My First Amplifier Design - diyAudio
Go Back   Home > Forums > Amplifiers > Solid State

Solid State Talk all about solid state amplification.

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 1st February 2015, 02:19 PM   #1
cwtim01 is offline cwtim01  Hong Kong
diyAudio Member
 
Join Date: Aug 2012
Default My First Amplifier Design

I finished reading the 2 books from Cordell and Self lately, will need to re-read both a couple more times, but I think it'll strengthen my understanding by an actual practice, so here I am with my first attempt to design an amplifier.
  • It's a classical LIN type of topology, naturally as that's what the 2 books are about;
  • I want to try something slightly different from a standard Blameless, so I've taken a push-pull VAS published by Samuel Groner couple years back, and bundle it with a EF3 from Cordell's book;
  • Target: 100W into 8R, 200W into 4R, 26dB close loop gain, simulated performance similar to a typical Blameless;
  • I hope it'll end up as something useful in my living room and sounding alright, instead of a giant oscillator busted into flames, so I've targeted conservative phase margin and gain margin. I also included VI limiting for more protection.
Attached are the schematic and some simulation results. In terms of simulated THD, figures are:
  • THD1K@8R: 10.2ppm@1W, 10.1ppm@10W, 10.0ppm@100W
  • THD20K@8R: 9.7ppm@1W, 23.9ppm@10W, 16.0ppm@100W
  • THD1K@4R: 10.1ppm@1W, 9.9ppm@10W, 9.6ppm@100W, 10.0ppm@200W
  • THD20K@4R: 31.4ppm@1W, 40.0ppm@10W, 33.1ppm@100W, 46.7ppm@200W
I believe these THD numbers are OK but not great, especially when sub-ppm designs are widely available in this forum. I was excited by some ppb level THD1K numbers with the circuit at some stage, but when I added safety features like zobels at the OPS transistors, base stoppers, feedback DC blocking capacitor ... etc, the THD degraded to the above at the end. I tried to mess with TPC, TMC and OIC to improve the performance, but trial and error didn't get me very far and I ended up sticking with Cdom compensation. Anyway, my novice construction skill will be the limiting factor of any actual build, so I decided to settle with these THD figures for now.

I have an IT background so I'm fairly technical, but I knew very little about electronics when I started this DIY hobby a year ago, so many of these things are entirely new for me. Hence, it's quite probable that I made a few mistakes somewhere, so your comments, corrections and suggestions are really welcomed.

Thanks and regards.
Attached Images
File Type: png My_Amp 0.4b.PNG (256.1 KB, 661 views)
File Type: png Frequency Response.PNG (53.8 KB, 612 views)
File Type: png Loop Gain.PNG (66.9 KB, 541 views)
File Type: png PSRR.PNG (34.1 KB, 511 views)
File Type: png Clipping.PNG (16.6 KB, 490 views)
File Type: png Output Impedance.PNG (22.0 KB, 82 views)
File Type: png Slew Up (with input network).PNG (35.1 KB, 77 views)
File Type: png Slew Down (with input network).PNG (35.5 KB, 78 views)
File Type: png VI Limiting 40V into 2R.PNG (28.4 KB, 84 views)
Attached Files
File Type: zip My_Amp.zip (7.7 KB, 45 views)
  Reply With Quote
Old 2nd February 2015, 03:10 AM   #2
Struth is offline Struth  Canada
diyAudio Member
 
Join Date: Aug 2012
Location: books at londonpower.com
Hi Guys

Some really nice plots there.

In practice, the 1uF supply bypass should be increased a thousand-fold, a bypass should be added for the output stage, and the zeners should have caps across them for the input stage. Zeners tend to be noisy.

The 100uF cap in the feedback loop is way undersized. It is contributing significant THD to all numbers. Same with the input cap. usually for simulation, those caps are left out to be able to assess what the active part of the circuit is doing. There has been enough independent study made about capacitors to kiow the range of values and dielectric types to use in the various circuit points. For example, an electrolytic is fine in the feedback network provided it is not used to shape the low-frequency response, i.e.e it should be several millifarads. The input cap should be polypropylene not polyester, as the latter introduces significant third-harmonic THD where PP does not.

You should try a sim using a standard up to date Lin-Thompson front endjust for reference. It is pedestrian but exhibits very good performance. When I first started learning to use LTspice, the Lin gave best performance but my favourite circuits are fully symmetrical. I now have the latter outperforming the Lin. I'm sure you will achieve the same end with your circuit with a little persistence.

Have fun
__________________
London Power / Power Press Publishing
www.londonpower.com
  Reply With Quote
Old 2nd February 2015, 04:16 AM   #3
diyAudio Member
 
Join Date: Nov 2006
Location: Adelaide, South Australia
Blog Entries: 27
cwtim,
Good on you for jumping in and having a go.

You wanted to get away from the standard blameless, well you certainly have. There is little of the concepts of blameless in there. Notably simplicity, which is where I become rather uncomfortable on your behalf.

Can I suggest that as an alternative to simulation you turn your hand to a bit of prototyping and building. I expect you will learn a couple of very important lessons:
-1- In simplicity is beauty
-2- In a feedback amplifier stability is overwhelmingly important, and controlling this is best managed through simplicity (refer 1.)
-3- It is possible to hack into a cad package something that is awesomely complex, but actually works worse in the real world than much simpler approaches (refer 1)
-4- Achieving ppm level THD is not easy in reality - though simulators will make all manner of promises. Build a prototype and measure the THD - the measurements alone I promise will challenge you.

(OPINION WARNING HERE: I challenge you to hear the difference between 1 and 10ppm, probably 100ppm and maybe even 1000 in some cases).

I have seen plenty of people showing simulations of sub PPM designs. Actualising these in the real world is a completely different kettle of fish:
- Many of these rely on insane levels of loop gain and bags of feedback.
- Physical layout, power and grounding issues etc can easily swamp this - I have seen simple wiring layout add 10 times this
- measuring this is really (really!) hard in the real world

What is my point?

If this is your first effort, the place for you to start is being able to design something that just works. If you start with such a complex, and in my mind potentially marginal, design it will be really frustrating.

Don't let silly numbers and idealistic goals of "zero" distortion and suchlike distract you from the essential need for a stable, manufacturable design that won't burst into seven sorts of instability and fry your speakers.

Build something and feel the pleasure at seeing it work. You will quickly realise that you don't need to have dome something massively complicated to achieve that outcome. Indeed.... in simplicity is beauty.
  Reply With Quote
Old 2nd February 2015, 06:41 AM   #4
cwtim01 is offline cwtim01  Hong Kong
diyAudio Member
 
Join Date: Aug 2012
Quote:
Originally Posted by Struth View Post
... You should try a sim using a standard up to date Lin-Thompson front endjust for reference...
Thanks Struth, Iíll try your suggestions, I Googled a bit but donít managed to find the Lin-Thompson front end you referred, could you share with me the relevant schematic or send me a link.
  Reply With Quote
Old 2nd February 2015, 06:44 AM   #5
cwtim01 is offline cwtim01  Hong Kong
diyAudio Member
 
Join Date: Aug 2012
Thanks googlyone, all wise words and I actually agree with most (if not all) of what you said.

I also like simplicity, the only amplifier I actually built (designed by a forum member) has 4 transistors on each channel, and itís indeed satisfying to listen to it, but itís a bit low on the power department (10W Class A). I then acquired the parts to build a Pass F6, but a F6 still quite doesnít give the power to drive the 4R speakers that I want to drive, so I turned to try designing one.

I did prototype, I put one channel of Blameless (OPS just 2 TO-92 transistors) on a breadboard and it worked, I also put one channel of the above VAS on a breadboard (again TO-92 OPS) and it oscillated at 26MHz, only cured by adding C12/13. Whatís interesting was that such oscillation was predicted by a gain peaking at the same frequency in LTSpice. I understand the limitation of simulations, it can give all sort of wrong predictions, but it can still be useful if the right questions are asked and interpreted correctly Ė though admittedly Iím still learning how to do that.

Quote:
Originally Posted by googlyone View Post
... If you start with such a complex, and in my mind potentially marginal, design it will be really frustrating ...
Not disagreeing, but could you elaborate this point Ė I thought Iím playing safe by the gain and phase margin figures, and I gave up the more advanced compensations because of stability concerns Ė so I definitely want your further input here.
  Reply With Quote
Old 2nd February 2015, 11:19 AM   #6
diyAudio Member
 
Join Date: Nov 2010
Quote:
Originally Posted by cwtim01 View Post
... but could you elaborate this point –
I am very much of the same opinion so if I may comment.
Most of the ultra low distortion simulations are very suspect so I would not be too disappointed if your amp does not reach that level.
But if the objective is only, say 10 ppm then you don't need that much complexity.
To be more specific, you have a second section composed of an EF assisted VAS plus folded cascodes. That's a complex loop to wrap compensation around.
You should place a Tian probe within this loop to check stability.
There's more to stability than the one loop you have checked
Also the push/pull schemes are vulnerable to surprises if the two sides do not match perfectly.
Very easy to miss this in simulation, since the values will all be perfect.
You need to check this manually in LTSpice, try the effect of imperfect matches for transistors and especially Miller capacitors.
Bob Cordell touches upon this but does not emphasize it, he often uses other methods to avoid the problem.


Best wishes
David

I had the comments about the inner loop stability more or less written before you posted on the 26 MHz instability.
Should have replied quicker so I could prove it wasn't just "wise after the event".
You will need also to check stability as the amp approaches the rails, with an OPS, in both simulation and reality.
This has a major effect, as do difficult loads so they need checks too.

Last edited by Dave Zan; 2nd February 2015 at 11:37 AM.
  Reply With Quote
Old 2nd February 2015, 01:19 PM   #7
Struth is offline Struth  Canada
diyAudio Member
 
Join Date: Aug 2012
Location: books at londonpower.com
Hi Guys

Everyone calls what Self names the 'blameless' a Lin, which is incorrect. The Lin circuit is just a unity-gain output stage preceded by a voltage gain stage, so the last two-thirds of the blameless. I believe it was Wahab who pointed out that it was Thompson who added the diff amp in front of the Lin to create the ubiquitous circuit that is 99% of all amps produced.

I believe going after the lowest THD possible is a must if you want a transparent circuit. There are natural distortions of sounds we hear but those generated in electronics are quite a different thing and need to be minimised as much as possible. No one can tell you how accurate spice will be compared to the real circuit, since there are too many variables and too many assumptions built into both the models and the sim algorithms. Simpler circuits seem to sim faster and real measurements corroborate performance. More complex circuits take longer to sim and tend to deviate more from the real world.

Were you to assume that the sim is off by an order of magnitude being too optimistic, then a 100ppb becomes 1ppm, which is below the resolution of an Audio precision system. This fact does not make the distortion inaudible. It is incredible what our ear is sensitive to. You will be thrilled to have a working circuit with 10% THD but it will quickly lose its luster once you live with it for a little while, especially if you listen to music that is "busy" in the midrange. The guys who like those single-mosfet amps and other similar circuits tend to listen to rather spacious music. I found the JLH to sound dreadful with any kind of modern music, but low-THD amps sound much better. Everyone has their own take on this.

As the poster above said, wiring and layout errors easily swamp the inherent performance of the circuit. Generally, if the board is laid out as the schematic is drawn, assuming a properly drawn schematic as your are, then the performance of the circuit will be pretty good. A convoluted signal path guarantees instability.

You might find the THD is lower if you take out the middle part of the circuit and use a conventional bias circuit driven by the VAS. This will allow optimisation of the idle condition for lowest THD. The current sources used are not the best ones - the feedback type is more linear. Personally, I am not a fan of circuits where the audio signal is holding off the output stage, as here. Pull-up/down is limited by the current source. Signal-driven pull-up/down provides as much drive as is needed for any load and any device set.

Have fun
__________________
London Power / Power Press Publishing
www.londonpower.com
  Reply With Quote
Old 2nd February 2015, 04:19 PM   #8
cwtim01 is offline cwtim01  Hong Kong
diyAudio Member
 
Join Date: Aug 2012
Quote:
Originally Posted by Dave Zan View Post
... That's a complex loop to wrap compensation around ...
Thanks David for your various remarks, very useful. I'll check this loop, it's not immediate obvious to how me on how to check the inner loop stability, but I'll do my homework to try to figure out first. (As I said many of these are new to me - it took me a good night of Google searches to learn how to plot loop gain with the Tian probe in it's basic form, and I still haven't figure out how to use a GFT probe). I might come back with a question in case I'm stuck.

From an engineering perspective, this VAS is probably unnecessarily complex for what I'm trying to do, but I selected this VAS cause I think it'll lead to more learning than a standard Blameless architecture, from this perspective it's probably a good choice ... another VAS I considered is the SuperTIS, but it's way too advance for me and I might try that when there is more knowledge under my belt.
  Reply With Quote
Old 2nd February 2015, 04:49 PM   #9
cwtim01 is offline cwtim01  Hong Kong
diyAudio Member
 
Join Date: Aug 2012
Quote:
Originally Posted by Struth View Post
...Have fun
Thanks again, I think I'll be busy in the next few weeks to try out the various suggestions
  Reply With Quote
Old 2nd February 2015, 08:47 PM   #10
diyAudio Member
 
Join Date: Nov 2010
Quote:
Originally Posted by cwtim01 View Post
...I'll check this loop, it's not immediate obvious to me on how to check the inner loop stability...
Or to most people Two factors for this.
For a simple Miller compensated amp the inner loop is usually stable, so there are usually no problems even if full checks are not done.
So most people are a bit careless but you should not be if you want to try less conventional circuits.
Secondly, a push/pull amp is more difficult to check stability in the inner loop.
Requires a balun, I posted one in the tread about Paul's "lunch-break" amplifier.
You are welcome to copy it and use.
Dr P J Hurst has a nice article about this on the Web, recommended for a technically minded person like you.

Best wishes
David

Quote:
I believe it was Wahab who pointed out that it was Thompson who added the diff amp in front of the Lin to create the ubiquitous circuit that is 99% of all amps produced.
I recall that it was actually Michael Kiwanka.
He was criticised as a pedantic quibbler but seems to be correct on this one so it's only fair to cite him.

Last edited by Dave Zan; 2nd February 2015 at 08:59 PM.
  Reply With Quote

Reply


Hide 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
Trackbacks are Off
Pingbacks are Off
Refbacks are Off


Similar Threads
Thread Thread Starter Forum Replies Last Post
7.1 amplifier design help human304 Chip Amps 11 1st October 2013 06:52 PM
New Amplifier design kzeprf22 Solid State 28 5th April 2009 03:45 PM
Amplifier Design PassFan Tubes / Valves 5 16th April 2003 10:45 PM


New To Site? Need Help?

All times are GMT. The time now is 07:17 PM.


vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2015 DragonByte Technologies Ltd.
Copyright ©1999-2015 diyAudio

Content Relevant URLs by vBSEO 3.3.2