Hi!
I have some questions regarding the THS4524 opamp.
I plan to use a single +5VDC supply for the opamp. The input signals are differential (symmetrical) signals with a common-mode voltage of 0V. This violates the absolute maximum ratings in the THS4521 datasheet, but is this still ok? I've seen this configuration with single-ended inputs, but not with differential inputs.
The opamp is going to drive a PCM4204 ADC, which needs a common-mode level of 2.5VDC. The THS4521 datasheet states that if Vocm is left unconnected, the common-mode voltage will be set to midsupply (2.5V). Is this good enough, or do you recommend to use the Vcom output from the ADC and connect it to the Vocm of the opamp?
thanks
Hi,
The THS4524 (i.e. THS4521/2/4 family of parts) was designed with 0V input common-mode, single-supply applications in mind. This is the reason why the input common mode range includes down to the negative rail. At the DC operating point, assuming 0V at the inputs of the Rg resistors and 2.5V (output common mode) at the outputs, the voltage at the non-inverting and inverting pins will be the voltage due to the voltage divider of Rg and Rf. This will be a voltage between 0 and 2.5V, which is within the input common mode range.
For example, in a differential gain configuration of 2V/V with Rf = 1kohm and Rg = 500ohm, the voltage at the input pins will be 2.5V * 500 / (500+1000) = 0.833V. For higher gains, the input common mode will still remain above 0V and within the input common mode range: at gain =10V/V with Rf = 1kohm and Rg = 100ohm, the common-mode voltage at the input pins will be at 0.227V, which is still within the input common mode range.
The differential input case is ideal because the voltage at the non-inverting and inverting input pins will remain virtually fixed at the operating point voltage as the differential input and differential output swings due to the symmetry. Meanwhile, in the single-ended input case, the voltage at the non-inverting and inverting input pins will move up and down with the input signal due to asymmetrical operation (single-ended input to differential output).
Leaving Vocm unconnected from the ADC should be fine - the second paragraph on p. 31 of the THS4524 datasheet notes that no performance difference was observed with Vocm left open or connected to the Vcm output of the PCM4204. If you do leave Vocm open, I recommend placing at least a 0.22uF capacitor to ground on the Vocm pin to help filter out noise.
Thanks for the quick reply!
The reason we didn't want to use the Vcom from the PCM4204, is because we want to keep the circuit as simple as possible, and we thought we needed som buffers (voltage followers) between the Vcom output on PCM4204 and the Vocm input.
But yesterday I read the article slyt119. Do I understand it correctly if I say that when the common-mode voltage is at midsupply, then we don't need buffers? We'll probably leave Vocm open though.
Another question:
In the THS4521 datasheet page 32 you are using 40.2Kohm resistors on the outputs. First, is it a typing error, should it be 40.2ohm instead? And I don't quite understand the purpose of these two resitors.
Yes, if you want the output common mode of the THS4521 to be at midsupply, then you can leave it unconnected to the ADC common mode output since the THS4521 output common mode will default to midsupply if Vocm is left open.
You're absolutely right about the resistors, they should be 40.2ohms. These resistors isolate the amplifier outputs from the large capacitive load (the 2.7nF and 100pF), which might otherwise introduce frequency response peaking and time domain ringing if connected directly to the amplifier outputs. The resistors also form a low pass filter with the capacitors to limit the noise bandwidth of the signal entering the ADC. I have noted down this typo and will change the labeling with the next set of revisions to the datasheet.
Thanks!
Hi again.
It turns out we have to use single-ended input instead of differential input. We have some problems with calculating the values of the resistors used to match the impedances. We try to use the formulas in document slyt310, but we don't get the same results as in the calculation example. The slyt310 refer to a worksheet used to calculate the values of Rt, Rg and so on. But we can't find the worksheet (FDA_input_impedance.xls).
Is this document still available? We need a gain of 1.75 and the single-ended source impedance is 50ohm. We have read that Rf should be in the range 300-500ohms.
Can you help us find the worksheet, or tell us how to calculate the correct Rt and Rg values?
thanks again!
Edit: Another thing, it seems like the land pattern and mechanical data is missing in the datasheet for the tssop38 (DBT) package. All the other packages are present.
Hello,
We have an online FDA component calculator for the single-ended to differential configuration at this link: http://www-k.ext.ti.com/srvs/cgi-bin/webcgi.exe?Company={5761bcd8-11f5-4e08-84e0-8167176a4ed9},kb=analog,case=obj%2812223%29,new
Please let me know if the link works. I will track down that Excel file as well.
I believe the mechanical data for the DBT package is on p. 48 in the current datasheet revision (SBOS458E). Is that sufficient? You can follow the land pattern guidelines for the TSSOP-16 (PW) package on p. 47, but note that the lead pitch is 0.5mm for the DBT (versus 0.65mm for the PW) and the leads are only 0.27mm wide on the DBT (versus 0.3mm on the PW).
To get the Excel, follow this link: http://focus.ti.com/general/docs/lit/getliterature.tsp?literatureNumber=slyt310&fileType=zip
There will be a copy of the .pdf article in the zip file. If you open the .pdf file, for example, in Adobe Reader, and click on the attachments icon (paper clip) on the bottom left of the window and you will see the Excel file as en embedded attachment. For some reason, the version of the app note through a direct .pdf link has the embedded attachments removed.
Thank you, both link works. But our circuit differs a bit from the one the calculators are based on. We have capacitor, blocking DC voltages on each input. The capacitor is messing up the impedance matching in the calculators. It's probably too much to hope for that you have a calculator with this option? We'll probably have to do the math ourselves...
Regarding the mechanical layout: I found the data on page 48 in the document you linked to, but this page is missing in the pdf I downloaded from the component page. It has the same document number though. Anyway, we have it figured out.
Even if you are AC coupling, you can still use the same resistor values as calculated. For example, see Figure 84 on p. 27 in the THS4509 datasheet for a DC-coupled, single ended-input circuit and Figure 86 on p. 28 for the AC-coupled equivalent. The DC blocking capacitors should be high enough value (say 0.1uF) so that at your frequencies of interest they are like shorts and the effective impedance matching is the same as in the DC-coupled configuration.
The calculators lump the resistors on the non-driven side into one resistor (i.e. Req = Rs||Rt + Rg). If you want to get a configuration like Figure 86, you can use separate resistors that are the same value as the Rs, Rt, and Rg on the input side (side being driven with input) and use capacitors to ground as the figure shows. The goal of this configuration is to keep the sides balanced by having essentially the same components on both sides of the FDA, using the same resistors/values, same capacitors, same parasitics, etc. You may not even notice a difference if you just keep the Req resistance lumped together as one resistor and use a single DC blocking cap to ground on the non-driven side (double value as cap to ground on Rt on input side).
Please let me know if my explanation is unclear. I can post a schematic example if that will help.
I just read back and saw that you are driving the PCM4204, so you are likely driving signals in the 1k-100kHz range. In that case, your DC blocking caps will probably need to be on the order of 10-47uF, not 0.1uF as I wrote in my previous post, depending on your signal frequency range.
Well, we finally figured it out, just one final detail regarding the frequency range: Ideally we'd like a signal range from 20-20kHz. (but frequencies below 100Hz is not that important)
We've read in slyt310 that THS4509 performs best with Rf in the range of 300-500ohm. Is this the case with the THS4524? To extend the signal range in the lower end, I believe we need to raise the value of Rf to above 500ohm. We don't want to raise the value of the capacitor much further due to physical size.
Rf of over 500ohms will be fine. If you look in the THS4524 datasheet, the Rf used for most of the example circuits (and used for characterization) is 1kohm.
Knut Olav,
Can you please post your schematic for the two cases you refer to when you say differential configuration and single-ended configuration? The app note shows some unconventional configurations that do not use symmetrical feedback and I want to make sure I understand what you are doing. Are you using symmetrical feedback resistors (same Rf and Rg on both sides)? If so, are you saying that you expect the gain to double when you switch from differential input drive (differential output) to single-ended input drive (still differential output)? In fact, the overall signal gain will be the same in both cases.
We are using symmetrical feedback resistors.
We want a total gain of 1.75. In the differential case everything work as expected in PSpice.
When we use the excel worksheet to compute the values of the resistors in the single-ended case, we need to specify a gain 4 times larger (1.75*4=7) to get the correct output when we simulate in PSPice. From the documentation we thought the gain should be 2 times larger. We just wanted to make sure that 4 times larger is correct, or if there is another error somewhere.
Knut Olav
Knut,
The definition of gain can get a little confusing when source impedances are considered. The calculators assume that gain is taken with the "input" as the voltage across the matched termination and not at the signal voltage source itself. If you are taking the gain as from the signal voltage source and not at the termination that is loading the source, then there will 6dB of attenuation (0.5V/V) at the input that you have to account for. In this case, you have to double the gain in the calculators.
The two configurations you show above are fundamentally different from each other. In the differential case, you have assumed a source impedance of 0ohms, which is why the gain can be directly calculated as 1.75k/1k = 1.75V/V. In the online calculator, I entered a gain of 3.5V/V, which is double the desired gain of 1.75V/V due to the 50ohm source termination, and I was given the values you will see in the attached TINA-TI circuit (note: TINA-TI is TI's handy, free Spice simulation software).
In the TINA-TI schematic, you will also see that I have used the exact same resistor values to implement a differential input amplifier. For the same values to be applicable, the resistor balance must be preserved -- the differential source impedance is 100ohms versus the single-ended source impedance of 50ohms. You will see that the gain for both configurations is identical at 4.85dB or 1.75V/V.
Where in the documentation does it say that the gain will be two times larger when the input is single-ended versus differential? Perhaps there is an unclear explanation of the FDA operation in that app note.
The discussion of input termination starting on p.15 of the following app note may be helpful, too: Fully Differential Amplifiers - sloa054