Diffrential amp design

[ali_asadzadeh]

Thanks for the good feedback, I want to replace it completely, so I'm not bond to any part or voltages, Just something came to my mind,Can I trough it all out and replace it with a decent 24Bit differential ADC, with an internal gain of 128 or more? something used for wight scales but with better common mode rejection. since I just want to measure a uV signal between two probes technically.

I want to achieve a good reading, either trough discrete op-amps, or better AMPs or another ADC, for Now I'm using AD620 and LPC1758 ADC to read the values, I know the LPC1758 ADC sucks, and it's not good enough, But we have made lot's of progress in recent MCU specially from ST and NXP,

[OwO]

depends on your requirements:
- CMRR
- Gain
- Noise floor (input referred)
- Input common mode voltage range
- Input differential mode voltage range
- Input impedance
- Bandwidth

In some cases cheap opamps will do. I don't think abusing the differential input of an ADC is a good idea though. I've never seen CMRR specified on an ADC. I also haven't seen an ADC (apart from audio ADCs) that have that much built in gain.

[Evan.Cornell]

I also haven't seen an ADC (apart from audio ADCs) that have that much built in gain.

A specific use-case, but ADS1232 (http://www.ti.com/product/ADS1232) has built-in gain of 128. OP hasn't said bandwidth required, so I have no idea if this would be suitable for the application or not.

[Zero999]

Oh I didn't realize it's using a gain > 100. In that case a discrete design with 3 opamps can achieve >100dB CMRR easily because the first stage provides 40dB of common mode rejection.

Sorry, I can't see how, without using very well-matched resistors, which would be expensive, unless there's something obvious I've missed.

...
That's what you might think, but the data sheets say otherwise.
https://www.analog.com/media/en/technical-documentation/data-sheets/AD620.pdf
http://www.ti.com/lit/ds/symlink/ina333.pdf

Input voltage noise from 0.1Hz to 10Hz is 1μV_PP for the INA333 and 0.28μV_PP for the AD620, at gains above 100, which the original poster's design has. The only advantage the INA333 has, noise wise is lower input current noise at 2pA_PP vs 10pA_PP, so might be better if the input is high impedance, but it's only rated to 5.5V single supply and the original posters design is +/-5V.

The problem with DS is that there is no 0.01 - 1 Hz line, and looking into similar IC from TI TLV333 , where noise spec is split into 0.01 - 1 and 0.1 - 10 ranges, I see 0.3uV and 1.1uV, that's more than x3 times.
Right, it's just analytics, but if I go father into <0.5 Hz than difference with ad620 where line is sharply rising would be more obvious.  Theoretically, I don't know if China have any part numbers, very unlikely anyway.
 Regarding voltage, OP was going to redesign circuits and why keep same voltage?

Yes auto-zeroing amplifiers theoretically have no 1/f noise below a certain point, but they do have more noise at higher frequencies, thanks to the interference from the oscillator. The data sheets don't seem to be showing it though, because as you said, there's no 10mHz to 1Hz data.

Thanks for the good feedback, I want to replace it completely, so I'm not bond to any part or voltages, Just something came to my mind,Can I trough it all out and replace it with a decent 24Bit differential ADC, with an internal gain of 128 or more? something used for wight scales but with better common mode rejection. since I just want to measure a uV signal between two probes technically.

I want to achieve a good reading, either trough discrete op-amps, or better AMPs or another ADC, for Now I'm using AD620 and LPC1758 ADC to read the values, I know the LPC1758 ADC sucks, and it's not good enough, But we have made lot's of progress in recent MCU specially from ST and NXP,

What's wrong with the circuit I posted? It did get two likes after all.

You could replace the AD620, with the INA821, which is better in nearly every way and slightly cheaper, according to Mouser. It also has the same gain equation, so the resistor values wouldn't need changing.
http://www.ti.com/lit/ds/symlink/ina821.pdf

I can't say much about the ADC, as it's not something I've done before, but having a stable, low noise, voltage reference is very important.

[ali_asadzadeh]

A specific use-case, but ADS1232 (http://www.ti.com/product/ADS1232) has built-in gain of 128. OP hasn't said bandwidth required, so I have no idea if this would be suitable for the application or not.

The Sample rate of the signal is around 500SPS and the signal is an alternating signal in the 30Hz range, it has a shape like this

What's wrong with the circuit I posted? It did get two likes after all.

It's a good Idea and thanks for sharing it, But maybe we could achieve something cheaper at the end of the day.

[Zero999]

What's your budget? How many are you making? Generally for small volumes, it often works out cheaper, to use more expensive components, due to reduced design and assembly times.

As I said above, the INA821 is cheaper, than the AD620 and better too.

[SiliconWizard]

What's your budget? How many are you making? Generally for small volumes, it often works out cheaper, to use more expensive components, due to reduced design and assembly times.

Agreed.

As I said above, the INA821 is cheaper, than the AD620 and better too.

Yep, it's less than half the price, has better noise and BW specs, half the power consumption, lower input bias current... looks like a no-brainer here?

[Marco]

INA821 seems like a no brainer, even with the BJT input it's not that far off from integrated JFET solutions. Cheap enough and you aren't going to do substantially better without going discrete.

[ali_asadzadeh]

What's your budget? How many are you making? Generally for small volumes, it often works out cheaper, to use more expensive components, due to reduced design and assembly times.

We will produce 20K units, so the lower the price the better.

[Zero999]

What's your budget? How many are you making? Generally for small volumes, it often works out cheaper, to use more expensive components, due to reduced design and assembly times.

We will produce 20K units, so the lower the price the better.

According to Digi-Key, the INA821 is $2.489 each, vs $5.60 for the AD620, in quantities of 2500.
https://www.digikey.com/product-detail/en/analog-devices-inc/AD620ARZ-REEL/AD620ARZ-REELTR-ND/617756
https://www.digikey.com/product-detail/en/texas-instruments/INA821IDR/296-INA821IDRTR-ND/10448327

Use the circuit I posted previously with the INA821, instead of the AD620. The op-amp doesn't have to be anything special, because the common mode rejection and raising the signal well above the op-amp's noise floor, is done by the INA821. If you have a higher voltage rail than 5V available or can live with the output being limited to +3V, the OP07 is a reasonable choice and is cheap too. The output can be helped up closer to +V by using a pull-up resistor or current source (this can be as crude as a single PNP transistor), at the cost of extra parts and current draw.
https://www.digikey.com/product-detail/en/texas-instruments/OP07CDR/296-14626-2-ND/555699

NOTE: The current source is pretty crappy, because it relies on the transistor's h_FE which is not a well controlled characteristic and is temperature dependant, but that doesn't matter, since all that's required is a pull-up of around 1mA or so. I chose the BC857B over just the BC587, because its gain can only range from 180 to 460, rather than 120 to 800 and it was the cheapest PNP BJT on Digi-Key, at a cent each.
https://www.digikey.com/product-detail/en/rochester-electronics-llc/BC857B/2156-BC857B-ITTR-ND/11512248

Of course if you're buying in quantities of 20k, then you want to go directly to the manufacturer, rather than a distributor, such as Digi-Key.

[ali_asadzadeh]

Thanks Zero999 , so the BJT  would allow for a 5V output swing?

[Zero999]

Thanks Zero999 , so the BJT  would allow for a 5V output swing?

It won't allow the output to completely reach 5V, but it'll get within a couple of hundred mV or so. Obviously it won't be able to swing to -5V. Oh and don't bother simulating this trick. It invariably doesn't work with most SPICE op-amp models because they don't accurately replicate the behaviour of the output stage.

That raises another point: what's the input voltage range of the ADC? Presumably it can't go below zero, yet both the circuit you posted at the start of the thread and mine will output negative voltages.

[ali_asadzadeh]

The ADC input range is 0-3.3V

[Zero999]

The ADC input range is 0-3.3V

Ok then the PNP pull-up isn't needed, but what happens when the op-amp's output drops below zero?

[ali_asadzadeh]

There is a Reverse Biased Diode parallel to the ADC input