Pretty respectable trace there for a 3Ghz bandwidth
Ok not being an RF guy, so shoot me
, would it not be better to have less capacitors with a larger value? Or is this a fundamental issue with this portion of the design?
Pretty respectable trace there for a 3Ghz bandwidth
Ok not being an RF guy, so shoot me , would it not be better to have less capacitors with a larger value? Or is this a fundamental issue with this portion fot he design?
The capacitors I used are each 10uF 50V X5R 1210. The idea of such a high value is to get a low frequency -3dB point.
12 x 10uF / 2 = 60uF
R= 25
F = 1/(2 x Pi x R x C) = 26.5 Hz
Regards,
Jay_Diddy_B
Hi,
Here are some TDR measurements made with a Tektronix 11801 and SD24 sampling head.
The Bias injector was place between two lengths of semi-rigid 50
cables:
From left to right
The 50 Ohm cable
the first 'wiggle' is the transition from the cable to the board, the SMA edge launch.
Then a short piece of 50 Ohm track.
two sections of 45 Ohms, about 25mm long, where the capacitors are
50 Ohm track
SMA connector
50 Ohm cable
I have marked the measured impedances on this drawing of the board assembly:
to raise the impedance where the capacitor are, I need to widen the gap between the traces and the ground plane.
Regards,
Jay_Diddy_B
In a very crude way that is the situation I was attempting to elude to
Great stuff Jay, really do not know where you find the time but top man
Nice demonstration of usefulness of TDR..
Great work (as always!!) Jay..
Regards,
Sinisa
I have marked the measured impedances on this drawing of the board assembly:
(Attachment Link)
to raise the impedance where the capacitor are, I need to widen the gap between the traces and the ground plane.
Damn! This is an optimization of my suggested change!!!
Respect!
(I had removed it because, afterwards, I thought it was something irrelevant... )
Hello All,
I am not sure if this is the best place to start a discussion about PicoTest Line Injectors. They are useful tools, but tend to be on the expensive side. I wanted to upload what I have to be archived, and for those who wish to design their own or at least are interested in such topics. I hope this helps someone, somewhere.
Picotest J2120A Line Injector
Price: $525
BOM:
(4) Capacitors
- C1||C2 = 35.7uF
- C3||C4 = 1.98uF
(1) Resistor
(1) nMOS
J2120A front and rear pics attached.
Picotest J2130A DC Bias Injector
Price: $525
BOM:
(24) Capacitors
- Parallel, Series network = 27uF
(1) Resistor
I do believe the equivalent network of the 24 caps is 27uF, I may wish to double check when I get a chance.
J2130A front and rear pics attached.
Ridley Injector Isolator Transformer
No specifications really. The manufacturer looks to be Avel Lingberg.
Model CA8518.
I hope this comes out clean, first time poster. Hell of a time with the pics.
*EDIT*
Added front picture of the J2120A.
Another nice one would be the "2180A" preamplifier to up the sensitivity of 8bit scopes. I wonder how they have done it in such a small package.
Could somebody explain what R1, R2, and C1 are being used for? R2/C1 look like some sort of high frequency roll-off, but at a surprisingly high frequency.
Thanks!
LTspice model J2111A
Could somebody explain what R1, R2, and C1 are being used for? R2/C1 look like some sort of high frequency roll-off, but at a surprisingly high frequency.
Thanks!
LTspice model J2111A
R1, R2 and C1 are needed to prevent HF oscillations. Its more or less an output damper.
Whats also interesting is that the circuit has a dead zone around 0V. Even if the current monitor output will be OK,
the sunk current is not. The manual did not show this until a few months ago.
https://electronicprojectsforfun.wordpress.com/measuring-a-picotest-j2111a-current-injector/
Thanks for the quick answer.
I'm simulating a simplified circuit (no diodes, just sink from positive rail, different op amp (OPA358)) to understand the circuit and trade-offs. However, I get peaking without C2, and with C2 I get a roll-off starting around 1MHz. I guess I need to pick a different op amp?
Yeah, the output network is a compromise between 3dB frequency and stability. It needs to be matched to the OpAmp, thats correct.
And dont forget to simulate with all kinds of DUT PSU output impedances !
This is what I have come up with so far; crossed out components are not part of the simulation.
Looks like R13/C1 is to compensate possible L1 wiring inductance going to the DUT.
I will simulate various DUT impedance configurations next.
I wonder whether I should add an additional op amp to optionally boost the Pr1 output gain. In general, I'm surprised by the signal level choices of the original design.
I will simulate various DUT impedance configurations next.
Doesn't look too bad - of course the wiring to the DUT has a huge impact at the high frequencies.
When you look at my webpage about it you see that the PicoTest risetime claims are only realistic if there is less than 2-3cm of wire length between output and DUT. This will not be possible in a lot of cases. IMHO, it is smarter to live with reasonable bandwidths (0-5MHz) and better stability. Most injectors are used to determine the response of switching regulators. Their regulator loop bandwidth (not the switching frequency) is normally far below 100kHz, so a few MHz of injector bandwidth is good enough.
I agree. This was mostly an exercise to understand the limits of this approach. It would be nice to observe most of the PDN performance at the supply vias of a chip. I don't have a good FRA/VNA anyways, so this probably will be good enough to be used with my RTB2004's K36 option.
Latest design attached (qucs-s). Maybe I'll get around to building it.
Testing PDNs is a tough test because of extremely low impedances here. Tests leads from your injector to the board are simply a no-no. Another challenge is that modern FPGAs draw 10s of amps at *very* low voltages. To make this move, you need to inject several Amps right at the DUT, asking for a very compact, solder-in solution for the injector and the same for the probe. This smells like copper-base PCBs, very modern switching MOSFETs in small cases, super-compact, all SMD designs, and so on. A funny challenge, but certainly not easy. An alternative for directly soldering the injector and the probes into the circuit is ultra-low inductance coax.
Dear All,
I was wondering if anybody made it to a next stage and actually made a real PCB of a current injector and did some measurements.
@Wolfgang, your remarks are pertinent and I wondered if you had the time and courage to improve the design?
Looking forward to continue reading!
Peter
Hi Peter,
I confess I did not do a lot in the field of power supply impedance measurement lately. The reasons are:
- finished my PhD in the field of precision oscillators and this consumed a lot of time
- For the PSUs I have, my injectors worked fine. I mostly work with low noise linear stuff.
- Making superfast, high amperage injectors is only possible with a bag of special tricks and components.
With my old eyes and shaky hands, I leave the development of these for the pros with appropriate manufacturing capabilities.
- If you have a specific question just ask. I try to answer as good is I know.
Best regards
Wolfgang
Thank you all for the information!
Thank you so much for the teardown information! Did you get pictures of other picotest modules?
Unfortunately no. I just have the injector. But I can only recommend to look inside their modules.
They are always good for some surprises