Alternative topologies

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Across some threads I've noticed that some people are missing the discussion of alternative topologies, (giving that nothing is completely original ... I hope not). This thread is open to all that are interested about different (not widely used) discrete amplifiers. To open the thread I show a CFA (Current Feedback Amplifier) based on the LH4118. Feel free to post your alternative amp, I am very interested on CFA's due it's unique properties.

Cheers
Arturo
 

Attachments

  • ampcc.png
    ampcc.png
    61.6 KB · Views: 893
  • SQUARE_100KHZ.png
    SQUARE_100KHZ.png
    17.8 KB · Views: 874
  • THD_20KHZ.png
    THD_20KHZ.png
    67.2 KB · Views: 806
Go Figure...

Prominent features:
Cheap BJTs tricked to behave like depletion devices with matched thresholds...
InsideOutAllison - No bootstrap caps...
Underbiased Schottky Square Law Crossing
Sliding bias would be the two emitter drops across R7.

Sorry, lost the original .asc file sometime back.

Might revisit this one sometime, I got ideas for improved emitter input impedance.
 

Attachments

  • NoKlass16.gif
    NoKlass16.gif
    56.2 KB · Views: 862
Last edited:
Helps to imagine ordinary input topology with 4 JFETs of 0.65V threshold.
OK, these virtual JFETs only have 47K emitter (er.. gate?) impedance...
But they are dirt cheap and come 2 matched in a DIP (4 BJT transistors).

Firstly, R15 tries to raise output bias.
But Q5+Q6 stop it right at the 1.3V required sliding bias...
Never enough to fully turn on all four cold Schottkys at once.
I must emphasize "cold", because hot power devices are never
allowed to shape this crossing. All shaped in overkill TO220
sized diodes that barely dissipate any heat.

The rest appears to be gravy to limit current and other mishaps.
Soft clipper at front end could and should have been done better.
I've learned a bit since drawn this more than a year ago...
 
Last edited:
fixing clipping ...

I changed the cascoded VAS to 2n5551/2n5401 completely because the clipping is much better than the prior circuit showed. A little lower THD's but nothing significant.

Cheers
Arturo
 

Attachments

  • ampcc.png
    ampcc.png
    61.8 KB · Views: 817
It will take me some time to understand it (and simulate it) ... a emitter coupled VFA, isn't it?
input impedance should be low I guess ...

Cheers
Arturo

You see through 0.65V emitter of Q1 to the base of Q8 in parallel with R3 47K.
So basically 47K per input emitter, 23.5K for both Q1 and Q4.

If we increase the impedance of the 47K pullups (replace w. current sources),
the emitter impedance of Q1+Q4 then begins to look more like Q8+Q9 bases.
 
Across some threads I've noticed that some people are missing the discussion of alternative topologies, (giving that nothing is completely original ... I hope not). This thread is open to all that are interested about different (not widely used) discrete amplifiers. To open the thread I show a CFA (Current Feedback Amplifier) based on the LH4118. Feel free to post your alternative amp, I am very interested on CFA's due it's unique properties.

Cheers
Arturo

Arturo,

Someday I will build a CFA power amp, keep going... I'm using a CFA opamp for my pre amp, 2000v/us and unmeasurable thd (at least with my equipment).

Ken
 
Hi Kenpeter,
A question: Q8,Q11 and Q9,Q14 drives M1 and M2, isn't the output impedance of this drivers enough low as M1 M2 are showing a capacitance maybe (including Miller) of 1 to 2 nF?, it seems that also Q15 and Q16 (that are limiting the current of the mosfets) also 'steals' some of the driving current. I am understanding OK?

Cheers
Arturo
 
About the CFA

Hi Ken, I am finishing a PSU system that will be common to all the amps variations I plan to test (besides 2 'traditional' topology i've build years ago) with instruments and auditions. For me there are 'good' and 'bad' amps with a range in between. I also don't believe on what should be the 'correct' sound (as a musician aficionado), I learned to hear several instruments and voice of different types in different environments (outdoor, at studio, chamber music, my 'music room' ... etc). So it is fair 'learn' hearing different designs and acquire a more broad listening culture, and learn the 'why's' for each case (if it is possible to link electric measures with audition). I am close to retirement so I will have a lot of time to do so.

Returning to the thread topic the CFA shows some nice properties among them a huge stability (f180° at 38MHZ with Dom. pole at 5.11MHZ with gives a PM of 142°) , a rapid response to high variations of current demand at the load, also it does'n has 'high impedance' internal nodes much susceptible to parasitics. The output stage topology (diode strings at the buffer) is a well known 'good' sound for me and has a decent distortion with 1.5 OHM output impedance drive, also has the ability to reduce the quiescent AB current from CLASS A at 30° to pure B at 100°, ~10W Class A when it is in AB region at 65°. I have found that stripping the degeneration resistors (of the VAS driver) also reduces the thermal drift to +/- 15 mv (from 0° to 100°), besides it reduces a bit the THD.

Cheers
Arturo
ampcc.png
 
Last edited:
Its a three way battle between push, pull, and sliding bias.
Resulting in two answers that drive the MOSFETs and keep
bias from running away....

Open loop impedance is too high, would not function open.
Low driving impedance only a result of the feedback loops.
There is enough available current to overcome Cgd Miller.

I want to try again with EKV+VDMOS hybrid .subckt per
Bob Cordell's book.
 
Hi Kenpeter,

I always being afraid about mosfets because of the self oscillating issue, and the use of gate stoppers that indeed works, but for me it is not a elegant solution, it adds distortion and noise, and the 'voltage drive' property holds only for low frequency as it rises the mosfets becomes current hungry, maybe the HF response compared to BJT is an indisputable advantage, but with BJT with 4MHZ Ft or higher I don't see clearly the advantage at audio, maybe the harmonic profile ??

Cheers
Arturo
 
So use bipolar Darlingtons, and ditch the Zeniers.
MOSFET is not a religion with me yet. (or is it??)

You asked alternative topology, all its about...

Bias and square law crossing the same either way.
Managed by the cold Schottky stack, not by which
style of output device does the brute work.
 
So use bipolar Darlingtons, and ditch the Zeniers.
MOSFET is not a religion with me yet. (or is it??)

You asked alternative topology, all its about...

Hi Kenpeter,
Oh yes, nothing to criticize about mosfets (despite I managed to blow some ...), I was arguing about my personal taste (truly biased), but you are right, I must open my mind (what I am requesting), nothing to do with religion, when they do the work OK that's enough, maybe I am getting old ...

Cheers
Arturo
 
No PCB yet

padamiecki :
Hi Arturo!
I found your topology best for my new amp,
I would only consider to use cfp pairs for the input devices...
could you show pcb you designed for your amp?

Hi Pawel, you will see I am at the very beginning with this amp. The protoboard shown (which is similar but not exactly the circuit posted) was mainly to check the concept and verify DC stability. Also picked up some music from the buffer (BD139/140) to a headphone (despite the mess and the 'wires' and the ground sounds good!!!).

I've tried CPF inputs but it makes almost impossible to bias the circuit (tested in practice), if you wish there are better signal BJT's than the 2sc945/2sa733 (unmatched pairs), I use them because is what I have on hand. Currently I am working on the PSU. Maybe in a week I will have the PCB design finished, but I must warn you that it will be a two layer PCB, I don't have any facility to do my own PCB's.

Cheers
Arturo
 

Attachments

  • proto.jpg
    proto.jpg
    211.6 KB · Views: 408
The thing that I like from CFA is bandwidth. That also the reason I am using CFA for my new VAS. 10MHz and up of bandwidth is easy to reach without compensator.

A733-C945 is Vbe matched, the datasheet shows many differences but their Vbe characteristics is matched. I think its pair is chosen by designer, and not offered by manufacturer, this is rare condition.


Add: Hi, artu. Did you tried the output stage with loudspeaker? I remember similar output stages in other thread. Something like current feedback classA thread.
 
Last edited:
Ontoaba:

The thing that I like from CFA is bandwidth. That also the reason I am using CFA for my new VAS. 10MHz and up of bandwidth is easy to reach without compensator.

A733-C945 is Vbe matched, the datasheet shows many differences but their Vbe characteristics is matched. I think its pair is chosen by designer, and not offered by manufacturer, this is rare condition.

Add: Hi, artu. Did you tried the output stage with loudspeaker? I remember similar output stages in other thread. Something like current feedback classA thread.

The A733/C945 was the 'best' choice a had, and yes this pair is not quoted as matched by the manufacturer because other characteristics differs widely, only same power and voltage rating, but it is also a good thing that this topology doesn't need strict matching components, even the VAS is unsymmetrical 2sa733/2n5401 upper leg and two 2n5551 in the lower leg (done to correct the shape of clipping).

CFA's are free from the 180° phase issue of VFA differential inputs designs, that's why you can reach higher frequencies safely, and I need only a 3.3pf FB cap. and a 10pf miller cap. in the cascode to shut up a 37MHZ instability, both completely immune to PCB parasitics.

About the output stage, yes it is very suitable to CFA because the way the current flows driving the OPS that makes easy to cancel DC OFFSET for a CFA which are known to be more difficult to achieve low offsets. Yes I know and heard, this output topology (buffer+OPS) with several different loudspeakers years ago and sounded very good. Also simulations reveal 0.066% THD at full power this stage alone, that is easily corrected by the NFB loop at 90db OLG. I've added a thermal compensation scheme that lowers the AB quiescent current inversely to the OPS temperature, as temperature raises the quiescent current lowers maintaining OPS within SOA.

Cheers
Arturo
 
Tube cad has an interesting circuite the xpp. I wonder if this cuircuite
would work using a DN2540 for the first "triode" and a IXTH20n50 for the
second "triode"

Probably not, MOSFETs lack "Mu" internal negative feedback that style of circuit would require.
Schading MOSFET to act like triode, complicated if you care to maintain high input impedance.
I think you selected a triode dependant circuit that doesn't translate terribly well to sand.
If you had a box of rare SIT's, yeah maybe...
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.