Blameless, Dx was wrong, it is a very good amplifier

[gaetan8888]

Hello SandyK

I did noted what he told me, here it is, but I don't know much;

----It turns the damping into quadrature at AF frequencies and thus makes crossover distortion worse because the amp is no longer capable of phase linearly controlling itself at high AF.----


Gaetan

[AndrewT]

Hi,
re post 98 schematic.
the VAS current is approx 4mA.
The maximum output current from the VAS to the driver bases is approximately +-4mA.
Have you looked at substituting a more realistic reactive load in lieu of 8r0 on the output?

I wonder if a two stage EF gives sufficient current gain for high peak current demand, when a low VAS current is designed into the amp.

[Giaime]

Hi AndrewT,

Originally posted by AndrewT
Hi,
re post 98 schematic.
the VAS current is approx 4mA.
The maximum output current from the VAS to the driver bases is approximately +-4mA.
Have you looked at substituting a more realistic reactive load in lieu of 8r0 on the output?

I wonder if a two stage EF gives sufficient current gain for high peak current demand, when a low VAS current is designed into the amp.

I tried 2ohm resistive load, it should be bad enough Even worse than the worst reactive models: these are useful to measure things such as the worst case dissipation in devices, resistors, etc etc. But the heaviest current (current, not power dissipated in devices) occur when the load is resistive and very low.

Yes, we all know about Cordell and Otala work about worst case loudspeaker current: Self cites the 6.6x factor with respect to an 8ohm resistor. A 2ohm resistor represents a CTF of 4, not far from worst case (of course this isn't an amp for continous 2ohm duty!).

Problems start to appear at about 300W on 2ohm, a power level that probably will never be met, with limitations from the PSU.

Of course I agree with you, I feel not so confident with 4mA VAS current, you're probably right, in practice (in the prototipe) I probably will have to increase current a bit, at least 10mA.

[AndrewT]

I probably will have to increase current a bit, at least 10mA.

is that why you chose 2sa/c for the VAS & CCS?

If the simulator thinks that the gain droop can cope with 2r0 then that is worse than my 8ohm*0.35 that I have recently started to use.

Transient signals into reactive loads needs gains of about 100 and 40 from the driver and output to generate 16A from 4mA. That is about 32Vpk if the 2r0 is a good modeler of your amplifier's drive ability.
But can your output device achieve 8Apk with 200mA of base drive?
Can the drivers give out 2*200mA for a base drive of 4mA?
I suspect not.

[BV]

schematic in post 98
Incorrect conection of Q72 (positive feedback). Colector should be conected to base Q68.

[Giaime]

Hello friends,

Originally posted by AndrewT
is that why you chose 2sa/c for the VAS & CCS?

If you're asking why TO126 devices instead of TO92 or similar, yes that's a reason, I plan to heatsink them. The Vbe multiplier is also TO126 because it's easier to mount on the same sink of the drivers (not the outputs).

If the simulator thinks that the gain droop can cope with 2r0 then that is worse than my 8ohm*0.35 that I have recently started to use.

2SC5200 have very good gain linearity, and so does 2SC3421. Yes, the simulator is a bit optimistic, of course: in real life probably I would increase VAS buffer current and live with less than 300W on 2ohm, which doesn't bother me

But can your output device achieve 8Apk with 200mA of base drive?
Can the drivers give out 2*200mA for a base drive of 4mA?
I suspect not.

Well, BD139-16 has a minimum guaranteed gain of 100, even at 4mA. Selecting good devices from a batch (original Philips) shouldn't be a problem. 2SC5200 has (at 8A Ic) a gain of about 60-80, so for clipping on 2ohm I need less than 2mA from the VAS buffer. I wouldn't use 2mA loading with only 4mA current in real life, I'll still prefer 10mA quiescent current, but hey, this value of current gives lowest distortion in SPICE. One can choose to trust not SPICE in distortion analysis: I choose (based on readings and little experience) to trust it, especially in those areas where little modifications involve only a resistor change, not redesign of the whole circuit.

And of course, the design is still to be tested. I will write it in te PDF: people should not try to build it now, I'm sorry if anyone misunderstood.

Originally posted by BV
schematic in post 98
Incorrect conection of Q72 (positive feedback). Colector should be conected to base Q68.

I would agree with you, certanly. The way I connected it, when the load is too heavy the output sticks to the negative rail, since I will provide a DC output control it shouldn't be a problem, but I think it's better the way you suggest. That way, the output simply clips, but I don't think it's enough to keep output devices into SOA: I tried increasing R80 to 100ohm, it barely works. Suggestions?

[Giaime]

THD @20kHz vs. VAS buffer current (keeping output stage quiescent current the same):

4mA quiescent current in the VAS buffer (R5: 150ohm)
Vin 0.14Vpk (1W on 8ohm): 1.24E-2
0.3Vpk: 1.48E-2
0.7Vpk: 1.51E-2
1Vpk: 1.43E-2
1.3Vpk (clip): 7.42E-1

7mA quiescent current (R5: 68ohm, less results in the same current )
Vin 0.14Vpk (1W on 8ohm): 1.30E-2
0.3Vpk: 2.55E-2
0.7Vpk: 1.51E-2
1Vpk: 1.44E-2
1.3Vpk (clip): 7.50E-1

Boh...

[AndrewT]

Hi,
R80 only protects the VAS from excess current.

The CCS is self protecting.

The output stage will need some form of I or IV limiting for protection. This should be supplimented with line fuses to rupture on sustained overload, before the sinks become too hot and then take out the output stage.
Q72 normally only limits when the output stage protection activates. All normal drive should never cause Q72 to trigger.

Yes, for a CFP output stage, the drivers Q19 & Q20 must be coupled to the Vbe multiplier (Q38). Three To92 in a row, glued to a short strip of thick copper or thicker aluminium is normally enough. But 3off To126 bolted will do almost as well.

[BV]

Giaime, can You post here schematic in .cir format? It is simplest way to do some changes..

[Giaime]

Hello BV,

Originally posted by BV
Giaime, can You post here schematic in .cir format? It is simplest way to do some changes..

I use Orcad Capture, I can't seem to find an option to export the circuit in .cir format. If you know a free program to draw circuits in .cir format, I can provide you a drawing so everyone can do his (or her) own tests.

BTW, I found out that reducing R10 and R11, collector resistors in the CFP driver stage, gives incredible distortion benefits, I almost halved the 20kHz THD... but still there is a THD degradation when you increase VAS buffer current to 9-11mA (from 4mA I'm using now).