Aleph-X builder's thread.

Official Court Jester
Joined 2003
Paid Member
best are the ones having least amount of interelectrode capacitance .... but cheapest ones are even better :clown:

it really depends of your intention - how many you want to put in each quadrant

if you prefer small number , IRFP150 is best choice , as I wrote

with little luck , you don't even need to match them , if staying with just one per quadrant

however , if you intend to feed speakers with less than nice modul of impedance , there is no replacement for displacement , as our friends from hot rod department use to say

X is bridged amp , each half seeing just half of load ..... so number of outputs is what is giving you muscles
 
no less than 2 mosfets per quadrant , so 8 per channel

only in case that you use IRFP150 , practically 2pcs of IRFP240 in one case , you can go with 1 per quadrant

Disagree. My Aleph-X monoblocks use 4 IRFP044’s with 15V rails and 2.5A bias per device, and have been in heavy use for around 10 years. It’s a matter of heat sinking. Remember, Nelson Pass tells us that Mosfets get more linear in response with higher bias right up until the magic smoke is released. The downside is that less devices results in lower damping factor, but with the aleph current source creating a sort of negative output impedance effect, the impact of the fewer devices is somewhat mitigated.

For guidance on where the boundaries are for per device power dissipation see Transistor Dissipation and Turbo Projects

Terry
 
Last edited:
I’m looking for the BOM for the original pcb, can’t seem to find any info in the wiki or am i just looking in the wrong corner?

Even when I'm not on vacation, Zen usually beats me to the punch. Have a look at my webpage for the Aleph-X amp - link is in my signature. I have the BOM for original 15v version as well as for the higher powered 22v version. Just about all of the information you need is there, though it's a little messy as the page evolved over a period of 7-10 years...

I also have a copy of the original wiki for this amp on my web page.
 
Member
Joined 2000
Paid Member
IRF9610 Alternatives

Hi,

Are there any alternatives to the IRF9610 mosfets? I didn't see much reference to updates or alternatives in this thread, unless I missed it.

A couple of the alephs received a face-lift in the input stage, or at least "better" mosfets were substituted. Can those articles be referenced to update the AX frontend?

Thanks,

Vince
 
Hello!

I have built an Aleph-X in 2007 and its a fine amp but temperature is a problem-too small heat sinks. Now I want to build it up again a „little“ bigger. Some friends are in the project, so have 18 Monoblocks in total!

Project basics:
- 24V Rails
- bias 3A / Rail
- 2 Pcs 0,19K/W per Channel
- 3 IRFPs in parallel
- own PCBs

Now some questions to avoid making mistakes on 18 Monoblocks:
- My old version was with IRFP044. Now I looked at some Datasheets and found the IRFP140N as the best solution: Better Curves than 044 for 24V, less capacitance than 150 which has much more current as really needed. 240 also has more capacitance.
- Input stage: like to replace the 9610 by 2 parallel J74 for my 2 monos. But I have not enough of them for all 18, so what do you recomment: the original 9610 or ZVP3310 or ...?
- For me the damping / definition in the low frequency range is the weak point of the amp. Has anyone tried to repace the input stages 390R by two CCS? Will give more gain at lower frequencies and I usually prefer the sound of „low-loaded-circuits“.
- The resistors back from the outputs to the input CCS (4k7) will ruin the common-mode-rejection due to low impedance loading. Therefore I put an K170 JFET in common-gate circuit between the Resistors an the CCS to give a high-impedance current source. Works well in LTSpice and a Prototype. Any experiencies to this topic?

Thank you,
Jean-Claude
 
Hello!

I have built an Aleph-X in 2007 and its a fine amp but temperature is a problem-too small heat sinks. Now I want to build it up again a „little“ bigger. Some friends are in the project, so have 18 Monoblocks in total!

Project basics:
- 24V Rails
- bias 3A / Rail
- 2 Pcs 0,19K/W per Channel
- 3 IRFPs in parallel
- own PCBs

If the heatsinks can take the heat I’d recommend upping the bias to 4A / Rail. Given your goals I’d also suggest more output mosfets in parallel, at least 4 and 6 would worth considering.

Now some questions to avoid making mistakes on 18 Monoblocks:
- My old version was with IRFP044. Now I looked at some Datasheets and found the IRFP140N as the best solution: Better Curves than 044 for 24V, less capacitance than 150 which has much more current as really needed. 240 also has more capacitance.
- Input stage: like to replace the 9610 by 2 parallel J74 for my 2 monos. But I have not enough of them for all 18, so what do you recomment: the original 9610 or ZVP3310 or ...?

Stick with the 9610’s. Going to the ZVP3310’s probably won’t provide much audible benefit, but you would need to add a cascode to the input diff pair to get deal with the power requirements.

- For me the damping / definition in the low frequency range is the weak point of the amp. Has anyone tried to repace the input stages 390R by two CCS? Will give more gain at lower frequencies and I usually prefer the sound of „low-loaded-circuits“.

That won’t really do anything for the damping factor or low end response. You’ll be better served by adding more output devices.

- The resistors back from the outputs to the input CCS (4k7) will ruin the common-mode-rejection due to low impedance loading. Therefore I put an K170 JFET in common-gate circuit between the Resistors an the CCS to give a high-impedance current source. Works well in LTSpice and a Prototype. Any experiencies to this topic?

I don’t think the common mode rejection is much impacted by this. If you match your devices well before building, those resistors can have higher value. Several of us have played around with this and found that the audible effects become unnoticeable at value >=10k. As to putting a K170 in common-gate circuit, I’ve never seen anybody do that, and it seems an added complexity using a high value part where simpler solutions exist.

My last thought: a big vote for Zen Mod’s Babelfish XJ - probably the best thought out Aleph-X effort anyone’s put together so far.

Cheers,
Terry
 
Hello,

thank you for feedback!

The basic project parameters are fixed, there was a long lasting optimation phase to the needs of all project partipiciants.
We wanted the amp as "small" as possible without losing significant sound quality at "normal" 8 Ohm speakers:
- 23V as the minimum at which the circuit works best
- 3.3A = corresponding tail current for Pmax@4R ~ Pmax@8R (80W).
- Drive the Power Transistors as high as possible for optimum relation between transconductance and capacitive load to the driver. Without losing reliability -> 25W per device -> 3 in parallel.

Attached I have the Input stage for discussion:

Background for the CCS J2/J5:
When designing the output stage there is a trade-off between low frequencies (transconductance=big devices) and high frequencies (capacitance=small devices). Loading the input stage with a CCS gives more OLG at low frequencies with no loss in performance at high frequencies, so why not? Additionally this brings back the OLG which was lost by JFETs instead of Mosfets and lowers the amplifiers output impedance at low frequencies.
Circuit: Load of the input stage is reduced from 390R to 2k2 (R73/R74). This retains a stable common-mode behavour. The rest of the needed current of the standard ~11mA is delivered by the CCS around J2/J5.
D5 / D6 are needed to avoid high voltage at the outputs gates when failure / startup.

Background for the circuit around J1: The performance of an differential amplifier depends strongly on the impedance of its current source. One reason why CCS are usually used and not resistors. The MacMillan Resistors disturb this - you can see it very well when using the amp unbalanced: The output voltage at OUT+ is significantly higher that those at OUT-, for me a reason why aleph-x responds so well on a balanced source. With the decoupled common-mode-feedback this problem is solved and even driven unbalanced the OUT+ and OUT- are (nearly) identical.
Circuit: The voltage caused by R12 is needed that the circuit will work. J1 is designed as ~4mA CCS by R11. The "MacMillan" Resistors (R31/R33) go to the source of J1, R30 is additional current from a servo. The JFET transfers the common-mode-signal into an high-impedance current.
For me its a nice way to get rid of the "problem" just by adding 1 Resistor and a small standard JFet :)
(R13 is only for measuring drain current, R10 only for safety).

Remark: The circuit shown can add +/- 4mA to the LTPs current which is enough with the 2*2k2+CCS load. In "standard" circuits with 2*390R load this current might be chosen somewhat higher.

Your thoughts are welcome :)

Regards, Jean-Claude
 

Attachments

  • Aleph_XJ_Input.JPG
    Aleph_XJ_Input.JPG
    51 KB · Views: 728
- 23V as the minimum at which the circuit works best
- 3.3A = corresponding tail current for Pmax@4R ~ Pmax@8R (80W).
- Drive the Power Transistors as high as possible for optimum relation between transconductance and capacitive load to the driver. Without losing reliability -> 25W per device -> 3 in parallel.

I think your math is off somewhere. 3.3A bias per side will give the 80W@4R, but 24V rails will give you ~130W@8R.

Also, your assumption that 23V rails are the min for best results is incorrect. My Aleph-X monoblocks run 15V rails and have put some pricey commercial amps to shame in comparison. If you look at Nelson Pass’ own commercial XA amps, he has models with rails as low as 12V.

For getting best linearity etc. from diving your output devices as hard as possible, you get greater benefit from increasing the bias current than increasing the rail voltage. You would be better served by lowering the rails and increasing the bias, while keeping the dissipation per device the same. This would also serve to give you better low end performance.

With regards to your arrangement of CCS’s surrounding the long-tail pair, I see what your intent is and the rationale behind your design choices, but I think you are using a sledgehammer to fix a “problem” that’s not really much of a problem. With the circuit you are proposing I’d be wary of relying only on simulations, and strongly recommend you prototype it to make sure it works as intended before you do any larger scale group building. I’d even more recommend looking at Zen Mod’s servo circuit as a more elegant solution that has been well proven. Also, the Aleph current source as-is already provides some serious low frequency grunt. I don’t think you’ll be disappointed.

Remember that one of the reasons Nelson Pass’ designs sound so good is that he uses very simple circuits where each part is optimized to get the best linearity out of the devices used. While he doesn’t avoid using feedback, he also seeks to use only as much as is required. The elegance and simple sophistication of his designs are a large part of why they sound great. Boosting the open loop gain so you can apply much higher levels of feedback may take away the Pass magic you want to achieve. But that is just my opinion.

On the other hand, this is DIY, and doing it your own way is half the fun, and if you share your experiences and results along the way then everybody gets to learn more about what works and what doesn’t. Just look at a couple of my development projects from waaaay back and see how many mistakes and iterations I made along my learning curve! It’s just plain fun to try stuff and see how it works.

Cheers
Terry
 
Hi!

Thank you for your input!

With 23V / 3.2A(6.4A total), the AXE-1 sheet (and LTSpice) gives me 110W@8R and 90W@4R with 53% AC current gain.
3A is a rounded figure, with 0R56 in the output CCS I expect ~3.3A. You are right, it eventually might be better to go to 0R47 in the current source which allows the option to bias up to (0,62V/0R47*3=) 4A in maximum. Its also a part of the project to add a bias-switch for either "high-end" or "casual" listening.

Regarding the front end: A collegue has already built up a "low power" Aleph-JX with 15V / 2A. Works well and sounds good, we think quite a bit nicer than my "stock" Aleph-X from 2006 (18V/2A). For the actual project I have a jumper on the layout to test both versions - CCS and the standard 390R.

You are 100% right - at least 50% of the fun is to bring in own ideas. And I never build up a identical component (loudspeaker, amplifier) a second time - its sooooo boring ;)

I like the pass concept of the simple circuits very much. But additionally I like to add my own taste: Use a simple circuit, and create an environment which allows this circuit to work best. I have no problem if it will get a little more complicated.

To my experience with different circuits I see benefits in low-loaded circuits. For me they are usually more dynamic and "airy" compared to heavy loads by resistors. The "problem" here is really not to produce too much open-loop-gain. In the Aleph-X I see no problem here since the open-loop-gain is only rised at low frequencies with the CCSs, from kHz upwards the open-loop-gain is limited by the output transistor capacity anyway.

With this in mind I developed the design modification in the input stage and chose the supply voltage: Higher voltages give significantly lower values of capacity in the output transistors and therefore a lighter load at high frequencies. Btw I would like to try out a cascoded output stage, but liked to keep it simple...:rolleyes:
I am aware of the fact that on the other hand a high current will improve the low end. But wasn that "Papa" Pass somewhere on the first pages of the threat which preferred higher supply voltages?

Personally, I find the input stage of ZenMod questionable. With the additional buffer its far away from the simple original circuit - much more that mine. And the buffer will increase open-loop-gain much more, therefore adds stability issues etc. If my version is "sledgehammer", what would you name this? As I see (and simulated) the Servo, it causes the same problems as the solution with the Macmillan resistors - poor common-mode-rejection and different output signal with unbalanced input (or are Im wrong ?) The "problem" here is that AC-common-mode-feedback (=not lowpass filtered) from the output also cancels out 2nd harmonic distortion.

Regards, Jean-Claude
 
weak bass, treble or room effects is why I kept the tone controls active on system.
some stuff sounds excellent flat, some requires loudness or treble boost to sound great.
As your system improves you will find great disparities in recording quality from LP to LP. the differences are not as stark on CD as the mastering phase for LP is not required.