Aksa Lender P-MOS Hybrid Aleph (ALPHA) Amplifier

Hi X,
I made a table last year,
here's the updated one with 0.15 sensor added, I settled on this version for 4R that I'll soon start building.

Code:
        | R131   | R132   | R128   |        |               
        | source | sensor | ccs-fb | bias   | speakers
--------|--------|--------|--------|--------|--------------
Alpha4R | 0.33   | 0.15   | 885    | 1.95A  | 4R-8R  
Alpha20 | 0.33   | 0.12   | 680    | 1.95A  | 4R-8R
Alpha20 | 0.33   | 0.11   | 640    | 1.95A  | 4R-8R
Alpha20 | 0.44   | 0.12   | 510    | 1.46A  | 8R  
Alpha20 | 0.44   | 0.11   | 480    | 1.46A  | 8R  
Alpha4R | 0.27   | 0.12   | 860    | 2.38A  | 4R  
Alpha4R | 0.27   | 0.11   | 790    | 2.38A  | 4R

If it is a 4R Alpha or a 8R Alpha is primarily defined by the R111 FB shunt resistor: 820R for 8R, 1k8-2k5 for 4R.
For a 4R you need minimum bias of 1.9A

Regards,
Danny
 
Whoa, those are really small changes in the resistor values and it makes my head spin! Are you measuring the resistors before installing them or are those the values listed on the parts you purchased/simulated? What kind of tolerances and temp coefficients are you using?

What do the different values of the CCS FB R128 do for the circuit?

I recall you posting a table (copied below) with different cap values for the different FB shunt R111. My guess is the cap values effects the gain in the VAS portion of the amp, with a higher value giving more gain. How did you come up with those recommend ratios and again what kind of tolerances are you looking at?

Gain settings: max. Vin for 17 Vout (max power):
Alpha20 with shunt fb 820r: 0.69 Vin
Alpha4R with shunt fb 1k8+68uf: 1.45 Vin
Alpha4R with shunt fb 2k2+56uf: 1.74 Vin (recommended, ratio H1-15/H1-4 = 99.74)
Alpha4R with shunt fb 2k7+47uf: 2.09 Vin
The FFT profile of the Alpha4R 2k2 looks a lot like the Alpha20 8R.
 
Shawnstium,

These are simulated values and they work well on the prototype. Not all these are tried, but they will always work well. In particular, the difficult value is R128, the fb CCS resistor. None of the values given in Danny's table alter the gain of the amplifier, which is set solely by R113 and R111 (22k/820R). These should be 1% tolerance.

R128 equalises the antiphase currents in the output devices so they show identical dynamic range, a few tens of mA to about 3.5A each at max output into 8R. Different for 4R load of course. These values are given, but it can be examined on a dual beam CRO at max output with a load. But they are easily derived using simulation, and they are accurate.

This is an open source amp. You are invited to build and play, and if you discover something significant, as Danny did, put it up and we will have a look at it. This amp is extremely tolerant of 'fiddling' and gives a consistent, almost invariant harmonic profile across many versions.

HD
 
Thank you for the feed back. I have started the build of the 4R using 22k/2.2k for R113 and R111. I’ll try a couple values for the bypass cap for R111 and maybe a few R in 128.
 

Attachments

  • 66FA5A44-67BA-4B9B-93D8-DFEA0DE1584E.jpg
    66FA5A44-67BA-4B9B-93D8-DFEA0DE1584E.jpg
    996.2 KB · Views: 465
Member
Joined 2014
Paid Member
Hello Alpha Builders,
I’ve remodeled my Alpha20 to utilize the new SLB psu and also added a regulated linear psu to power the PWM fan controller.
I’m happy to report that the “remodel” was well worth it! The amplifier is much quieter at idle, almost silent. Revisiting this project has me thinking that I’ll build another set of boards following one of Danny’s 1.95A, 4R/8R models. (Thanks for all the Alpha20 modeling you’ve done :))

Vunce
 

Attachments

  • 16350C83-C3E9-4B48-B004-DED133AA1D93.jpg
    16350C83-C3E9-4B48-B004-DED133AA1D93.jpg
    1 MB · Views: 651
  • 5F41A547-B598-4F77-9CD2-A3398B5ACE7A.jpg
    5F41A547-B598-4F77-9CD2-A3398B5ACE7A.jpg
    994.1 KB · Views: 445
  • 2533BCED-CD49-4043-AFBC-354C827A73BC.jpg
    2533BCED-CD49-4043-AFBC-354C827A73BC.jpg
    944.3 KB · Views: 448
  • A5DEAA3F-DFC3-4B66-9BED-B9C074E17C1B.jpg
    A5DEAA3F-DFC3-4B66-9BED-B9C074E17C1B.jpg
    978.8 KB · Views: 1,895
Hi Shawnstium,

Those values are simulated in ltspice, after confirmation by Hugh and X that the simulation model corresponds to the real amplifier.

The 0.11, 0.12 and 0.15 options are for which resistors you want to use, 0.11 is for two common 0.22 in parallel.

R111+C111 is an RC filter, when changing the resistor value you must also change the cap value to keep the filterpole the same.
But R111 is the one if you want to change the gain (feedback),
the 4R version needs some more feedback to get the same stability and distortion spectrum as the 8R version, do this by changing R11 to 1k8-2k5

Danny
 
One thing I noticed while testing KSA992’s, over several different projects, is that they are incredibly sensitive to temperature change. Just pulling them out of the strip to test them was resulting is very inconsistent results for me. I put alligator clip leads onto my transistor tester so I wouldn’t have to touch the transistors and throw off the measurement.
 
Member
Joined 2014
Paid Member
Nice work, Vunce!
Is that the nice super flexible 16ga silicone insulation RC model car wiring you are using between the SLB and the amps? Looks very low strain for how thick they are.

Hi X,
Yes it is!
In fact, it’s actually 14ga, 600v rated and very flexible.

Amazon.com: BNTECHGO 14 Gauge Silicone Wire Spool White 25 feet Ultra Flexible High Temp 200 deg C 600V 14 AWG Silicone Rubber Wire 400 Strands of Tinned Copper Wire Stranded Wire for Model Battery Low Impedance: Car Electronics
 
Well, I finally finished reading all 322 pages of this amazing build thread to get caught up on what looks to be my next project. Took me most of today on and off, and last night. :D

I just finished up a custom 384kHz TDA1541A based DAC and an Aikido HPA/LSA build that combines together a wonderful sounding headphone amp and selectable passive volume control or excellent active preamp in the same box, with multiple inputs/outputs. Now I need to move on to actually driving my speakers properly.

After reading through the thread, I have a bit of a laundry list of questions that don't seem to have been answered along the way, or there is conflicting information between posts. Hopefully AKSA and XRK can chime in on these questions, and I think the questions might help other new and old readers to get the current state of the build as of today.


  1. The 4-ohm version seems like the build to build, even for 8 ohm speakers. As far as cooling it, the 4U/400mm Dissipante, 5U/300 Dissipante, or 5U Deluxe seem adequate. Please correct me if I'm wrong here. I'm a tube guy by heart, so I don't care if my amps burn me :p. I just care about device longevity and having an entirely passive cooling solution.

  2. I really like the fact that C111 (C4) is no longer a 1000uF electrolytic in the 4 ohm version. This hasn't been explored much in the replies, but being able to use a 47uF film cap here ensures that there aren't any electrolytics in the signal path. I will do anything to avoid signal path electrolytics as my entire signal chain up to this point also avoids them - polypropylene is always preferred, but 47uF polyester would probably fit better here.

  3. There was a brief discussion on shunt compensation of the FET gate to supply in the CCS, was this abandoned for the non-BB builds?

  4. AKSA snuck in some DMMT5401 matched devices in his LTSpice sims. I'm not afraid of SMD and wonder if these prematched same-die devices are an advantage over matching my own KSA992's and thermally bonding them? Maybe I could drop some header sockets in for the LTP and try it out with a little OSHPark daughterboard...

  5. Are we still pulling C105? I don't see it in the LTSpice 4-ohm asc so I assume it's gone and the B&W schematics haven't been updated for it. I assume the ALPHA20-4R.asc is basically the schematic/BoM for the 4-ohm version.

  6. Is the AS-4220 still the recommended transformer for the 4-ohm build with the SLB supply or Schottky (SiC or normal) based rectification? There's also been no real discussion about inrush current limiting on the toroidal, especially for the BB designs. Catching the AC waveform during power on at -just- the right time with residual flux in big toroidals is bad news bears. Are even the 400-600VA ones still too small to actually care about this?

  7. There also hasn't been much discussion on input impedance. 22Kohm is actually relatively low. I'd like to have seen it up around 47K-100K (1M would be ideal, but these aren't tubes!). How much work would it be to raise this? Cursory intuition and very ignorant simulations say multiplying R102 (R3), R113 (R8), and R111 (R9) all together would raise the input impedance and keep the same gain profile/loading, with a very slight increase in THD -- it might even allow C111 (C4) to come down some. What does this do to compensation/tone? The reason for my input-Z concern is using a passive volume control and proper impedance staging. The volume control on my preamp is 20Kohm series/ladder hybrid and I'd really like to present a decently high impedance to it in passive mode if possible (...if necessary?). I know input impedance is one of the first things to decide in design, but I'm hoping a total redesign wouldn't be necessary to adjust it (at least the Aleph CCS might be the same!).


As always, I immensely appreciate the hard work and effort everyone has put in to these designs. This build (in any version) looks like a straight killer. Solid State SRPP on Steroids.

I'd like to also give a big shout out to JPS64 - all these boards are absolutely gorgeous. I can't get over the amount of effort he gives them and artistic skill he has. :worship:
 
Hi Raptor Lightning,

I will try to answer a few of your questions; XRK might chime in too. You are talking here of the ALPHA20 4R version, or the BB? I will respond with the 20/4R version, the smaller amp.

1.The 4-ohm version seems like the build to build, even for 8 ohm speakers. As far as cooling it, the 4U/400mm Dissipante, 5U/300 Dissipante, or 5U Deluxe seem adequate. Please correct me if I'm wrong here. I'm a tube guy by heart, so I don't care if my amps burn me . I just care about device longevity and having an entirely passive cooling solution.

The original design was set up for 8R, and in fact 4R output is somewhat compromised; the efficiency drops and thermal settings are far higher. The 5U is OK for two channels, but count on sinking 120W of heat on each heatsink for each channel. This is 2 x 24V x 2.5A = 120W plus 1st/2nd stages, and this stresses even large heatsinks. Consider forced cooling.


2.I really like the fact that C111 (C4) is no longer a 1000uF electrolytic in the 4 ohm version. This hasn't been explored much in the replies, but being able to use a 47uF film cap here ensures that there aren't any electrolytics in the signal path. I will do anything to avoid signal path electrolytics as my entire signal chain up to this point also avoids them - polypropylene is always preferred, but 47uF polyester would probably fit better here.

Modern electros with 100KHz ratings for SMPS are actually sonically very good, particularly if you have a few volts across them. In forty years the sonic properties have really improved.


3.There was a brief discussion on shunt compensation of the FET gate to supply in the CCS, was this abandoned for the non-BB builds?

During bench testing we avoided lag compensation and found shunt compensation (VAS to ground) was not effective. X and I finished on 2x18pF, one from VAS to fb node (phase lead) and the other across the 22k fb resistor. The global fb of this amp is set at 220k/22k, 20dB, a low value for any SS amplifier. There is a heavy nested fb between first stage and VAS, but since the output stage operates in Class A and is very linear the global fb is very low and this confers excellent sound quality as the harmonic profile is monotonic with very low, higher artefacts.


4.AKSA snuck in some DMMT5401 matched devices in his LTSpice sims. I'm not afraid of SMD and wonder if these prematched same-die devices are an advantage over matching my own KSA992's and thermally bonding them? Maybe I could drop some header sockets in for the LTP and try it out with a little OSHPark daughterboard...

I ruefully admit these very small SOT363 devices are tiny, and I use them on my production SAKSA 85. We will need to match 992s for both Vbe at 1mA AND beta to match there DMMTs, but the result will identical sonically.
I am almost at my limit fitting these devices, so TO92s are a better match if you over 45 and can't see as well....... ahem.........

5.Are we still pulling C105? I don't see it in the LTSpice 4-ohm asc so I assume it's gone and the B&W schematics haven't been updated for it. I assume the ALPHA20-4R.asc is basically the schematic/BoM for the 4-ohm version.

You should not pull C105. This is phase lead and very effective.

6.Is the AS-4220 still the recommended transformer for the 4-ohm build with the SLB supply or Schottky (SiC or normal) based rectification? There's also been no real discussion about inrush current limiting on the toroidal, especially for the BB designs. Catching the AC waveform during power on at -just- the right time with residual flux in big toroidals is bad news bears. Are even the 400-600VA ones still too small to actually care about this?

One channel can use a 250VA toroid and two can be supplied from one 500VA toroid. If you are in US, ask XRK about this (he uses Antek) but in Oz I suggest you use Tortech in Sydney, who make excellent toroids for SS amps and I use every time. A 500VA does not an inrush limiter for 240VAC supplies, but does for 120VAC mains and a thermistor is suggested by XRK.

7.There also hasn't been much discussion on input impedance. 22Kohm is actually relatively low. I'd like to have seen it up around 47K-100K (1M would be ideal, but these aren't tubes!). How much work would it be to raise this? Cursory intuition and very ignorant simulations say multiplying R102 (R3), R113 (R8), and R111 (R9) all together would raise the input impedance and keep the same gain profile/loading, with a very slight increase in THD -- it might even allow C111 (C4) to come down some. What does this do to compensation/tone? The reason for my input-Z concern is using a passive volume control and proper impedance staging. The volume control on my preamp is 20Kohm series/ladder hybrid and I'd really like to present a decently high impedance to it in passive mode if possible (...if necessary?). I know input impedance is one of the first things to decide in design, but I'm hoping a total redesign wouldn't be necessary to adjust it (at least the Aleph CCS might be the same!).

You can increase Zin by increasing the input and fb resistors, but a design is needed to ensure the amp is completely stable. Another method is to use a bootstrap on the input, a la Doug Self who discusses this in depth. 47k would increase the noise levels a little, a few dB, what is your preamp Zout? Anything better than 5k SS or tube should be fine into 22k Zin, BTW.
If you amend the Zin, you will have to examine the stability issues, but the amp is presently running well, fully designed and debugged, and while you are encouraged to let us know on the forum what you have found, please realise that benchtesting is behind us now as we are moving ahead with many other designs! A 20k SA is fine with this amp; I have seen this on other amps of same impedance and the levels have sufficient steps in all listening situations.

I commend you to this amp. It sounds absolutely wonderful and offers huge efficiency compared with other designs because of the brilliant NP Aleph J output stage.

Cheers,

Hugh
 

Attachments

  • AlphaBBschem.png
    AlphaBBschem.png
    193.1 KB · Views: 329
Last edited:
Hi Raptor Lightning,

I, like yourself will be building an Alpha 20 so I asked similar questions way back when. It is important to know the impedance curve of your speaker. My impedance nadir is 4.8 ohms (and only over a limited bandwidth, i.e. 100-300 Hz), as such I can get by with a bias of 1.96A which helps from a heat dissipation standpoint as I can get away with passive cooling. I would double check the impedance curve of your loudspeaker as there are several versions of the Alpha 20 you can build from a low bias of 1.35A to higher biases of 2.4A. The frequency range over which your speakers are dropping near 4 ohms will also be very important to ascertain. If it drops to 4 ohms at frequency extremes (i.e. 20Hz or 20khz), it obviously doesn't matter. I would take a close look at the impedance curve in the musical region (i.e. 50hz to about 5+khz) and then build the right version of the Alpha 20 thereafter. Obviously the 2.4A versions need an enormously large heatsink but a better option is active cooling.

Using the schematic on post 2780, as well as the changes mentioned on post 2789 and reading through Danny_66's post 3201, I determined that for my purposes, I can build an amp that has about 94+ watts of heat dissipation/ch (+/- 24V @ 1.96A) and can be passively cooled using a 5U/400 chassis. I will be using an SLB supply as you are planning. For a soft start I have Neurochrome's excellent ISS board I have already built.

Regarding the input impedance, please understand that 20k ohms is not really that low (especially in the solid state world ;) ), and in my application, I am using a TAPX preamp which is an autoformer based volume control that provides all the advantages of a passive pre along with a very low output impedance (typically less than 1 ohm in my use, given that my dac has >2V RMS output and my speakers are 96dB efficient). I would look into it if I were you or drive the amp directly from your DAC (if it has its own volume control). Of course the other option is to use an active preamp.

But as Hugh said, the amp as currently designed, particularly with the input stage and feedback circuit is a done deal and a redesign would mean a rebuild as well as further measurements and testing.

Happy Father's day to all! :cheers:

Best,
Anand.
 
Last edited: