My Current Project - Design Stage

[jkuetemann]

I am still playing around with some designs trying to come up with my next project. I have built amps in the past, all fairly generic Lin-type layouts, using a single LTP and single ended VAS with different features. I am looking to step up a little. I'd like to submit to the community a candidate for construction.

I would like to try a 'balanced' design for the IPS and VAS and a triple EF OPS. I've based this design on my 20 year old college electronics schooling and books by Self, Slone and Cordell. I'd like a little peer review. This seems to simulate nicely in LT Spice. It is not esoteric by any stretch.

[CBS240]

Hi

Slight change in terminology, a balanced amplifier typically means there are two output stages driving the speaker at oposite phases, in full bridge. This circuit would be generally referred to as complementary symetrical. It looks like the infamous complementary diff input stage with mirrors from Slone. It appears that there could be 22mA bias in Q5 & Q12, in combination with Q19 & Q20 acting as two, transistor ring CCSs. Of course with 50Vcc this much bias current would lead to over 1W of power on the 2N5550/5401 devices and they would release their magic smoke rather quickly. The big problem with Slone's complementay differential input stage current mirrors is that the VAS current is undefined, which is a significant problem. Bias would be very touchy and temperature dependent as well. There are ways to fix this, some simpler than others, but I don't see Q19 & Q20 doing anything but current limitation of the VAS. Some folks have gotten the circuit to simulate fine but will have problems in the real world. The simplest way to fix the problem is to replace the current mirrors with fixed resistors. There have been many discussions in the forum for years over this circuit.

[jkuetemann]

Yes, I used the word balanced and shouldn't have. Complimentary Symmetry is a better description.

R6 and R7 serve to stabilize the VAS quiescent current, a modification I saw in Cordell's latest book. As Slone presents the circuit the quiescent current is indeterminate and therefore not a practical circuit. Differentially loading the LTPs will reduce their gain somewhat but will allow a stable and predictable VAS current that is a function of LTP current and VAS degeneration by R11 and R12. In the real world the values of R6 and R7 may need to be lower if the transistor pairs in each LTP aren't fairly well matched. Also to give a good match between LTP tail currents one of either R5 or R23 will likely be or will incorporate a trimmer pot.

I realize this is likely to be pretty plain vanilla for many but is somewhat more amp than I've done in the past. Any other comments? Thanks.

[Piersma]

Please take a look at Samuel Groner's paper and his "triple" OPS.

[CBS240]

Originally Posted by jkuetemann

Yes, I used the word balanced and shouldn't have. Complimentary Symmetry is a better description.

R6 and R7 serve to stabilize the VAS quiescent current, a modification I saw in Cordell's latest book. As Slone presents the circuit the quiescent current is indeterminate and therefore not a practical circuit. Differentially loading the LTPs will reduce their gain somewhat but will allow a stable and predictable VAS current that is a function of LTP current and VAS degeneration by R11 and R12. In the real world the values of R6 and R7 may need to be lower if the transistor pairs in each LTP aren't fairly well matched. Also to give a good match between LTP tail currents one of either R5 or R23 will likely be or will incorporate a trimmer pot.

I realize this is likely to be pretty plain vanilla for many but is somewhat more amp than I've done in the past. Any other comments? Thanks.

Ah yes, I missed R6 R7. Is there a significant advantage to using mirrors instead of plain fixed resistors? Have you compared sims? Would be interesting to see what if any differences would be.

[jcx]

complementary symmetry diff pair inputs are only "eye candy"

the symmetry looks good on the schematic but doesn’t do anything useful for circuit performance compared to ”spending” the extra Q elsewhere “enhancing” the basic Lin topology - Self is quite clear on the logical failure of its "improving" input stage performance

if you do so well everywhere else that the input does start to limit performance then bootstrapped cascode may help some and towards the back Cordell shows some cascomp options

Self does posit that push-pull VAS may have theoretical advantages - which could be had by mirrored diff pair VAS and ordinary diff pair input – shown in the original JAES Cordell Mosfet amp paper

but it also appears that SE VAS performance is seldom limiting, at least after "beta enhancement"/input buffering, output buffer/triple load
and possibly cascode

2-pole compensation then takes better advantage of the improved loop gain

the output appears to be the major audio power amp limiting stage - Class A bias is great for the distortion #
A or AB both "want" heavily paralleled output Q - preferably higher frequency, sustained beta types running with peak I below the Ft peak
Boost/filter input/VAS supplies, clamp Vswing to keep >5 V Vce on output devices

There really doesn’t appear to be much further to go in the direction of “accurate amplification” with the full panoply of enhancements to the Lin topology shown in Self, Cordell books

It is far more likely to have implementation problems limit performance - not the circuit topology

[jkuetemann]

Originally Posted by CBS240

Is there a significant advantage to using mirrors instead of plain fixed resistors? Have you compared sims? Would be interesting to see what if any differences would be.

I have and there seems to be less of every harmonic, especially when the 'helper' transistor is present, or if using a Wilson current source arrangement. The better the LTP balance the better the overall performance. This is also why the mirrors are heavily degenerated. I suppose it's possible to get the LTP to share current evenly between each leg using resistive loading, but if there is any change to operating conditions the good balance is lost, the current mirror dispenses with such effects and enforces good sharing of current.

[jkuetemann]

Originally Posted by Piersma

Please take a look at Samuel Groner's paper and his "triple" OPS.

Any 'proven' examples of this configuration? I gave simulation a try and it oscillates. I'd be curious to see a working configuration.

[CBS240]

Hi

My latest amp uses common mode feedback to balance current in a complementary balanced bridge. It works absolutly wonderful with perfect balance but I have yet to see anyone else do something similar. Doesn't discourage me because I know it works. It is not affected by temperature and the bias is always steady. This circuit is for a balanced amp, but the concept works for a single symetrical topology too, as was the previous design. It is very different from the typical LIN topology and after building this circuit IMO LIN is not the best.

Oh yeah, its pretty fast too....

[jkuetemann]

After inserting some base stoppers the configuration suggested by Piersma began to simulate nicely. I'll take some time to check this out as an option.