Monitor Amps - Class A Monster Madness ?

[alexc]

And in the meantime, I just started to get some parts for a decently apportioned bench PSU.

I've gone for this ebuy module  http://www.ebay.com/itm/LV30B-5A-Variable-Positive-Negative-Voltage-Regulator-5A-1-5-30V-DIY-Kit-/222252198964?hash=item33bf436c34:g:~CQAAOxyhodRuzWJ

Its dual output, based on LM1084 LDV regulator for not too many dollars.

Supposedly can do a max of 5A and can do parallel/linked/bipolar operation depending on jumpering.

I'll hook it up to some 34-0-34 Vac secondaries and set it up for use as an adjustable up to +/-37Vdc at a max of 3A or so.

Or at a lower voltage and current most of the time.  In a simple box with an old computer fan and a 200VA toroid etc.

It has power on delay relay protection, which is handy when flinging this much juice around in the hope some makes it to the speaker! Just the thing for bench testing. I'll be interested to see what ripple one gets at 1.5A  per rail. 

It's worth the bother to get a better bench supply - I can see that discrete class A is where I'll be playing for a while, and the power gets chewed up pretty quickly ...

 

[alexc]

And for the rack unit I'm doing up, these JLH 1969 modules ...

http://www.ebay.com/itm/TIP41C-JLH1969-single-ended-class-A-Power-amplifier-kit-2-channel-/322270574723?hash=item4b08d2b883:g:tbgAAOSw4shX5PLq

Cheap as chips (literally!) and the next level up in power handling TIP41 package.

Supposed to be able to do 10Wrms into 8ohms with sufficient heatsinking and psu - more like 5W peak before meltdown.

All good fun and as I'm finding out, the performance is surprisingly good for 2 power bjts and 2 signal bjts per channel.

 

[alexc]

and lastly for the day's work, some of these are on the way .... for my Hot House Class A monoblocs 


Six of these,  to be precise! - the single biggest spend of this series, weighing in a 115usd delivered way, way down down.

The toroid is a bit less than that, as are each of the 3  individual chassis-es

 

[alexc]

A bit more testing  ....

Did some driving of Tamradio 600-600 ohm 1:1 output transformers from an old Ramsa japan desk.

Pretty good results, really. THD 0.005% driving +16.3dBu with a signal-hum margin of 93dB.

Again done with the L and R unbalanced RCA  inputs to the pcb driven by the + and - phases of the Motu balanced out.
ie. 'mono balanced bridged' mode inputs

That's good news for me, because it means I can use these cheap and compact modules to drive low Z traffos with good performance.

I have yet to get a feel of the performance using these modules in 'stereo single ended inputs and outputs' mode.

I'm pretty sure that I'm getting a lot of the low THD and noise from the use of stereo channels as a mono fully symmetrical gain stage.

It's food for thought as it ideally quadruples power into a load ..  and all that. I've not done this thing before with power amps, believing that the THD would be quite a bit higher  .... due to the reduction of the Zload effectively seen by each amp.

But, it's looking like a really good way of getting lower THD, whilst maximising the pathetic efficiency of the CLass A thang.

 

[alexc]

I've also gone for this module from  ebuy ...

http://www.ebay.com/itm/DC-Version-NE5532-OP-AMP-Preamlifier-Pre-amp-with-ALPS-27-Pot-Shield-Cover-/182196144801?hash=item2a6bbc8aa1:g:kjYAAOSweWVXe4yN

Its a stereo, NE5532 dual opamp chip minimal  preamp pcb with fader. Unbalanced ins and outs. I think non-inverting, but no spec.

I'm going to use it to do some comparisons on these Class A modules - the NE5532 is as we all know, a very good performer for a dual opamp package.

I like this module for the shield surrounding the chip, which apparently reduces noise.  It's significant, because I plan to do some fooling about with opamp subs. I doubt I'll get 'better' performance than the NE5532, but interesting to try.

First sub is the mighty TI LME49860 AN  ....  it is a specification star :

• Easily Drives 600Ω Loads
• Power Supply Voltage Range: ±2.5 to ±22V
• Optimized for Superior Audio Signal Fidelity
• THD+N (AV = 1, VOUT = 3VRMS, fIN = 1kHz)
• Output Short Circuit Protection – RL = 2kΩ: 0.00003% (Typ)
• PSRR and CMRR Exceed 120dB (Typ) – RL = 600Ω: 0.00003% (Typ)
• SOIC or PDIP Packages
• Input Noise Density: 2.7 nV/√Hz (Typ)
• Slew Rate: ±20V/μs (Typ)
• Gain Bandwidth Product: 55MHz (Typ)
• Open Loop Gain (RL = 600Ω): 140dB (Typ)
• Input Bias Current: 10nA (Typ)
• Input Offset Voltage: 0.1mV (Typ)
• DC Gain Linearity Error: 0.000009%

The people who can design and manufacture something like this, with a one-off 5usd price tag are clearly and wonderfully, total geniuses   

Texas Instruments have to be the rockingest chip house on the planet. I used to love programming their floating point TMS dsp chips back in the day ...

ANYway ... it will be good to 'get back' to some opamp based 'reality checks'.  I've been working on tubes and discretes for a long time. One can get a little 'class A' kool-aid happy at times 

Now I just need to find an inverting buffer module.

 

[alexc]

Found an inverting opamp pcb that suits me - tick

Now for a bench power amp of the Class AB variety - I need a high efficiency super low distortion module so I can drive tube finals' in my performance measurements. And, of course, cheap!

It will run off my bench PSU which is to be +/- 52V dual rails, low ripple and up to 3A max continuous with a 5A peak.
Using the LDV regs of the LT1084 type ...

Gone for a super simple module based on the (formerly National) TI LM49811 'power amp driver' device on a pcb with moderate on-board heatsinking, with BJT power devices for current gain.

http://www.ebay.com/itm/Dual-DC-20-80V-24V-48V-LME49811-Mono-Channel-100W-HIFI-Audio-Amplifier-Board-Kit-/122014073238?hash=item1c689b0196:g:~UMAAOSwr41XHuhq

They will be mounted on a suitable medium large heatsink with a switchable fan - for a bench amp this is no problem.

What it needs is super low THD. This TI LM49811 is supposedly capable of forming a >100Wrms module with THD+ of around 0.0004% (+/-50V supply, unloaded, 20Vrms out, 30KHz BW).

Now that's what I'm talking about - about the same as my Motu 828Mk2 balanced analog loopback test, with invididual harmonics at the limit of REW resolution of 0.0001%.

For 20usd per channel  delivered way down south 

......

my current bench amp is a Yamaha A-500 'Natural Sound'   

A good consumer hifi amp back in the early 90s, rated at approx 100Wrms into 8ohms.

Here's a spectra from it into an 8ohm dummy load
->  0.1% THD at 93.6Wrms into 8ohms, with noise level 72.5dB below fundamental.

That's not too bad, all in all - but far short of modern 'hidef audio expectations' 

I want to beat this by several orders of magnitude!  We shall see 

 

[alexc]

So now that I have the modules necessary for 'high def audio'  for the  functions of :

      debalancing, adjustable gain, inverting and balancing
    100Wrms capable amplifier

in addition to a 1RU rack with XLR connectors,  a JLM 5 Rail adjustable PSU  for low-voltage and toroidal power transformer

I can get real about 'measurement accuracy' to the next level

Once all that is setup, I'll resume my Class , Class B and Class AB investigations 

 

[alexc]

While I wait for parts, here are some baselines that I use all the time and am looking to 'better' the performance of  ...  in my DIY noodlings ....


First up -  my music room amp - a Pyle Pro PT-2000

which is a 2 channel power amp of the classic Class AB type, done discrete with BJTs all round.

Its quite a decent amp, weighty in both .. weight  *and* power  -  it is sturdily constructed - mostly utilitarian but with a couple of nice features.

Here's the front panel ...  it cost me around 200usd back circa 2000, including long distance shipping to my door!

I've used it regularly since then to power my Tannoy System 8 nearfields without incident.

I did replace the fan for a modern, quiet unit a few years back - it is near silent.

 

[alexc]

and the inside pic ...

- *very* big psu traffo
-  single large amp pcb with some reservoir capacitors and a dual speaker turn on delay relay
-  single quite modest sized heatsink
-  small fan at back
-  large VUs with surprisingly accurate and responsive movement and 2 sensitivity settings on a front panel switch
-  simple passive headphone output

It has a good array of connectors, balanced xlr inputs, which appears  to use active debalancing  - not quite sure since I don't have the schematic.

All round a very compelling package for such an 'affordable' unit 

 

[alexc]

Here's the basic  spec ....

purports to make 'a maximum' of  1000Wrms per channel at 4ohms, both channels driven.

That would make a peak capacity of 2KVA I guess, at the psu traffo! Golly gosh  ;D  .  In any case, it surely is a very large  unit - an E core type.

The primary fuse is rated at 240VAC 9.5A. I think it is a fast blow type.

The PSU is completely unregulated on the 65Vac CT - it has a 20A metal bridge rectifier,  heatsink mounted and  well greased. A pair of 10,000uF 80V 105C resevoir caps and that's it. 

The small signal discretes  look to have a discrete regulator circuit.

One other notable features, again simple, but well executed  (I know cause I trip it regularly when testing!)  is the  'short-circuit output protection' and 'heatsink overheat protection'. Both of which shut down the current, without noises and return the current flow when the fault condition is lifted.

The quoted figure for 8ohms 120Wrms continuous is probably a bit conservative but realistic.

The psu traffo measures  +/- 61Vdc  and it has an auxilliary winding to supply the meter opamp and lights etc.

 

[alexc]

In music usage, it never sees much more than 20Wrms   All that capacity is never much used.

Typical usage is more like 16Wrms of music - it's not a big room .

The goal is to replace it it with some 25Wrms 'Class A heat monsters'   

SO - this guy is becomes my 'bench amp', for testing and the like ... - I want to utilise the supply rails to feed an external  regulator unit at my bench -  that I can use for my more 'powerful'  diy noodlings .....

The +/-61Vdc ripple I measure is something like 7mVpp at idle, increasing to 1.7Vpp at heavy load.

At the speaker terminals, that measures something like 5mVpp at idle, which is about the same as the ground hash. Under heavy load, that increases some.

Again, I'm hoping to improve on these performance figures with my diy psu regulator unit.

SO - here is the first snapshot  .....    all pretty relaxed at 32Wrms into 8ohms    0.0031% THD with 0.0132% THD+N.

The noise floor is a bit higher than I'd like -  with ripple of 7mV on the rails and a similar amount on the ground lines.

 

[alexc]

At the maximum reasonable end of the performance range, here is the snapshot for 134Wrms into 8ohms


Pretty good really - 0.0106% THD    and  0.0169% THD+N. 

H2 dominating H3 and H4+  -  with a nice even, decreasing ramp shape.

So that's what I would rate this amp as 135Wrms into 8ohms per channel .

 

[alexc]

So - how low can I go with class A ? 

I'm hoping to get to  around 0.009 % at 25Wrms into 8ohms. 

We shall see 

 

[alexc]

And just for laughs, here is the TL071 opamp  ..  in single-ended supply, diff amp config.

Also known as the JLM VU buffer  (sans rectifying diodes in the load)

Pretty good THD eh?  Not much there!

This is an example of what I am hoping to beat in my class A small and med power category 

And surely I should get better result with some of the new fangled lme opamps and driver chips .

 

[alexc]

and finally, this is the 'max rated power' THD snapshot for my own el34 edcor monoblocs with just 7dB of nfb  ....

15.5Vpp into 8ohms  load box  -> 30Wrms    with  0.187% THD+N    and    102.5dB signal-hum margin

 

[ruffrecords]

and finally, this is the 'max rated power' THD snapshot for my own el34 edcor monoblocs with just 7dB of nfb  ....

15.5Vpp into 8ohms  load box  -> 30Wrms    with  0.187% THD+N    and    102.5dB signal-hum margin

Is this a PP amp? I guess it must be to have 30W output. In which case it is strange that the even harmonics are not canceling.

How do you know it has 7dB of NFB?

Cheers

Ian

 

[alexc]

Yes, it is a 30Wrms el34 pp amp and yes it is not cancelling the H2 as much as one might expect in a pp output stage.

I think it's being generated in the voltage amp and phase inverter sections.

The circuit was pretty standard stuff, except mine is quite low sensitivity. I wanted to try a minimal nfb setup, hence the 'only 7dB nfb'  I mentioned.

I did basic measurements of the open loop gain and closed loop gain at 1kHz and adjusted the network with a resistance substitution box and wound up with the 7dB figure.

Fairly arbitrary choice really -  based on just the idea of having just enough nfb  to bring the hum and noise down to desired levels ....  rather than attempting to get the lowest possible THD by trading off excess gain.

I have a spare half on the first 6922 position - I could wire it in to make much more gain to trade and reduce the H2, but as I said, an exercise in less nfb.

I did a lot of listening and watching on the CRO etc as I tweaked the nfb network, which admittedly is quite primitive

Only a resistor, so no frequency shaping in the nfb. 

I haven't done much in the way of transient and stability analysis either, beyond the super simple nature of the 3 stages of amplification in the classic way. It has direct coupling of the va to pi to minimise the poles and so forth.

It's never exhibited anything that I would call 'instability' and I'm confident it sounds good and measures quite well.

I ended up keeping the nos orange amperex 6dj8 in the end, despite the slightly higher THD -  as compared with a modern eh6922.

Mostly a case of 'audiophoolery' than anything else - it  surely sounds good however, and the noise was lower (due to the lower gain of the nos compared with a good eh).

The best measurements I've made have surely been with modern jj5751 and eh6922, with just basic microphonics by vendor.

But I did go down the '6922 rabbit hole' of nos purchases in the hope of catching me some extra mojo. I exited reasonably intact, with only a couple of hundred extra spent 

Anyway, here's the spectra of the config I have kept ... since I did the last measure-and-tweak last summer.

 

[ruffrecords]

Good to find someone who actually knows how to measure the amount of feedback!!

It is surprising how little NFB you need in tube designs. The good thing about that is it makes stability simpler. Its large quantities of NFB that lead to instability.

Cheers

Ian

 

[alexc]

Surely true what you say Ian. 

Although the nfb can improve nearly every aspect of performance, thd, noise, hum, bandwidth etc, excessive nfb can cause phase shifts sufficient to cause instability pretty easy.

I'm not enough of a boffin to know how much nfb  and what degree of 'response tailoring' 'sounds best', so I keep it simple generally.

Whenever I start to go further into stability/transient analysis and measurement, I end up losing interest and wind up playing the guitar instead!

Mr Peter Baxandall made some interesting points regarding his analysis that too little nfb makes for increased higher order distortion products - he seemed to advocate > 20dB of nfb on that basis.

Others, such as Nelson Pass, I think, wrote on the subject too, taking the position that such an analysis is too simplistic, typically done on a single active device in non-representative circuits.

My own limited experience and preferences in this area leads me to the minimal nfb requirement to improve an already good implementation.

I can definately say I like the 'life-like' impression that zero global nfb gives in my minimalist se hifi amp. PSU performance is the limiting factor, tho. So far, brute force filtering has been my approach. I'm only now getting to regulated PSUs in my tube work.

Also I do typcially  use variable nfb usually from about 5dB to 20dB with a potentiomenter in th nfb network. It usually stays at the lesser end of the dial.

There's no doubt in my mind that for solo instruments, less nfb sounds most compelling. It's less obvious to me for music program material, however.

I'm just starting to look at the issue in bjt circuits where the amount  nfb is mostly much higher. Here I am experimenting with more local nfb vs global and so on.

I'm intrigued by Wayne Kirkwood's implementation of Class A power amplification with  defeatable local/global nfb options and will give that a go at some point.

 

[madswitcher]

Hi there,

I built the JLH Class A amp about a year ago, but the higher output version based on some work done by Geoff Moss.  It gives a good healthy 50 watts into 8 ohms, 7 Hz to 80 KHz at about 0.1% distortion just before it clips, and needs some hefty heatsinks - the enclosed photo is a monoblock with a separate psu.  It sound pretty good, but there are a couple of things you can do to drop the distortion down even more - I am experimenting with them at the moment and am getting about 0.0035% second harmonic - a drop of about 26dB or so on the original design.

One of the best things you can do is to use a capacitance multiplier after the main smoothing caps - and don't use a regulated PSU.

Cheers

Mike