other forum members can share their experiences here

Perhaps some other forum members, who use MIDI more frequently than me, can contribute to this topic.
I recommend excluding Thunderbolt out of this discussion, it has the same "quality" like PCIe.

The question is, whether the RayDAT MIDI quality is simply fine enough to drive external devices with MIDI input like a keyboard or synthesizer without having to worry about jitter.

This seems to be the only requirement here, right? Is this correctly understood?

Maybe one additional thought about the MIDI cabling.
Regarding 16x analog I/O (Pulse 16) I assume you want to connect at minimum 8 devices with stereo outputs or even more if some devices are mono.
This means also, that every device needs a MIDI through ports, so that you can build a serial chain from the 1st to the last device.

I am not so deep in MIDI .. but I would question myself
- whether all devices need to be connected and how much devices this is
- whether all devices have a MIDI Through connector
- whether such a needed "MIDI through" plug has a little latency

So it might be an advantage to have a RayDAT offering 2x MIDI I/O so that you could distribute the 8 or more devices to the two different MIDI ports of a RayDAT instead of having only 1 MIDI Port.

There is another solution possible, but around €3000.
One of the smaller Sonnet Cases with 2 slots for around €350  and the HDSPe MADI card from RME.
This would have the advantage that you can save the 4-8 TOSLINK cables (depending whether you want to use D/A conversion on the Pulse 16M. You would only have one MADI cable (OM3 or OM4 multimode fiber cable).
The HDSPe has only one daughter card and 2x MIDI and Word Clock are on board.
MADI would offer the possibility for further expansions (64ch at single, 32ch at double speed).
But then you also need the little more expensive Pulse 16MX.

The cost is definitively higher (out of budget) but not really excessive if you compare with the possibilities that it might offer to you for the future, just in case you would need more ports or a  preamp.
With such a solution you could even add a 2nd Pulse 16MX and even record in double speed.

Pulse 16 MX ~ €1249
HDSPe MADI ~ €1399
Sonnet ~ €350
Total ~€2998

Was thinking about that too! Saw it is discontinued already. Such a bummer, because I could definitely use the CV outs to send CV to my Eurorack modular. Gonna look for a used one indeed!

Regarding the Raydat and Madi HDSPe solutions: It is solid advice, but I decided to go with the digiface route. The reason for that is that I will already have rock solid clock using this route. Users of the USAMO report back very solid, jitter free clock from it generally, so that will be probably a non-issue; even though I won't be using the superior Raydat or Madi HDSPe.

Expendability is indeed important, but I will be more than happy using only 16 inputs and outputs. It really is already a big step up from what I had, and probably better than any of my peers are using. I'm producing techno music and ambient, and usually conversion is something that is considered as an afterthought in this scene. People rather spend money on new sound generation devices or dynamics/coloring devices/eq's. I'm revamping my whole studio, and splurging hard and if I can save on something, this means I can use this money to buy another ''crucial'' piece of gear, like a nice compressor.

About clocking a possible future pre-amp, I reckon analog gear does not need to be clocked to the converter, right? I could also not really find evidence to support this claim. And if I would have to clock it, god forbid, I'll be able to use the word clock out of the Ferrofish. I will be buying a Midas Venice 160 or 240 or f/u-series to use for pre-amps and eq's and this does not have a clock in anyways.

About the subject of midi: I won't be daisy chaining my devices, but I'm already using a Kenton midi through box, for 5 devices. If needed I can buy one with more ports. For now it works fine! Daisy chaining indeed adds latency, therefore I stopped doing this a while back. Even with the MOTU M4, I have a pretty stable midi clock signal. I have very little latency as is. In my DAW, Ableton Live, it's possible to configure delay compensation for midi and audio latency, so it'll work.

For matters of stage performance, I will be mostly using my MOTU as a DAC for my macbook. The MOTU has actually a pretty good DAC and it will be more than fine for these instances. So using a Sonnet with a HDSPe or PCIe with a big 19 inch converter would be overkill!

Thanks for all the advice. It won't be lost on me. It's just that I have to make budget choices right now. If ever I need to expand, I'll just sell the digiface, upgrade the Ferrofish to MADI and go for one of the proposed solutions, but I'm thinking that will be many years from now

Allright! I get what you mean, that is using a pre-amp that acts as a converter and sends channels over ADAT, MADI or AES. In that regard I would definitely miss a WC if running at double speed!

But the way I was thinking about adding pre-amps, is adding pre-amp DI before hitting the Ferrofish. This is purely about an analog signal chain before hitting the Ferrofish converter. Like using e.g. a SPL Charisma 8 channel, which doesn't have and need a WC connection.

In my case, if I'm correct and read your article correctly, clock will be send over the ADAT stream, and because the Ferrofish has ADAT IN, it will receive clock over ADAT stream from the Digiface. Or I could alternatively slave the Digiface to the Ferrofish, but this is non-preferred.

The more I think about it ... start small with Digiface USB and make your experience.
Even the Pulse 16 allows to be upgraded to MADI later, check that, if this is still the case.
Getting the Pulse 16 with MADI and CV from used market would be an option.
Based under assumption double speed and if you need preamps, then connected analog through Pulse 16..

To be honest, I don't understand what you are trying to say or what new aspect you would like to introduce into the discussion once the picture has been clarified.
It's clear that you can use the Digiface USB as a clock master anyway and that's not the point at this stage.
In the thread, some considerations were made as to whether the setup should not be designed differently in order to cover all requirements (including double speed) but also possibly to be able to do without the Expert Sleepters device for MIDI or to maybe connect a preamp in the future through ADAT or better MADI, to have the needed amount of channels.
In the end it turned out that he would connect mic preamps analog via the Pulse 16. This meant that Word Clock would not necessarily be needed. This also made the idea of using a RayDAT a lower priority.
Current status is, that his initial idea of using the Digiface USB, triggering the Expert Sleepers via CV capable phones output is the best solution to not spend too much money.

Ah, this you mean.

Well, I wrote this in case that he might get a preamp without digital inputs (who knows),
because not all models have a digital input for clock synch purposes.

Then the only solution would be to make the preamp to act as clock master.

But this is not so comfortably from an operational perspective.
Better to set sample rate in the project and all connected slaves learn through the master.