| zox2003 |
1.After playing with different input configurations for self-oscillating class D I would like to get your opinion and facts on difference between having Comparator (LM311) on input of amplifier versus High speed OPAMP (LM6132,LT1220)...What is a difference?Which one is better?
2. I have seen that some design's (IR design,...) are using feedback prior to output coil.Some other use after output inductor.What is a main difference and which one is recommended?
Most of the amps with OPAMP on input are using FB prior to output inductor.WHY? Can it be modified so FB is taken after(if is better)?
Regards.... |
|
|
| Ouroboros |
The IR reference designs all use pre-filter feedback. Post -filter feedback requires a different amplifier architecture, although to an unpractised eye, the pre and post-filter self-oscillating designs look very similar. You can build a post-filter (UCD style) amplifier using the IRS20124 or IRF20955 chips, but the recent IRS2092 with the modulator included in the driver chip can't be used for a UCD style of amplifier.
To get an understanding of the UCD amplifier with post-filter NFB, try to find an on-line copy of Bruno Putzeys AES paper. (I don't have a link for this but I'm sure someone else here has).
The modulator stage really needs to be a high-speed comparator. The LM311 or LM319 will work, but something like an LM361 is better if your MOSFET driver stage will accept TTL-level inputs. Even high-speed op-amps don't make good comparators. |
|
|
| zox2003 |
| So, it is better to have comparator on input then OP AMP, I guess...? |
|
|
| Ouroboros |
It might not be that simple. If you look at the latest IR reference design (IRAUDAMP5), you'll see it uses the IRS2092. The 2092 has an input stage which is basically an op-amp, followed by a separate comparator. The two together form the modulator, as in a sigma-delta amp (like this one), the comparator works on the combined signal from the op-amp. The op-amp isn't doing the switching, simply merging the input and feedback signals and applying the required frequency compensation.
In a UCD-style amp, the comparator has one input from the incoming audio, another from the NFB network, and simply switches (with no hysteresis required) when one input to the comparator passes the voltage on the other. This switching needs to be as fast as possible.
So basically you need to see in a particular design where the transition between the linear signal and the PWM signal is happening. This stage always requires to be a comparator of some form. Although an op-amp will work happily as a comparator at low speed, at higher speeds the output is unlikely to slew fast enough, and if you run an op-amp open-loop, it is also likely to have a noticable delay when coming out of saturation. As you will want to be switching in the 300 to 400kHz region for a full-bandwidth class-D amp, then I would have thought that an LM311 is about the slowest device you would think of using. |
|
|
| lumanauw |
| quote: | | , but the recent IRS2092 with the modulator included in the driver chip can't be used for a UCD style of amplifier. |
How about inverting UCD with IRS2092? |
|
|
| Ouroboros |
| quote: | Originally posted by lumanauw
How about inverting UCD with IRS2092? |
It might be possible. The input op-amp is a transconductance amplifier (OTA), so the link between the - input of the comparator and the output of the OTA is at a high impedance. You possibly could insert the post-filter NFB signal in at this node using a suitable network. It would be easy to model the input OTA using normal SPICE building blocks so I might have a go at seeing if I can get a similar architecture to work properly. |
|
|
| Ouroboros |
Checking the 2092 data sheet shows that the input OTA has a transconductance of 100mA/V. You can't put NFB locally around this stage as it would then make the OTA into a voltage amplifier, which would have a low output impedance, requiring a resistor to be inserted between the amplifer output and the comparator (which is not possible).
It might be possible to ignore the input OTA altogether (by connecting both its inputs to 0V). It will then not be trying to source or sink current to/from the input of the comparator, allowing a normal inverting UCD style connection to the comparator (as you show in your diagram).
I've got some DIL IRS2092 samples, but no time at the moment to try this out unfortunately. |
|
|
|
