I'm considering doing these as a group buy potentially with kits if there's enough interest. I wanted a nice compact high performance amplifier so I thought of the LME range. The amp uses the LME49811 and the ALF08NP16V5 for the power stage the keep things compact. I have PCBs to build and test (after christmas I can get some THD measurements). The boards measure just 55 x 22mm.
The smallest SM part is 0805.
So would people be interested in something like this? I think I could get the board coast down to about 7.5USD?
Here's some pics.
The smallest SM part is 0805.
So would people be interested in something like this? I think I could get the board coast down to about 7.5USD?
Here's some pics.
Should be a great little project. I've used the LM4702 with lateral mosfets and it sounds very nice indeed.
Would suggest that these drivers do like some heat sinking -- so you might want to factor in some space so that a small hs can go on the driver chip.
If you panelize the boards -- i.e. -- design slots so they can be snapped apart -- you can bet a lot more square centimeters per pfennig. i know that my board house charges more for this, but it helps get unit costs down.
Would suggest that these drivers do like some heat sinking -- so you might want to factor in some space so that a small hs can go on the driver chip.
If you panelize the boards -- i.e. -- design slots so they can be snapped apart -- you can bet a lot more square centimeters per pfennig. i know that my board house charges more for this, but it helps get unit costs down.
Nice, work!
As the PCB design is already finished, it is too late to comment on it, so here some ideas if someone gets inspired to design something similar and even smaller.
Put the IC on the same heatsink as the mosfet. To the opposite edge on the PCB to the mosfet. Solder half the pins underside the PCB and half on top side.
Use IC and mosfet as the attachment to the heatsink.
Solution also solves a lot of other issues that otherwise need careful thinking. Such as the placement of the bias trimmer and compensation capacitor. Also the input and feedback component areas "automatically" become concentrated and super miniscule - which is nice...
As the PCB design is already finished, it is too late to comment on it, so here some ideas if someone gets inspired to design something similar and even smaller.
Put the IC on the same heatsink as the mosfet. To the opposite edge on the PCB to the mosfet. Solder half the pins underside the PCB and half on top side.
Use IC and mosfet as the attachment to the heatsink.
Solution also solves a lot of other issues that otherwise need careful thinking. Such as the placement of the bias trimmer and compensation capacitor. Also the input and feedback component areas "automatically" become concentrated and super miniscule - which is nice...
Isn't good idea to mount the LME49811 heat-sink close to C4 and C5 capacitors. That way the location of the components is not the best. Keep trying.
If you mount LME49811 to the output transistors heat-sink, the transistors in the input stage will be influenced by heat transistors from the output circuit of the amplifier. Even if a LME bias circuit will work properly - we believe this theory, basically is everything else - this mode of operation is not 100% beneficial. For this reason get in the other case: that the LME to use separate heat sink. But this case is not good because bias circuit not follow temperature output power transistors. So, we are two cases that we can not really choose. An alternative would be to use a external bias circuit instead of the trimmer. Try this.
For this reason, solutions kind LME49811 or other also I do not see viable, or in other words, let them. Use audio amplifiers schematic have only discrete components. It's much better that way.
If you mount LME49811 to the output transistors heat-sink, the transistors in the input stage will be influenced by heat transistors from the output circuit of the amplifier. Even if a LME bias circuit will work properly - we believe this theory, basically is everything else - this mode of operation is not 100% beneficial. For this reason get in the other case: that the LME to use separate heat sink. But this case is not good because bias circuit not follow temperature output power transistors. So, we are two cases that we can not really choose. An alternative would be to use a external bias circuit instead of the trimmer. Try this.
For this reason, solutions kind LME49811 or other also I do not see viable, or in other words, let them. Use audio amplifiers schematic have only discrete components. It's much better that way.
Mounting the LME that way is not good practice. The board isn't thick enough, the height of the board is then determined by the IC and it's subjected to (most probably) a higher heat than one with it's own sink. And the design won't be much smaller.
This amplifier is designed t work on 30V max rails so with the constant curretn of the LME this means it will dissipate just 0.9W. There is a gap behind the LME for a sheet of aluminium to fit which should be clipped on or, preferably, glued on with heat conductive glue. Estimated temp rise is just 20C, which is nothing.
And I wouldn't say discrete component designs are much better - most of the time they will be bettered in every aspect by integrated solutions.
Your correct however they will only be subjected to more than this when the power is applied so I'll see how they fair and replace with diodes is necessary.
Nope, Dual Differential (as in consuming two amplifiers) or what is often called a fully differential or a balanced amplifier. Nelson Pass would call it SUSY. Here is a good example. Your boards could be swapped for the LME49600 parts described here. PRODUCT HOW-TO: Differential line driver with excellent load drive | EE Times
And Twisted Pear has their Sympatico amplifer http://www.twistedpearaudio.com/docs/amplifiers/sympatico_schematic.pdf
Notice the feedback loops from amplifier outputs back to the differential inputs.
And Twisted Pear has their Sympatico amplifer http://www.twistedpearaudio.com/docs/amplifiers/sympatico_schematic.pdf
Notice the feedback loops from amplifier outputs back to the differential inputs.
Last edited:
Hi!
Your 1W 1kHz performance is excellent!
Its almost 2x better than they were able to squeeze out from the Nationals test circuit - good job!
However, do you have an opinion why the rest of the figures does not correlate better with the Nationals results?
1kHz 10W should be around 0.0007% in your amplifier.
The 10kHz performance should easily be quite similar as the 1kHz results.
And your 20kHz results are already 3 times worse than the chips specifications makes us expect - which is odd, if you compare it to your 1W/1kHz result???
What equipment did you use to test your amps?
I suppose in 1W/1kHz noise dominates the figures, but at 10W THD becomes dominant?
At 10kHz and 20kHz THD is dominant in all figures?
Why is that?
Do you have any scope shots of the signals?
I tested them dead quick one evening after work with an Audio Precision System One. You do have to remember National/TI only specify the performance of the chip itself and not with the output devices in the loop. Just a basic follower has quite high distortion (in the 1+% range) and so the lower open loop gain makes the distortion noticeably higher at higher frequencies than what is specified in the datasheet.
I did forget to mention I haven't done any fiddling with it yet including changing any components so I hope to get the figures much better than what they are here.
Thanks
It will be difficult to keep them stable Carl. The LME49811 has a min gain of around 26dB I believe, to maintain stability that means you will have to keep the second closed loop gain high or you can add a potenctial divider in the loop to 'waste' the gain of the LME however this will impact noise figures negatively. Also the high amount of feedback in an existing loop will cause problems for stability too, phase will be your enemy.
Hi!
Im sure you will get MUCH better performance from it by experimenting with some component changes or firstly by figuring out what causes the THD rise with higher frequencies.
Your 1kHz/1W shows the performance capability of that amp.
Good job.
But find the cause of the performance decrease in higher freqs.
Gate resistors OK?
Are you still sure you dont inject noise from Mute input?
Is compensation capacitor OK? What value did you use in it?
Im sure you will get MUCH better performance from it by experimenting with some component changes or firstly by figuring out what causes the THD rise with higher frequencies.
Your 1kHz/1W shows the performance capability of that amp.
Good job.
But find the cause of the performance decrease in higher freqs.
Gate resistors OK?
Are you still sure you dont inject noise from Mute input?
Is compensation capacitor OK? What value did you use in it?
I'll try applying some bypass caps across the mute resistor. I used 100R for the gate resistors. I stated off with 30pF comp and ended up with 22pF, any lower and it wasn't stable. I will have a play tomorrow hopefully.
I had failed to consider that. I forgot that gain was a problem when experimenting with the LM3886 gain blocks as well.
Did you also test with 330R on Nch and 220R on Pch as Alfet has on their recommendations?
Could it be that the 49811 does not have the muscle to drive that FETcombo? And thats why it is not performing better with the higher freqs?
That combo does not look that hard to drive at all...???
It does drive some other mosfets like childplay...
I was thinking about testing a balanced (/bridged) amp with LM4702 driving two of those Alfets. But I would like it to have 3 leading zeros on its figures. If you dont get 4 zeros to your figures, it might be impossible for me to get 3 to mine...
My 20KHz is perfectly good enough and is what to be expected. Considering the FETs in the loop and the fact National conducted their THD measurements with a 30KHz filter (I used 80KHz) means they didn't measure any harmonics at 20KHz - just noise. Then look at the line on the graph for 80KHz bandwidth and the THD rises fast when getting to 20KHz, this with the FET in the loop and the reduced open loop gain make the THD rise fast to the perfectly good results I achieved.
Don't get me wrong I will be doing some more dev work to try and get it lower but I doubt it will be possible. Also National will use the best chip they have to conduct their datasheet measurements.
Don't get me wrong I will be doing some more dev work to try and get it lower but I doubt it will be possible. Also National will use the best chip they have to conduct their datasheet measurements.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.