LM1875 nested Gainclone

linuxguru · 17th February 2009, 07:11 AM

Here's a first attempt to apply an LME49860 op-amp as a front-end along with an LM1875 as the power stage, with a nested-feedback schema. The topology is vaguely reminiscent of Technics/Class S circuits, but has some improvements. The initial simulation results are excellent - the THD20 numbers are much better than the LM1875 alone exhibits.

Edit: Thanks to Pedja for his LM1875 model.

linuxguru · 17th February 2009, 07:17 AM

Here's the simulated THD20 FFT at 10W into 8 ohms. H3 is < -140 dB. There's no easy way to fix the H3 dominance, it's an internal artifact of both the opamps.

linuxguru · 17th February 2009, 07:21 AM

Here's the closed-loop AC small-signal response. With C7 = 10 pF, as shown, the response rolls off rapidly above 1 MHz. Increasing the value of C7 to 22 pF or 47 pF will roll off the response at a lower frequency, with a small cost (a few dB higher) in THD20.

lumanauw · 17th February 2009, 03:46 PM

Hi, Linuxguru,

Very interesting. What is the purpose of R12=39k? Is this schematic sensitive for resistor value changes (is there somekind of balanced resistor configuration) ?

Is the LM1875 give all the current needed, or opamp still give some %output via R3=100ohm?

linuxguru · 17th February 2009, 04:06 PM

R12 = 39k is part of the nested inner feedback loop for the LME49860, which linearizes it substantially. I determined the values of the feedback resistances empirically by simulation, but these are by no means optimal.

There are some stability constraints, however. R7/R2 sets the gain for the LM1875, and that has to be greater than 10. R8/R5 has to be almost exactly equal to R7/R2, to minimize the feed-forward current contribution of the LME49860 through R3 (100 ohms). R3 cannot be too low, or U2 will go into current limiting (U2 contributes a small amount of feed-forward correction current directly to the load through R3).

The Boucherot RC cell has typical values found in similar designs, but it might aid stability if the R is reduced further to 2.2 or 4.7 ohms (this will increase the dissipation in the Boucherot cell, however).

There is some scope to play around with some of the feedback resistor values, and it is likely that these can be optimized further by simulation.

It has also been brought to my attention that my topology is broadly similar to Mauro Penasa's LM318/LM3886 gainclone - and indeed, the main ideas are the same: a front-end op-amp, with nested feedback (around the front end, and globally). There are some uncanny similarities (notably, the substitution of the Zobel inductor by a small resistor in both topologies), but my topology is simpler and includes a key element missing in Penasa's gainclone: feedforward error-correction from the front-end opamp (the LME49860) directly to the load. I have avoided the use of reactive bridging or compensation where possible, and there is only one externally introduced pole - from C7 (10 pF).

lumanauw · 17th February 2009, 04:50 PM

Is it possible to arrange so the current in R3=100ohm always near 0 at all condition? I mean LME49860 only give voltage swing, all current to load are handled by LM1875.

linuxguru · 17th February 2009, 05:11 PM

Ideally, we'd like to minimize the feed-forward current, but it's not possible to keep the phase of the signal from the LME49860 the same as that of the signal from the LM1875. The main reason is that there are internal poles, including a dominant pole, in the LM1875, which will cause a slight lag in output waveform. Hence, there will be a small difference between the two signals at all frequencies.

One can omit R3 totally (similar to Mauro Penasa's design), of course, but THD will increase without the feed-forward current component. It will still be better than the LM1875 alone, because the front-end LME49860 is still within the feedback loop.

lumanauw · 18th February 2009, 03:04 AM

You're right. If we omit R3, the whole thing become Walt Jung's composite opamp/buffer.

lumanauw · 18th February 2009, 03:13 AM

Hi, Linuxguru,

I don't clearly understand how this bridge works in patent #4,716,378. Amplifier A is the main amplifier, provide voltage swing. Amplifier C is current amplifier.

What is the gain working on amplifier C? Is it unity gain? If it needs unity gain stable amp, we cannot use chipamp for amplifier C.

lumanauw · 18th February 2009, 03:15 AM

Here's an example. IC352 is the amplifier C above. The bridge is around R373-R375/C359-R377-R379.

Is IC352 working on unity gain mode or with some gain?

Thread Tools	Search this Thread
Show Printable Version Email this Page	Search this Thread: Advanced Search

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Gainclone troulbe (LM1875 distorsion issue)	haft	Chip Amps	10	25th February 2007 08:29 PM
My LM1875 gainclone... point2point...	w00t	Chip Amps	21	23rd March 2005 09:12 AM
lm1875 gainclone lack speed	tonitze	Chip Amps	16	8th March 2005 09:36 AM
2 general gainclone questions, and an lm1875 question	Adam M.	Chip Amps	11	4th March 2005 11:53 AM
Just fired up my new dual mono lm1875 Gainclone.	Shoog	Chip Amps	5	19th January 2004 06:09 PM

New To Site?	Need Help?
Register to Participate Search Privacy Statement Contact Us	Frequently Asked Questions Did you forget your password? Mark Forums Read


	Please consider donating to help us continue to serve you. Ads on/off / Custom Title / More PMs / More album space / Advanced printing & mass image saving

17th February 2009, 03:46 PM	#4
lumanauw diyAudio Moderator Emeritus Join Date: Oct 2002 Location: Bandung	Hi, Linuxguru, Very interesting. What is the purpose of R12=39k? Is this schematic sensitive for resistor value changes (is there somekind of balanced resistor configuration) ? Is the LM1875 give all the current needed, or opamp still give some %output via R3=100ohm?

17th February 2009, 04:06 PM	#5
linuxguru diyAudio Member Join Date: Sep 2007	R12 = 39k is part of the nested inner feedback loop for the LME49860, which linearizes it substantially. I determined the values of the feedback resistances empirically by simulation, but these are by no means optimal. There are some stability constraints, however. R7/R2 sets the gain for the LM1875, and that has to be greater than 10. R8/R5 has to be almost exactly equal to R7/R2, to minimize the feed-forward current contribution of the LME49860 through R3 (100 ohms). R3 cannot be too low, or U2 will go into current limiting (U2 contributes a small amount of feed-forward correction current directly to the load through R3). The Boucherot RC cell has typical values found in similar designs, but it might aid stability if the R is reduced further to 2.2 or 4.7 ohms (this will increase the dissipation in the Boucherot cell, however). There is some scope to play around with some of the feedback resistor values, and it is likely that these can be optimized further by simulation. It has also been brought to my attention that my topology is broadly similar to Mauro Penasa's LM318/LM3886 gainclone - and indeed, the main ideas are the same: a front-end op-amp, with nested feedback (around the front end, and globally). There are some uncanny similarities (notably, the substitution of the Zobel inductor by a small resistor in both topologies), but my topology is simpler and includes a key element missing in Penasa's gainclone: feedforward error-correction from the front-end opamp (the LME49860) directly to the load. I have avoided the use of reactive bridging or compensation where possible, and there is only one externally introduced pole - from C7 (10 pF).

17th February 2009, 04:50 PM	#6
lumanauw diyAudio Moderator Emeritus Join Date: Oct 2002 Location: Bandung	Is it possible to arrange so the current in R3=100ohm always near 0 at all condition? I mean LME49860 only give voltage swing, all current to load are handled by LM1875.

17th February 2009, 05:11 PM	#7
linuxguru diyAudio Member Join Date: Sep 2007	Ideally, we'd like to minimize the feed-forward current, but it's not possible to keep the phase of the signal from the LME49860 the same as that of the signal from the LM1875. The main reason is that there are internal poles, including a dominant pole, in the LM1875, which will cause a slight lag in output waveform. Hence, there will be a small difference between the two signals at all frequencies. One can omit R3 totally (similar to Mauro Penasa's design), of course, but THD will increase without the feed-forward current component. It will still be better than the LM1875 alone, because the front-end LME49860 is still within the feedback loop.

18th February 2009, 03:04 AM	#8
lumanauw diyAudio Moderator Emeritus Join Date: Oct 2002 Location: Bandung	You're right. If we omit R3, the whole thing become Walt Jung's composite opamp/buffer.