THS4271: Distribution amplifier 10MHz with Dliyannis structure filter

You are driving the input C of those next stages, try changing  R39 to 20ohm, you could also probe across R27 to check the supply current draw. 

So I did run this in TINA with and without 8pF added as a load, normally - if that is an issue a sharp noise spike will appear - I am not seeing any change? 

Looking into the model (circa 2011) - this is very idealized output stage so maybe not simulating accurately, 

The THS4271 does pull a lot of quiescent current - about 28mA so the quiescent power is almost 300mW - probably will run a little hot

I took a look for a better alternate op amp, you might consider the LMH6609 simulated here - neither op amp gave a very good filter shape actually, Those RC values might could use some improving, 

Filterpro delivers this, kind of bad with those 1pF caps, but here I scaled everything by 22 for 22pF caps and divide the R's by 22, not hitting the response shape very well - maybe due to input parasitic C on the op amp model, 

Hi,

looks stable:

woj_ths4271.TSC

woj_ths4271_1.TSC

But what me worries is R39. It could interfere with C38. The isolation resistor R39 should not be zero. Read the section "Driving capacitive loads" of datasheet. And C38 should not be 100nF. Make it much smaller or add an additonal isolation resistor.

Kai

I might have been missing a connection in my sim? Kai's looks great, here is his file witht he LMH6609 dropped in. the higher curve to the right was the LMH6609 

Adding the cap load does not look bad, but again the THS4271 model looks way simplified

Dropping the LMH6609 into this, now a lower phase margin is showing up, 27deg. Yes, I think trying 20ohm for R39 might be useful. 

Thanks a lot for the advice. Tomorrow I will change the resistor R39. I forgot to mention that the SJ9 and SJ6 jumper are not short-circuited. All components behind the SJ9 and SJ6 jumper are not soldered. The filter circuit and all LMH6723 amplifiers with components are installed.

Hi Wojciech,

you could measure whether the signal is still ok at the output of THS4271 and gets corrupted only at the output of LMH6723. Use such a scope probe to avoid nasty resonances:

And, very important, never put the probe directly to the pins of an OPAmp. Always insert an isolation resistor or 50...100R or so which you solder directly to the pin of OPAmp to be measured. This is especially important when probing the output or the -input of an OPAmp.

This brings me to my next point. I don't see any isolation resistor or termination resistor at the output of IC7 and IC8. What load do you want to drive with these two OPAmps?

Another tipp is to avoid the use of high frequency voltage followers (gain = 1) in the signal chain, especially at the output or the circuit where difficult loads may have to be driven. Some HF OPAmps don't like that at all and are notoriously instable. It's better then to run them with a gain of 2 and to install a 2:1 voltage divider at their input.

Another tipp is to play a bit with the supply voltage decoupling. You could increase the decoupling capacitance by taking a 2µ2/10V/X7R/0603 ceramic high cap. But don't connect uneven decoupling caps in parallel as this can result in nasty resonances ruining the performance of supply voltage deoupling. A single 2µ2 cap (instead of the single 100nF cap) at each pin should be sufficient.

You could als vary the filter resistance from 2R to 20R and/or add a ferrite bead. I prefer the "FBMH1608HM601".

And, finally, the most important tipp is to use a solid ground plane, if you don't already use it

Kai

Thanks a lot for the tips. Instead of R39, I soldered a 10 Ohm resistor. Unfortunately, I don't see much improvement. I took some photos:

This is input signal on RF_IN:

This is filter out, on R39 resistor.

This is out of RF_DRIVE_A (out of IC7)

And this is out of IC4

And out of IC4 when i put finger on filter (THS4271):

This brings me to my next point. I don't see any isolation resistor or termination resistor at the output of IC7 and IC8. What load do you want to drive with these two OPAmps?

IC7 is to drive the amplifiers from IC1 to IC5, and IC8 is to drive the amplifiers from IC9 to IC13.

I will try change amplifiaction of LMH6723 to 2 and solder isolation resistor on SJ1

Hi Wojciech,

looks as if some switching noise is superimposed on the signal? Are the supply voltages stable, clean and noisefree? Are you using switchers (DC/DC converters) to generate the supply voltages? Can you measure the supply voltages and show the scope plots?

Now I understand that you are designing a distribution amplifier? Well, your  distribution amplifier might suffer from the parasitics of copper traces from the output of IC7 to the inputs of inputs of IC1 to IC5. If they are longer, effects due to improper terminations of characteristic impedance of these lines can be the consequence.

There are many possible remedies which you can try out. In any case I would install a 100R series resistance directly at each +input of LMH6723. This resistance helps to dampen resonances of the line and the input of LMH6723. Sometimes the input stage of a HF OPAmp can oscillate without being noticable at the output of this OPAmp.

Then, there must be at least one 50...100R resistance directly at the outputs of IC6 to IC8. Sometimes it can help to install such a resistance into each line going from the "master" OPAmp to the "slave" OPAmps. This resistance must be installed directly at the output of IC7. The same is true for IC8 to IC13.

In any case the lines from the master OPAmp to the slave OPAmps should be as short as ever possible.

If this does not help, you may need to think about using microstrip technique in combination with proper line termination.

The jumpers can help to find out whether the many unterminated master slave connections are the issue here. Disconnect as many OPAmps from the master as possible and see whether the interterence is changing. It would be good if the jumpers were mounted near the output of master OPAmp. If this is not the case, you might want to cut the line by the help of a sharp knife near the master OPAmp.

Kai