Pjotr said:Maybe it is already said and have I missed it in this long discussion:
If you use the inverting topology for audio the op-amp MUST be unity gain stable!
You can get rid of this by placing a sufficient low resistor at the input, according to the attached schematic, to keep closed loop gain above 20 dB under all input conditions.
The 'proper' way to do this is to hang that extra 10k resistor directly between the + and - input terminals of the op-amp to set the noise gain. Doesn't anyone read the app notes for basic op-amps anymore? You can get a LOT of design ideas studying those old data books.
Overall, the inverting configuration is considered 'better' (from a theoretical standpoint) because the common-mode voltage is zero regardless of closed loop gain and input voltage. If the input stage LTP uses just a resistor for bias instead of a CCS this advantage is not insignificant - it will result in better linearity. Some op amps are also better behaved with zero CM voltage.
Every time you mention inverting phase or phase
splitting to the semi technical audiophile, they go into
shock.
I have another hypothesis which occurs when the stage
is not being used for much gain:
It is my experience that at low and unity gains, a lot of
op amps are on the edge of oscillation. Over the years
I have made a number of bad sounding circuits (no, not
ones you've seen) and many of them were bad sounding
not because they were oscillating, but because they were
near the edge. If I opened up the feedback gain, they
got better as they moved farther from the edge.
As a result, whenever I use op amps in an audio circuit, which
is not often, I make sure that I either use up or throw away
about 20 dB of open loop gain. If the circuit has a gain of
10 or greater, this is automatically achieved.
If the circuit has less than 20 dB gain, I toss some away.
With an inverting amp, a resistor from the - input to ground
will do it. If not, then resistance from both + and - to
ground is used.
Works like glue....
I know this thread is 10yrs old now but this comment (and others) got me thinking.
How valid a technique is this as a method of compensating an opamp ?
I have posted some real scope shots and measurements here,
http://www.diyaudio.com/forums/anal...opamp-compensation-technique.html#post2707345
Thanks...
I see the corners getting sharper on your scope, so it seems that you're repairing the inverting amp dullness problem? Oh, good job!
Could you do me a huge favor while you've got that scope and a handy power op amp?
It would be:
Put two of 220uF (one on each rail--low impedance, low ESR types) and one of 2uF (solo cap across rails--250v ecap or 60v~100v polyester dip) and two of 3.3nF (one on each rail) for power smoothing, all at very close proximity to the amp. It may be necessary to either move any ringingly large caps away from the amp or select low impedance types. That amp could be non-inverting STK, ST or Inverting NatSemi. The scope may show something very interesting if there was before and after photos. Disclaimer: The little power circuit example is great for the amp that needs it but terrible for the amp that doesn't.
Could you do me a huge favor while you've got that scope and a handy power op amp?
It would be:
Put two of 220uF (one on each rail--low impedance, low ESR types) and one of 2uF (solo cap across rails--250v ecap or 60v~100v polyester dip) and two of 3.3nF (one on each rail) for power smoothing, all at very close proximity to the amp. It may be necessary to either move any ringingly large caps away from the amp or select low impedance types. That amp could be non-inverting STK, ST or Inverting NatSemi. The scope may show something very interesting if there was before and after photos. Disclaimer: The little power circuit example is great for the amp that needs it but terrible for the amp that doesn't.
Last edited:
With a non-inverting amplifier, a frequently ignored topic is the current in the feedback resistor and the feedback shunt resistor.
When observed at exactly the same gain setting. . .
It seems that when the feedback is a mirror image or "sample" of the output, you can simulate overrun or underrun conditions via changing the current in the feedback.
The sound effect differences? Maybe that is from changing the current.
P.S.
Overrun non-inverting amplifier, even slightly, even just overruning the pre, not the whole thing, will make the support circuit look really crazy or normal values not working as expected. For example, the op-amp seeming to need really really big caps right at it or else the sound is bad? Well, that thing is suffering overrun. But, that's a typical gainclone.
When observed at exactly the same gain setting. . .
It seems that when the feedback is a mirror image or "sample" of the output, you can simulate overrun or underrun conditions via changing the current in the feedback.
The sound effect differences? Maybe that is from changing the current.
P.S.
Overrun non-inverting amplifier, even slightly, even just overruning the pre, not the whole thing, will make the support circuit look really crazy or normal values not working as expected. For example, the op-amp seeming to need really really big caps right at it or else the sound is bad? Well, that thing is suffering overrun. But, that's a typical gainclone.
Last edited:
Mooly,
Do you mind if i ask whether you pursued this concept and if so with what results?
The honest answer is that I wasn't entirely convinced by the sonics when listening for an extended listening trial of several days daily use but I didn't try incrementally small changes.
Since writing that thread I have become a champion of the LM4562 opamp which seems to me to be pretty much perfect when used correctly.
Mooly is using that to fine tune his op-amp based preamp.
His post on multi-compensation, aka supplementary compensation, is interesting because, technically noise gain would worsen the THD performance; however, the part used was possibly insufficient on tonal/stability performance and very high spec on THD performance (it is a high-end part). This rather impractical imbalance was corrected with compensation. The op-amp then performed more practically without taking sizable hit on THD performance. Conveniently, he shows how a "fine tuning" dial can be added.
P.S.
On that thread, I also discussed a very similar sounding effect from the LTP soft clipper, and it was a bit hard work to defeat it because the tone went a little forwards on the orthodox simpler version. Conceivably, if one wanted the compensation without the limiter function, either a capacitor or a short could be used instead of the diodes (no idea which--needs simulator). The main fb shunt resistor can be a multi-turn potentiometer (the soft clippy circuit hooks up to wiper). The series 470 can be a variable resistor. I had added that 470 to actually defeat the (accidental) compensation that had (accidentally) made blockbuster big imaging at the expense of a slightly more forward tone.
That sort of circuit is supposed to bring an LM3886 up short before it sounds off with the spike system or bring a discrete amp up short before it loudly sticks its outputs. However, it does also decrease the gain, and thus a compensation difference (controlled by that series 470, or rather audible in the orthodox version that doesn't have the extra series resistor).
So, there's something else to play with.
His post on multi-compensation, aka supplementary compensation, is interesting because, technically noise gain would worsen the THD performance; however, the part used was possibly insufficient on tonal/stability performance and very high spec on THD performance (it is a high-end part). This rather impractical imbalance was corrected with compensation. The op-amp then performed more practically without taking sizable hit on THD performance. Conveniently, he shows how a "fine tuning" dial can be added.
P.S.
On that thread, I also discussed a very similar sounding effect from the LTP soft clipper, and it was a bit hard work to defeat it because the tone went a little forwards on the orthodox simpler version. Conceivably, if one wanted the compensation without the limiter function, either a capacitor or a short could be used instead of the diodes (no idea which--needs simulator). The main fb shunt resistor can be a multi-turn potentiometer (the soft clippy circuit hooks up to wiper). The series 470 can be a variable resistor. I had added that 470 to actually defeat the (accidental) compensation that had (accidentally) made blockbuster big imaging at the expense of a slightly more forward tone.
That sort of circuit is supposed to bring an LM3886 up short before it sounds off with the spike system or bring a discrete amp up short before it loudly sticks its outputs. However, it does also decrease the gain, and thus a compensation difference (controlled by that series 470, or rather audible in the orthodox version that doesn't have the extra series resistor).
So, there's something else to play with.
Last edited:
indeed, will readily agree with that.
my thought was to use the technique so as to be a bit less heavy handed in the capacitive element of power amp NFB. Start with, for example the shine7 BPA-300 (below -- one half of the bridge shown), then lighten up the capacitive compensation and place something like 10k2 from -IN to +IN. It's a somewhat crude example but you get the idea.
To lighten up the capacitive compensation so it isn't as heavy handed, perhaps a resistor series with C3 (that may also need a different value for C3).
yes, that's one of the aspects of it that i had in mind when i referred to as somewhat crude. my source is balanced and designed to feed a 600 ohm load, so in my application a BPA300 could have a very low feedback shunt R allowing for a very reasonable feedback series R and lots of flexibility in how i handle compensation in the NFB loop.
To lower the impedance levels, you can use a T network for the feedback resistor.
From the output, a 10k and 499 in series to ground, and a 10k from their junction to the inverting input. This will give close to 20x gain.
yes, i must say that this works quite well and i had used it in both of my unbalanced (ie:RCA input) chipamps before i switched them to BIGC (ie:buffered). it's slightly noisier than a conventional design but well worth it if you want to avoid the buffer yet need a high-Z input. i used 10k1 -- 681//1k -- 10k1 all 1% RN60s.
Last edited:
Yes, the best way to do this is to make the T resistor connecting to the negative input as large as possible, and the other two as small as possible.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.