• WARNING: Tube/Valve amplifiers use potentially LETHAL HIGH VOLTAGES.
    Building, troubleshooting and testing of these amplifiers should only be
    performed by someone who is thoroughly familiar with
    the safety precautions around high voltages.

Moskido Hybrid Aikido-MOSFET Amplifier

If I had to restart my hifi hobby and could keep only one component that I own now - it would be my DIY Moskido amps which sound better than any amplifier I have ever heard. I completed my Moskido amps about the time the Moskido thread disappeared from DiyAudio which may have left a few persons stranded in the middle of their Moskido projects. The purpose of this thread is for the DiyAudio community to help each other complete their Moskido amps.

For those of you who are new to the Moskido amp design – building one of these is for the advanced audio DIYer. All the Schematics for the Moskido can be found on Bob Prangnell's excellent Mad Scientist Audio web site in the DIY section. The heart of the Moskido is the Aikido gain/driver stage. Aikido boards/kits can be purchased from Glass-Ware through the tubecad.com web site.

Bob Prangnell's schematics are for a 50 watt/channel class A amp using a 6SL7->6SN7 Aikido and a double pair of complementary MOSFET power transistors. I originally built a 10 watt/channel Moskido using a 6SN7->6BX7 Glass-Ware Aikido and a single pair of MOSFET power transistors. I recently designed a 7193->6BX7 push-pull Aikido PCB which is discussed in a separate thread (http://www.diyaudio.com/forums/anal...6bx7-push-pull-aikido-preamp.html#post4653656), but has a voltage gain of only 7.7x which works great with my 98db efficient speakers. Judging from the heat produced by my 10 watt Moskido, the 50 watt units must be serious room heaters.

Edit: The new Bias PCB has been working great for 2 years now. The Obligato Gold coupling caps sound amazing. Vocals have never sounded so natural on my Moskido amps! The Bias PCB eliminates the nasty project step of bread-boarding Bias boards from a schematic. I can also email the Eagle files to anyone who is interested in them, however, this PCB includes at least one device that I created in a lbr file. The BOM and instructions are complete. Because I am not allowed to change/add attachments, please go to last post for more information.
 

Attachments

  • Moskido front.JPG
    Moskido front.JPG
    369.2 KB · Views: 2,249
  • Moskido side.JPG
    Moskido side.JPG
    399 KB · Views: 2,137
  • 7193 pcb assembled.JPG
    7193 pcb assembled.JPG
    244.4 KB · Views: 2,134
  • 7193 bottom Moskido.JPG
    7193 bottom Moskido.JPG
    736.2 KB · Views: 2,107
Last edited:

Murdoc

Member
2010-11-17 11:02 am
Germany
Hi Carlsor,

thank you for your thread!

It's interesting for me that the old Moskido Thread isn't any more or got closed, because before I began to build Pass/Firstwatt Amps, I also stumbled across the Aikido Amps and saw the Mosfet Output.

At the moment I'm looking here and there for a new Pre, and thought about Tubes as Pre. So this comes in handy, and I also appreciate it a lot that you started this to give information for all stuck in finnishing their apt builds!
Perhaps if a PCB for the Bias Network is available I'll start this as my new project! :D:spin:
What's also interesting for me is that you're using BUZ900/BUZ905 Mosfets. Did you use others and can say something about the differences in sound? They would be a drop-in replacement for all the mentioned pass/firstwatt amps, and if they change the sound to the better, it's always worth a try.

Regards,
Matthias
 
I only tried the recommended BUZ900/905 MOSFETS so I can't compare them with anything else. The Moskido amp sounds so good that I am not tempted to try other MOSFETS. The BUZ900/905 MOSFETS were designed for audio whereas many of the others were designed for high speed switching.

The beauty of this hybrid is that it is really two amplifier stages joined by a capacitor. You can take the Aikido gain/drive stage or the MOSFET current buffer stage and combine it with some other circuit.

I will be posting my bias board design soon and make PCBs for myself and others who are interested.
 
Moskido Bias PCB

Below are pictures of the schematic, Eagle PCB.brd, and an OshPark picture of what the top of the Moskido Bias PCB will look like. This PCB is 73 x 90 mm. This is a dual mono design.

My bias circuit is a bit different from the one developed by Bob Prangnell. I’m sure others have been successful in bread-boarding Bob’s bias circuit, but I had oscillation problems with the MOSFETs.

1. I use the servo circuit to adjust the pos and neg voltages to the TL431 Vreg circuit whereas Bob injects the servo output into the resistor network after the TL431 Vreg.

2. I connect the TL431 Vreg output directly to the MOSFETs driven by a single Mundorf silver/oil 3.9uf film coupling cap from the Aikido to the neg side of the bias Vreg plus a 100uf cap across the bias pos/neg outputs. The concept is that the MOSFETS will have a more difficult time oscillating with such close coupling to the Vreg. The TL431 Vreg is forced to dance up and down with the music signal, but it sounds great anyway. In Bob’s circuit there are two 390K resistors between the bias Vreg outputs and the MOSFETs and two coupling caps from each Aikido output to the MOSFETS.

I designed the Bias board to have options for the user.

1. Through-the-hole or 3216 SMD resistors can be used in all resistor locations except for the +15/-15 voltage regulation to the TL071 servo opamp. IMO, Susumu RG resistors sound clearer than the Dales, however, this bias circuit is not directly in the signal path so this difference may not be audible.

2. I added a provision for resistors between the Vreg and the Bias outputs to the MOSFETs. There are multiple hole locations for installing the capacitor of your choice across the Bias outputs. This will allow for connections more like that shown in Bob’s Schematics using two output caps per Aikido.

I will be submitting the PCB design for OshPark to make a set of boards this Sunday night for me to populate and test. Please PM me with any suggestions, questions, and criticisms. Now is the time to include any additional design/layout ideas.
 

Attachments

  • Bias Schematic.jpg
    Bias Schematic.jpg
    590.4 KB · Views: 1,912
  • Bias PCB.jpg
    Bias PCB.jpg
    798.4 KB · Views: 1,187
  • Bias PCB Oshpark.jpg
    Bias PCB Oshpark.jpg
    111 KB · Views: 539
It might be a bit academic but I'd add a couple of the small 0.1uF ceramics across the supply pins of the tl071

Is there a way to combine the offset of the tl071 with the offset of the 2n pair?

As there's a bit of space to spare, perhaps you might add a couple of 3 terminal regulators near the V+ and V- inputs just to help isolate this circuit from any of the others (to-92 size) or maybe simple C-Multipliers with bc transistors like the 'flea' supply.

Perhaps, shift the 2n3904/06s together to couple together for temperature

Quite possibly, none of this is necessary as it's not on the signal line, but ...
 
Bias Board Design Points

It might be a bit academic but I'd add a couple of the small 0.1uF ceramics across the supply pins of the tl071
I added these to the TL071 for optional use. I prefer polyester caps for this application. IMO, because the TL071 is adjusting its output at less than 1hz there is no need for quick energy release caps.

I also changed from Zeners to 7815/7915 Vregs for the TL071 and added electrolytic caps to the incoming V+/V- power.

Is there a way to combine the offset of the tl071 with the offset of the 2n pair?
No - these are two independent DC offset sources.
If you can accept up to a 3mv DC offset at the output to the speaker you can leave out the 100K pot and 1.5K resistor.

As there's a bit of space to spare, perhaps you might add a couple of 3 terminal regulators near the V+ and V- inputs just to help isolate this circuit from any of the others (to-92 size) or maybe simple C-Multipliers with bc transistors like the 'flea' supply.
I don't understand what you are asking for here. Send a drawing.
Expanding the PCB image makes the available space and board look bigger than it is.

Perhaps, shift the 2n3904/06s together to couple together for temperature
No benefit from doing this. The servo correction takes care of all sources of DC offset except for TL071 input offset error.

Below are my latest amateur schematic and PCB board layout.
 

Attachments

  • Bias PCB.jpg
    Bias PCB.jpg
    838.1 KB · Views: 528
  • Bias Schematic.jpg
    Bias Schematic.jpg
    642.8 KB · Views: 652
....just me going a bit overboard, as usual!

The idea of adding a regulator, or C-multiplier, to the boards V+, V- is to provide some extra isolation from the amps supply, if used, particularly at turn-on or turn-off - it's no bid deal and can easily just apply an external R + C network (if required)

There's probably 50 different ways to layout any board and this one of yours looks to have covered pretty much everything including the ground plane.

You can use the smaller to-92 size 15 volt regulators instead of the to-220 size.

A question about the speaker terminal - Does the 'SPEAK IN' connection go to the ground plane, that is connected to the power supply of V+,V-, etc, etc and a potential ground loop

You could turn R13 vertical (to fit between IC3 and C2) and if R21 trimpot could also turn anticlock 90*, could run track to mid pin to R13 outside and avoid the top layer thru hole connection

Similarly, if you turn 2n3906 anti clock 90*, widen the pins a bit and pass the base connection thru the centre gap and avoid the chance of bridging the gap and shorting the pin to the ground plane - you could also do a similar thing with the tl431 by moving the C10 above the tl431 (between tl and R16) and run the track between the 'a' and 'c' pads of the tl431 (if the donuts were spaced apart a bit more).
[If the Eagle program doesn't have a larger footprint for this, could just use a trimpot footprint and a short line to indicate the flat face of the to92 device on the top screenprint layer]

As you can see, I'm not a great fan of using thru hole pcb 'vias' for component connections and generally use the component legs to act as 'links' instead, particularly with electro caps and narrow track/holes clearance holes of many pcb design programs.

Anyway, my comments are just from my perspective and no criticism intended, plus, I'm no pcb guru at any of this
 
If your interested in looking at alternatives for your Schematic and PCB layout (as well as another source to have the boards made) look at </title> <link rel="profile" href="http://gmpg.org/xfn/11"> <link rel="pingback" href="http://www.expresspcb.com/xmlrpc.php"> <link rel="shortcut icon" href="http://www.expresspcb.com/wp-content/themes/flat-responsive-child/favicon.ico" /> <title>Exp . Its what I use for my smaller projects and prototyping.

So far what your doing looks great and @jameshillj suggestions are spot on too.
 
More Bias Board Changes

Thanks for the careful PCB review. Keep the questions and suggestions coming.

This version of the Bias board includes the 7815/7915 voltage regulators and puts the 2N3904/2N3906 offset adjust/servo control after the +/-15 volt regulators.

There are 3 possible ground locations - and yes - using more than one ground connection can create a ground loop. Pick the one that fits your project best.

The TO220 7815/7915 have much better line and load regulation than the TO-92 versions that can handle 35 volts.

This is a slow and steady low performance electrical circuit so I am not worried about connections through component leg "vias" which have never given me problems in the past. I can't see any more ways to keep connections from going through the bottom of the PCB. Eagle pin gaps have yet to let me down.

This PCB has to work for me before anyone else should try it. The proof is in the smoke and fire test. I will issue a BOM and detailed set of instructions for testing and adjusting the bias board.

I can email the Eagle files to anyone who wants them, however the SMD/Through-the-hole component is a device I created.
 

Attachments

  • Bias Schematic.jpg
    Bias Schematic.jpg
    633.8 KB · Views: 865
  • Bias PCB.jpg
    Bias PCB.jpg
    824.2 KB · Views: 845
Here I am thinking again....this could be dangerous...LOL

If we changed the +/- rails of the MOSFET end to +48 / -48 VDC could we not just have one power supply and us a Aikido LV ( Aikido LV ) as the front end?
The Aikido LV could use the +48VDC rail referenced to the 0 volt center tap for its power and with some minor changes to the power input of carlors's bias circut be run off of the same single power supply.
I am not sure if the Aikido LV has enough drive to run the MOSFET's directly, but using a complementary transistor driver stage they could receive their drive signal.
This would eliminate the need for the High Voltage and make the power supply design simpler and possibly safer and make a few more output watts too, say around 100 watts into 4 ohms.
 
Moskido Bias Board Working

I am listening to one mono Moskido amp with the new bias board in it. It has a new parts sound but otherwise very good. Bass is tight. See pictures below.

Bench tests showed no MOSFET oscillation. Assembly, test, and adjustment procedures underway and will be put into a post. The servo correction seemed a little over aggressive so I will change a couple of component values in the second bias board which has already been assembled. BOM almost done.
 

Attachments

  • Assembled Bias PCB.JPG
    Assembled Bias PCB.JPG
    681.4 KB · Views: 600
  • Bias Installed.JPG
    Bias Installed.JPG
    737.8 KB · Views: 587