A 50W lateral MOSFET amplifier based on David Tilbrook's AEM6000

[Suzyj 56]

Way back, I wanted to build some amplifiers to drive a lovely old pair of Infinity RS-5B speakers that my PhD supervisor gave me. I found the original articles for the AEM6000, and set to updating that to use available (at the time) transistors.

 

Since then I've done some other versions of this amp, mostly working to shrink the PCB somewhat, optimise the layout, and achieve respectable noise, THD, and slew rate specs while keeping the quiescent current under control.

 

This latest version uses one pair of Exicon ECX10N20/10P20 lateral MOSFETs, for around 50W RMS into 8Ω. It has reasonably good distortion, of around 0.0005% at 1KHz, and features rather lower quiescent power consumption than a lot of lateral FET amplifiers, because I care about that. It also does just 8nV/√Hz input referred noise, so you can enjoy your music without hiss.

 

It's a nice little amplifier, easily my favourite. I've built a bunch of them and use them for all sorts of stuff.

 

All the files for making one, including gerbers and a comprehensive construction guide, are available on my Google drive.

 

It requires surface mount assembly (MELF and SOT-23) but don't you think it's time you tried that?

 

All the components are available now.

 

 

 

AEM6000 Based 50W Amp Construction Notes.pdf

[Janus77 386]

I read through the attachment Suzy, too technical for me  though I'd love to try something like this one day. 

What are the cost of the parts roughly out of interest?

[Suzyj 56]

The build price is dominated by the MOSFETs, at around $15 each, and the JFET, at around $8.

 

I’d guess something like $60-80 per amp.

[Monkeyboi 721]

11 hours ago, Suzyj said:

Way back, I wanted to build some amplifiers to drive a lovely old pair of Infinity RS-5B speakers that my PhD supervisor gave me. I found the original articles for the AEM6000, and set to updating that to use available (at the time) transistors.

 

Since then I've done some other versions of this amp, mostly working to shrink the PCB somewhat, optimise the layout, and achieve respectable noise, THD, and slew rate specs while keeping the quiescent current under control.

 

This latest version uses one pair of Exicon ECX10N20/10P20 lateral MOSFETs, for around 50W RMS into 8Ω. It has reasonably good distortion, of around 0.0005% at 1KHz, and features rather lower quiescent power consumption than a lot of lateral FET amplifiers, because I care about that. It also does just 8nV/√Hz input referred noise, so you can enjoy your music without hiss.

 

It's a nice little amplifier, easily my favourite. I've built a bunch of them and use them for all sorts of stuff.

 

All the files for making one, including gerbers and a comprehensive construction guide, are available on my Google drive.

 

It requires surface mount assembly (MELF and SOT-23) but don't you think it's time you tried that?

 

All the components are available now.

 

 

 

AEM6000 Based 50W Amp Construction Notes.pdf 1.82 MB · 11 downloads

 Very nice work Suzy and well documented too.  Very tempted to build a pair of these.

 

Cheers,

Alan R.

[Rrobot 63]

Fantastic work Suzy! The name David Tilbrook took me back to my youf - dad built a pair of large Philips equipped 3-ways from ETI in 1980 that were my first intro to hif as a very young child. Tilbrook was the designer there too. He seems to have been quite prolific in the 1980s. 

 

Looking at all those SMDs gave me chills but it looks like a very fun and rewarding project. Extremely thoroughly documented - you've probably covered this in other posts but have you attempted to sell these as kits or finished units? They look really well designed. 

 

Thanks for sharing this with the community! 🙂 

Edited December 10, 2020 by Rrobot

[itsparks 59]

On 08/12/2020 at 4:06 PM, Suzyj said:

Are there any pcb's available to buy ?

 

 

 

AEM6000 Based 50W Amp Construction Notes.pdf 1.82 MB · 20 downloads

 

[Suzyj 56]

The idea is you send the Gerber files to a pcb manufacturer, and they send you pcbs.

 

[Suzyj 56]

On the basis that more is always better, here’s a 100W version:

 

100 watt design files

 

Same thing, but with twice as many MOSFETs.

 

I’ve only built a few of these, as the little one makes more sense for my usual active crossover speakers, but they perform quite well.

 

Edited December 11, 2020 by Suzyj

[itsparks 59]

9 hours ago, Suzyj said:

The idea is you send the Gerber files to a pcb manufacturer, and they send you pcbs.

 

That was my "BAD" i didn't see the part that you attached the gerbers till the end  sorry about that sir ! I downloaded them already.

8 hours ago, Suzyj said:

On the basis that more is always better, here’s a 100W version:

 

100 watt design files

 

Same thing, but with twice as many MOSFETs.

 

I’ve only built a few of these, as the little one makes more sense for my usual active crossover speakers, but they perform quite well.

 

Thanks for the 50 & 100 watt versions im going to order a few boards to build! Are the smaller transistors all still available ? 

[Leinster Lad 230]

17 hours ago, Suzyj said:

On the basis that more is always better, here’s a 100W version:

 

100 watt design files

 

Same thing, but with twice as many MOSFETs.

 

I’ve only built a few of these, as the little one makes more sense for my usual active crossover speakers, but they perform quite well.

 

Hiya Suzy, was wondering when this bigger version would show up 🤐

Really great that you decided to share with the DIY community

 

Just one quick question please.

 

The connectors used for the pwr and speakers ??  I've looked on mouser but there are so many that are similar, but not quite the same.

I've measured with calipers and get 2.8mm x 0.9mm x 10.4mm long.

 

I cannot for the life of me find the connector to suit ( the female one that would have the cables connected to it )

 

Could I trouble you for a part number please ?

 

The little input ones as well if possible?

 

cheers heaps

David

 

[Suzyj 56]

Hi David,

 

They're a connector I quite like, a Faston 110. They're a shrunk version of a Faston 250 (general purpose spade connector).

 

The particular flavour I use is 2.8x0.81. Here's a link to the relevant Mouser page for terminals:

 

https://au.mouser.com/Connectors/Terminals/110-FASTON-Series/_/N-5g5u?P=1ywwwf4Z1yphpxn

 

The uninsulated ones generally have a U shaped part where the wire goes in. A Hero FRH-07 crimp tool is just the ticket for crimping them (or of course you can just solder to them). I got mine from Eastern Beaver (and still chuckle whenever I read the name):

 

https://www.easternbeaver.com/Main/Elec__Products/Tools/tools.html

 

Make sure you put some heatshrink over the top.

 

The little ones are a Harwin M20, which is a 2.54mm pitch connector with 0.64mm square pins.

 

Three pin socket housing:

 

https://au.mouser.com/ProductDetail/Harwin/M20-1060300?qs=Jph8NoUxIfWw8TD5O8shNw%3D%3D

 

And crimp terminals:

 

https://au.mouser.com/ProductDetail/Harwin/M20-1180042?qs=pYcASEcTDE4r0B52MhViPw%3D%3D

 

I use a Hozan P-706 crimper to crimp these little guys. (also available from Eastern Beaver)

 

Cheers,

 

Suzy

 

[Leinster Lad 230]

Thanks heaps Suzy, now to start planning !  ( well I am a between jobs (SAP) planner )

[itsparks 59]

20 hours ago, Suzyj said:

On the basis that more is always better, here’s a 100W version:

 

100 watt design files

 

Same thing, but with twice as many MOSFETs.

 

I’ve only built a few of these, as the little one makes more sense for my usual active crossover speakers, but they perform quite well.

 

I'm ordering a pile of these boards  for my self ! I LOVE smd amplifier boards !  

 

Do you have any recommendations for good PSU ? Schematic ? 

 

Funny my Aleph turbo Is all SMD and used bottom mounted fets also !

[Suzyj 56]

I did a write-up of some of the development process:

 

http://www.suzyj.net/2017/01/measuring-mosfet-power-amplifier.html

 

In testing, I used a 2x40VAC 300VA toroidal (MT2140 from Jaycar) for the supply, with a BY224 (4.8A) bridge rectifier, and 2 x 6800µF 63V electrolytics for the rails.

[Suzyj 56]

13 hours ago, itsparks said:

That was my "BAD" i didn't see the part that you attached the gerbers till the end  sorry about that sir ! I downloaded them already.

Thanks for the 50 & 100 watt versions im going to order a few boards to build! Are the smaller transistors all still available ? 

 

Yup, all the parts are available. See the parts list.

 

Also, I'm not a sir.

[stereo coffee 405]

Suzy and I have been having a discussion within messages, concerning the circuit description she provides. The description I understand, being at odds with the term cascode. (1), (2) (3) and (4)

 

History https://en.wikipedia.org/wiki/Cascode#cite_note-3

The use of a cascode (sometimes verbified to cascoding) is a common technique for improving analog circuit performance, applicable to both vacuum tubes and transistors. The name "cascode" was coined in an article written by Frederick Vinton Hunt and Roger Wayne Hickman in 1939, in a discussion on the application of voltage stabilizers.[3] They proposed a cascode of two triodes (the first one with a common cathode setup, the second one with a common grid) as a replacement for a pentode, and so the name may be assumed to be an abbreviation of "casc(aded triode amplifier having characteristics similar to, but less noisy than, a single pent)ode".[4]

 

 

Suzy's circuit description partial.

 

“Referring to the amplifier schematic, Q1 forms a differential pair, with the amplifier input connected to one side and the feedback point (divided by 16 via R48 and R49) on the other. Q7 and Q8 form a 5mA current sink to bias the input stage. 800μA flows through each JFET, and the balance is used to bias a cascode, comprising Q3 and Q4. The cascode stage ensures the input JFETs only have 12V across them regardless of supply voltage, and also speeds the input stage up, by virtue of reducing miller capacitance effect. The current source comprising Q5 and Q6 source current into the cascode bases.”

 

 

Certainly there is stabilization at Q3 and Q4 bases, and by virtue then at Q3 and Q4 emitters  by use of the zener diode ZD1, in my opinion zeners are poor regulators (3) – however a cascode as we know it, I consider is wrongly termed for Q3 and Q4 in Suzys description as neither (being located above Q1a and Q1b) are common base ( grid or gate ) form. As a result the proper use of the term cascode appears placed in error. Whilst some small bias is allowable in a cascode & is sometimes seen with RF implementations, accompanied by a bypass capacitor to ground for RF purpose (6 )  , the usual circuit is typified by base ( grid or gate ), effectively grounded by DC voltage source

 

( (4)This upper transistor is referred to as the cascode device. Because at high frequencies the cascode transistor's base/gate is effectively grounded by DC voltage source _{Bias )}

 

Rather in your description and schematic, 12v positive is shared on both bases of Q3 and Q4, provided by current source at Q5 and Q6 to bias ZD1. There may be indeed some good attributes of the cascode present, by implementing stable voltage, as the emitter of Q3 and Q4 then assists the drain of the lower devices, however as I understand, should not be termed a cascode, as the circuit lacks the bases being effectively grounded by DC voltage source. Rather your circuit description referring to the schematic, provides a stable DC voltage source certainly, but is at 12v potential & not grounded by DC voltage source, neither is a capacitor present. 

 

I look forward to understanding ,why the term cascode is used ( rushed to perhaps ) in your circuit description. I note

the original designer did not use the same implementation (5)

 

 

(1) A cascode describes an amplifier stage consisting of a common emitter (source or cathode ) stage followed by a

common base ( grid or gate ) stage. page 89 Designing with field effect transistors Siliconix Inc second edition

(2) https://en.wikipedia.org/wiki/Cascode#cite_note-3,

(3) https://www.diyaudio.com/forums/parts/35821-noise-measurements-leds-zener-diodes.html

(4) https://wiki-stage.analog.com/university/courses/electronics/text/chapter-10?s[]=cascode

(5) https://archive.org/details/ETIA1981/ETI 1981-01 January/page/n23/mode/2up

(6) Basic Radio Tepper  2nd revised edition 4-78

[Suzyj 56]

I suspect the whole diffamp thing is confusing you.

 

Can you agree that the node at the junction of R4 and R5 is a virtual ground due to the symmetry of the differential amplifier?

 

If you do that, then split the whole input diffamp down the middle and look at one side. Q1a source goes to virtual ground via a degeneration resistor. Q3's base is held at 12V wrt our virtual ground, due to the 12V zener. The node formed by the emitter of Q3 and the drain of Q1a is at 11.3V.

 

Our load resistor (R6) hangs off Q3's collector. It's a straightforward, uncomplicated cascode.

 

It's perfectly valid to do cascodes (and folded cascodes) in differential circuits. Here's a nice set of slides explaining how folded cascodes work for opamps: http://www.seas.ucla.edu/brweb/teaching/215A_F2014/opamps.pdf

 

Regarding other assertions:

 

Yes, zeners are noisy (indeed I've used them as RF noise sources professionally). When used as shown in the amplifier, it's noise contribution is negligible because it's common-mode, and the amplifier has high CMRR. The proof is in the pudding. The actual measured input referred noise of this amplifier is about the same as an OP134, at 8nv/√Hz.

 

Why are you talking about the ETI477? It's a fine amplifier, but it's completely different to the AEM6000, which was published in Australian Electronics Monthly in 1986.

 

 

  

[andyr 6,058]

4 hours ago, Suzyj said:

 

The proof is in the pudding.

  

 

You know your technical stuff, Suzy (much more so than I do!   ) - but the aphorism is ... "the proof of the pudding is in the eating "!  

 

Andy

 

[Suzyj 56]

Oh no! I’m usually such a grammar nazi.

[stereo coffee 405]

17 hours ago, Suzyj said:

I suspect the whole diffamp thing is confusing you.

 

Can you agree that the node at the junction of R4 and R5 is a virtual ground due to the symmetry of the differential amplifier?

 

If you do that, then split the whole input diffamp down the middle and look at one side. Q1a source goes to virtual ground via a degeneration resistor. Q3's base is held at 12V wrt our virtual ground, due to the 12V zener. The node formed by the emitter of Q3 and the drain of Q1a is at 11.3V.

 

Our load resistor (R6) hangs off Q3's collector. It's a straightforward, uncomplicated cascode.

 

It's perfectly valid to do cascodes (and folded cascodes) in differential circuits. Here's a nice set of slides explaining how folded cascodes work for opamps: http://www.seas.ucla.edu/brweb/teaching/215A_F2014/opamps.pdf

 

Regarding other assertions:

 

Yes, zeners are noisy (indeed I've used them as RF noise sources professionally). When used as shown in the amplifier, it's noise contribution is negligible because it's common-mode, and the amplifier has high CMRR. The proof is in the pudding. The actual measured input referred noise of this amplifier is about the same as an OP134, at 8nv/√Hz.

 

Why are you talking about the ETI477? It's a fine amplifier, but it's completely different to the AEM6000, which was published in Australian Electronics Monthly in 1986.

 

 

  

 

Thanks, indeed R4 and R5 are a virtual ground, mainly providing Q1a and Q1b sources potential & ZD1 's anode ability,  at its cathode then, to zener at its avalanche voltage.  The test I consider to validate cascode,  is no signal phase inversion inferring cascode , or inversion not cascode at q3 and q4 collectors, relative to q3 and q4 emitters   I might have to build the board to test, just to see. 

[Leinster Lad 230]

On 12/12/2020 at 9:33 AM, Suzyj said:

They're a connector I quite like, a Faston 110. They're a shrunk version of a Faston 250 (general purpose spade connector).

 

The particular flavour I use is 2.8x0.81. Here's a link to the relevant Mouser page for terminals:

 

https://au.mouser.com/Connectors/Terminals/110-FASTON-Series/_/N-5g5u?P=1ywwwf4Z1yphpxn

 

Hi Suzy, I must be having an old bloke look, but for the life of me I cannot find the Faston 110 terminals that go into the board.

 

Would it be possible to trouble you for the part number please ??

 

Thanks heaps

David

[itsparks 59]

1 minute ago, Leinster Lad said:

 

Hi Suzy, I must be having an old bloke look, but for the life of me I cannot find the Faston 110 terminals that go into the board.

 

Would it be possible to trouble you for the part number please ??

 

Thanks heaps

David

Just solder wires into the holes ?  

[Leinster Lad 230]

1 hour ago, itsparks said:

Just solder wires into the holes ?  

I like the flexibility that connectors give.

[itsparks 59]

Just now, Leinster Lad said:

I like the flexibility that connectors give.

I guess, im really not a fan of pin mechanical connectors my self. 

[Janus77 386]

Jason, i went bungy jumping in Nanaimo once, nice place..

[itsparks 59]

Just now, Janus77 said:

Jason, i went bungy jumping in Nanaimo once, nice place..

I love the island! Much different than Vancouver city !

[Janus77 386]

I'd like to get up to Mt Cain for a ski one day, a long way from everything.

 

(sorry for the OT)

Edited December 16, 2020 by Janus77

[Suzyj 56]

2 hours ago, Leinster Lad said:

 

Hi Suzy, I must be having an old bloke look, but for the life of me I cannot find the Faston 110 terminals that go into the board.

 

Would it be possible to trouble you for the part number please ??

 

Thanks heaps

David

TE 63756-1

 

https://au.element14.com/amp-te-connectivity/63756-1/terminal-pcb-tab-1-4mm/dp/2841308

[Leinster Lad 230]

Thanks Suzy, that part number at mouser is in 1000 piece lots !

[andyr 6,058]

16 minutes ago, Leinster Lad said:

Thanks Suzy, that part number at mouser is in 1000 piece lots !

 

Maybe try RS Components?  Ask them what spades, 488-1241 connectors fit onto.  (I use these to fit onto SLA battery lugs.)

 

I think the right part number is 719-7031.

 

Andy

 

Edited December 16, 2020 by andyr

[Leinster Lad 230]

No worries Andy, no need to buy 1000 of them as  I can get them individually at Element14

[itsparks 59]

Has anyone built the 50watt version of Suzy's amps yet ? I'm ordering PCB's for both soon and just curious !

[soundbyte 1,478]

.

Edited December 20, 2020 by soundbyte

[Suzyj 56]

I look forward to seeing the results of your simulations and measurements of your proposed changes on a board.

[soundbyte 1,478]

Why bother, trying to help but your mind is obviously made up that nothing can improve the circuit, nothing I can say or do will change that, a pity.

Moving on.

[Suzyj 56]

Okay then, I'll go along with this. Here's what I've done:

 

1. In simulation: Insert a 10mV 10KHz AC voltage source in series with the zener, and simulate the circuit. Measured AC on output is 450nV, giving me 87dB rejection of voltage noise at this point. Like I said, good CMRR.

2. In reality: Stick a 12V zener in series with a 1K resistor. put that across my lab supply and put 5mA through the zener. Measure the voltage noise density across the zener with my 3585B. I measure 79nV/√Hz, flat across the band. For reference my 3585B does 11nV/√Hz with it's input shorted. It's a real beast for noise measurements.

 

So, combining these, the noise contribution of the 12V zener used for biasing the first stage cascode is 3.6pV/√Hz output referred, or 220fV/√Hz input referred (gain of 16).

 

Compare this with the 8nV/√Hz input-referred contribution of the input J-FETs.

 

Regarding the 2 -transistor current source oscillation stuff. I've read about this dozens of times, but am yet to see an actual measurement of this happening. I've simulated and measured real versions of these things to death trying to find anything, and cannot.

 

I'm all for making improvements, but they have to be grounded in reality, not lore.

 

There is a lot that can be done to improve this circuit, and I'm all ears for good ideas. Over on DIYAudio there's a thread for just that.

 

Thus far I've found the biggest improvements are to be found by inserting a folded cascode for the first stage load, and using a current-mirror load for the first stage (higher impedance gives better OLG). I've had a play at trying to come up with a current mirror load for the second stage, and it works well in simulation, but I'm not convinced it's anything but a simulation fluke. I use the folded cascode current mirror load on another amplifier (google DIYAudio cascaded diamond) and it's an absolute ripper there - it improves the PSRR, and makes that amplifier work on any power supply from about +/-5V to +/-20V. I've also played a lot with VAS stages. I'm playing with a cool 200W version with a class-H output stage ripped off from a NAD3240PE, but done with FETs, that should make an easy 250W RMS into 8Ω, with ~0.001% THD at 10KHz, with a quiescent dissipation of around 10-12W (vs ~35-60W for a normal class-AB latfet of this power).

[andyr 6,058]

Wow, formidable (spoken the French way)!  

 

I'm hoping you are also an LTspice whizz (as I have some Qs   ).

 

Andy

[itsparks 59]

Anyone know where we can get the SST404 dual matched N-JFET units. "Q1" 

[Suzyj 56]

There are two manufacturers that I know of:

 

CA Logic, distributed by future electronics, and

 

Linear systems, distributed by micross.com

 

[itsparks 59]

10 hours ago, Suzyj said:

There are two manufacturers that I know of:

 

CA Logic, distributed by future electronics, and

 

Linear systems, distributed by micross.com

 

Yeah i found them at future Electronics,

 

Looking for the EXICON ECW20N20 and EXICON ECW20P20  units are 15$ each Euro.  

 Any substitues for these instead of something that is really hard / expensive to get.