|
|
Post by gjaky on Nov 4, 2015 12:27:46 GMT -5
Again one step closer.  |
|
Deleted
Deleted Member
Posts: 0
|
Post by Deleted on Nov 4, 2015 15:49:13 GMT -5
Wow! Crazy man. Looking very good. Close to testing how it works I guess.
|
|
|
|
Post by gjaky on Nov 4, 2015 16:47:15 GMT -5
Unfortunately I'll have very little spare time in the foregoing weeks, but I hope I can build the power supply section at least...
|
|
|
|
Post by gjaky on Nov 7, 2015 16:21:18 GMT -5
I built the first (prototype) board from the higher bandwidth version. The power supplies needed quite a bit of work to makr them stable, they oscillated even without load... In the end I could make them work just as expected. Then added the AD834 + BFT95 part of the circuit (without output opamp) that part of the circuit worked right away. After that added the output opamp (THS3201), at first it seemed to work, but with time it tend to oscillate, switched the opamp to CLC449, it looked more stable, but still was some instability, then changed the BFT95s to the slower 2SA1765s ("only" 1GHz) transistors, and the oscillation went away. So with AD834+2SA1765+CLC449 I got a rise time about 2ns on the scope screen (the generator was a 300MHz ECL pulse generator). If take in account the limited bandwidth of the generator, the oscilloscope and even the FET probe, this would mean more than 400MHz analog bandwidth of this circuit at 2Vpp output! Actualy this is about the same to what I measured on the old M9000 VIM, which employing only the 2N3906 transistors instead. I haven't tested the 2SA1765 + THS3201 combo, maybe even that could stay stable, and even could get a little more bandwidth, but frankly this is already over the top. Also I am amazed that this PCB I designed handles the high frequency this nice.  |
|
|
|
Post by Casethecorvetteman on Nov 7, 2015 17:01:48 GMT -5
Good stuff  |
|
|
|
Post by stridsvognen on Nov 7, 2015 18:08:39 GMT -5
Now we want screenshots.. Nice work. |
|
|
|
Post by gjaky on Nov 8, 2015 2:59:45 GMT -5
Now we want screenshots.. Nice work. Screenshots are still a dream. So far I made my little board work on the bench. Next step would be to mate this with the VNB, and then the modified VNB with the NEC. This later step is risky because if something goes wrong the tube my get spot killed. So I'll install it in a burnt tube PG xtra (that's what I have) but the tube is the same and the control signals are also similar. If the PG xtra (and the new neckboard) survives the test then we can talk about screenshots  |
|
|
|
Post by Casethecorvetteman on Nov 8, 2015 3:15:47 GMT -5
Well hop to it Gábor!! |
|
nashou
Unmoderated Off Topic
Tech in Training.....
Posts: 1,239 
|
Post by nashou on Nov 9, 2015 22:39:39 GMT -5
I built the first (prototype) board from the higher bandwidth version. The power supplies needed quite a bit of work to makr them stable, they oscillated even without load... In the end I could make them work just as expected. Then added the AD834 + BFT95 part of the circuit (without output opamp) that part of the circuit worked right away.
After that added the output opamp (THS3201), at first it seemed to work, but with time it tend to oscillate, switched the opamp to CLC449, it looked more stable, but still was some instability, then changed the BFT95s to the slower 2SA1765s ("only" 1GHz) transistors, and the oscillation went away. So with AD834+2SA1765+CLC449 I got a rise time about 2ns on the scope screen (the generator was a 300MHz ECL pulse generator). If take in account the limited bandwidth of the generator, the oscilloscope and even the FET probe, this would mean more than 400MHz analog bandwidth of this circuit at 2Vpp output! Actualy this is about the same to what I measured on the old M9000 VIM, which employing only the 2N3906 transistors instead. I haven't tested the 2SA1765 + THS3201 combo, maybe even that could stay stable, and even could get a little more bandwidth, but frankly this is already over the top. Also I am amazed that this PCB I designed handles the high frequency this nice. I remember talking to Greg Eisemann about this transistor. He used the NXP BFT93 and said you need to add more caps and an inductor. Here is an email he sent me when I asked about some issues I had.
To be honest the best input change too is the
Wideband transistors
NXP
5-7 GHZ . I use them on sections of the marquee
and all over barco.
Hi,
Remember when adding a 5 GHZ you to take full
advantage of the BW for video you need to eliminate
the harmonics and noise.
You need to add extra CAPS on the power to the
transistors along with a bias resistor.
You also should add a Inductor to reduce the harmonics
of RF that will leak into the 5 gHZ..
I.E the signals of your cell phone for example will
leak into the circuit at 5 GHZ and one harmonic 2
order or 3 order could be the right frequency at even
200MHZ to show up on the picture. So if you have a
quality Inductor small and put in the signal path
before the 5 gHZ then you will reduce harmonics and
allow the full BW and a much better quality picture.
At that BW you need to have BW isolation so only the
larger signals will get though.
He also used the LMH6702 op amps and for those he said . Did you remove the peaking caps on the 6702 the old chips were peaked and you have to remove the peaking caps.This will remove the oscillation.Just thought id add that info. Nashou |
|
|
|
Post by gjaky on Nov 10, 2015 1:39:24 GMT -5
Thank you Nashou. In fact I stumbled upon an application note on the AD834, where the circuit used in the 03p VIM is proposed, and its theory of operation is detailed. The use of 2N3906 transistor is coming from there, there is also an explanation why. For now I want this circuit to work stable, and ditto for the VNB, once I see the construction works, and worth the effort, then I might look into further performance improvement. I also saw on the scope screen that the the current setup on my add-in board a little undercompensated, so there is still a little more bandwidth in it, but even now it works better than expected. Let's see if it works in the NEC first Attachments:AD834_AN-212.pdf (590.36 KB)
|
|
|
|
Post by gjaky on Dec 1, 2015 12:33:22 GMT -5
Small update. At my work I made new friends at the RF division, so in the lunch break we measured my circuit's transfer function with an Agilent E8358 network analyzer. this stuff can measure the frequeny response from 300kHz to 9GHz., clearly not a bottle neck. The result is interesting. The circuit crosses -3dB point at ~380MHz, then coming back for a peak at 500MHz, then it falls eventually. It does not seem as smooth as I expected, however from 10MHz (20MHz pixel clock:~480P) to 300MHz (600MHz pixel clock!) it stays within 2dB which is in fact quite nice result in an electro engineer's "book".  |
|
Deleted
Deleted Member
Posts: 0
|
Post by Deleted on Dec 1, 2015 14:05:03 GMT -5
Do you have a chance to compare this with standard unmodified boards from Marquee or Nec or Barco?
|
|
|
|
Post by gjaky on Dec 1, 2015 14:28:49 GMT -5
Do you have a chance to compare this with standard unmodified boards from Marquee or Nec or Barco? Not really, not directly. I don't want live up the RF guy's goodwill, that Agilent stuff is crazy expensive... Otherwise my little board has a one function only, this will be part of a system. I stayed with almost the exact circuit that is in the 03p VIM, only the couple transistor is altered, but I don't think it make a big difference after all. But again, the 03p VIM altogether is a far more complex circuit than this. High MHz numbers may sound good here but there are other contsrains in the circuits always which are holding back the performance here or there. |
|
Deleted
Deleted Member
Posts: 0
|
Post by Deleted on Dec 1, 2015 14:49:53 GMT -5
Yes it might even be that a component has higher bandwidth but because the response is not linear this could result in lower bandwidth at the 1 on 1 off pixel pattern? If you happen to shift a bumb on that frequency the test pattern will look best Am I right? Complicated. |
|
Deleted
Deleted Member
Posts: 0
|
Post by Deleted on Dec 1, 2015 14:56:41 GMT -5
One other question can you somehow relate the 2ns measurement to this picture? Is that close to the 380MHz? I am also very curious if this shows in the picture as some kind of peaking.
|
|
