Full bridge feedback stability issue

Member
Joined 2003
Paid Member
Hi.

Im struggling with a very basic full-bridge setup. Feedback is a UCD-style back to both inputs of an lm311. No matter what feedback arrangement i try, i get the same odd waveforms when looking at the 2 half bridges individually. When looking at the actual differential signal between the half bridges it looks better, but still odd.

The behaviour is load-dependant. With light/no load the waveform are very good and sine-like. With load and low levels the waveform are also ok. Based on simulator models I have a strong feeling that the change on the sine-wave is where the output inductor goes from continuous to discontinuous mode. Any ideas on what to look for? My control-theory is light-years away from doing a full analysis of this.

First picture is output from the 2 half bridges. Second is a zoom on the point where the sine changes. I realize it's a bit hard to fully see

Can supply schematics later, but it's very basic.

Not looking for ultrahifi. Just stable performance for a sub. -ans some learning:)

Kind regards TroelsM
 

Attachments

  • IMG_20230604_174243892_HDR.jpg
    IMG_20230604_174243892_HDR.jpg
    342.4 KB · Views: 67
  • IMG_20230604_174305918.jpg
    IMG_20230604_174305918.jpg
    335.1 KB · Views: 63
Moderator
Joined 2011
Is the scope floating, isolated from the mains ground?
Is the scope probe ground clip connected to the power supply negative terminal?
Is the power supply itself floating, or grounded to the AC mains ground?
Charge pump gate supply?

In general, an unbalanced scope probe will not give proper results unless certain conditions are met.
 
Last edited:
Member
Joined 2003
Paid Member
Hi Rayma.

Scope is grounded to PE. Amp is floating relative to PE.
AC-part of wave-form is the same regardless of probe-gnd going to amp-gnd or amp-Vcc.
Charge-pump-supply appear to be sufficient.

PSU is +/-30V regulated.

Simplified LT-spice schematic is attached. While not giving identical results compared to prototype, the simulation clearly shows that the waveforms from each half-bridge are not pretty sinewaves, - but the difference-signal is ( at least in the simulation).

Fidling around with the feedback components and will alter the waveforms, but not the overall result: waveforms from each HB appears a bit erratic/random,
I realize that as long as the differential waveform is good, thats all the speaker will ever "see", but the erratic HB-signals appears non-stable, and I like my amps to at least appear stable :)

Actual output-circuit is IR2184 + 2XIrf530. While not "good" in any way I have used that exact combination before (in half-bridge) with very fine results.

UCD  bridge simple 1.png


This is some of best waveforms I can get from the simulator. Proto-type waveforms are worse. - but simulator do not include the large deadtime from IR2184, so there ARE of cause differences.


output 1.png
 
Last edited:
Instead of scoping both channels separately, you should consider the differential that will feed the speaker load.
Otherwise you may see some common mode distortions that are irrelevant because they cancel on the differential output.
To achieve this you better use a virtual scope - for instance REW software and a soundcard with symmetric input.
Contrary to the scope plot this will depict the relevant signal trace, bandwidth limited to audio frequency range
 
  • Like
Reactions: 1 user
I do not see a post-filter feedback in your simu. So, yes, there may be some influence of magnetizing current through ouput inductor. Disappearing magnetizing current stops charging output capacitance and directly influences output voltage ramping. IR2184 is way too slow imho.
 
If you want a real bridged amp to behave predictably, the "first" output needs to be completely independent of the second output. The second output can be an inverted slave of the first or independent. When you have interdependencies, bazar things happen like common mode oscillations and distortion. Mixing the "balanced" input with feedback will not stay "balanced".
 
Member
Joined 2003
Paid Member
Steveu: sorry, i don't fully get what you mean. I my sim and prototype the 2 outputs are not completely independent as one is always the inversion of the other.
As far as I can gather from other schematics on the www, that's Standard.

I've tried different couplings (k-factor) from one output-coil to the other and while it does affect performance, the phenomenon is also present with 2 separate coils.
 
It is probably the dead time exhibiting. Bellow is that of TPA3255 I found on YouTube. As you mentioned one is the inverted of the other, in other words the carrier phases are opposite. This can only function for AD modulation. If you keep the phases the same but invert the input signal, then you can apply BD modulation after which it can be convertd to my ABD mod. For that you need to insert a resistor grounding the output capacitors and the dead time distortion vanishes.
 

Attachments

  • dist 3255.JPG
    dist 3255.JPG
    41.2 KB · Views: 36
Hi.

Im struggling with a very basic full-bridge setup. Feedback is a UCD-style back to both inputs of an lm311. No matter what feedback arrangement i try, i get the same odd waveforms when looking at the 2 half bridges individually. When looking at the actual differential signal between the half bridges it looks better, but still odd.

The behaviour is load-dependant. With light/no load the waveform are very good and sine-like. With load and low levels the waveform are also ok. Based on simulator models I have a strong feeling that the change on the sine-wave is where the output inductor goes from continuous to discontinuous mode. Any ideas on what to look for? My control-theory is light-years away from doing a full analysis of this.

First picture is output from the 2 half bridges. Second is a zoom on the point where the sine changes. I realize it's a bit hard to fully see

Can supply schematics later, but it's very basic.

Not looking for ultrahifi. Just stable performance for a sub. -ans some learning:)

Kind regards TroelsM

To begin, it is advisable to start with a functioning half bridge circuit using the comparator LM311. Over the years, numerous projects have been completed using this particular comparator. You can even refer to various service manuals available, as there are plenty to choose from that provide reliable instructions.

Careful with some loose designs on the forum claiming stable outcome for they are flawed in many ways.

I would recommend you place solder footprints around the various stages to make up "future feedback network alterations" once you have it working.
 
The behaviour is load-dependant. With light/no load the waveform are very good and sine-like. With load and low levels the waveform are also ok. Based on simulator models I have a strong feeling that the change on the sine-wave is where the output inductor goes from continuous to discontinuous mode.
Half-bridge circuits are bidirectional converters and therefore always operate in CCM. Even at light loads, excess energy in the output capacitor can still get back to the DC bus.

This maybe easily visualised by looking at the converter in direction of reverse power flow (output to input). The observant person should see a boost converter.
Any ideas on what to look for?
Try reducing the dead time to see if the distortion reduces. The blips occur exactly where the load current changes polarity.