Another buffer conversion

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
moamps said:



Hi,

Had you looked into the matter in greater detail, you would have noticed that this circuit is an interesting example of how to get such a response (Q=2 or so) with only one reactive element (C) by simply pushing a chip beyond its frequency and phase limits (or intended use). Impracticable as it may be...

Regards

That doesn't sound like a bad short analysis to me. So there are other reactive elements, since it takes two or more. They are, if I may use a term that has arisen lately, erh... imbedded. :)

Don't forget these chips are internally compensated, with a pole starting well below 100Hz, giving rise to 90 degree phase shift up to about 200KHz, and then deteriorates above that. That 90 degrees shift (and more) is fed back into the loop. That's got to have the potential to react or inter-react with a cap in that proximity. BTW, the more severe shift is avoided by aiming for higher gain than even 20dB. I say 30-33dB gain is desirable when looking at the fact that it is close to the open loop gain at 200KHz. We have to limit feedback to within the 90 degrees, IMHO! Anyway, that has been my thinking all along.

Could some of you guys try it out, set it up to 20dB, listen to it, then up it to 30-33dB and compare it. I would be grateful.

Conclusion? So can we rule out transposing the two resistive elements? I think so! Unless we can find that ideal opamp.

Joe R.
 
Pedja said:
Hi Joe,

I checked a few chips more using the same circuit and haven’t had success. It is shown behavior of chips without the shunt cap and with 68pF, 680pF, 6.8nF and 68nF caps.

Note that the OPA chips (having shorter bandwidth) set to gain of 20 anyway have roll-off inside the audio band.

Pedja


Pedja said:

I did look at 549 and you are right, it has less bandwidth, the comp pole is below 1Hz? I think this is some what lower than 1875, 3875 etc. Also 90 degree phase is only maintained up to less than 100KHz compared to almost 200KHz in 3875.

Look at the diff in open loop gain @ 20KHz, 3875 circa 50dB & OPA around 33dB. Not enough feedback to keep it flat @ 20KHz at 20dB gain. I think the ultra low comp is the reason. You just don't know what the true open loop is. Looking at the phase response, flat to only 90KHz? 1875 flat to near 200KHz?

Can I ask a favour? Try with the 1875 model you have there, set gain to 33dB, is it flat to 20KHz (17dB feedback)? I would like to know?

Joe R.
 
Pedja said:
LM1875 and gain set to some 33.4dB (feedback 470k/10k) and 20kHz roll-off is 0.135dB (33.425 - 33.290), matching phase shift at 20kHz is some 10 degrees.

Pedja

I suspected as much. Thanks. The extra open loop gain means that at my preferred increased gain, the response is just about flat to 20KHz.

Is it possible for you to compare the sonic performance at this higher gain, that is putting 470K in your Fet-IGC? This is one aspect that I am yet to get independant feedback. Buffer, yes. LPF, yes. Higher gain (lower feedback), not yet.

Thanks again.

Joe R.
 
Hi Joe,

I had a plan to check this, at least because I in fact need more gain with 0.8VRMS DAC’s output, but to change Rfb I should dismantle almost entire amp (just look at the pic of my amp, and it is the only GC I have at hand, what a digging it would be…). I do have a possibility to change input resistor, FET shouldn’t have some particular problems to drive, say, 4.7kOhms, but… not a way to get objective insight, IMHO.

But this is for sure in my “to do” list. Not a promise, but I’ll probably have something to report about this soon.

Btw, I did checked the stability (of LM1875) with 220k and 470k Rfb driving 1uF in parallel with 8 Ohms load, ringing at the output is the same in the absolute level, which means it’s relative level is halved in the case of the 470k Rfb; also, the second ringing cycle is somewhat better damped in that case.

Pedja
 
Pedja said:
Hi Joe,

I had a plan to check this, at least because I in fact need more gain with 0.8VRMS DAC’s output, but to change Rfb I should dismantle almost entire amp (just look at the pic of my amp, and it is the only GC I have at hand, what a digging it would be…). I do have a possibility to change input resistor, FET shouldn’t have some particular problems to drive, say, 4.7kOhms, but… not a way to get objective insight, IMHO.

But this is for sure in my “to do” list. Not a promise, but I’ll probably have something to report about this soon.

The 4K7 should work, but coupling cap might need to be bigger? When you can, thanks.

Btw, I did checked the stability (of LM1875) with 220k and 470k Rfb driving 1uF in parallel with 8 Ohms load, ringing at the output is the same in the absolute level, which means it’s relative level is halved in the case of the 470k Rfb; also, the second ringing cycle is somewhat better damped in that case.

Pedja

I would consider reasonably positive, thanks again.

Now it's near 12 mid-night and my bed is awaiting my presence. :sleep:

Don't you just love those little smilies.

Joe R.
 
AX tech editor
Joined 2002
Paid Member
moamps said:


Hi, Jan

From somewhere :"Ringing and overshoot are the first signs of instability, and oscillation is the final sign, ..... "
There is no discrete reactive elements around, but something is ringing, like bells...... ;)
This Pedja's picture show some resonance..IMHO
I saw this on my scope today too (network 10k-6n8-680R-220kFB)
Overshoot was on ca 23kHz with only 2dB (LM3875)

Regards

The point I was trying to make a bunch of posts ago was that this has nothing to do with resonance. The LPF used, by Pedja I think it was, simply decreased the feedback factor at high frequencies, and therefore, as expected, the gain at high frequencies. Since a square wave has a lot of frequencies, when you increase the gain of the higher freqs, you see overshoot. Plain old circuit theory, nothing at all to do with resonance.

Since the chip also runs out of excess gain at high freqs, there is ultimately a limit to the max gain you will see. On the other hand, the higher gain at higher freqs does improve the stability, (you move closer to open loop), also standard plain old opamp theory. So, the peaking of the square wave doesn't indicate less stability, on the contrary, the limited peaking with the lower feedback factor shows improved stability.

I probably didn't do a good job of explaining all this, sorry about that. But there's so much to say and so little time....

Jan Didden
 
Joe Rasmussen said:
The 4K7 should work, but coupling cap might need to be bigger?

IMO there is one important problem here: at the moment I have 1k/4.7nF low pass filter (and btw 33nF before that – despite of my doubts, it works – so in total, 2nd order passive filter :nod: , I’ll post more about this soon). Swapping input resistor for 4k7 (instead of 9k I have now) I will get somewhat different roll-off curve. Not drastically different, but different enough for possible misleading conclusions. So I decided to go with Rfb swapping in another GC (soon).

Pedja
 
_________________________________________

Found in OP27 datasheet: "...a pole is created with Rf and the amplifier’s input capacitance, creating additional phase shift and reducing the phase margin."
_________________________________________


Attached screenshots show the response if one of three resistors in circuit is varied. Where not varied, values assumed are: for the first input resistor 10kOhm, for the second input resistor 680 Ohms and for the feedback 220kOhms. Shunt cap is 6.8nF. Opamp is LM1875.

Pedja
 

Attachments

  • lm1875_stepping.gif
    lm1875_stepping.gif
    12.9 KB · Views: 371
Pedja said:


IMO there is one important problem here: at the moment I have 1k/4.7nF low pass filter (and btw 33nF before that – despite of my doubts, it works – so in total, 2nd order passive filter :nod: , I’ll post more about this soon). Swapping input resistor for 4k7 (instead of 9k I have now) I will get somewhat different roll-off curve. Not drastically different, but different enough for possible misleading conclusions. So I decided to go with Rfb swapping in another GC (soon).

Pedja

OK, that is probably the wisest. I suppose it's not just about gain but really what is the optimum amount of feedback. Let the gain lie where it falls. Like I believe you said, you can use the extra gain too. My SACD/CD player is about -11dB below 2V RMS, so 33dB is handy.

Joe R.
 
Pedja said:
_________________________________________

Found in OP27 datasheet: "...a pole is created with Rf and the amplifier’s input capacitance, creating additional phase shift and reducing the phase margin."
_________________________________________

And after all the fuss, I thought it won’t be bad to post this. It is from LM6172 datasheet.

Pedja
 

Attachments

  • lm6172_ds.gif
    lm6172_ds.gif
    11.2 KB · Views: 279
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.