Capacitor size for avoiding voltage drops

I am using a GSM module (A7 GPRS, www.ebay.de/itm/182256206840) and although my power supply is rated well above requirements (it supplies 2A) I experience voltage drops.

I thought about soldering a capacitor directly onto the board between VCC and Ground, but have no idea what size to use (the biggest one I have 2mF). Or should I use a big one and a small one? What's the best practice here?

Depends on how big and how long the voltage drops are. What are you running? Do you know how many ma each component consumes?

I run an Arduino Nano, a temperature sensor and the GSM module. Not sure how much the GSM module draws, but when it starts sending it seems to draw a lot (I am in a remote area with bad cell phone coverage, so I guess it sends with full power).

When transmitting, GSM modules can draw more than 2 amperes. The power supply is inadequate.

The old 'rule of thumb'was 2000uF/Amp... And your 2A psu will be at it's limit for the peaks, but plenty for an average of about 1/8 that.

And ( sadly) the current draw doesn't reduce enormously with reduced power levels.

I seem to remember that even with 0dBm output they still took several hundred mA..

It's down to the PA design.

Allan

jremington:
When transmitting, GSM modules can draw more than 2 amperes. The power supply is inadequate.

The transmission is typically for 1 out of 8 slots, and the slots take about 0.58ms, so the capacitor
would need to hold 1.2mC without drooping too far (lets say 0.2V), so 6,800uF might do. Larger
if using multiple transmit slots (upto 3 can be used IIRC).

C = dQ / dV = I dt / dV

Never bothered with big C's in GSM mobile phone design...

1/ the pa was rated down to 3.4v , so even a largely discharged Li battery could manage that
2/ a fast power control loop held the power constant during the pulse, even with sagging battery volts, as well as controlling the rise/fall of the pulse for compliance and emission reasons

of course there's a limit - when the battery's dead, it's dead. And a large cap wouldn't have helped much - and it would take up loads of space and have significant cost.

We always hated that b****y SIM card - took up huge amounts of PCB space

regards

Allan

allanhurst:
And ( sadly) the current draw doesn't reduce enormously with reduced power levels.

I seem to remember that even with 0dBm output they still took several hundred mA..

It's down to the PA design.

Allan

power in depends output stage current which is roughly proportional to output amplitude,
output power depends on amplitude squared - so efficiency drops markedly at lower powers
if the output stage voltage is fixed.
If you have power converter driving the output stage supply voltage to match the amplitude
of the output, you can much better efficiency at lower powers (modern phone chipsets undoubtedly
pull this trick among others to optimize battery life).

The PAs used in GSM phones were 3-stage FET amplifiers.

The reason their efficiency was so poor at low power levels was that you had to keep all 3 stages biased on to get any output. Their input was at about +7dBm, so at low powers they were attenuators, not amplifiers....

If you have power converter driving the output stage supply voltage to match the amplitude
of the output, you can much better efficiency at lower powers (modern phone chipsets undoubtedly
pull this trick among others to optimize battery life).

They might - but in general 3/4G use linear ( class A) PAs which have much worse efficiency than the class C PAs used in 2G (GSM).

Using fast power convertors for greater efficiency.has considerable problems

1/ you have to switch much faster than the symbol rate for the psu to track the power required.
2/ biasing the FET transistors becomes more complex as the PA HT varies.
3/ You have to keep all the switching artifacts out of the transmitted signal to meet adjacent channel interference specs

I don't say it can't be done, but when I tried this for EDGE some years ago I couldn't get it to work.

Big PAL/NTSC tv transmitters used 2 psu's for their klystrons - a higher voltage just for the sync pulse.But they were terribly inefficient - 1MW dc in for 110kW out.... the power bill was significant!

regards

Allan