Daniel, i tried your suggestion and it worked. But let me explain .
Psu= +/- 37V
I have install Rf=22K, Ri=680R Gain set to 33.
Rin= 680R Rg=22K
Rm= 39K
Dc offset = 30mV.
I change Rin= 330R Rg=10K everything else remain the same.
Dc offset = 6mV.
Since I was happy with the result, i didn't try further values for Rin + Rg, but i install a Cm cap of 100uf, as suggested by the datasheet, since i want to use the mute function with a little delay.
With the Cm 100uf installed and no other changes Dc offset = 20mV.
I try some smaller values for Cm and found that 22uf gave the best results of Dc offset = 4mV. (Rm was 39K)
Further degrease of Cm to 10uf result in Dc offset = 30mV.
So i kept the Cm to 22uf and increase the value of Rm from 39K to 49K and Dc offset reduced to 1.7mV!!!!!
According to the Datasheet RM≤(|VEE| − 2.6V)/I8, Mute pin has to be feeded with minimum of 0.5mA.
In my first try I had originaly install Rm= 39K = Im = 0.88mA= Dcoffset = 4mV.
With Rm = 49K = Im = 0.70mA = Dcoffset = 1.7mV
All the above tests was done with and without Ci = 220uf. Dc offset was the same with and without Ci installed.
So the 10K - 330R is a good combination for the input. In addition to that it seems the degrease of Imute current can further degrease the Dc offset.
I will try to do the same tests with another chip in order to clarify the results.
Psu= +/- 37V
I have install Rf=22K, Ri=680R Gain set to 33.
Rin= 680R Rg=22K
Rm= 39K
Dc offset = 30mV.
I change Rin= 330R Rg=10K everything else remain the same.
Dc offset = 6mV.
Since I was happy with the result, i didn't try further values for Rin + Rg, but i install a Cm cap of 100uf, as suggested by the datasheet, since i want to use the mute function with a little delay.
With the Cm 100uf installed and no other changes Dc offset = 20mV.
I try some smaller values for Cm and found that 22uf gave the best results of Dc offset = 4mV. (Rm was 39K)
Further degrease of Cm to 10uf result in Dc offset = 30mV.
So i kept the Cm to 22uf and increase the value of Rm from 39K to 49K and Dc offset reduced to 1.7mV!!!!!
According to the Datasheet RM≤(|VEE| − 2.6V)/I8, Mute pin has to be feeded with minimum of 0.5mA.
In my first try I had originaly install Rm= 39K = Im = 0.88mA= Dcoffset = 4mV.
With Rm = 49K = Im = 0.70mA = Dcoffset = 1.7mV
All the above tests was done with and without Ci = 220uf. Dc offset was the same with and without Ci installed.
So the 10K - 330R is a good combination for the input. In addition to that it seems the degrease of Imute current can further degrease the Dc offset.
I will try to do the same tests with another chip in order to clarify the results.
Member
Joined 2009
Paid Member
Full marks - I love this idea. The matching of the input sections on the i.c. are probably pretty good because it's a monolithic device and they likely track well with temperature.
I think this approach is going to become de rigueur for chip amps !
Makes me consider building one (Daniel has been trying to persuade for awhile).
I guess the reason this is not used often is because Voffset varies not only with each LM3886 (requiring individual trimming) but also with mute current, temperature, etc.
BTW, I get less than 4 mV output offset on the MiniRef 3886 with a JRC4580 in the outer loop and no other special trimming.
BTW, I get less than 4 mV output offset on the MiniRef 3886 with a JRC4580 in the outer loop and no other special trimming.
And the next thing is to find which two resistors will give your preferred gain setting AND 0mv offset. There's some combinations (for feedback current) that are either better or worse.
Just notice that there's a wide variety of resistor selections that could result in a gain of about 33. You're not locked in to the datasheet's suggested 22k feedback resistor.
If you can get two chips to do the exact same thing then your stereo should be pretty good; but, otherwise, the imaging could be off a bit on one side.
Also, you're going to want a speaker protector kit for when the chip has an accident, because you don't want a worn out chip to take the speakers out with it. And, DC coupled amplifiers have a long history of both shorter lifespan and sudden breakage. So, be sure to protect the speakers in some way.
The remaining problem of the DC coupled amplifier is one of inferior dynamics, but hopefully you can make up for that in the preamp or buffer circuit, and just helping the source a bit with a buffer could really help level the tone too.
Last edited:
I have a little bit of evidence, but not proper and thorough testing, that aiming for a low output offset by balancing all the input pair conditions results in lower output offset drift.
If I can build up an input pair that has an offset of 1mV, then I find it hardly changes over a big range of temperature.
But if that same LTP is deliberately mismatched then the output offset drifts a lot more.
That mismatching could be any combination of different: Ic, Ib, Vce, Vbe & Tj.
But do keep in mind my lack of proper testing !!!!
If I can build up an input pair that has an offset of 1mV, then I find it hardly changes over a big range of temperature.
But if that same LTP is deliberately mismatched then the output offset drifts a lot more.
That mismatching could be any combination of different: Ic, Ib, Vce, Vbe & Tj.
But do keep in mind my lack of proper testing !!!!
Maybe your rails are unequal dc voltage or have more noise (ac content) on one rail than the other?
Question:
Did you measure DC offset with a load connected, such as a ~6 ohm resistor in the speaker jack? The idle load of the amp would increase, symmetrically and possibly swamp the offset.
The bass on LM3886 is great, but the mids can be a pest, and without a totally thorough answer for getting rid of harmonic distortion. SO, I would suggest to start with LM1875, which is very easy, or TDA7294, which is quite good when the gain current and amount is set within the limitations of the inbuilt miller comp (aka, you'll need a preamp).
Noise on the rails is a low current voltage source and easily felled by a load on the rails, or the RC that you've got powering the mute circuit will remove some noise from just one of the rails.
We probably need to examine the power supply and scrape off some of that noise. A nice head start could be RC's across the transformer secondaries.
What bridge rectifier do you have, and do you have a CRC power supply?
Yes, I measure with a 8,5ohm resistor i had in hand. Offset with the resistor was -5,4mV.
I also change the input resistors from 330 & 10K to 680 & 22K, the offset remain unaffected.
It looks that the change of values in components or even the installation of Ci or Cin has little or no effect at the dc output, which remains negative on all the channels (4).
I also measure noise output of 0.1mV AC, as Andrew suggested in another thread
Thanks Marc and sorry for highjacking your thread, dut i still dont like the negative offset.
I assume Daniel is wright, it has to be the psu responsible for this.
Up to a +/- 100mV offset shouldn't really be a problem at the output of the amp. Maybe this is all you will get without even matching the input stage resistors.
While experimenting on a circuit I have had up to 500mV offset on the speakers with no audible change in sound AND it certainly is perfectly safe for the speaker.
I should pull out my 3886 board and see what the offsets were.
I do remember trying with no nfb cap and making no other adjustments. There was a film cap at the input. If the source is dc decoupled already then an input cap isn't required either.
However if you use a volume pot at the input of the chip amp it's varying resistance will cause the offset to change depending on position but might still be within acceptable limits. But dc through the volume pot isn't recommended.
While experimenting on a circuit I have had up to 500mV offset on the speakers with no audible change in sound AND it certainly is perfectly safe for the speaker.
I should pull out my 3886 board and see what the offsets were.
I do remember trying with no nfb cap and making no other adjustments. There was a film cap at the input. If the source is dc decoupled already then an input cap isn't required either.
However if you use a volume pot at the input of the chip amp it's varying resistance will cause the offset to change depending on position but might still be within acceptable limits. But dc through the volume pot isn't recommended.
Yes, Marc. The mixed coupled LM1875 was a large group project some years ago. Here's the basic details:
Chip LM1875, amp power caps 470u//100n, power supply board 8x 2200u per rail, transformer 18,0,18vac 4a, input load resistor 10k, feedback resistor 27k, feedback-shunt resistor 820R, input cap 2u and small signal groundlift 2R. It has a high gain of 34, usually 0mv offset and a level frequency response.
I have not made another DC coupled or mixed coupled amplifier after discovering that the dynamics are inferior.
The projects that followed, all had something very much like: 10k input load, 330R stopper, 2u input cap, 2k2 feedback-shunt resistor, 68k feedback resistor, 220u or 330u nfb cap, 2R or greater small signal groundlift.
For amplifier board power caps, TDA7294's want 220u~330u and no mute cap, LM1875's want 330u~470u, and and LM3886's want 680u~1000u and 47u mute cap. Lower figures for the amp board power caps, are facilitated by selecting a just right rail to rail cap power filter, else use the higher figures. Rail2Rail cap for LM1875, LM3886 is Mallory SEK 4.7u 250v from Allied Electronics, and just one per each amp board. TDA7293 and TDA7294 wants a 2u bargain grade high loss polyester for its rail2rail cap. That filter dumps power noise and makes more laid back mids, which eases selection efforts for all other caps. Cleaner power (and cooler amp heatsink) is always worth something.
The NFB cap took some selection effort and usually some bypass cap work to restore a level frequency response; but the extra effort there, paid off in more power, longer life, speaker protection and bigger dynamics. NFB caps are easiest to select when they're big enough for 5hz and when the input cap is not big enough for more than 10hz. Refer to AndrewT on the specifics.
However, a NFB cap does not necessarily balance the load evenly for IN+ and IN-. See the following schematic with 10k-470R load at IN+ (Q15) and 10k//4u7-470R load at IN- (Q14):
Circlophone 2012 Remix (this newer edition not yet proofed)
Both in+ and in- are balanced equally. You've got an offset trimmer circuit on the in+ side. On the in- side the 10k//4u7 has been added to the feedback loop. For sure it wasn't easier than simply using the NFB cap.
P.S.
For the LM3886 that won't stop doing DC offset even if you have it fully ac coupled (have input cap and have NFB cap), then I would suspect a ground loop as the offset causing culprit. Try making the amplifier board power umbilical cables and the small signal ground cables all exactly the same length to minimize differences that cause voltage differential. Also, try to retrofit small signal groundlift resistors to decrease the current of ground loops. A retrofit isn't as seamless or effective as strict adherence to star grounding on new builds; however, the goal of minimizing differences is the same. It is even better to decrease the number of grounds, because if there was just one ground then there's no ground loop.
I measure dc voltage directly on the pins of the chip, both rails are exactly the same (+36.2V, -36.2V), i also change the input resistors and the mute resistor and cap.
Offset remains unaffected.
My psu has a dual bridge and 47.0000uF per rail,
i install two bypass caps of 100uf & 100nf, and an RC network after that (R=1ohm, C=680nf) Cs caps directly on the chip are 1000uf.
Offset remains the same -6mV.
I also try to play some music, it sounded loud and clear.
So to summarize this i started this trying to achieve similar to Marc resulst but with a different way and i end up with a negative figure that i cant get rid of it.
Although this is acceptable i want to find the cause of it. Any advise?
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.