using Multiplying DAC to generate variable height pulses

[eliocor]

I need to generate rectangular pulses with width between 1 and 10us: their output peak voltage can range from +5V to -5v in (at least) 5 mV steps.
The original pulse (generated by a PWM microprocessor pin) have a pk-pk output voltage of 5V:


What I need is a variable attenuator with at least 10 bits of resolution (better 12 or 14 for linearity reasons).
I was thinking to use a Multiplying DAC like the AD5446 (14bit) set in 4 quadrant configuration: 13 bit for attenuation and one for sign [see Bipolar Operation, pag. 16].
This device is reasonably cheap for my requirements and its reference bandwidth is 12MHz.

The output results could be summoned in the following two pictures:

output set to 2.7V from the original 0-5V pulse

output set to -3.3V from the original 0-5V pulse

The output pk-pk voltage need to be changed not more than once per second.

- - - - -

Based on your experience, do you think my approach can be worthful?

Have you different suggestions/solutions on how to solve my needs?

[jahonen]

I can't say if it is any simpler, but maybe another option would be a normal DAC (connected to have a bipolar output) and an analog MUX to select either zero voltage or the DAC output according to the MCU PWM output?

Regards,
Janne

[eliocor]

It seems interesting: I have to look at the commutation speed of the analog switch, but it should work.
Thanks for the suggestion.

[free_electron]

Jahonen is right.
Sending this kind of signal through a multiplying dac will be trouble... The bamdwidth limitation of the dac will heavily distort the edges...

[Zad]

The AD5444 has a 3dB bandwidth of 12MHz, I can't see 1uS pulses being a huge problem.

[eliocor]

Yes, 12MHz bandwidth can be enough, but I like the suggestion of using a "standard" DAC:
my project will be open sourced, so I have to care about the possibilities of using different hardware.
A "standard" DAC will be easier to get than a Multiplying DAC.
In the next days I hope to get some time to test this configuration and see if it can be better than my initial thoughts:
my request for comments was for identifying the cheapest and safest way to generate those pulses.

Thanks in advance for all of your welcome comments!!!