Charge Pump Conundrum

Thread Starter

BilltheBlob

Joined Jul 6, 2020
3
Hello!

I'm having trouble deciding if I should use a charge pump for my project. I'm making an EMG front end circuit (See schematic attached) and used a charge pump (LM2776DBVR) and an LDO (TC593002ECBTR) to create a negative supply for the op-amps.

When I assembled the circuit though, it didn't really work as expected. Alone, the circuit created a -3.0 supply, but when it was connected to the op-amps the line was set to around 2.68 V. After playing around with the circuit for a bit, I got rid of the LDO, connected the charge pump directly to the op-amp supply line, and put a 100 ohm load in parallel.

The charge pump has a built in setting where if the current is less than 40 mA, it goes into a pulse skipping operation, so I'm guessing this was the issue.

Anyways, my question would be:

-would a different topology like a buck boost converter work better?

Adding the load in parallel causes the circuit to work but it's an extra 30 mA which is wasted as heat. Charge pumps definitely take up less space and are cheaper, but my end goal is to make some sort of wearable out of this, so having wasted power is not ideal.

Cheers
 

Attachments

ronsimpson

Joined Oct 7, 2019
2,985
I think you need a negative supply because you need a good signal at "0 volts". So the exact voltage is not critical. I worry that noise from the charge pump will leak into the signal. (power supply rejection ratio at the charge pump frequency) I would remove the LDO regulator and insert a inductor. C13, L1, C14 As crutschow said make the caps bigger. L1 & C14 make a low pass filter.
 

WBahn

Joined Mar 31, 2012
29,976
How sensitive to noise is your circuit? Charge pumps (and switchers in general) are not good matches for low-signal analog circuits, unless they are synchronous with the sampling circuitry (turns pseudorandom switching noise into fixed-pattern noise).

If you are wanting to make something wearable, be very careful and get someone to do the design that knows what they are doing with regards to electronics that interact with human beings -- there is a huge degree of legal liability you are opening yourself to if you are not very careful about doing things right, both from an electronic standpoint and from a regulatory/legal standpoint.
 

cmartinez

Joined Jan 17, 2007
8,218
I've quite successfully used the ICL7660 chip in the past to generate negative voltages for analog-sensitive devices. I suggest you take a look at its datasheet, see if it can help you.
 

Attachments

Thread Starter

BilltheBlob

Joined Jul 6, 2020
3
I think you need a negative supply because you need a good signal at "0 volts". So the exact voltage is not critical. I worry that noise from the charge pump will leak into the signal. (power supply rejection ratio at the charge pump frequency) I would remove the LDO regulator and insert a inductor. C13, L1, C14 As crutschow said make the caps bigger. L1 & C14 make a low pass filter.
Thanks! I'll try using the LC filter.

The signal is low passed with a 2nd order filter with a cutoff frequency of 750 Hz. The switching frequency of this charge pump was 2 MHz, would I still have to worry about the noise in the signal even if it's filtered?
 

WBahn

Joined Mar 31, 2012
29,976
Thanks! I'll try using the LC filter.

The signal is low passed with a 2nd order filter with a cutoff frequency of 750 Hz. The switching frequency of this charge pump was 2 MHz, would I still have to worry about the noise in the signal even if it's filtered?
Depends on what the noise spectrum looks like, what the rejection ratio is, what other coupling modes exist, and how much noise the analog sections can handle.
 

ronsimpson

Joined Oct 7, 2019
2,985
Power supply rejection (INA827):
You can see that supply noise gets into the output. Note it is very dependent on frequency and gain of the amp. At 2mhz it will run to the output.
1594091897499.png
Some amps will not give this graph but will only give the number for 60hz.
I know you made low pass filters but this noise will go to the output and may cause strange readings. You might add a RC low pass filter at the very last thing. You filter is 750hz so add a simple RC filter set for 2khz or higher.
Remember the 2mhz of your charge pump. It may cycle skip and cause 500khz noise. (or lower)
 
Top