LMH6553: Single power supply and single-end input

Hi Na na 78,

the idea of RT is to make the input of this circuit look like an exact 50R load. This is necessary if you want to connect a cable to the input. In high frequency circuits the cabling is usually down by the help of 50R technique. This means, that the driver and the cable must also have an impedance of 50R.

Kai

Hi,

To answer your questions:
1. For single ended input and single power supply circuit, I would recommend to use the circuit in Figure 58. In Figure 58, the RT is absorbed into the Rm, where the RM = RT||RS. You may want to take a look at figure 56 as well that defines what each resistor values are.

When driving the LMH6553 in single-to-diff configuration, it is usually desired to match the source impedance (Rs = 50-ohms in this case) with the active input impedance of the device in-order to minimize any power reflection at high frequencies. The RT is used primarily to provide this matched active input impedance. So, I think it is required to implement the Rt in customer's system.

2. Yes, there can be a scenario when a different value Rt would be required than what is specified in Table 1 and 2. The value of Rt is primarily determined by three conditions: 1) Device Gain, 2) RF or feedback resistor value and 3) Source impedance (Rs). If you can provide us with these three conditions from the customer, then we should be able to provide an appropriate circuit using the LMH6553.

3. Do you mean the common mode range of the LMH6553 inputs or the VCM pin that controls the output CM voltage? On a single +5V supply, the input CM voltage at the LMH6554 inputs IN+/- can typically be 2.5V+/-1.2V or +3.7V/1.3V. Similarly, the VCM pin voltage that controls the output CM voltage can typically be from 2.5V+/-0.81V or +3.31V/+1.69V.

Best Regards,
Rohit

Hi Kai

Thank you for your reply.
We will consider 50ohm termination.

BestRegards

Hi Rohit

Thank you for your reply.

I understood 1, 3.

Regarding to question 2, I add infomation about a circuit the customer is configuring now.

As shown in this figure, it offers single-end input to single-end output configuration, with 5V single supply.

Gain  ( single end) : 7.96

Feedback resistor : Rf = 430ohm

Source impedance : Rs = 50ohm

Vclamp=3.5V

Vcm=2.5V

Could you please check this configuration and please suggest modification if any?

BestRegards

Hi Na na,

there are some mistakes in your last scheme. Please have a look at your first scheme again. The - input of LMH6553 sits virtually at signal ground. So, the upper 59R resistor and the upper 255R resistor are in parallel, forming an input impedance of about 50R.

In order to keep the wiring of LMH6553 symmetric, the feedback loop arround the + input of LMH6553 imitates the impedances seen by the - input. So, the 49.9R resistance in series with the 100n cap imitates the hidden source impedance of the driver, which sits in parallel to the upper 59R resistance.

Or by other words, the + input and the - input of LMH6553 must see the same impedances to signal ground. In the first schematic this is fullfilled. In your last schematic this is not fullfilled.

Kai

Hi Kai

Thank you for your reply.

I modified following circuit based on Figure 56.

I believe this will offer Av=16 (8V/V with single-end out) and it will match for 50ohm system.

Could you please review it?

BestRegards

Hi Na na,

I don't think that such a high gain is possible with this chip in a 50R system, because when chosing Rg=50R and omitting Rt, Rf must be much higher than the maximum recommended value (325R).

Let's see what Rohit means...

Kai

Hi,

I think the circuit you have posted should work with termination resistor Rt = 135-ohms, and Rf = 700-ohms providing an input impedance match of 50-ohms.

However, one thing I wanted to point out is that due to the input impedance matching, you will lose 0.5V/V at the input. As a result, the total single-ended gain from input to single-ended output will be 4V/V. If the input frequency is not very high and the input interface trace is not very long, then you should be able to reduce the Rs to 10 or 20-ohms, and subsequently change Rm = Rs||Rt. This change will achieve higher gain from input to single-ended output and keep both paths symmetrical. For example, with an Rs = 20-ohms, the Rm will need to be 17.5-ohms.

Also, if the input and output are single-ended, then is there a reason to choose a fully differential amplifier such as the LMH6553? I think an op-amp can be used here instead and would be a cleaner solution. Is it primarily because of the LMH6553 clamping feature that the device is being considered here?

Best Regards,

Rohit

Hi Rohit

Thank you for your reply.

We can consider to change Rs.

> Is it primarily because of the LMH6553 clamping feature that the device is being considered here?
Yes. It is intended to control output swing level using clamping function even if the input is large.

So if 1Vpp and 50MHz input pulse is supposed, we expect the single-end output could be limited 1Vpp with Vcm=2.5V and Vclamp = 3.5V.


BestRegards

Hi

Please tell me your view about following situation.

Based on the following circuit, we saw an output that had slow changing at falling edge against input pulse.

Input: Upper figure, Output : Lower figure

I could not get why it has slow change only at the falling edge.  

As you expressed, is it cause of input termination? Or any other cause?

What kind of cause do you think? Please tell us anything thinkable.

BestRegards

Hi Na na,

was it measured with your original and faulty circuit from 26 march?

Kai

Hi kai

It was measured with the circuit which is not along with Figure 1 or Figure 58. (It is the last figure I attached.)
The circuit has not modified yet to meet recommendation.

BestRegards

Hi Na na,

TINA-TI simulation shows that a capacitance at the input can be responsible for that behaviour:

Cable capacitance could do this. This shows how important proper cable termination is in high frequency applications.

Kai

Hi,

I would recommend you to modify the circuit for input termination as provided in my previous post. As Kai mentioned, not having proper input termination at high frequencies could result in long settling due to input reflections caused by the cable.

Best Regards,

Rohit

Hi

Thank you for your cooperation.

I have to apologize that I offered wrong information.

Let me indicate the measured point using following circuit.

The measured wave forms were obtained at the point ① and ②.

A wave measured at point ① :

A wave measured at point ② :

The wave at the point 1 seems to be not so dull.

Input to the LMH6553 is a pulse from I/V amplifier output.

And a cable is not used at AC coupled input. (10mm wired on the single board)

BestRegards

Hi Na na,

do you see the long settling at measuring point 2 with the scope being connected to measuring point 1 at the same time? Or by other words, are both scope probes simultaneously connected to the circuit?

Kai

Hi Kai

It was not measured at the same time, it was obtained individually.

BestRegards

Hi

Let me consider recommended feedback resistor.

I changed Rf from 430ohm to 275ohm with the gain around Av=16 ( Av=8 in single-end out). 

Please refer following simulation result.

1. Rf=430, Rg=27 

2. Rf=275, Rg=15

Smaller Rf is, dulling at the output seems to be mitigated.

In addition, I can see the recommended feedback register is 275 or 325 by the package, in the datasheet page 19.

Questions 1

Is there possibility to make output dull if larger Rf is selected?

Question 2

Can you offer recommended Rf and Rg to set gain as 16?

Questino3

Is there maximum gain for LMH6553?

BestRegards

Hi Na na,

yes, current feedback OPAmp can only work properly with feedback resistors lying in a limited range. So, increasing Rf above the recommended value can result in improper performance like increased settling time.

Kai

Hi Kai

I apologize for this late reply.

What do you think of limitation of gain settings?

I wonder if it is feasible to set 16V/V gain ..?

BestRegards

Hi Na na,

the gain limitation in datasheet to 12dB (table 1) has mainly to do with the termination needs when connecting a 50R cable to the input. So, without having a 50R cable to connect to the input you could try to increase the gain like shown in figure 62.

But keep in mind, that the LMH6553 will come to its limits when increasing the gain so much. I think it would be better to use two cascaded OPAmps and split the gain onto two gain stages.

Kai