Hi All,
is there a way to make the National inverted Chipamp circuit, unity gain stable?
Can the stability margins be adjusted to allow various reactive loads to be driven?
Is there a universal method that applies to the whole range of National Chipamps, or does each implementation need it's specific topology/values to ensure that stability?
Can any of the above apply to the Non-Inverting versions of the National Chipamps?
is there a way to make the National inverted Chipamp circuit, unity gain stable?
Can the stability margins be adjusted to allow various reactive loads to be driven?
Is there a universal method that applies to the whole range of National Chipamps, or does each implementation need it's specific topology/values to ensure that stability?
Can any of the above apply to the Non-Inverting versions of the National Chipamps?
Hi Andrew. Compensating an amplifier implies to make it unity gain stable. Niquist criterion tells that an amplifier can oscillate when closed loop is 0 or n * 360deg phase rotation and gain is 1. As you know, all kind of amplifier has internal phase rotation caused by stray capacitances, inductances, and resistances (The last indirectly). So, first you must do that the amplifier will be unity gain stable per se, and then adjust it to the load it will drive. Example: suppose that an amplifier is internally compensated, like, TDA2003. When you attach a reactive load (a moving coil, a transformer, a line), they may be capacitive or inductive at some frequencies. With the internal output resistance, it forms a phase rotation that may cause amplifier to get oscillations, this is the why, the zobel network. So, you must compensate the amplifier itself, and then with the load. Then, give to you a cookbook to this situation is too difficult.
Seeing as a chipamp is a decompensated power opamp, you might find this application note helpful : http://www.ti.com/lit/an/snoa486a/snoa486a.pdf
a problem is how slow audio power chips amps are - the GBW of the audio power chips amps is speced with 4:1 range, Lm3886 GBWmin = 2 MHz / Gain_min 10 = 200 kHz
conservative engineering from the data sheet # puts the noise gain (zero) corner frequency barely above audio
a detail of the noninverting noise gain circuit is that you need low Z at the non-inverting input at all frequencies that the compensation works over - 20 KHz to beyond 2 MHz for the compensation curve to be well defined
conservative engineering from the data sheet # puts the noise gain (zero) corner frequency barely above audio
a detail of the noninverting noise gain circuit is that you need low Z at the non-inverting input at all frequencies that the compensation works over - 20 KHz to beyond 2 MHz for the compensation curve to be well defined
I am using inverted LM3875s and LM4780s all the time, with gain just a bit over -1. No problems so far. At the moment I am listening to pretty basic P2Ped 2 x LM3875 with P2Ped 6N1P preamp. No zobel (though short speaker wires used). Both ICs on CPU heatsink (which is small, I know), but it is really just warm, not hot.
wrong - do you have a 'scope?
the stability gain is the "noise gain", for inverting amps it is 1 + |Av|
the LM4780 datasheet open loop frequency response plot shows that it is possible that the shown (presumed "typical") response could be stable at Av = -1, noise gain of 2 (look for the phase margin at the ~ 6dB gain interecpt frequency)
but the "worst case" chips, or with heavier capacitive loads you could expect to experience oscillation - given that the manufacturer spec on GBW has a 4 :1 range considerable conservatisim is recommended
engineers follow the manufacturer's recs, with margins to avoid problems in production, over time, with differing batches of parts
diy builders can violate the limits - but it is a good idea to understand what you are doing, have the equipment, know how to make the verifying measurements
it is irresponsible to encourage other diyers who may not have the skills, equipment that such practice is "safe" given your own necessarily limited experience with the part
Last edited:
I am not encouraging at all.
I am just saying that it works for me. And it is not like only a pair of chips. I have built LM4780s with gain slightly less than 10, LM3875s, also tried paralleled config with low gain. In all these amps I used quite low value resistors, short wires, low impedance input source, P2P method, that probably affects performance a bit too, or maybe it is just a luck.
I am just saying that it works for me. And it is not like only a pair of chips. I have built LM4780s with gain slightly less than 10, LM3875s, also tried paralleled config with low gain. In all these amps I used quite low value resistors, short wires, low impedance input source, P2P method, that probably affects performance a bit too, or maybe it is just a luck.
In this datasheet from National Semiconductor, there's some unity gain formulas that I can't read but the schematics are really easy. http://www.ti.com/general/docs/lit/getliterature.tsp?genericPartNumber=lm675&fileType=pdf
Perhaps the circuit for LM3886 will look more like 3 steps forward and 3 steps back, resulting in unity gain, but that is a guess.
Perhaps the circuit for LM3886 will look more like 3 steps forward and 3 steps back, resulting in unity gain, but that is a guess.
Hi all. Some time ago I tried with a LM4780 in inverted mode, 30dB gain. The chip started to oscillate heavily in the moment I removed the source from input. I found the explanation later: when the inverting input is in the air, the gain wasn't 30dB anymore, but 0dB. NFB loop is only between chips output and inverting input, affecting also the bias of the input stage. The point is that in inverting mode the source must be always connected and have low output impedance.
These days with an inverting topology (which is my preferred mode of operating chipamps) I put a series RC between the -ve in and GND to ensure that even with no source connected, the chip sees a minimum noise gain at HF.- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.