Thus seems I may get some minor money to renovate a laboratory devices in shelf.
So the first thing where I need it harshly is osciloscope for ZVT-H bridges circuit faults control.
Very often I have a problem where say at 300V it works ideally, but as soon I lift up the Vcc to standard 530V then 1200V igbt are damaged. Most surprizing, it is damaged at small power conditions without even beating out the network 5-Ampere fuses, but those damaged devices are at least 320 Amp bricks. Never was able to understand why it happens.
Thus, I need to cach a needles, say in 5 minute interval, by monitoring four voltages and 4 currents in upper and lower right and left shoulders, and probably the load current and voltage as well, and if Vcc then ideal, or other combination 4 voltages 4 currents and 4 gate voltages - again 12 channels.
So, the 12 channels are mandatory, all channels mandatory must be optoisolated one from other at least with 1 kV or better 2 kV. The needle proposed time-scale is ca 10 nanosecs. Thus the BW must be near to 1 GHz or welcome if over. The memory must be enough to record those 5 minutes and replay it later, may be used or stand-alone-DSO or USB-DSO.
So the question for SMPS guru - what a model of DSO I ought to choose for 10-50 kEur budget?? May You name at least one candidate?

Bit explored what are boarders of state-of-the-art components. Practically for 5 USD are more than one optoisolator choices at scale 5-10 nsec, ADC choices are much weaker, 8-12 bits for 10-20 ns costing near a thousand per piece, but stays written it may blow a 400 MHz output via SPI (however it seems a bit too tight figure) to the computer RAM or HDD, as such data massive may need near a Terabait. All this means that component basis to fulfill my demands principally are available. However such device I had no success to find nowhere. What devices YOU are using to solve a similar task on H-bridge??

 

OK, today I found an oscillo capable to work with 16 channels uin demanded freq range and it cost about 30K or in used sector 19K. But for galvanic decoupling exists a term fiber optical isolated probe. However the price 49 000 x 16 pieces seems astronomic even if choose a most slowest and cheapest LeCroy HVFO-103 (4500 Eur per piece). It means we must erect a special surplus project for elaborating the good FOIP. At hackaday exists one working up to 30 MHz what seems bit too slow in nowadays when some opto transducers may work over 3,8 GHz as well their specific genuine opamps. So, now about ideology how to organize it.
First that seems we ought to use the double fibre for each channel in differential maner, to eliminate the some potential instability of current-light characteristics.As two channels always are having common node (i.e. curent and voltage) may use the 4-fiber line transducers.
OK, time press me go off, will tell later.

 

OK, Im back.

One of oscillo choices is DPO70000SX (at de.tek.com) and other even better is Keysight.com having Infiniium V-series
as well (especially) the 86100 series, with memory deepth depending on external HDD used for it. So, 1 TBit is no a problem for it.

However I didnt clearly recognized do the 90 GHz able optical inputs are capable for analog signals or have only logical 1 and logical 0 states. Anyway, the analog part is over the 3 GHz so it will serve for long time while the everyday`s semiconductors will overshoot this.

Overall ideology for the fiber optically isolated probes are well stated at https://hackaday.io/project/12231-fiber-optic-isolated-voltage-probe yet their success is so sad as 30 MHz (however for many applications even this is rather good).

The opto-converter capabilities ends at bold end to about 400 Gbps, but the four-digit price. So, wiser would be to go below the 25 GBps where user-friendly pricing begun available, as an example 11 Mbps at 87 USD or 10 Gbps even for 19 USD. So, name the Silicon Labs with their SI5040, but still the further appreciation of the best choice is needed. For example, 3,8 GHz works Finisar FTLF8524 for 23 Eur as well the Avago/Broadcom AFBR series and HFBR series and it have a "normal" emballage to differ from "nude" case-less SMD wonders. I am little bit thinking about computer dedicated PC networking opto-input dongles like TL-SM311LM or newer, but I am afraid that to convert this logic nil/one dedicated circuit to linear analogue mode will be no so simple task as I would will.

So, lets take for exploration the AFBR-5710Z (400 picosecond fronts). In that datasheet is reccomended a specific driver circuit HDMP1687 (sic! last not able to work with analogous signalling). Sure must exist a better alternatives looking like ordinary opamp input with ECL differential output, I have seen it time ago. Aah, it was LMH6552 with 1,5GHz BW-product and LMH6555 with 1,2G. Even better are 2 GHz opa 691, 2691, 3691 as well ada4927 but outstanding in the 2 GHz group is the THS3202 (monopolar output) by slew rate 9V/nsec). However the current ability is only 115 mA for goal of 300 mA and I look for it uncurable. Actually tere is another candidate with 200 mA but 4,5 V/nsec the THS4511 with full differential output. First is giving settling to 0,1% the 19 ns and second 16 however the rise/fall for first is 0,45 ns.

So, summing the said, this moment seems the all fast opti links are dedicated digital only without of possibility to let the laser works analog way. Thus the fastest are HFBR 1505 digital version capable (maybe) work analog (10/13 nsec) and HFBR1528 (sure capable analog 10/13 ns) and AFBR1531 (13/7 nsec). The input cascade sure THS3202 and output cascade the same. Sad I cannot see any faster analog-able optolink.
P.S. wait wait, what about GFD3500 (6/8 nsec), only there the recomended HFE7000 is not valid and must be substituted seems with HFD3023 (6+6 nsec) or better the nude diode HFD3033 (1,2/1,2 nsec) then be married with THS - only the fault is the obsolete without of leaving ANY honorable successors (recomended one is 10-fold slower). Or, as alternative, use as receiver the identical led as transmitter, just two afbr1531 looks one inside each other via those fiber.
P.S. Here are two interesting opamp candidates LMH5401 (*GHz 17 V/ns) and LMH6401 (4,5 GHz 18,2 V/ns).

Next week I`ll be in the Space exhibition at Bremen, probably I shall gather some now-know-hows about better component labels. I shall tell then here.