THS4524: I am confused by some of the examples in the data sheet

The converter needs to be driven below full scale

The max input differential voltage is a fault condition, normal operation will have a zero differential input as the loop drives that to zero. 

If you look way down on the ADC web folders you will find complete example designs discussing many issues. 

The THS4524 is the only quad FDA out there, the single and dual versions were massively upgraded in the THS4551 and THS4552. But no quad there as it is a bit impractical. 

Hello Shawn,

I think a lot of the confusion here is coming from the understanding of the "max differential voltage" spec. This is not the difference between Vin+ & Vin- but the difference between the non-inverting pin & the inverting pin of the amplifier. 

In regards to the 0.27 V/V gain, this is correct for the DC gain assuming there is a resistive load at the outputs. Resistors R41 & R42 will affect your output common-mode since the FDA will by trying to set this voltage at its outputs rather than at the circled nodes. The FDA is always trying maintain the non-inverting pins & inverting pins at the same voltage. So, when there is zero input both sides of the amplifier are around the same voltage. When a differential input is added, the non-inverting & inverting pins are maintained around the same value... the only way to do this is by adding a differential output = input * 0.27 V/V at the circled nodes, whether or not R41 & R42 are included or or not. 

Hopefully that made sense, if not you can try simulating this circuit in TINA and see the effects of including R41 & R42. Be sure to add a resistive load to this circuit to get proper results. You can also find more details on how FDA's function in this series.

Best,

Hasan Babiker

Also, this particular circuit is what I call an imbedded integrator Cload driver. not what you want to do unless absolutely necessary, most ADC drivers have a cap across the ADC inputs and simple series R isolation eliminating those feedback C's. 

But yes, those series output R's are inside the loop and do not effect the DC gain, 

Hasan Babiker31 said:

Hello Shawn,

I think a lot of the confusion here is coming from the understanding of the "max differential voltage" spec. This is not the difference between Vin+ & Vin- but the difference between the non-inverting pin & the inverting pin of the amplifier. 

Yes, it was late and I confused by pin names. I was talking about the "gap" in voltage between the inverting and the non-inverting pins I did not mean to point at the power supply to the opamp.

However, IIRC, that is still only a 1V max differential. Did I decode that part of the spec sheet correctly?

Hasan Babiker31 said:

In regards to the 0.27 V/V gain, this is correct for the DC gain assuming there is a resistive load at the outputs. Resistors R41 & R42 will affect your output common-mode since the FDA will by trying to set this voltage at its outputs rather than at the circled nodes. The FDA is always trying maintain the non-inverting pins & inverting pins at the same voltage. So, when there is zero input both sides of the amplifier are around the same voltage. When a differential input is added, the non-inverting & inverting pins are maintained around the same value... the only way to do this is by adding a differential output = input * 0.27 V/V at the circled nodes, whether or not R41 & R42 are included or or not. 

Hopefully that made sense, if not you can try simulating this circuit in TINA and see the effects of including R41 & R42. Be sure to add a resistive load to this circuit to get proper results. You can also find more details on how FDA's function in this series.

I only downloaded TINA yesterday. I'll give it a whirl very soon.  I don't think I am confused about the part of this being "differential". If I were looking at a plot of the voltage curves, there would be a centerline around which my signals would mirror each other (equal but opposite in amplitude). If one leg trended up to 0.5v, the other would trend down to -0.5V.  The offset of that centerline to ground is my common mode voltage. That applies separately to both the input and the output sides of this circuit.

Maybe I am misreading the spec for the ADS1278?  I just don't see which voltage value I need to match to get a "full range positive" digital output of 0x7FFFFF and a full range negative output of 0x800000 (two's compliment signed binary).   Figure 88 of the ADS sheet shows a Vref of 2.7v using a REF5025.   So, if I use that as my target, the gain of my THS4524 would need to turn my 1V (max) input differential into a 2.7 (max) differential to the ADC to get a full scale reading?  or can the THS4524 accept a higher input differential without clipping/distorting/failure?

(if I got that part right...)

What's confusing me is that the sample circuit in the THS4524 sheet (fig 85) shows a unity gain. However the description of it says it could push the ADS1278 to within 0.5dB of it's full scale. If I presume the same REF5025 was in use here (it should be due to the 27MHz internal clock reference was being used for High-Precision mode) then how did it nearly reach full scale.  By my calculation a reduction of 0.5dB from a 2.7V reference is still 2.54V (way more than the THS4524 was providing with unity gain)

Thanks!

Shawn

Michael Steffes said:

The converter needs to be driven below full scale

The max input differential voltage is a fault condition, normal operation will have a zero differential input as the loop drives that to zero. 

If you look way down on the ADC web folders you will find complete example designs discussing many issues. 

The THS4524 is the only quad FDA out there, the single and dual versions were massively upgraded in the THS4551 and THS4552. But no quad there as it is a bit impractical. 

Below full scale, yes. But if I get as close as practical with my upper input limit I have as many values as practical (output codes) in my data stream. I don't want to overdrive the ADC and suffer value clipping.

I don't understand your second sentence. Are you trying to say that the common mode voltage of my input lines should be as close to 0 as practical before arriving at the opamp (where it will get a portion of the Vocm voltage due to the divider formed by the feedback and gain resistances) ?

In your last paragraph did you mean there are no octo (not quad) FDA's because those would be impractical?

Hello Shawn,

I didn't think you were referring to the power supplies. You can provide an input greater than 1V to the amplifier, however this is not set at the inverting & non-inverting pins. Using the same figure as a reference:

In this case your differential input should be connected at "Audio inputs" circled in red. The "1V max differential" spec is for the nodes circled in blue. The FDA will always try to keep the non-inverting & inverting pins (in blue) the same & so in regular use case the differential voltage between these two nodes is close to zero.

Best,

Hasan Babiker

You were very right to have me go play around in TINA.  I played around with R41 & 42 and C21 & 22, did some voltage probing, found the AC Analysis table with its network analysis landmarks.  That was when things finally clicked into place.

I also think I understand why R41 & 42 do NOT contribute to the gain. Yes, they are part of the feedback loop (C21 had 2V across it on a 1V input signal) but they also act as a voltage divider which cancels out the extra gain so the voltage to the next stage is really Rf/Rg (which for one leg of your circuit is R33/R23 = 230/1000 = .23V/V )

Armed with that nugget of info, I went back to the spec sheet (I did originally refer to the right nodes) and located (section 7.1)  that my max input signal is limited to smidge more than my supply voltage (Vs+0.7V) but that's not going to sound clean. An amp can't drive much (if any) past its supply voltage. That same limit applies to the output voltage.  So... (this is where I am doing a self-check. Please let me know if I am right).

if Vs is 2.7v, could  I bring in a .125V differential signal without a low-pass filter on the feedback loop, and have a gain of 20V/V to max out at 2.5V headed to the next stage?

But, If I did insert the low-pass filter (let's use the top-half numbers of your diagram), I couldn't make the ratio of R33/R23=20 because then Vout- would be well past Vs+0.7V .  Let's say (for ease of math) that R33 and R41 are the same value.  Vout- would actually see a voltage of 5V. So I would need to limit myself to a max gain of about 10-ish?  I can adjust  the voltage across C21 by varying the size of R41 (and that determines my actual Vout).

Does this compare with your experience? (Did I interpret what TINA was showing me correctly?)

I know how frustrating it is when a new person asking questions gets things all mixed up in their head.  I greatly appreciate the time you are spending to help me get over this bump on the learning curve.

Regards,

Shawn

No problem Shawn,

The "Absolute Maximum Ratings" are the voltage at which the device will be damaged if it is exceeded. This is not meant to specify the limits for normal operation. For your input range you will want to look at the "Common-mode input voltage low" & "Common-mode input voltage high" specs in the datasheet. Note that this is not the differential voltage but the voltage limits at the non-inverting & inverting pins (circled in blue earlier).

For your output range you will want to look at the "Output voltage low" & "Output voltage high" specs. Again this is the voltage range for each output and not the max and min differential output voltage. 

I've included the specs for a 5V supply, there are also ratings for 3.3V.

Best,

Hasan Babiker