Monitor Amps - Class A Monster Madness ?

[alexc]

Hi

Anyone else caught the 'solid state Class A' amplifier  'bug'  out there ?

I've 'been there, done that' with tube amps of the class A variety, up to around 20Wrms 'music power'

BUT

now, I have an 'itch to scratch' for a solid-state version

Completely stupid, in terms of cost-heat metrics, and even performance wise. But, as PT Barnam remarked about 'a fool' and 'their money' likely being 'parted', it's my next silly audio obsession.

Just started buying parts now for a pair of monoblocs and seperate  humungous power supply

  - chassis with large heatsink for rectifiers
  - 500VA toroid,  with option of adding a second one as time and madness reach fruition
  - monte  bruto forza filtering, bipolar, unregulated C-L-C-R-C
  - psu output of around +/- 35Vdc at 5A continuous per polarity from 35Vac secondaries
  - monster can caps in pairs of 22000uF, 60V  for a total of 66000uF each rail
  - stupidly dimensioned inductors with 'permite' core under alu frame with15 awg wire for 15mH each
  - vehicle trailer heavy duty cable with 5 way conductors and heavy duty connectors
  - heavy duty mains filter
  - high current schottky rectifiers, full wave, one for each secondary winding
  - quenching cap and a few poly film filtering caps for the secondaries as well
  - 50W  dropping resistors to get from 40V to 35V on each rail under load (5A)
  - slow start module to prevent blowing the house fuses
  - voltage and current meters

That's going to run me something like 650usd

THEN, the actual amps .....

- JLH 1996 25Wrms modules with additional tweaks, including 4 power BJTs per channel
- option of adding additional amp module per monobloc for 50Wrms power
- monobloc chassis with monster heatsinks able to sink heat up to 50Wrms each
- speaker protectors and fusing
- alps Level potentiometer
- transformer balanced inputs with Jensen 10K/10K

Amps will be around 350usd for the pair


SO - around 1Kusd for the pair at 25Wrms each channel and an additional 200usd to double the output power to 50Wrms per channel and/or allow biamping 

Only a complete loon would biamp with such grossly inefficient amps as class A!

Talk about daft, eh?    BUT, when you've (almost!) done all the tube amps and limiters you are going to , what other over-the-top vistas are there ?

And not crazy expensive, considering US made commercial types like this can run to many-many thousands!

Of the professional builders,  Nelson Pass comes to mind .... with the Zen Class A and more/less  refined variants ...

Anyways, it'll take a while, prolly a year or so,  to really get finished  .....

I bet there are some other crazy DIY class A guys out there - am I right?   

----

Here's a pic of the psu chassis - in gold and red  

The monoblocs are the same but silver livery with black heatsinks.

They aren't absolutely huge, being only 240mm across  and 260mm high on the outside.

And that suits the 2x monoblocs + remote PSU approach very nicely.

 

[alexc]

Oh yes - and I live in an area that is predominantly good and cold 9 months of the year, being one of the most southern civilised places in the world  ..... at the bottom of Tasmania, Australia.

In my music room, a little (or a lot) of heat is welcome! Saves me using the heater so much 

---

Here's the John Linsley-Hood Class A 1996 amplifier with mods by Geoff Moss and Tim Andrews, circa 2003

More info at the 'Class A Site' http://sound-au.com/tcaas/jlhupdate.htm, created and maintained by Geoff Moss.

Thanks, of course to Mr J L Linsley-Hood, Geoff Moss and Tim Andrews 

 

[alexc]

In keeping with the DIY nature that I love, I'll be building the speakers to go with ..

I have some wonderful late 60s japanese 60Wrms 12" alnico woofers, paper cone and cast alloy frame.

I'll be teaming them up with some Fostex 6" full range drivers for mid and high frequencies. No additional tweetering required!

I'll make up some crossovers after I've prototyped the frequency using cheap commercial active crossover units.

I think that will run at around 600usd including solid timber boxes (!) and 60s Sansui timber-lattice grills 


This time, this WILL be my last new build

 

[emrr]

Awesome!

You should look at this:

http://www.proaudiodesignforum.com/forum/php/viewtopic.php?f=6&t=679

 

[ruffrecords]

I remember reading the original article. I seem to remember he used 2N3055s in the output and a huge heatsink.

Cheers

Ian

 

[ruffrecords]

Awesome!

You should look at this:

http://www.proaudiodesignforum.com/forum/php/viewtopic.php?f=6&t=679

Link gives me error 404.

Cheers

Ian

 

[alexc]

Yep - that's the one. Trusty old 2N3055s in the 1969 version that he published at that time in Wireless World.

I remember boxes of them original Motorola 2N3055 devices lying around, at my first job - elec eng in the public railways in the early 80s.   

If I known they would eventually be worth 10usd a piece, I would have commandeered some  !

Some are using more modern devices -  this version, the Geoff Moss and Tim Andrews' mods, uses MJ15003.

These are (only) going to dissipate something like 25W of heat continuously each, as I will have doubled up on the JLH 1996 design's pair for the same Iquiesscent. 

So thats 4x MJ15003 per channel -  somethink like 100W on each side heatsink of the monblocs for the 'fully blown' version.

200W per monbloc of heat, for a total of around 500W for the whole extended system.

WELL, my electric space heater does 1000W at idle and it makes very little sound!

 

[alexc]

Thanks Doug - that link thread surely parallels what I'm feeling right now, alright 

"it's like looking into a living snow mirror" 

Plenty of great discussion there for me to work thru from the esteemed Wayne Kirkwood and our own JR and Ricardo.

Brilliant - I think I will find it hard to resist doing the Wayne's DCOA2 module as part of my own 'research' 

All his work on the subject for a very, very reasonable price 

Just awesome - I was flagging a bit on my diy efforts of late. Now I have a renewed vigour in this area of  heaters-that-make-sound.

 

[ruffrecords]

Works here.

Odd, it works here now???

Cheers

Ian

 

[alexc]

To go with these forthcoming JLH based amps, I'm knocking up a studio utility box using similar tech ....

Using recycled parts and going for minimal cost, I'm doing up a 2ru box with these uber cheapy ebay modules :

- JLH single ended Class A based stereo pre amp  stage with  active baxandall (ne5532) 3 band eq
http://www.ebay.com/itm/Class-A-NE5532-2-3300UF-Tone-Board-Finished-Board-Volume-Control-Board-/321872846500?hash=item4af11ddea4:g:5iQAAOSwFnFWAPtH

- JLH single ended Class A based stereo line amp stage with additional traffo bal outputs
Still deciding which

- discrete regulated dual output PSU module
Still deciding which

I add an  8 input passive summing network to stereo busses to feed the above modules.

Its going to be used to passively combine 4 unbalanced sources (into a transformer for balancing), and sum that that with 4 balanced sources.

Box total should be around the 200usd mark

 

[alexc]

and here's the first spectra of the JLH Class A module, configured as a headphone amp, typical listening level driving 60 ohms headphones with the Motu interface across the outputs.

This is single ended version, Vcc +9V, 100mA approx Iquiescent.

Pretty low distortion products    - the THD is 0.001%, the THD+N is higher at 0.0325% (20KHz BW) mostly due to the LM317 regulation.

My box will have quite a bit less ripple etc (hopefully) from the use of class A discrete regulation. Also, I'll be driving higher impedance outputs, as well as higher +Vcc and control over Iq, so that should improve things moar

Can't wait for the parts to start getting here  

 

[alexc]

With some moderate heatsinking, using TIP41 bjt for output devices. Can be good for up to 3Wrms into 8ohms in this setup.

The switchable output traffos will be there for adding THD   

Here's my Motu 828Mk2 interface with analog balanced 'loopback' baseline ....  20KHz measurement bandwidth.

Its pretty low THD+N, at 0.0003% at +1.0dBu output levels ...

It has +16dB of headroom to clip, where the THD+N rises to 0.0004% and finally to 0.0008% at fully clipping  a further 1dB beyond onset.

I think the JLH headphone amp I'm fooling with so far, will get down to 0.0007% or so when I have it powered properly.

ie. output level is set for 'normal listening' into a 60ohm headphone set, and the REW measurement set's Motu 10K inputs are across that.

Interestingly enough, I have the JLH stereo headphone amp set up as balanced 

ie. the Motu's balanced out + signal and -signal have very symmetrical voltage waveform with respect to ground, and of opposite phase.

So I drive the Left and Right Inputs (rca phono jacks)  each with the + and - signals.
The outputs are fed to the respective + and - signals of the Motu balanced inputs.

The Motu and JLH stereo pcb symmetry throughout is pretty impressive - the THD+N figures are very low.

It's not always the case that IC 'balanced' signals have voltage waveforms that are symmetric wrt ground.  It sure makes testing/messing about  easier when they do.

So - all good - one can happily use an unbalanced rca stereo JLH headphone amp pcb as a mono balanced line stage of good driving capability with *alot* of fidelity, at low-ish amplitude signal levels.

 

[ruffrecords]

and here's the first spectra of the JLH Class A module, configured as a headphone amp, typical listening level driving 60 ohms headphones with the Motu interface across the outputs.

From the graph it looks like the actual output level is about -10dBu or about 245mV. You said the phones are 60 ohms which means the actual power delivered is only 1mW. Is that right?

Cheers

Ian

 

[alexc]

In the REW spectra above, the hphone amp pcb has a fader at the intput RCA jacks,  which is turned most way down.

I'm still making up my 'test jig' for this kind of testing more 'proper'

ie. device-under-test 
                ->  a stereo headphone amp with a pair of stereo unbalanced RCA input and output jacks

        test-set
                -> Motu 828Mk2 with balanced inputs, outputs and REW


Until I get my test-harness right, I am driving the headphone amp L, R unbalanced inputs with each phase of the Motu balanced output wrt ground.

Same deal at the outputs.

The gain structure, excluding the 'bridging effects'  is something like this :

REW send level :  -20dBFS -> +1dBu 
HPhone AMp    :  -28dB (fader) + 18dB(fixed voltage gain) + unity gain (power BJTs)
REW return        :  -30dBFS -> -9dBu

So yes, about -10dBu is about correct for the return level.

The spectra is done this way :

Source

- connect both stereo RCA input jacks to one Motu bal line Input in 'L-R bridged mono mode'  (bal)
- send -20dBFS 1KHz test signal (equivalent to +1dBu)

Load

- connect stereo headphones (unbal) as load
- run music and set fader for typical listening -> fader set to -28dB (attenuation)
- disconnect headphones
- replace with one Motu Line Input in 'L-R bridged mono mode' (bal)

- run spectra with -20dBFs (+1dBu)
- take snapshot


SO - I haven't worked the numbers properly yet. I'm not sure what the 'bridged' mode of the load comes out to. 10K ?

And I haven't yet checked the voltages on my  CRO for the 'bridged' modes of operation.  I know it's calibrated when  using balanced connections, and my load box is calibrated for speakers.

I figured to wait until my PSU is up to the job with the much lower ripple of a dedicated class a module.

Also, I havent yet messed with +Vcc and measured Iq ; again, waiting for some psu parts to get all that going right.

So, that was the basic test. Turning the fader up quickly clipped my Motu, with onset of clip doing 0.1% THD at somethink like 8Vpp (off +9V supply).

So - early days yet in terms of getting fully familiar even with this simple circuit. 

Once I get my bearings, I will make up a balancing/de-balancing/inverting/amplifying  test-set based on this.

The transformers I'm using for test purposes have a reasonably high THD at high output levels, like 0.1% or so. If I can better that, it will be worth the efforts.

Also, I have to work out my 'dummy load box' settings for a range of power resistors  from 16ohm .. 600ohm - at present I only have scaling and loads for speakers 4,8,16 ohms at power.

Thanks for the interest -  I will get my  act together in a while 

----

Heres the spectra of an Edcor XSM 600:15K step up - it fares pretty well on  THD. This is what I need to beat in a test peice doing bal/unbal, gain/atten/invert duties 

 

[ruffrecords]

I am even more confused now. So you disconnected the headphones to make the measurement so there was no longer a 60 ohm load on the output. So what was the load when the measurement was made?

Cheers

Ian

 

[alexc]

Yep confusing, which is why I kept the original JLH amp spectra a bit 'qualitative'.

I'm just doing a further test.

So ..

-----

Here is the loopback spectra balanced input to balanced output - baseline for further tests.

NOTE - there is a +21dB factor in converting from the application level 'dBFS' readings to 'dBu'

I've 'verified' this in my 'calibrations' with various applications' reported dBFS figures and absolute voltages as measured to the limit of resolution of my CRO/DVM

 

[alexc]

And here is the JLH 1969 ST Headphone amp pcb with 'bridged' inputs and outputs baseline .... this is just the Motu interface as load for the headphone amp pcb.

Again, I'm not sure of what the load impedance is in this 'stereo unbalanced to mono balanced bridged' mode is.

Probably each channel of the headphone amp pcb sees 10K to ground .. I think.

The voltages wrt ground are 1Vpp input to each channel, and 7.6Vpp out of each channel into each phase of the Motu inputs.

PSU hash is  5.27mVpp on +15Vdc

NOTE - there is a 6dB factor (x2)  converting from 'phase-ground' voltage measurements done with CRO/DVM to Motu's balanced phase-phase application level dBFS readings.

I have 'verified' this in my 'calibration' of my Motu's reported dBFS figures in various applications, as well as against absolute voltages measured into 600ohm loads to the limit of resolution at the CRO/DVM.

So 7.6Vpp CRO measured 'phase-ground' notionally becomes 15.2Vpp  'phase-phase' or 5.37Vrms or  +16.3 dBu (which I annotated) and finally to the -4.3 dBFS reported in the REW spectra above (by subtracting the aformentioned 21dB!)

Just to clarify for the eagle eyes out there
.....

So -  into balanced 10K loads (most likely), and pushing output levels to the onset of clip of my audio interface, its a pretty hi performing widget.

Signal-hum-margin of  +97.7dB,  THD at 0.0036%, THD+N  at 0.0057% driving +16dBu into 10K balanced load

(the noise floor could be lower with proper PSU)'

and frequency / phase response are :

20Hz    -0.41dB and +0.4 degrees
20KHz  -0.17dB and  -1.2 degrees 

There is a filter at each input - a 470uF electro, a film cap and a  inductor. I think it is limiting the BW at the top end.

One could expect some cancellation effects going on due to the 'Stereo-unbal  to Mono balanced bridged' mode of operation. That would be most obvious in the H2 harmonic.

I am yet to test each amp in isolation, but I would think it would be a little higher in THD.

 

[alexc]

So now onto a lower load, say nominal 60ohms, which is the specified impedance of my AT headphones.

Same conditions, but with 60.5 ohms loads ( a pair of paralleled 120ohm, 0.75W metal film resistors).

So that's 60ohms load, across each output phase to ground, with the Motu's 10K input impedance in parallel, also each phase.

Motu's high input impedance  shouldn't cause any additional loading effects.

The pcb input fader is all the way up now and applied signal is a bit higher too, to get to the onset of clip at Motu balanced In.

1.32Vpp  in  at each rca jacks wrt ground.
7.91Vpp out at each ST 1/4" phono jack pole to ground

THD and THD+N have increased by an order of magnitude, as expected, with the unit going all out into low Z load

and

hum floor unchanged
output level is mostly unchanged.

I estimate the output  power levels to be approx 115mW into 60ohms each channel.

The spec says 200mW into 120ohms which would be some 4.9Vrms.  I think it would come close to that, but I will check later.

Pretty decent fidelity at a reasonably high output level into a fairly heavy (low impedance) load.

I think I can improve that some, with better PSU
ie. higher, cleaner supply will increase the low THD performance

Of course, to get there requires more and more Iq and power etc.

 

[alexc]

OK - finished the edits and trying to make sure I don't confuse anyone unnecessarily 

Some confusion however is both a reality AND a requirement! Those would be the necessary unnecessaries.

Its the necessarily unknown necessaries that confuse me.

Next test will be  'THD  into 10K load  at onset of of JLH clip'.

That requires me to scale the output levels down from the mighty JLH 1969 amps to the Motu's puny input voltage range. 

Time to rejig my load box

 

[alexc]

here's a snap of the Motu 828Mk2 analog loopback at the onset of clip, which is +18.2dB .... apparently

I'm just going to recalibrate and make sure everything is where I expect it to be .... do a quick look at some historicals.

while I await some parts.

....

I've also a mind to do a mod of this circuit to use the 'front end' from the neve-e 438 amp stage. Like taking the first two transistors in the 438 and substituting them at the front of the jlh 1969.

I would like to know more about applying a balanced signal to the base and emitter of the pnp with nfb to the emitter of same. Its not too different in the jlh as it is, but sans the -ve input.

Also, as one might expect, staple a JLH amp onto the back of a neve-e n1272 type of thing.

Funnily enough, I have some such to recycle after I broke up my  units #0001 and #0003.

Their time had come - my strict '3 strikes and you get broken up' policy to thin out the herd and stop me wasting an inordinate amount of time in maintenance 

So #1 will become  #1_a  and get some big ass heatsinks before hitting the street (my music room) again with attitude.