What do uncompensated amps behave?

[AndrewT]

Hi Dan
what is mandatory?
An open loop bandwidth wider than audio?
Or open loop bandwidth beyond the audible range is unecessary?
If beyond then how far beyond?

 

[AudioFreak]

Well ideally you should have closed loop FR an order of magnitude outside of the band you intend on using. So if you want 20Hz - 20kHz, you should be thinking something around 2Hz - 200kHz.

 

[MikeB]

Hi Tom !

Have you tried the resistorload in vas ? It's a very simple change,
just add a resistor in vas to gnd, or symetrical, two, one on each
side of vbe-multiplier.

Mike

 

[MikeB]

Well if you want decent phase response anywhere near the edges of the audible range then it's pretty much manditory.

My experience is, phasehifts within closedloop are evil, the less
phaseshift the better dynamics and details. Sadly i was not able
to get below 20° at 10khz with the amp beeing stable.
(the 20° with opened loop,phasedifference from input to output)
But, with "normal" designs, you easily reach values above 50°.
(like ccs-loaded vas, driven from currentmirrored LTP)

Do you mean -3db at 200khz ?

Mike

 

[traderbam]

But in every single amp i've constructed it was absolutely necessary to keep it flat, or sound was "trash".

There are two possibilities: either having flat open loop phase within the audio band is important, or the thing you changed to achieve this changed the sound for an entirely different reason.

Well ideally you should have closed loop FR an order of magnitude outside of the band you intend on using.

Ideally there should be no distortion whatsoever. In practice the O.L. bandwidth is not very important of itself. If you are using feedback then you want as much as possible: typical designs use Miller compensation to create a first pole at only a few hundred Hz, causing the O.L. phase to drop to near 90 deg during most of the audio band. But this doesn't really matter. I'd say you want to aim for a C.L. bandwidth of 40kHz and any more isn't important. What can happen is that when a designer alters the circuit to create a 200kHz bandwidth they also alter other factors and it is these other factors that dominate the sound.

 

[MikeB]

There are two possibilities: either having flat open loop phase within the audio band is important, or the thing you changed to achieve this changed the sound for an entirely different reason.

Yes, i was thinking about that too, adding these resistors lowers
Z-out of the vas, making it less sensitive to reactive loads.
But my latest amp did not need these resistors for reasonable flat
OLBW, and it sounds as it should, like the others with resistors.

Mike

 

[AudioFreak]

Do you mean -3db at 200khz ?

Yes. Although commercially -3dB at 100kHz is more practical there are a few situations where you can really benefit from -3dB @ 200kHz if you pay close attention to stability etc.

 

[thanh]

My LM3886 chip amp sounds extremely harsh

Have you ever listened the sound of TDA7295?

 

[tmblack]

I haven't tried the TDA7295.
Is that a smooth sounding one?

The current mirror seems to make the sound unnaturally crystallized.

Could this be the transistors used(2SA1015)?

Using resistor loads in LTP and current source sounds very natural.

Tom

 

[MikeB]

Hi Tom !

Can you show schematic of your amp ?

Mike

 

[tmblack]

Here is the schematic of the modded amp.
A well proven LIN design but with current mirror.

Tom

 

[MikeB]

What transistors are you using ?

 

[thanh]

I haven't tried the TDA7295.
Is that a smooth sounding one?

I dont also know!
A magazine said it like that. Who have listend its sound?

 

[tmblack]

I have used common Motorola transistors, MPSA06/56 for inputs,MJE15030/31 and MJ15003/4 for output.

The TDA7295 uses DMOS outputs and might sound smoother.
The cost of the TDA7295 is reasonable to try.

Tom

 

[MikeB]

And what did you use for vas ?
You can speed up your LTP by ~3 by not using mpsa06/56, these
devices are incredible slow...
You could use mpsa8099/8599 or 2n5551/5401.

Mike

 

[tmblack]

I use MJE15030/31 for VAS.
You must mean MPS8099/8599. Not always in production but 2SA1015/2SC1815 is low noise, fast and available.

Tom

 

[MikeB]

Sorry, yes of course mps8099...
The 2n5551/5401, should be easy to get and are cheaper than
japanese transistors.

Mje15031 in vas ? Unbuffered connected to a currentmirror ? Ouch !
Mje15031 has very big inputcapacity, giving large reactive load
to the currentmirror having nearly infinite zout. This combined with
the 100pF cdom gives enourmous dynamic phaseshifts.
You could replace q5 with a normal signaltransistor and cascode it
with a mje15031 (or better suitable), this removes a lot of dynamic
phaseshift.

Mike

 

[tmblack]

I agree and pondered on MJE350 for VAS.
No room to cascode or add a buffer, but how about 2SB649A video transistor for VAS?

Tom

 

[ilimzn]

You can speed up your LTP by ~3 by not using mpsa06/56, these
devices are incredible slow...
You could use mps8099/8599 or 2n5551/5401.

2N5551/5401 are fast and high voltage but have a relatively low gain. If the 90V Vceo of the MPS8099/8599 is suitable, they would be a far better choice.

 

[tmblack]

I am using MPS8099 is LTP now.
Question is what is the best for VAS?

I have examined popular op-amps and all use current mirror loads and many with single transistor VAS.

So maybe there is hope in the design.

Tom