Thanks for bringing us back to the core of the issue. This has been an interesting conversation where I have learned a lot, but it also seems clear that there isn't an alternative that doesn't bring it's own problems.
I agree with KSTR....
With an offset adjustment on LM318 you could even avoid C9 altogether.
All you need is check the DC out from time to time and adjust acordingly.
This is a DIY project and not a commercial offering, that should work all over the globe. And it even has a well functioning DC sensing circuit.
If you have your amp in a place without excessive temperature swing, you should be ok.
I have in a number of other amps , removed (shorted it to be more correct)this cap without any problems but I still have to try it in this.
The same is true IMHO with the input capacitor. Not needed and I still dont think the amplifier sounds better with one installed.
Koldby
With an offset adjustment on LM318 you could even avoid C9 altogether.
All you need is check the DC out from time to time and adjust acordingly.
This is a DIY project and not a commercial offering, that should work all over the globe. And it even has a well functioning DC sensing circuit.
If you have your amp in a place without excessive temperature swing, you should be ok.
I have in a number of other amps , removed (shorted it to be more correct)this cap without any problems but I still have to try it in this.
The same is true IMHO with the input capacitor. Not needed and I still dont think the amplifier sounds better with one installed.
Koldby
I would agree. Remove C9, use offset trim and should you find the output DC too drifty for your tastes, replace trim by a DC precision JFET opamp like OPA132 as standard integrator feeding via 1M or higher into one offset pin, loading the other with equal resistance to GND. Tune those R's for reasonable pull range, then tune corner freq. The closed loop gain the servo sees from the main amp will be very low which will result in a rather fast time constant for the integrator, this helps for using a much smaller integrator cap that original C9 which is what we're after. There could be issues, i.e. maxing out the servo at very low freq as max pull range must be kept rather low, maybe less than 10mV (input referred). Shorting input C or making it to large would be notably risky, therefore.
If you want microvolts, add a manual offset trim to the servo opamp ;-)
that's nothing but a different way to build a "servo" (that is, an automatic control system). Only much more complicated...
in principle, you're right. The offset is small, and in most cases it's definitely not a problem. Even without C9.
For a while I have personally listened to a my_ref with C9 short-circuited (and neither a servo nor any manual offset trimming). It have been used that way for months without any problem.
Just adding manual trimming may make that solution viable even for people using the my_ref with delicate high-efficiency speakers.
Yet, eventually we find out that (for some unclear reasons) our my_ref sounded better with a good ElCap for C9 rather than with a short circuit in its place! (go figure...).
Now, we know that the original designer decided to use a servo for his own "Evolution". Perhaps he had some good reasons to do that.
Comparative listening tests indicates that while the my_ref Rev.E/FE (with proper BOM) sounds even better than the Evo overall, the "Evolution" still sounds a bit better in the bass department.
Of course, there could be a large number of different reasons for this (and there have been yet no similar comparison of a a my_evo with a Rev.E/FE with C9 short-circuited to exclude the most obvious), but until further tests I can not dismiss the suspect that the servo may have it's own role in this result...
sure. See above.
indeed. But only as far as the source connected to it is AC-coupled or at least have no DC offset of its own...
Last edited:
Of course, but who uses sources with DC out? It must be defective then or badly designed.
Koldby
That's true if you are thinking about large DC offsets. But small ones (<0.1V) are fairly common in sources and preamplifiers with DC-coupled output. At least a few mV are almost unavoidable in such case.
It always puzzles me why people are afraid of DC offset.
Even a several 100mV offset at the output of a poweramp is of almost no consequence. In some instances it could even be a benefit as the zero crossing is not @ zero output and if the amp has a small problem @ zero crossing, it could be masked. Sure you can get a little pop when turned on, but who cares?
Even some loudspeakers can benefit from a little DC bias.
Of course if you have a source with 100mV dc offset, it will be amplified to an unacceptable value, but it is very rare. A few mV at the source is IMHO not a problem.
Koldby
While most speakers can indeed tolerate some 100mV of DC offset without problems, I can't see how displacing the coil off it's center (effectively reducing the max available displacement) could be beneficial. Not to mention some delicate high efficiency drivers which can be easily damaged even by a relatively small DC offset.
For what regards the offset at the amp input, the problem is that it will alter the biasing of the input stage. In some designs, such as the one we are talking about, that is extremely critical and have a deep influence on the overall sound quality.
For what regards the offset at the amp input, the problem is that it will alter the biasing of the input stage. In some designs, such as the one we are talking about, that is extremely critical and have a deep influence on the overall sound quality.
Have you tried to listen if you can hear a difference with 100mV DC at the output, even in this amp?
Actually you can improve the sound of some opamps by placing a current source at the output to shift the zerocrossing.
But I think we have a differnce of opinion here, that is not easily solved, so maybe we should just leave it there. Peace.
Koldby
Actually you can improve the sound of some opamps by placing a current source at the output to shift the zerocrossing.
But I think we have a differnce of opinion here, that is not easily solved, so maybe we should just leave it there. Peace.
Koldby
If you've read the previous posts you know that I have listened to a my_ref without C9 (which end up with an offset more or less of that order of magnitude).
With the (mid/low sensitivity) speakers used that was not a problem. But that's just one particular case. I would not connect it that way to something like, say, a delicate and expensive Lowther speaker (I would not use any DC-coupled amplifier with such a speaker anyway, but there are people who do).
BTW, I have to say that my main concern here is at the input, rather than the output. We need to avoid C9 without altering too much the biasing of the input differential of the LM318 which have been chosen by Mauro. He clearly stated (and it's not hard to believe) that the biasing of the input differential of the LM318 is extremely sensitive to DC bias changes and does have a deep impact on the overall SQ.
well, that's a whole different thing. That way you are shifting only the output stage, not the whole OpAmp. By doing that (properly), you are basically using the output stage like a "single-ended" class A stage (basically using only one-half of the push-pull output stage). Something which may have its advantages in terms of linearity, but of course also drawbacks (you drastically reduce the OpAmp output dynamic...).
OTOH, you will never want to "unbalance" the whole OpAmp, particularly the input differential (which is extremely critical, being the heart of the NFB mechanism. And it's very sensible and delicate, being designed to operate only with the very small signal levels of the NFB "error" signal..).
of course, always!
I only tried to explain a different POW: I never intended to be polemic, let alone aggressive. Please forgive me if I gave you that impression.
Last edited:
I know you dont want to get polemic, but this could be a never ending exchange of opinions, thats why I wanted to stop now. If you had read my comments, you would know I found that My-ref sounds better without C9.
I do NOT agree with you that the input bias is such a big deal.
A long tailed pair is just as happy with a little difference in one of the legs as long as it dosent upset the hole thing, but that is not the kind of offset we are talking about here.
In the case of the opamp with a currentsource at the output, What do you think makes the DC out @ 0 V? The feedbackloop.... So you upset the input bias in order to maintain 0 at the output...
But I think this is getting OT that is why I wanted to stop it. I could go on forever, but I don't think it will interest anyone else.
Koldby
I do NOT agree with you that the input bias is such a big deal.
A long tailed pair is just as happy with a little difference in one of the legs as long as it dosent upset the hole thing, but that is not the kind of offset we are talking about here.
In the case of the opamp with a currentsource at the output, What do you think makes the DC out @ 0 V? The feedbackloop.... So you upset the input bias in order to maintain 0 at the output...
But I think this is getting OT that is why I wanted to stop it. I could go on forever, but I don't think it will interest anyone else.
Koldby
oh, I see. IME I've got to the opposite conclusion... that's likely why we end up with such a different POW.
Mmmh, here is an interesting idea to add the offset null pins to the SPICE model of (any) OpAmp:
Re: OpAmp Vos Trim (Null) Doesn't Work (pins 1 & 8) - Discussion Forums
Of course that's far from accurate, and possibly will screw up a few other things. But perhaps it can be used here to help designing the servo... :-?
P.S.: some of the argument previously discussed in this topic can be found in this paper: Analog Devices tutorial MT-037: Op Amp Input Offset
Re: OpAmp Vos Trim (Null) Doesn't Work (pins 1 & 8) - Discussion Forums
Of course that's far from accurate, and possibly will screw up a few other things. But perhaps it can be used here to help designing the servo... :-?
P.S.: some of the argument previously discussed in this topic can be found in this paper: Analog Devices tutorial MT-037: Op Amp Input Offset
Last edited:
Mmmh... though by no means realistic, can we use the collectors of Q1 and Q2 to (roughly) model pin 1 and 5 in the 318?
http://www.linear.com/docs/4196
http://cds.linear.com/docs/en/software-and-simulation/LM318.txt
http://www.linear.com/docs/4196
http://cds.linear.com/docs/en/software-and-simulation/LM318.txt
Code:
* Version 2.0 Copyright � Linear Technology Corp. 10/19/04. All rights reserved.
*
.SUBCKT LM318 3 2 7 4 6 ;(+IN -IN V+ V- OUT)
* INPUT
RC1 7 80 7.074E+02
RC2 7 90 7.074E+02
Q1 80 2 10 QM1
Q2 90 3 11 QM1
*
C1 80 91 300E-12
RXC1 91 90 1E3
CXC1 91 90 15E-12
C2 1 8 5.000E-12
*
RB1 2 102 1.0000E+00
RB2 3 103 1.0000E+00
DDM1 102 104 DM2
VZ1 104 103 5.5
DDM2 103 105 DM2
VZ2 105 102 5.5
RE1 10 12 6.209E+02
RE2 11 12 6.209E+02
IEE 12 4 6.000E-04
RE 12 0 3.332E+05
CE 12 0 2.632E-13
*
GCM 0 8 12 0 1.414E-08
GA 8 0 80 90 1.414E-03
R2 8 0 1.000E+05
GB 1 0 8 0 5.318E+01
RO2 1 0 7.4000E+01
*
RS 1 6 1
ECL 18 0 1 6 3.172E+01
GCL 0 8 20 0 1
RCL 20 0 1E3
D1 18 20 DM1
D2 20 18 DM1
*
D3A 131 70 DM3
D3B 13 131 DM3
GPL 0 8 70 7 1
VC 13 6 3.6473
RPLA 7 70 1E4
RPLB 7 131 1E5
D4A 60 141 DM3
D4B 141 14 DM3
GNL 0 8 60 4 1
VE 6 14 3.6473
RNLA 60 4 1E4
RNLB 141 4 1E5
*
IP 7 4 4.400E-03
DSUB 4 7 DM2
* MODELS
.MODEL QM1 NPN (IS=8.000E-16 BF=1.818E+03)
.MODEL QM2 NPN (IS=9.347E-16 BF=2.222E+03)
*
.MODEL DM1 D (IS=1.000E-19)
.MODEL DM2 D (IS=8.000E-16)
.MODEL DM3 D (IS=1.000E-20)
.ENDS LM318
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.