Good progress so far. I think you are right, design a Pcb where you can control the layout and parasitics better, maybe even a 0.8mm Pcb to keep tighter what needs to be tight. Might need to lose the headers also and solder direct, just keep everything on the topside and a large 0Vdc groundplane on the bottom.
@@IanScottJohnston Well, I understand that it is tempting to be able to tweak without a socket, but at the same time the chip has heatsinking on the bottom, so the heat sink should be on the other side
A large groundplane below the inputs increases the capacitance the feedback network will see. This low pass effect in the feedback loop could lead to instability, so you have to be careful with that.
Good evening. I definitely agree with others: to obtain top level performnce you have to design an "RF board" for your current feedback opamp. Use of smd components to reduce parasitics is of paramount importance. Minimize trace length is equal important, of course. Good luck and regards from Italy.
very good choice. I had done just that recently using two THS3091 in parallel because I saw that keysight uses two THS3091 in their modern 3362xA (120MHz) arb generators. So I made a similar daughter board like yours (slightly bigger) with two THS3091 on it and heatsink replaced that hybrid in 33250A. I just left that BW-SW out and the unit was able to calibrate and worked 🙂 Instability is because of the messy vero board assembly. YOu have to design a proper layout and make the board according to TI recommendation. This is very critical in high BW and high power current feedback opamps
also check the square /pulse waveforms. I could not quite achieve the overshoot specs but the rise time was ok. you should get even better rise time I guess but the overshoot and ringing I am not sure. you have to check that too
@@feedback-loop no I left it out as you did. I dont think OP37 was only used for DC correction, i think the original hybrid had some freq response issues at below 1MHz (plus the varying dc offset issues) and they somehow compensated those problems with Op37 circuit which by the way has a positive feedback and I still cannot figure out why it has positive feedback. I didnt check if the actual circuit is negative feedback and the schematic has an error or not. However, with the modern ICs like what you used I believe Op37 circuit is not necessary. The only thing that maybe makes a dc offset compensation necessary can probably be temperature variations. although the manual does not state that as one of the reasons for U1903
Awesome work and it's always a pleasure to watch you, the only thing I could add to the discussion is to design a PCB. This would be fascinating whatever the outcome and I belive you have a great chance at improving things further....cheers.
The breakout board pins and sockets present as a capacitive load, causing instability. Try adding a few ohms (2 or 3) between the output of the opamp the the breakout pin. It'll help stability. Minor impact on "50 ohm" matching. It's pretty amazing you got a 900MHz current feedback opamp working without massive oscillations on a thru hole breakout board, with thru hole resistors... Very nice.
If a few ohms doesn't help enough, you can actually add much more, 10ohms maybe, and just increase the gain Rf to account for the voltage loss into the 100ohm total load. It'll greatly increase stability.
Building this on perf board is just asking for trouble. Good grounding on a solid plane piece of pcb, building dead bug and everyone knows to put bypass caps mounted on the chip for Vcc but not many know to put bypass caps on the ground pins on the chip and wire to Vcc. You want the lowest inductance grounds possible ie short and use small package smd caps.
Did you test for performance with other waveforms? I'm curious if a 80MHz square wave with all its extra harmonics would cause the oscillations even if the 80MHz sine wave doesn't.
At 80MHz square wave looks practically the same as sine wave. And that includes the unmodified instrument. So, yes, I tried, and the result is the same.
The Cadeka(Exar) KH563 output amplifier used in Tabor WW2572A AWG and it claims 150MHz bandwidth for 24dB gain. Maybe it is a good replacement for THS3491.
Trying to workout a mystery box from the outside, ie, not knowing what's inside, is a pita. You've probably already done it, but I guess if you have a working one you can probe it's pins.
you should measure phase margin and gain margin , you can not just put one opamp without stabilsation criterium , test with square wave you will get nice oscillation🤣🤣
![ปฏิเสธไม่ไหว (Crush on you) - LIPTA feat. No One Else [Official MV]](http://i.ytimg.com/vi/NBw1jF342vw/mqdefault.jpg)
So, basically, you created an asic from scratch. You are amazing.
A beautiful investigation, adaptation, journey and result.
Good progress so far. I think you are right, design a Pcb where you can control the layout and parasitics better, maybe even a 0.8mm Pcb to keep tighter what needs to be tight. Might need to lose the headers also and solder direct, just keep everything on the topside and a large 0Vdc groundplane on the bottom.
Why not the other way around? Ground plane on the top side with a heat sink on it. The chip does get warm.
@@feedback-loop just thinking you can modify if all components topside. Removable heatsink on top of that.
@@IanScottJohnston Well, I understand that it is tempting to be able to tweak without a socket, but at the same time the chip has heatsinking on the bottom, so the heat sink should be on the other side
A large groundplane below the inputs increases the capacitance the feedback network will see. This low pass effect in the feedback loop could lead to instability, so you have to be careful with that.
This is quite a bit more complex than the usual projects. This is really quite the learning experience.
Good evening. I definitely agree with others: to obtain top level performnce you have to design an "RF board" for your current feedback opamp. Use of smd components to reduce parasitics is of paramount importance. Minimize trace length is equal important, of course. Good luck and regards from Italy.
very good choice. I had done just that recently using two THS3091 in parallel because I saw that keysight uses two THS3091 in their modern 3362xA (120MHz) arb generators. So I made a similar daughter board like yours (slightly bigger) with two THS3091 on it and heatsink replaced that hybrid in 33250A. I just left that BW-SW out and the unit was able to calibrate and worked 🙂 Instability is because of the messy vero board assembly. YOu have to design a proper layout and make the board according to TI recommendation. This is very critical in high BW and high power current feedback opamps
also check the square /pulse waveforms. I could not quite achieve the overshoot specs but the rise time was ok. you should get even better rise time I guess but the overshoot and ringing I am not sure. you have to check that too
Interesting. How did you do load sharing? With 0 ohm resistors? :)
Did you use DC correction (did you connect pin 6)?
@@feedback-loop no I left it out as you did. I dont think OP37 was only used for DC correction, i think the original hybrid had some freq response issues at below 1MHz (plus the varying dc offset issues) and they somehow compensated those problems with Op37 circuit which by the way has a positive feedback and I still cannot figure out why it has positive feedback. I didnt check if the actual circuit is negative feedback and the schematic has an error or not. However, with the modern ICs like what you used I believe Op37 circuit is not necessary. The only thing that maybe makes a dc offset compensation necessary can probably be temperature variations. although the manual does not state that as one of the reasons for U1903
Awesome work and it's always a pleasure to watch you, the only thing I could add to the discussion is to design a PCB. This would be fascinating whatever the outcome and I belive you have a great chance at improving things further....cheers.
The breakout board pins and sockets present as a capacitive load, causing instability.
Try adding a few ohms (2 or 3) between the output of the opamp the the breakout pin.
It'll help stability.
Minor impact on "50 ohm" matching.
It's pretty amazing you got a 900MHz current feedback opamp working without massive oscillations on a thru hole breakout board, with thru hole resistors...
Very nice.
Pages 28/29 in the datasheet.
If a few ohms doesn't help enough, you can actually add much more, 10ohms maybe, and just increase the gain Rf to account for the voltage loss into the 100ohm total load.
It'll greatly increase stability.
Very interesting. One of your best videos.
Nice modification, thanks for sharing. 🔌💡
Great video
Building this on perf board is just asking for trouble. Good grounding on a solid plane piece of pcb, building dead bug and everyone knows to put bypass caps mounted on the chip for Vcc but not many know to put bypass caps on the ground pins on the chip and wire to Vcc. You want the lowest inductance grounds possible ie short and use small package smd caps.
Doing an FFT analysis on a square wave might be worth doing too, to see how it handles the rise and fall times.
Very nice repair progress, the instability could be related to the power supply voltage/current, have you tried altering those to see if it helps ?
Job well done😊
Did you test for performance with other waveforms? I'm curious if a 80MHz square wave with all its extra harmonics would cause the oscillations even if the 80MHz sine wave doesn't.
At 80MHz square wave looks practically the same as sine wave. And that includes the unmodified instrument. So, yes, I tried, and the result is the same.
Nice job.
The Cadeka(Exar) KH563 output amplifier used in Tabor WW2572A AWG and it claims 150MHz bandwidth for 24dB gain. Maybe it is a good replacement for THS3491.
where is it available?
@@feedback-loop I bought two many years ago in ebay. I can send one left to you if you find nowhere.
@@vincentjudyjaneric Well, thanks, but I would like to come up with a solution that is reproducible.
Trying to workout a mystery box from the outside, ie, not knowing what's inside, is a pita. You've probably already done it, but I guess if you have a working one you can probe it's pins.
I have a question.
Are you a God?
Did you try to take heatsink off the original hybrid? Maybe they have some commercial opamp in the hybrid that could be repaired?
No, I didn't. Very much doubt it is repairable.
@@feedback-loop Would be still educational to see insides :)
you should measure phase margin and gain margin , you can not just put one opamp without stabilsation criterium , test with square wave you will get nice oscillation🤣🤣
I am trying to stand on the shoulders of giants from TI :)
Hm, you have a Russian accent. Are you from Russia?=)
yes
Great video.