The current pump relies on accurate matching of the ratios R8:R9 to R5:R6
R3 is the change made for the improved current pump. This now requires R9:{R3+R8} to be ratio matched.
R3's precise value is not critical to the proper operation of the current pump.
http://www.ti.com/lit/an/snoa474a/snoa474a.pdf
Note the 1% tolerance for R13 in figure 6.
I wonder if the slight change in R3 resulted in a more accurate ratio match?
Could that have given rise to the apparent improvement?
Last edited:
Andrew,
I have been explaining in the wider context, with the current pump and it's amplification factor as part of the composite amplifier. R3 is the single most important part in defining the current pump's transconductance.
Then this transconductance is working on the actual load impedance.
By changing R3 one changes the open loop gain.
Ciao, George
I have been explaining in the wider context, with the current pump and it's amplification factor as part of the composite amplifier. R3 is the single most important part in defining the current pump's transconductance.
Then this transconductance is working on the actual load impedance.
By changing R3 one changes the open loop gain.
Ciao, George
I have been trying to follow this progression of mods, but I still am not clear on a couple things.
Can R3 be changed from .47 to .33 to make an improvement without any other component changes?
Can the comp parts be removed and C32 value be changed without changing out the LM318? I do not object to leaving C9 in the circuit. I realize that is the ultimate goal of all this modification, but I have no problem listening to music with C9 in the circuit. The last time I messed around with C10 and C34 values according to Dario's recommendations, I did not like the results at all. I don't want to go back through all that or start another debate about comp values. Does simply removing those comp parts make the amp sound better? In what way? Then why were they there to begin with? Because of the LM318?
Peace,
Tom E
The Howland Current Pump relies on an accurate ratio match of the 4 resistors, ("ratios R8:R9 to R5:R6").
The improved current pump adds in a fifth resistor, ("requires R9:{R3+R8} to be ratio matched").
That must now be included in the ratio matching.
If you change the fifth resistor value, then you unmatch the ratio and you NEED to RETRIM.
Last edited:
The starting point for all mods is to apply My_Evo Rev. A compensation which means, as stated in build guide:
- Change R3 from 0.47R to 0.33R
- R39 not populated
- C10 not populated
This configuration works technically better according the original designer Mauro Penasa and sounds better according all people that actualy tried it.
After My_Evo Rev. A mod is applied, to change opamp you will need to remove C34 and change C32 (OPA827 only from what I've understood, Joseph K could help here)
Are you serious Andrew?
The resistor at the denominator is 47KOhm... and a change of 0.14Ohm should make a difference?
Last edited:
Yes. I have just taken the smaller, 22kohm value of the bridge as reference. The 0,14 ohm shift with all values rounded up results in a 1:100000, that is 0.001% change in the balance of the bridge.
Now the prescribed precision for the bridge resistors is 0.1%.
That is 100 times, two orders of magnitude higher uncertainty than that of caused by the change in tbe current sensor element.
Andrew should go home and revisit his homework.
Now the prescribed precision for the bridge resistors is 0.1%.
That is 100 times, two orders of magnitude higher uncertainty than that of caused by the change in tbe current sensor element.
Andrew should go home and revisit his homework.
Yes, I am serious. Go back and look at the Howland Current Pump notes. These show the use of tighter tolerance resistors for the ratio setting AND they put in a low value trimmer to allow precise tuning of the RATIO. This is very important.
I will go further:
The final tuning of this ratio should be done on the assembled PCB. This requires the resistors to be fitted and soldered and tapping points attached to apply voltage and take readings of the RATIO match of the assembly
I attempted to describe this years ago in a Rev C Thread. Few took any notice.
Last edited:
I remember your posts about it and since them I've always found your approach at least extreme and not needed in this application of the HCP, 0.1% resistors are enough to ensure stability and, as calculated by Joseph K, at worst the change from a different R3 is 0.001% way under the specified tolerance.
Andrew, I think there are loads of errors in your logic here..
I would quote the article linked at by You:
What we can see here that they are principally suggesting to use 0.1% resistors (and this is what we are doing here!) and maybe, "even" 0.01%... So even the author knows the impossibility of that mission.
Then, what you miss completely is the output impedance of the bridge. The levels of precision are calculated for getting at least 250kohm output impedance.
Mauro instead very clearly had defined his target for output impedance: 500ohm.
So, instead of irrelevant studies not satisfying for our application at hand, what if You would go back and re-read Mauro Penasa?
And, finally, I have to say I really feel sorry for being polemic with You but I feel an extremely stubborn resistance from your part.
Stubbornly being in the wrong.
Ciao, George
I would quote the article linked at by You:
What we can see here that they are principally suggesting to use 0.1% resistors (and this is what we are doing here!) and maybe, "even" 0.01%... So even the author knows the impossibility of that mission.
Then, what you miss completely is the output impedance of the bridge. The levels of precision are calculated for getting at least 250kohm output impedance.
Mauro instead very clearly had defined his target for output impedance: 500ohm.
So, instead of irrelevant studies not satisfying for our application at hand, what if You would go back and re-read Mauro Penasa?
And, finally, I have to say I really feel sorry for being polemic with You but I feel an extremely stubborn resistance from your part.
Stubbornly being in the wrong.
Ciao, George
Last edited:
Just to clarify this point, the aim of this modification is to replace the LM318 with a modern, higher performance part. Doing this results in a clear step forward in the sound. Linking out C9 is a side benefit that's worthwhile, but not as large an improvement as the main opamp substitution..
Tom, if we were talking about technical aspects, it would be easy(*) to show the results here, as George have already done.
But! How could anybody else tell you whether you will or will not like the results, subjectively? Clearly, there's no other option than to try it and judge yourself.
FWIW, I have listened to many successive "evolutions" of George's FE boards and -IMHO- these mods are definitely well worth it. A lot of quite sensible improvements, on all the "parameters" of perceived SQ.
(*) it's gettin' such results which is not easy at all... but that's another story. Fortunately we have George for that.
in short, yes. Though there are many ways to skin a cat. Mauro himself did change those details several times, while keeping the LM318. The "Evo Rev.A" have been introduced by him here:
ARCHIVIO SCHEMI E PROGETTI AUDIO -> Evolution Revisione A.pdf
yes, the change of the main compensation cap (C32) value is required only for some OpAmps, among which OPA827.
The ADA4627 (one of the recommended types) will work just fine with exactly the same ("Evo Rev.A") compensation originally designed for the LM318.
(OTOH, yet some other - currently non recommended - OpAmp types may require even deeper changes to the compensation scheme to work in the circuit).
yes, this would be quite important... should your goal be that of producing a true voltage controlled costant current source. That is, to obtain an output impedance as high as possible. But that is not the goal in this application!
I would (partially) disagree here. Indeed, the OpAmp replacement was a goal in itself, and it constitutes a quite sensible improvement (both WRT measurements and perceived SQ).
Yet, IMO/IME, gettin' rid of C9 also provides a "huge" improvement in terms of SQ. YMMV.
Last edited:
The 4898 could be yet another good option but, unfortunately, there are some drawbacks. First of all, to get best performances (surely in terms of SQ, and likely also in terms of measurements) you need to pair it with a very good (active) preamp or buffer. Which must have DC-coupled output, with no (very little) offset and very low output impedance (also at DC). That's the only way to keep a reasonable "balance" of the impedances seen by the two inputs of the OpAmp, allowing it to work at its best (and to avoid C9).
In terms of SQ, I have listened to the FE he have assembled with that OpAmp and I can tell that it (paired with his DCB1) does sound really well - that is, about as well as other boards using the 827 or 4627. But I am afraid that there have been no chance to do a direct comparison on the exact same hardware, thus it's not possible to tell precisely the differences.
Did you know Caddock makes and Mouser sells a 0.ZERO33 ohm MP-930? I did not, but I know now. 
I don't know why anyone would need that small value, but I certainly don't need the two I ordered by mistake. If anyone wants them for the cost of postage, let me know. Meanwhile, mods must wait.
Peace,
Tom E
I don't know why anyone would need that small value, but I certainly don't need the two I ordered by mistake. If anyone wants them for the cost of postage, let me know. Meanwhile, mods must wait.Peace,
Tom E