LMH6703: "step" in rise before the plateau - why and how to solve?

Well Pinhas, that is a pretty fast edge - and yes the LH6703 does have a very high slew rate, but I wonder if the step is where it is running out of slewing on the rising edge. Any chance you could slow the input edges a bit and see if that goes away?

Hi Pinhas,

Pinhas Rosenfelder said:

I am seeing a "step" before reaching maximum voltage.

And I see a ringing

Such a ringing is usually taking place at both edges, the positive and the negative. So, even if there is a step, there's also a ringing superimposed on the step very probably.

In this frequency range nearly everything can be responsible for the ringing. Even the way the coax cables are laid on the rable.

I have some tips:

1. In this frequency range use only highest qualit semi-rigid coax cables. Good semi-rigid coax cables can be very expensive unfortunately. I guess you have them already.

2. Exactly follow the layout example shown in the datasheet.

3. The supply voltage decoupling is critical in this fequency range. If you have two uneven decoupling caps paralleled, the bigger of them should have some ESR like it can be seen with tantals. e.g. Otherwise a nasty resonance can occur which can directly be seen superimposed to the output signal at frequencies where the PSRR has almost decreased to zero. See figure 18 of datasheet.

One remedy is to replace the parallel circuit by only one 2.2µF/X7R ceramic high cap in 0603 package. Another remedy is to mount a small resistor in series to the bigger cap (if it is a ceramic high cap) simulating the ESR of a tantal. Something between 0.22...1R should work. Or use a tantal for the bigger cap.

A resonance can also occur at the resonance frequency of LC supply voltage filter, if the ferrite bead and/or the decoupling caps have too liitle ohmic resistance for the dampening. In such a situation it can help to add some series resistance of 1...4.7R in series to the ferrite bead.

Additionally, the LC filters could be improved by modifying them to Pi-filters. Just mount a decoupling cap from the input of each ferrite bead to signal ground. Each OPAmp should have its own Pi-filters.

4. Experminent with all the component values of resistors. 

Kai

I could try to add capacitance after the previous stage to slow the edges a bit - but I DO NEED the ~1ns rise at the output - were you thinking that a 400-600ps rise at input would be better?
why would that be?

Kai,

Thank you for all the tips.

Indeed I see a ring too - I have yet to determine if it is "real" or an artifact of the "test point" output to the scope.  but it is the second order problem right now.

Of course - maybe the step is also from that - but I am afraid the step is real.

It seems to me all the issues you describe refer to sources of ringing - but could any also cause that "step"?

regarding the large caps - my instinct also would have been tantalum for the higher ESR - but TI recommends ceramic, as well as used them in practice on the TI eval board? (6.6uF ceramic - but I doubt the 4.7uF I used is the problem?)

I did replace the ferrites with 3ohm resistor - no real significant change.

my layout is very similar to the eval board, except i have two stage close together.

the output is to a good quality test cable. 50 ohm in series to the cable to the scope.  there is an effective stub on the real line (no via the 50 ohm) - with a few mm, then a 1uF series and another 2mm and a 1M load and some series inductance.  All things which might introduce a bit of ring too I suppose - but the "step"?

wondering about some series resistance right at output - but on the eval board I used 2-3ohms and all was fine

Ideas?

thanks

Hello Pinhas,

I have been quite busy this week; I will take a look and see if I have any ideas also.  I appreciate your patience and curiosity.

Best,

Alec

Hi Pinhas,

Pinhas Rosenfelder said:

regarding the large caps - my instinct also would have been tantalum for the higher ESR - but TI recommends ceramic, as well as used them in practice on the TI eval board? (6.6uF ceramic - but I doubt the 4.7uF I used is the problem?)

You way want to read this thread:

https://e2e.ti.com/support/amplifiers-group/amplifiers/f/amplifiers-forum/1129355/opa2810-bulk-capacitance-guidance-for-sharing-amongst-amplifiers

Kai

Hi Alec, Kai, and Michael,

1 - The step was still there - quite clearly.  when I extend the pulse it shows little evidence of ringing, maybe small ringing.

2 - primary problem to me now is the STEP - it seems to "run out of steam" on the slew at about 4-4.5V..ANY SUGGESTIONS?

do I need to add a small cap on the supply filtering?   might i need more bulk capacitance?  

BTW I put 10 ohms instead of the ferrites on the 9V supply..make sense?

gain now 6 on second stage (82 and 420 ohms approx).

divider beetween stages 33 and 100 (70%)

I also added 20 ohms series on the output instead of 3.3 ohms, and looking with scope via a 950 ohm series resistor and 1uf (100nf?) capacitor in series instead of 50 ohm - lighter load?

the part driving it burnt out while probing before I could take a photo of the scope 

WOuld appreciate advice

thanks

P

Hi - seems this never sent - and just sent now...in the mean time I have had some progress.

1 - the step seemed caused primarily by some loading or reflection from another line capacitively coupled and diode clamped - I swapped in a high impedance ferrite instead of a small inductor and the step is gone..

2 - I still have a problem of a "knee" where the slew slows for the last 500mV-1V of the ~6V slew. the first 5 volts take sub 1ns, and the last volt or less takes another 1ns (I am slightly into saturation, I think.).  This doesn't happen in the simulation.

Any idea how I can get a faster edge all the way to saturation level plateau?

gain now 6 on second stage (68 and 340 ohms approx).

divider between stages 40 and 100 (70%)

I also added 10 ohms (though no better or worse than 20 - just less voltage drop)  series on the output instead of 3.3 ohms, and looking with scope via a 450 ohm series resistor and 1uf (100nf?) capacitor in series instead of 50 ohm - lighter load?

attaching a photo of scope plot - the output is via 453 ohms so scaled 1:10

the faster signal is the inverse of the input signal

Hello Pinhas,

It could be either slew rate or low frequency distortion, I may need to run some simulations to see which it could be.  

Best,

Alec