while im blathering if you put mirrors into old circuits or update them you maybee affecting the bias of the next stage and if your simulating LTPs do it with a load or the current thats pumped has nowhere to go and plot the load and leg currents VS input voltage if you really want to see whats going on. Sorry to all who a;ready know this.
For those of you looking for very high performance current sources: TI's REF200 has nomally 100Megohms Zout. Try to match (!) that in discretes...!
http://focus.ti.com/lit/ds/sbvs020/sbvs020.pdf
Jan Didden
http://focus.ti.com/lit/ds/sbvs020/sbvs020.pdf
Jan Didden
AndrewT said:
Hi Andrew,
Take a look at the IPS-VAS I use in my MOSFET power amplifier with error correction at www.cordellaudio.com. The LTP drives a balanced differential current source, which in turn drives a cascoded VAS based on a differential pair. This may address your concern. However, you might object to the fact that the differential output of the VAS drives a current mirror on the other rail. I don't think that causes any degradation, and of course the measurements are very good.
Cheers,
Bob
rayfutrell said:
Hi Ray,
Identical to a Baxandall pair
http://www.diyaudio.com/forums/showthread.php?s=&threadid=25172&highlight=
The current mirror is essentially an instrumentation technique; it's used in servo and control technology and there is a need for very high accuracy. I use an asymmetric CM in a FF power supply which meters the current draw, produces a much diminished analogue of this current and uses it to control the regulation.
In a SS amplifier, the main emphasis should be the sound quality. The distortion per se is not the only emphasis, we need to make a psychoacoustic assessment.
If we match the LTP transistors for Vbe and beta and we design the bias networks for identical impedance, then in the conventional Bailey amp the two currents will be within 5% of matched, depending only on LTP matching. However, if we use a CM, something interesting happens.
Let's use a unipolar VAS of beta 150 running at 10mA - keep it simple. This will need a bias of 67uA. The LTP stage current is one milliamp, 500uA running down each side. This 67uA will subtract from the LTP stage current; only 933uA will be passed to the CM.
The CM will set up, in theory anyway, identical currents in its two legs. With a shorted C/BB on the drone side, the bias requirement for beta 150 on the CM devices will be 6.7uA; half this will pass to the drone side, the other half to the active side.
The CM will pass identical currents through its degeneration resistors; this will total 933uA with the remaining 67uA passed to the base of the VAS. Thus each CM device will pass approx 466uA, while the output LTP device will be passing 466uA + 67uA = 533uA. 533uA on the output LTP device will add to the 466uA of the drone LTP device to give us our stage current of 1000uA.
IOW, the LTP is intrinsically unbalanced by the CM; one side is 466uA, while the other is at 533uA - even though the CM is perfectly balanced. This is substantial, plus 6.6% on one side, minus 6.6% on the other. This will introduce H2 and H3 distortion.
If a resistor from joined bases to the rail could pass the difference in the 6.7uA and 67uA bias currents, then this imbalance would be corrected.
For the values given, and degeneration resistors of 100R, the resistor value would be 11K6. This would suck another 60uA from the drone side, and correct the imbalance, restoring lowest possible distortion.
Hugh
In a SS amplifier, the main emphasis should be the sound quality. The distortion per se is not the only emphasis, we need to make a psychoacoustic assessment.
If we match the LTP transistors for Vbe and beta and we design the bias networks for identical impedance, then in the conventional Bailey amp the two currents will be within 5% of matched, depending only on LTP matching. However, if we use a CM, something interesting happens.
Let's use a unipolar VAS of beta 150 running at 10mA - keep it simple. This will need a bias of 67uA. The LTP stage current is one milliamp, 500uA running down each side. This 67uA will subtract from the LTP stage current; only 933uA will be passed to the CM.
The CM will set up, in theory anyway, identical currents in its two legs. With a shorted C/BB on the drone side, the bias requirement for beta 150 on the CM devices will be 6.7uA; half this will pass to the drone side, the other half to the active side.
The CM will pass identical currents through its degeneration resistors; this will total 933uA with the remaining 67uA passed to the base of the VAS. Thus each CM device will pass approx 466uA, while the output LTP device will be passing 466uA + 67uA = 533uA. 533uA on the output LTP device will add to the 466uA of the drone LTP device to give us our stage current of 1000uA.
IOW, the LTP is intrinsically unbalanced by the CM; one side is 466uA, while the other is at 533uA - even though the CM is perfectly balanced. This is substantial, plus 6.6% on one side, minus 6.6% on the other. This will introduce H2 and H3 distortion.
If a resistor from joined bases to the rail could pass the difference in the 6.7uA and 67uA bias currents, then this imbalance would be corrected.
For the values given, and degeneration resistors of 100R, the resistor value would be 11K6. This would suck another 60uA from the drone side, and correct the imbalance, restoring lowest possible distortion.
Hugh
if using high gain transistors in the LTP mirror (as in the LTP)
say BC550C or BC560C or similar
we can have a gain of 400 ... even at lower voltage/currents
the mismatch, if only count for gain
would be
400 uA ... 401 ua
800 uA ... 802 uA
in other words 0.25%
and this is quite satisfactory for most DIY amplifiers
there are other parameters that cause your amplifier more problems, than this
like power supply and wiring and resistor tolerances / temperature issues
So, just because the petimeter people find it interesting to find 'the ideal' ultimate current mirror here,
this does not mean that normal average diy members
should start to think this is someting that is going to ruin your Diy Amplifier sound.
And feel worried about these things.
Because there is not much to worry about.
Lineup .. thinks some abstract discussions can be good
.. but have not often much relevance in reality. Not for DIY Audio.
say BC550C or BC560C or similar
we can have a gain of 400 ... even at lower voltage/currents
the mismatch, if only count for gain
would be
400 uA ... 401 ua
800 uA ... 802 uA
in other words 0.25%
and this is quite satisfactory for most DIY amplifiers
there are other parameters that cause your amplifier more problems, than this
like power supply and wiring and resistor tolerances / temperature issues
So, just because the petimeter
people find it interesting to find 'the ideal' ultimate current mirror here,this does not mean that normal average diy members
should start to think this is someting that is going to ruin your Diy Amplifier sound.
And feel worried about these things.
Because there is not much to worry about.
Lineup .. thinks some abstract discussions can be good
.. but have not often much relevance in reality. Not for DIY Audio.
AKSA -Post 105
Several Sydney members have been using for many years, variations on what Hugh is proposing, to achieve complete front end balance. (before the Halcro patent of the added diode!)
The result is a marked improvement in soundstage, and what appears to be a S/N improvement. My present method is a series 1N5819 Schottky diode (vertically mounted) and 200R 20T trimpot, inserted in the unloaded side of the CM to give virtually identical collector voltages of the very well matched LTP transistors. If Hugh's method works as hoped, it would be a more elegant way of achieving the same results. Instead of just simulating it, why not try it ? You may be surprised.
SandyK
Several Sydney members have been using for many years, variations on what Hugh is proposing, to achieve complete front end balance. (before the Halcro patent of the added diode!)
The result is a marked improvement in soundstage, and what appears to be a S/N improvement. My present method is a series 1N5819 Schottky diode (vertically mounted) and 200R 20T trimpot, inserted in the unloaded side of the CM to give virtually identical collector voltages of the very well matched LTP transistors. If Hugh's method works as hoped, it would be a more elegant way of achieving the same results. Instead of just simulating it, why not try it ? You may be surprised.
SandyK
cbdb said:
Sounds like you have all of the answers. We should be building your design.
Seems to me you can have both, low distortion and low noise. Why settle for one or the other?
cbdb said:
Why not one of those entire front ends from National the LME49-whatever? Those are only $14 (here). Everything is well matched and running at the same temp on those, right?
Why not skip the whole deal and just buy an amp ready made? Save even more hassle.
AKSA said:
Current balance in the LTP (without a current mirror) has nothing to do with Vbe matching of the LTP transistors. It is determined by how accurately the resistor connected between the base and ther emitter of the VAS transistor forces the LTP's active leg current to equal 1/2 the LTP tail current.
Secondly if an emitter follower is used to buffer the VAS the imbalance caused by the VAS/EF base current drops to insignificance (even with a LTP tail current as low as 1mA, which would typically give a mediocre slew rate, and doubly so without a current mirror).
Cheers,
Glen
you can have both (what is low?) but a lot of things in design are trade offs this is one
Why not use technical advances? do you shun the new high ft output trans. fron ON
(if your saying you dont learn anything from using these blocks I agree, but if your looking for the best design possible then I dont)
Why not use technical advances? do you shun the new high ft output trans. fron ON
(if your saying you dont learn anything from using these blocks I agree, but if your looking for the best design possible then I dont)
I happen not to agree with you Lineup, mirrors and current sources are an important aspect of amplifier design and like everything else we should try to find ways of improvement. They certainly have a effect on sound quality and this certainly is relevant to audio. Its often in these small details one finds improvement in sound quality with same topology.
cbdb said:
Hi,
I have a little, amateurish amp here that uses some pretty high ft devices, the first version of it used the LM394 as the differential pair. I don't shun technology.
Why make the front end so it costs as much as the rest of the amp? What gains are made? How much noise is reduced?
Define "best design". There's a thread here and I haven't seen any winners.