Go Back   Home > Forums > >
Home Forums Rules Articles diyAudio Store Blogs Gallery Wiki Register Donations FAQ Calendar Search Today's Posts Mark Forums Read

Solid State Talk all about solid state amplification.

The Blomley Class B amplifier
The Blomley Class B amplifier
Please consider donating to help us continue to serve you.

Ads on/off / Custom Title / More PMs / More album space / Advanced printing & mass image saving
Reply
 
Thread Tools Search this Thread
Old 25th May 2018, 05:36 PM   #31
Antoinel is offline Antoinel  United States
diyAudio Member
 
Join Date: Jan 2012
Location: Pennsylvania
The Blomley Class B amplifier
Hello Ian and thank you for your above post which overflows with a generous dose of valuable knowledge, expertise and guidance.

The CSD power output stage in your [.asc file] is the one to use. In addition to the independent and proven results of your research, its resultant pulse currents are identical in magnitude for the upper and lower Pout half-stages. With the added bonus of a stay-on current [your term] once the pulse disappears. Meaning the output transistors do not turn off in operation This is a real accomplishment and a leap forward for improving this and other amps.

I'll buy the Linear Audio Vol. 13 which has your article on CSD. And also read-in the links at the bottom of your posts. There's a lot of education therein and in Linear Audio.

I have a suggestion which I hope that you consider; because of our intense interest, enthusiasm and evident skill in this subject matter and more. You've already done a lot of research towards the following suggested power amplifier, and I feel that you are having a lot of fun in this thread.

1. Take your Cube-Law Class A amplifier.
2. Remove the cube-law and linear drivers.
3. Plug-in a CSPR at the vacant location of the above drivers.
4. The valuable enhancements which precede and follow the CSPR driver stay.
5. The PCB for the Cube-Law Class A amp exists. An epi-board for CSPR can hopefully sit on top for the time being. That'll be a beautiful sight to see.

I will continue to look forward to your theoretical and practical output. This nonstandard topology [thanks to Blomley] is truly in evident great hands [yours]. I am glad for this open source diyAudio which continues to offer new possibilities for other amplifiers to happen.

Best wishes

Anton
  Reply With Quote
Old 3rd June 2018, 08:38 AM   #32
haiqu is offline haiqu  Australia
diyAudio Member
 
haiqu's Avatar
 
Join Date: Oct 2005
Location: Brisbane
Quote:
Originally Posted by Mooly View Post
There was a recent article on the Blomley over on 'allaboutcircuits'. Its a great read.

My 40-Year Love Affair with a Remarkable Amplifier—A Class B Amplifier for Audiophiles

I also did a sim for curiosity. You will need to put your own models into it and alter voltage source V1 to get the desired quiescent current:
Where do I find "discrete models.txt" required by this sim?
__________________
[VK9VK] With great power comes great current squared times resistance.
I spent most of my money on fast cars, hot girls, hi-fi gear and booze. The rest I just wasted.
  Reply With Quote
Old 3rd June 2018, 09:29 AM   #33
IanHegglun is offline IanHegglun  Australia
diyAudio Member
 
Join Date: Mar 2017
Quote:
Originally Posted by haiqu View Post
Where do I find "discrete models.txt" required by this sim?
I got around this in Post 10 sim by adding:
.include Cordell-models.txt
.model 2N2955 ako:MJL21193C

and remove the ".inc discrete models.txt"
  Reply With Quote
Old 3rd June 2018, 09:59 AM   #34
haiqu is offline haiqu  Australia
diyAudio Member
 
haiqu's Avatar
 
Join Date: Oct 2005
Location: Brisbane
Thanks Ian.
__________________
[VK9VK] With great power comes great current squared times resistance.
I spent most of my money on fast cars, hot girls, hi-fi gear and booze. The rest I just wasted.
  Reply With Quote
Old 3rd June 2018, 01:37 PM   #35
Mooly is offline Mooly  United Kingdom
diyAudio Moderator
 
Mooly's Avatar
 
Join Date: Sep 2007
The Blomley Class B amplifier
Have you got the sim working now ?
  Reply With Quote
Old 4th June 2018, 04:19 PM   #36
N101N is offline N101N
diyAudio Member
 
Join Date: May 2018
Unfortunately, the suggested solution is based on some erroneous assumptions. Whenever the dynamic (signal) energy level is in excess of the static (DC bias) energy level, there are two possible scenarios: either switching to higher static (DC bias) energy level or remaining in saturation (nonconducting state).
  Reply With Quote
Old 5th July 2018, 04:07 PM   #37
logician is offline logician  United Kingdom
diyAudio Member
 
Join Date: Oct 2006
Default Hugh, thank you and...

Quote:
Originally Posted by AKSA View Post
Logician,

(1)...The FFTs reveal a profile very similar to a well compensated, modern 'competent' design. That is, it has H3 higher than H2, H5 higher than H4, and H7 higher than H6. There is a long series of harmonics to the 200th created by the heavy global feedback.

(2)...I have not built this particular circuit but I have discovered that there is a relationship to harmonic profile and sound quality, almost regardless of THD. I would expect this amp to sound 'hifi' but not musical, unlike many of Pass' DIY amps, particularly the Zen series, and the JLH 10W which despite high THD sound wonderful. I agree that the high pole at 3MHz and very good phase margin is outstanding, but these do not make good sound quality.

(3)...People disagree about THD and sound quality. I would expect the Blomley with its non-switching output stages to deliver very low high, odd orders; but from Ian's FFTs these objectionable harmonics do appear, much like most AB modern amps.

(4)...Perhaps global feedback is part of this? But why does the non-switching advantage reveal very low levels of high, odd orders?

HD
...

please accept my apology for the time taken in responding to your post.

Part of the delay is due to the Summer holiday season here in Northern parts. The remainder arising entirely from those difficulties attendant to providing accurate and meaningful responses to your excellent points and questions.

...(1) In each of my two earlier posts, only the first FFT is from the non switching version. The second is provided merely for comparison. i.e. the same output stage but without the non switching circuitry in operation. Also, I wouldn't read too much into the absolute shape of the mathematical noise floor as this is quite mutable over wide ranges of various input parameters.

...(2) I am most grateful for the considered emphases you have made in this part of your response and confident that you are quite correct in these assertions. To me, the most significant of your points here being that, perceptually, there exists a substantive decoupling of harmonic profile from absolute distortion level. I remain very keen to develop a clearer understanding of how this may apply in its limits. Particularly in the light of Ian's response above, do you have any thoughts on an outline numerical framework for this?

...(3) I accept your first point here without reservation and concur with your very reasonable expecatation. It is both disappointing and at first somewhat puzzling that this clearly does not apply. I will conjecture below as to why this may be so. Further by inference it is also accepted that some moderately uniform predominance of even over odd harmonics may well be preferable to a listener. Although again, at present, I have no clear insight as to any numerical basis for this. I would be most grateful to learn of any thoughts you may have on this matter.

...(4) Having worked to investigate your first question here, my current conclusion is that whilst global NFB is certainly an important factor it may not be the prime or indeed the preferred correlate for solution of inter harmonic series differential mismatches in non switching output stages. As may be seen by way of only one data point taken from the first of each of the above sets of FFT's. In extrapolation, it appears that these odd to even ratios are largely maintained in their relative proportions almost regardless of any sensible amount of NFB i.e. at the very least, enough to provide sufficient control of output impedance etc.

So, from the above and after a great deal of effort to confirm the following:

Although I would dearly wish to be proved wrong about this. My findings to date are that; the non switching amplifier topologies which are presently well understood will in all likelyhood consistently deliver this unsatisfactory odd even harmonic profile imbalance.

This, it would appear, is generally because they are, as with any conventional AB or fully balanced amplifier, in essence, made up of two 'separate' near identical output units. Which I suggest with a suitably requisite measure of NFB will, of necessity, act together to cancel and thereby reduce the even harmonic series.

I am continuing to work, so far with some notable successes, on developing a class of non switching output stage that offers every benefit of these but without any of their all too apparent disadvantages. Which, in your post, you have been kind enough to detail.

Please consider this my personal note of thanks to you for providing the impetus to look deeper into addressing and identifying those means and methods necessary to overcome these last remaining issues.

One example of which I include below for comments.

Cheers, ian
Attached Images
File Type: png ANSA NSOP.png (33.2 KB, 219 views)
__________________
...Cæteris paribus; scientia est potentia. Sed compos scientia creat potentia magnam.
  Reply With Quote
Old 5th July 2018, 05:45 PM   #38
logician is offline logician  United Kingdom
diyAudio Member
 
Join Date: Oct 2006
Default Thank you Ian...

Quote:
Originally Posted by IanHegglun View Post
Unfortunately, my Linear Audio article in Vol.4 on harmonic weighting is behind a paywall, but I briefly cover it's application in my free Supplement on p17, mentioned in Post 10. The key point:
Why is the harmonic roll-off rate an important parameter? Our hearing sensitivity curves roll-up at
approximately +40 to+50dB/dec in the 1kHz to 3kHz range when multiple tones do not cause
masking. This means we want any harmonics from crossover distortion to roll-off faster than
-50dB/decade so harmonic distortion from crossover distortion cannot be heard.

Standard Class-B amplifiers have difficulty reducing the level of the higher order harmonics from crossover distortion due to:
1) Our ear boosts the higher order harmonics at about the same rate as they reduce so all are nearly equally apparent, and
2) The negative feedback factor of most Class-B power amplifiers rolls-off at 20dB/decade in the same frequency range where our ear boosts the higher order harmonics and this makes it hard to suppress crossover distortion and make it inaudible...
After writing my Vol.4 article I realised that the rate of roll-off of harmonics in -xdB/decade is an easier way when doing simulations to assess whether high order harmonics like from crossover distortion than adding a special weighting filter. You can do it by placing a ruler on the FFT plot and slide the ruler to say 0dB and start of a decade and read off the dB drop at the next decade. Easy as.

If you read anything less than -50dB/decade then crossover distortion will most likely be the dominant distortion in the sound of the amp -- then you need to check if the weighted THD is below about 1%, eg by using the double differentiator I used in Post 10 and RMS add all harmonics to the 100th with a 100Hz fundamental; the ErrorLog THD using ".four {freq} 100 V(D2) ". You probably want a 20dB safetey margin, that's a target of ~0.1% for D2 100 harmonics with 100Hz to be safe enough to say any HD will be inaudible at this power level. Note a double differentiator is +40dB/decade not the +50dB/decade of our hearing so it under estimates what we can hear, but this is partly offset by taking harmonics to 10kHz (rather than 3kHz where our hearing sensitivity to falls).

Effectively the double differentiator magnifies all the higher harmonics so they can appear on a FFT plot of V(D2) at the same relative audibility as the lower order harmonics. And when D2 is displayed on an o'scope IF you can see any harmonic distortion then it's most likely you can hear it as well from your amp.
---
The two plots you gave with 100Hz, the first with NS and about -60dB/decade harmonic roll-off is very good. Also the first low order harmonics are -120dB below fundamental, so low that they would be hard to measure let alone hear.

The second without NS has about -30dB/decade, but with the first low order harmonics at -120dB below fundamental you have a safety margin of +80dB (below 1% needed for audibility). So even if you weighted the higher harmonics they are unlikely to get to -40dB to become audible (at least at the output level used here).

You can conclude that 1) NS does in fact reduce high order distortion, and 2) that the distortion even without the NS is not likely to be audible.
---
Regarding the Blomley approach, which uses 1) current drive of the splitters, and 2) non-switching in the power stage. The use of current drive of the splitters increases the amount of high order distortion compared to ordinary voltage drive splitting. Current drive of the splitters causes a faster transition through the zero-cross and a doubling of the rate when weighted with say +40dB/decade (double differentiator) gives 4 times the level of higher harmonics, that's a net increase of +20dB/decade. So unless current drive of the splitters can reduce the level of higher harmonics by -20dB/decade at the same time then you are not getting ahead by using current drive of the splitters.

An exception might be in the extreme case where the switching rate at the zero crossing is so fast that the majority of the higher harmonics are pushed above 20kHz where they don't affect the audible range (assuming there's not much other nonlinearites to created demodulation into the audio band by intermodulation). Antoinel's Class-aP (Post 5) would seem to fit into this category if the switching rate is fast enough. I'd be interested to see FFT plots and some D2 values for any of Antoinel's Class-aP amps.

Cheers
... for the comprehensive scope and extensive scale of your tremendously diligent and very kind response. For which, myself, and I reasonably anticipate many others are extremely grateful.

I respectfully offer to you the same apology and explanation for delay as with Hugh above.

The clarity and thouroughness of your answers have both informed and inspired my efforts in seeking to overcome the issues he has indicated.

Their substance has provided me with the confidence to approach these problems from different directions and explore some 'outside the box' possibilities knowing now that there is a well bounded function space within which to place any potential solutions.

Thank you once again,

Cheers, ian
__________________
...Cæteris paribus; scientia est potentia. Sed compos scientia creat potentia magnam.
  Reply With Quote
Old 6th July 2018, 08:33 AM   #39
keithj01 is offline keithj01  United Kingdom
diyAudio Member
 
Join Date: Mar 2014
Location: Southern England
I experimented with a version of Blomley's design in the '70s. Using very ordinary slow transistors, I saw crossover distortion. Not surprising, really, as several parts of the circuitry need to be very agile. Now, with modern devices, there's every opportunity to revisit the concept and end up with an excellent amplifier. Major advantages are freedom from the usual problems associated with class AB designs and relative simplicity.

I subsequently happened to bump into Peter several years later, and he said he'd moved onto a much better solution! Didn't say what it was though...
  Reply With Quote
Old 7th July 2018, 01:47 AM   #40
AKSA is offline AKSA  Australia
diyAudio Member
 
Join Date: Sep 2001
Location: Melbourne, Australia
Quote:
So, from the above and after a great deal of effort to confirm the following:

Although I would dearly wish to be proved wrong about this. My findings to date are that the non-switching amplifier topologies which are presently well understood will in all likelyhood consistently deliver this unsatisfactory odd even harmonic profile imbalance.

This, it would appear, is generally because they are, as with any conventional AB or fully balanced amplifier, in essence, made up of two 'separate' near identical output units. Which I suggest with a suitably requisite measure of NFB will, of necessity, act together to cancel and thereby reduce the even harmonic series.

I am continuing to work, so far with some notable successes, on developing a class of non switching output stage that offers every benefit of these but without any of their all too apparent disadvantages. Which, in your post, you have been kind enough to detail.

Please consider this my personal note of thanks to you for providing the impetus to look deeper into addressing and identifying those means and methods necessary to overcome these last remaining issues.

One example of which I include below for comments.
Thank you Ian, a great post.

I cannot offer substantial proof for my assertions in my early post. This is because I design and manufacture amplifiers for the audiophile market and work on critical listening tests, an attempt to create a better mousetrap but without the usual corporate R&D team with expensive labs. My R&D costs are minimal and based on many thought experiments. I have tried to involve local Universities in psychoacoustic study, but of course they are interested in corporate investment. I have spend money and time I am ashamed to recount in developing many amps, starting with the original AKSA, then the Lifeforce, the Maya I, the Soraya, the NAKSA, the SAKSA and the latest the Maya II. All are quite unrelated but represent a steady improvement of sound quality over 18 years.

I moved to mosfets in 2011. Modern mosfets are so fast, so easy to drive, so electrically robust and thermally strong that I feel they cannot be ignored, particularly as they are now inexpensive.

I have worked with local engineers in the non-switching area, and one, Paul Bysouth, has been quite successful with one NS amp which sounds very good, on par with my flagship amp which is not NS in fact. But his amp is more complex, and the H5/H7/H9 artefacts appear to be about the same as my non-NS design so on balance I do not see much advantage. In light of the last poster it could be related to the speed of the correction transistors; like error correction this technology requires actives which are perhaps a speed order of magnitude higher than the outputs, which is problematic.

I have found that a linear, decreasing harmonic profile sounds very natural, but measures badly. If H2 starts at -90dB, and increasing orders drop by 10dB finishing with H5 at -120dB, the sound is very clean but cannot be criticised. OTOH, if H2 starts around -68dB, with increasing orders dropping 10dB and finishing with H5 at -98dB the sound is markedly warm, natural and rich. While we recognise a 'sound effects box' it is difficult to ignore such an amplifier, it sounds wonderful, not the same as a tube amp (with it's typical H2 starting around -45dB) but not dissimilar. All this indicates subtle, unexpected psychoacoustics, almost the negation of the zero THD yore we all look for.

My recent work is an attempt to move the switching point off the crossover point. If the passive device in an AB PP output stage can switch off a few volts well into the active, device duty cycle the death throes of the passive device will be presumably lost in the high current prevailing in the active device.

The final comment concerns feedback, which in global fb amps create corrections to corrections, out to higher orders. Something seems to happen beyond about 35dB of global fb where the front to back, 'depth of image', seems to shrink, giving only 2D stereo image. Many effects influence depth of image, but fb seems to be pivotal and I cannot figure this out. This phenomenon is very clear with tube amps - notably many do not have global feedback - and it attracted my interest some years ago. My amps deliver staggering depth of image because of a radical design in the feedback. Significantly I do not deliver low THD; it's typically 0.06%, but the profile is as elegant as your last plot, linear decreasing with H5 and beyond below -100dB. A measure of excellence seems to be the percentage of the THD which comprises H2, H3 and H4. Anything over 99% such amps sound very, very good.

HD
__________________
Aspen Amplifiers P/L (Australia)
www.aksaonline.com

Last edited by AKSA; 7th July 2018 at 01:49 AM.
  Reply With Quote

Reply


The Blomley Class B amplifierHide this!Advertise here!
Thread Tools Search this Thread
Search this Thread:

Advanced Search

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off

Forum Jump

Similar Threads
Thread Thread Starter Forum Replies Last Post
Class D Vs Class I (Balanced Current Amplifier) rmsaudio Class D 30 23rd February 2018 02:54 PM
Questions about how to decrease the PASS A-40 amplifier into a class AB amplifier. novtango Pass Labs 4 14th February 2018 05:11 PM
Modify a class D subwoofer amplifier in full range class d amp MarioRestucci Class D 8 27th February 2017 08:21 AM
class D vs BD Rockford class mono amplifier... mixalos spl Car Audio 5 3rd November 2014 06:13 PM


New To Site? Need Help?

All times are GMT. The time now is 02:38 AM.


Search Engine Optimisation provided by DragonByte SEO (Pro) - vBulletin Mods & Addons Copyright © 2018 DragonByte Technologies Ltd.
Resources saved on this page: MySQL 15.00%
vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2018 DragonByte Technologies Ltd.
Copyright ©1999-2018 diyAudio
Wiki