My amplifier

Discussion in 'DIY' started by Alan0354, Mar 29, 2017.

  1. Alan0354

    Alan0354 Super Member

    Messages:
    2,087
    Location:
    Bay Area
    Hi

    I have been working on my first revision of my amp and been spending a lot of time on it the pass few weeks. It is still in working progress, but at least today I reach a mile stone of doing the first FFT test. I am still working on lowering the distortion and still have to put into the chassis and test in real listening.


    Chassis.jpg
    This is my two chassis, I first bought the smaller one, but determined the heatsink is too small, so I got the second one which is a clone of Krell 50W class A chassis. The power dissipation is about 70W/ch, so I need all the heatsink I can get. The amp is only going to be about 70W 8ohm or 140W 4ohm. I sacrifice peak power for big class A region. I get about 16W of class A for 8ohm or 8W for 4ohm load

    Second run pcbs.JPG This is the set of pcb. I separate the OPS board from the IPS/VAS board so I can swap to different designs in the future. I have a board for voltage regulation to get rid of the ripple from the transformer and separate filter board for each channel. I use separate bridge rect for each channel. Each channel has total of 80,000uF (40,000uF for each rail)


    Test setup.jpg This is my test setup, I use +/-31V bench supply to power the amp. I still use 20,000uF per rail. The bias is quite stable, it drops about 10% from stone cold to fully warm. I tested by driving with large input signal, then remove the signal to verify the current does not change.


    Blameless.JPG
    I use burst pulses to keep the power down and I can use large signal. This picture shows a positive pulse from 0V to 15V( as seen on the display). I use +/-30V to power the amp. I also show the width of the pulse of 9.335uS to compare to the next picture. You can see the rise time is less than 700nS and very little overshoot and no ringing. I don't know how important is group delay. The top trace is the input signal, from the picture, the delay from input to output is definitely less than 0.2uS, maybe even shorter.


    Blameless clipping recovery.JPG This picture shows way over drive. You can see the recovery is about 1.2uS. The display show the original 0.9335uS width for a non clipping pulse, you can see it just get about 1.2uS wider, there is not funny wave shape on the clipping on the top.


    Blameless slew rate rise time.JPG
    This picture shows the slew of the rising edge during clipping. This shows 25V/uS.


    Blameless slew rate fall time.JPG This shows slew rate of falling edge, about 25V/uS.


    The closed loop gain is 21. I tested the stability by lowering the closed loop gain down to 9 and show it's stable. I also put a pole at about 300KHz to take away 45 deg phase margin at around the -3dB crossover point. It passed both test. And you can see the square wave looks good.
     
    Last edited: Mar 29, 2017
    dlucy and Hyperion like this.
  2. Alan0354

    Alan0354 Super Member

    Messages:
    2,087
    Location:
    Bay Area
    This is the FFT result. I ran 1KHz, 5KHz, 10KHz, 15KHz and 20KHz. The load is 4ohm, not the conventional 8ohm, so I expect distortion will cut in half at 8ohm.

    I ran two amplitudes, one is 22Vpp which is 15W output. Then I ran 40Vpp which is 50W output. The order is 1KHz 22V, 1KHz 40V, 5KHz 22V, 5KHz 40V, 10KHz 22V, 10KHz 40V, 15KHz 22V, 15KHz 40V, 20KHz 22V, 20KHz 40V.

    Don't read the Vrms in the display, I use a step attenuator to optimize the THD. Being a cheap analyzer, there is a window of sweet spot to show the lowest THD, I have to adjust the step attenuator to give the output amplitude from the amp to get lowest THD. But you can read the THD shown as THD L: in the middle upper of the screen. I labeled the 2nd and 3rd harmonic and you can see the number below the fundamental in the display.

    THD is still high for 10KHz and higher at 40Vpp. I am hoping I can find something wrong with the design.

    Still have a long way to go, I'll spend this week to see whether I can improve it, then starting next week will be building the amp. Hopefully I can get this done in a month. Then I can think of what is the next project.

    1KHz 22Vpp.JPG 1KHz 40Vpp.JPG 5KHz 22Vpp.JPG 5KHz 40Vpp.JPG 10KHz 22Vpp.JPG 10KHz 40Vpp.JPG 15KHz 22Vpp.JPG 15KHz 40Vpp.JPG 20KHz 22Vpp.JPG 20KHz 40Vpp.JPG
     
    Last edited: Mar 29, 2017
    OMGCat! and Hyperion like this.
  3. Alan0354

    Alan0354 Super Member

    Messages:
    2,087
    Location:
    Bay Area
    Forgot to post the schematics.

    OPS 3EF.JPG
    This is the OPS pcb with 9 pairs of output transistor. I use MOSFET as SS relay. I use 3 pairs of MOSFET to lower both the resistance and lead inductance of the MOSFET. I use 3EF with opposite polarity pre-drivers. I do this to get more headroom. I also bolt the pre-driver and the driver transistors together on a separate heatsink so their Vbe temperature drift cancel each other ( PNP and NPN placed opposite direction cancel out the temperature drift). 3EF literally eliminate loading of the VAS to get the maximum loopgain.

    Speaker protection power up delay.JPG
    This is the power up delay and speaker protection in case there is a DC of more than 1V.

    Blameless.JPG
    This is the IPS/VAS pcb. I use current mirror load for the LTP to increase the gain of the IPS stage. I use cascode to lower distortion and eliminate the Miller cap of Q9. I use darlington VAS to avoid loading the IPS. The input is really pseudo differential as the whole circuit is referenced to the ground return of the preamp. Only R22 tie the circuit ground to the power amp ground. R22 is used to limit the ground current if a ground loop is formed.

    I pull out all the stops to increase the open loop gain of the amp to eliminate distortion. The philosophy of this design is to have very high feedback gain to lower distortion. I also use clamping diodes to prevent any transistor from going into saturation at all points. This is to speed up the overdrive recovery time of the amp and make sure the amp behave nicely when clipped. You can see the clipping recovery picture that it recovered quite nicely after clipping.

    EDIT:

    The THD at 40Vpp at over 10KHz is still too high, I am hitting the wall. If anyone have suggestion, let me know.

    Thanks
     
    Last edited: Mar 29, 2017
    rogerfederer and Hyperion like this.
  4. Alan0354

    Alan0354 Super Member

    Messages:
    2,087
    Location:
    Bay Area
    I think I am at the end of the road. I suspect it's the wiring that caused the distortion, so I measure the THD of 20KHz at different points inside the amps as shown in the first picture:

    Points of measurement.JPG This picture show the points I measured. Labeled transistor where this is the output of the amp before the SS relay and all the cables.


    20KHz 40Vpp transistor output.JPG This is measured at transistor output, THD=0.00821%, almost don't see any harmonics. I think it's measuring noise. Or the QA401 is funny and have ghost level.


    20KHz 40Vpp SS relay output pcb.JPG This is measured after the SS relay show in the top picture. It shows 0.0075%. The number jump back and fore, but this is the ball park. I use 3 pairs of very low resistance MOSFET, as indicated, the SS relay did not add noticeable THD.

    20KHz 40Vpp inside banana connector.JPG This shows measurement on the other side of the internal cable. Look at the THD is 0.014%, almost double from the other end of the cable. This is two 10 gauge cable that are only about 8" or so long. Double the THD.

    20KHz 40Vpp.JPG This is measured at the banana connector going to the 4ohm load. THD=0.019%, Just going through from inside of the connector to the out side added 0.005%.

    I am using double 10 gauge wires ( the two big red wires). I question how people measure and publish their THD spec. This also show how important is the speaker cable. Look at the FFT at the output of the transistors, I don't even have much harmonics.....period. I don't think I can get better than this. I think with the cheap spectrum analyzer like QA401, this is it.

    Feel free to comment. How come I never see people talk about this. I am sure I am not seeing things. I did not even solder onto the circuit, I just hold the input to different points back and fore and it's repeatable. The result is almost as good as page 73 of Cordell's book. I think his is theoretical best. His has cascode VAS, but I really don't think it improves that much.

    Based on this result, I highly question when people claim they get better than 0.005% THD at 20KHz into 4ohm. Maybe it's a lot more forgiving to drive 8ohm. I have to try it as my load is set up to be 4ohm and easily goes to 2ohm, not to 8ohm.

    Alan

    EDIT

    Now I am just ranting!!! Maybe that's the reason people say low distortion does not matter. The damn cables add so much distortion that all the effort to lower distortion got masked by the cables. Maybe if people have giant cables and short cables, then a low distortion amp might shine. Maybe this calls for Monobloc amp so you can place the amp right next to the speaker. Just make sure the preamp can drive a long coax.....which I don't think is a problem at all for the new generation preamp using opamps. All good opamps can drive cables.

    If I am right, people really need to rethink this whole thing.
     
    Last edited: Mar 30, 2017
  5. loudnoises

    loudnoises Escalates Quickly Subscriber

    Messages:
    934
    Location:
    Pondside, NH
    I'm barely able to keep up with you here Alan, but this is great.

    THD has always been a ghost i chase. The math guy in me wants the lowest possible number, but the audio guy in me keeps questioning where the numbers relate to what my ears hear. I bounce between old setups with THD ratings claimed .1% and "newer" gear pushing it down to .01%. But they both can sound great to my ear. Your measurements are showing me that these ratings are not only suspect by nature, but almost moot compared to the distortion and variables wires, heat, time (etc) introduce to a system.

    So where to go from there? At what point do you stop measuring and see how miles davis sounds? My ears tell me that s/n ratio has more of an immediate effect on what i hear with components aside from the amp, but a low-THD amp almost always sounds better to me than one with higher "published" ratings. I'm curious what these little tweaks do to the sound.

    :lurk:
     
  6. Technics-goy

    Technics-goy Active Member

    Messages:
    258
    Location:
    KW, Ontario
    Hi Alan,
    Well I can certainly understand your frustration, I whish I could help. However I may be getting flak for this but I do not want to sound offensive in any way so please forgive me but I speak my mind freely. From some of your postings I getting the impression that you do have a reasonable background in electronics and that you do have good understanding of some of the theory involved but I think you may not have enough experience in analog to know that such a project is not that trivial.
    In some of your postings you are changing subject faster than anyone could possibly keep up with. It leaves me with the impression that you are in a bit of panic mode or totally stressed out or both. Also you mentioned that this is your first amplifier design / build. I don't know what your expectations were but did you really believe you can design and build your first amp from scratch and seriously get THD right on the money on the very first try while doing this on the side ? Seriously ?
    When you change subject so fast in your postings it seems you do not really care if a reader can follow you, how do you expect anyone to follow and offer useful help when the impression is that at the other end just a headless chicken bounces around. You are bombarding the reader with everything you read in the books but you not fully understood.
    You need to be much more relaxed, leaned back and most of all well focused. You want to focus on a single subject dial in on it and be analytical about it. In short, doing engineering stuff...
    In my humble opinion you should write a whole lot less but read a whole lot more.
    I agree that a wire mess and rats nest, which unfortunately seems very common in the DIY audio world, does not help things.
    Designing from scratch requires a very deep understanding circuit and semiconductor theory but also a tonn of experience and perhaps a PhD or even higher education.
    As I hinted in another post that when even people which mastered to become professors have some times not understood a subject to the extend they should before writing papers and teaching stuff then it becomes evident that it takes a whole lot more than just been educated to some level.

    When people tried what you tried and realize this is not easy and fail to meet the design goal then all of the sudden the design goal of having exeptional low THD is watered down and relativated and explained away as not been that important.
    True innovative analog power amp design is a specialty only very few people have mastered to a level that deserves recognition. The Japanese for sure. The Germans ? Not so much. The Americans ? Hmm...

    Also I think since this forum is mostly about vintage stuff you might find additional help over DIYaudio.

    Just my 2 cent worth of opinion
     
    roger2, tmtomh and botrytis like this.
  7. botrytis

    botrytis Trying not to be a Small Speaker Hoarder Subscriber

    To the OP - listen to the above poster - he is giving very wise advice.

    The amp looks impressive and GL with it.
     
  8. Technics-goy

    Technics-goy Active Member

    Messages:
    258
    Location:
    KW, Ontario
    Hi Alan,
    I hope your are not getting me wrong on my above post.
    I have had another look at your scope screen shot. What stands out at me is that the rising / falling edge does not look clean and straight. It appears wavy to me, suggesting that some component(s) seem too slow or unable to perform, it seems that the GNF takes over to force things back into line.
    This wavyness, which does not belong, screws up your spectrum and THD. I suggest you look for the component(s) (likely the transistors) that are slow. Look at the datasheets ft parameter. Perhaps start to systematically replace transistors and see if that cleans up the edges.
    Also the slight overshoot may need to be looked at but perhaps this overshoot goes away or will get less once you find the offending lazy transistors (or diodes) and get a nicer edge.

    2017-03-30 13.12.20.jpg
     
  9. Alan0354

    Alan0354 Super Member

    Messages:
    2,087
    Location:
    Bay Area
    Hi Technics-goy

    I hope I don't sound defensive, if it is, pardon me. I am not trying to blow my horn, just want to explain my situation.

    Actually, I am quite familiar with this kind of circuits, half of my career was designing with transistors, I was a IC designer designing bipolar opamps type of circuits for Exar. It was in 1984 where we literally designing with individual transistors inside the IC. I studied semi conductor theory from IC designs. Power amp is just a big discrete opamp. The so called blameless is just a typical internal of an IC opamp like 741. That's the reason I like Cordell so much because he too was an IC designer. I was on DIYaudio and I really talked to him a lot before. To me, his book is an extension of the famous IC design book by Grey and Mayer that was the industry standard at the time.

    I worked for LeCroy designing the front end of digital scopes in early 80s. Opamps were no where close to the speed we want, we designed discrete opamp and S/H circuits with transistors into hybrids( SMD circuits into a DIP package) that kicked the butt of Comlinear stuffs. I designed front end of ultra sound medical scanner for Seimens which was all discrete transistors circuits. It was not until early 90s then the IC opamps are fast enough and low enough noise and we moved onto IC circuits.

    Yes, I am very aggressive, I expect the first design to work......Just like real life jobs that you don't have a learning curve and you are expected to get things done the first time. I do pride on moving from one field to another and did good, from starting with MPU design, programming, to digital scope, to IC design, to Ultrasound scanner, to mass spectrometer, all completely different field, learn and master them and move to another field. I do have expectation for my first amp to at least good.

    I know you talked about your mentor and how he defies all the professor and scientist. I walked on both sides. In fact in my mass spectrometer and semiconductor metrology field, I am like one of those. That's the reason my former company called me back to do the research and development on the new generation equipment 10 years after I retired from them. They are writing a patent that I'll be part of it on my invention. I worked with a bunch of PhDs for years and I started to see the value of theories. A lot of my co-workers are PhDs, there is a middle of the road where theory and hands on meet and you get better result than if you rely on one or the other only. My experience is theory does work, and if you model it correctly, you can predict the practical result. I spent my first 20 years on the bench, but since I became the manager of EE, I pulled away and let younger engineer under me do it and concentrate on idea, theory, and invention. That worked very well.

    I see stupid PhDs too. Just because of the degree does not mean anything on the ability in design and creation. I always joke, those can do, do. Those cannot do, teach!!!

    I think people have different ways to achieve their goals. I believe in studying, think, experiment and make the first one work. My first version is at least as good as the Nakamichi Stasis designed by Nelson Pass. Right now I am trying to confirm that the speaker cable is hurting the performance as I show in the result. I expect to make this one to achieve low THD with high loop gain. Then I am going to move on and go for other topologies like the New Class A or others. I am not going to repeat another one with same approach to achieve lower THD, has to be another way.
     
    Last edited: Mar 30, 2017
  10. Alan0354

    Alan0354 Super Member

    Messages:
    2,087
    Location:
    Bay Area
    This is a picture of recovery from hard clipping on the top using burst pulses that is a lot harder on amps. If you look at the first picture that the pulse is only 15V high, not clipped, it's a lot smoother. Driving transistor to clip change the Ccb and others a lot, usually it shows the worst behavior.

    I experimented a lot on this, the slight non straight is small ringing caused from feedback from output back to input. I discovered there are two types of instability, one is the closed loop feedback stability where it's the easy part to tame. The second is the higher frequency of like 3MHz where I proved it's from output to input. At worst condition, I can hold a screw driver to the output and grab the coax of the input and make it go into oscillation. I conclude it's the high loop gain and high current at the output stage that radiates back to the input ground.

    If you notice, the input return is reference to the source, the return tie to ground through R22. This improve hum quite a bit. BUT the down fall is it's more susceptible to feedback. When I first used 50ohm, it's more prone to feedback oscillation. I lowered to 22ohm and improve a lot. I even use 2200pF in parallel with the R22 to help.

    I am pushing the open loop frequency response, I am using the fastest transistors I can find, raising the open loop gain to the highest I can get. That's the price of it. I did do stability test, the closed loop gain is 21, I lowered to 9 (adding in R28 with 15K resistor) and proved it's stable with 4ohm. Then I created a pole on the feed back path ( the optional R29 and C27 in the IPS/VAS schematic) at about 300KHz to cut 45 deg of phase margin and see whether the amp remain stable. I finally did put both in and still it's not bursting into oscillation.....it did ring like hell, but not oscillation. My amp have a lot of margin in stability, that I am sure.

    I want to get enough feedback gain at 20KHz to lower distortion, so I extend the -3dB quite high.

    Stability testing.JPG
     
  11. MarZutra

    MarZutra Super Member

    Messages:
    4,461
    Location:
    New Scotland, Canada
    Good LORD.... You guys are fabulous. I'd be a lump of quivering fried smoking puss if I ever attempted such a project. Love the choice for the chassis. Wish either of you lived handy.... I'm sure I could use your skills... Whoa. Kudos!!
     
  12. Technics-goy

    Technics-goy Active Member

    Messages:
    258
    Location:
    KW, Ontario
    Hi Alan

    Alright you do know your stuff and I did get my flak... o_O
    I am not familiar much with your circuit I would have to sim it myself to see what you see but I have not much time these days.
    No offence, are you familiar with error cancellation concept ? It was very successfully used in some power OpAmps. Perhaps you may want to explore the possibility to add that to your driver and OPS to reduce the burden of your feedback loop.:)
    Taking care of the major part of distortion at the drivers and OPS improves things by about a magnitude.
     
  13. whoaru99

    whoaru99 Epic Member

    Messages:
    38,164
    Location:
    LoTL
    FWIW, at least a couple brands I'm aware of, Bryston and Levinson, do/did suggest long ICs and short speaker cable approach is preferable to the other way around.
     
  14. Alan0354

    Alan0354 Super Member

    Messages:
    2,087
    Location:
    Bay Area
    I want to emphasize that I am not bragging or anything, I just want to tell my history and where I come from. I am comfortable with the circuit. BUT I am new in audio. I have no idea what make it sounds good or what. I am feeling my way through all these. Like some say THD is important, some laugh at it. I don't know, I can only express what I observed. Have you seen my amp comparison that I found my all low THD amp sounded identical to the Nakamichi Stasis that the design philosophy are polar opposite. I am more question perception and reality. I am still feeling my way into all these.

    I definitely am going to look into error cancellation concept like your New class A or other. Getting into class A definitely lower the THD a little. I cranked up to 2A and THD is noticeable lowered. I just want to finish this chapter as it's so close and let it be a mile stone. I am open to anything after that.

    I was on DIYaudio a while back, everyone seems to concentrated in the more primitive design like this one, the two books pretty much use this type as bench mark. So I decided to do this one first.

    The issue to me right now is my observation of THD going from 0.008% to 0.019% just from the 8" of doubled up 10 gauge wires and through the connector!!! How are people going to compensate for the speaker cables. I took down my setup after the finding last night, may be it's a mistake, I should measure on the other side of the speaker cable at the load end to verify. Today, I am on matching power transistors to build the second board.

    This is my wild guess, based on the THD got worst with higher frequencies where loop gain is low at like 20KHz. Most offensive is the odd harmonics which is symmetrical on both ends. The emitter resistor might be the problem. The more you drive the output away from the base line, the more current it takes, the more voltage drop across the emitter resistor of the output transistors. It's a natural generation of odd harmonics. So if I am right, lowering the emitter resistors will lower the odd harmonics. I am using 0.27ohm. I want to experiment using like 0.18ohm and let everything else the same and see what happens. In order to do that, I need very well matched transistors. So this is where I am at, find a set of match transistors so I can use 0.18ohm resistors.

    I hope to hear more input from you. I am just thinking out loud.

    Thanks
     
    Last edited: Mar 30, 2017
    dyche01 likes this.
  15. Technics-goy

    Technics-goy Active Member

    Messages:
    258
    Location:
    KW, Ontario
    Hi Alan,

    that's cool, I am glad you did not get me wrong. You do have a very healthy dose of self-esteem I might add. I have been working for the last 16 years or so on Radar stuff for a well known company, multi kW S-Band and L-Band. Generally in this business we are concerned with signals that the well below the noise floor, 20-30dB or so below. This makes most of the circuits involved precision electronics, well exceeding the capabilities of the finest test equipment money buys, in that sense we can not tolerate the slightest hint of distortion with signals been well in the GHz and kW range. There is no test equipment in existence that can actually measure parameters of concern, only the system itself once it is build. In my world we do not rush things, we may need several revision until we fully understand an issue that need to be corrected, getting it right the first time is managements wet dream but not reality. I understand you live in the commercial world and do not have such luxury. But for hobby amp stuff, take your time.
    Anyway, I don't want to bore you with this. I want to add I am by no means an expert in anything and I might be, if at all, of limited help to you. I am trying to help you and point you to things that are relevant in some sense and I trust that you see what I am trying to show you and that you pick up on these things since you do have background.

    Please, for your own sake, slow down a bit and take a breather, take a little bit time to think about stuff and reconsider.

    With your statement above, you misunderstood me, the Technics New Class A is not, repeat is NOT, utilizing error cancellation concept at all !
    It is more like a kick a$$ exceptional well designed Class A/AB. For now put it on the back burner for another day as you focus on the amp at hand.

    With regards to your statement above, I think the emitter resistors are fine, anywhere in this ball park 0.22 to 0.47 is useful as it gives local negative feedback on each transistor also they help with equalization and load sharing, provided the transistors are somewhat matched.
    I am not sure if a blanket statement such as lowering these resistors lowers the odd harmonics is true. For now I would leave these resistors as is.

    The amplifier topology chosen may be already maxing out the THD capability as is. I suspect that changing component values will only marginally improve things if at all. I believe you need to look into error cancellation and apply this capability to your circuit at the drivers and OPS.

    Remember when I mentioned a certain symmetry is good ?

    This is what I meant.

    cancellation.gif
    This figure above shall be just a general representation of the principle of Error Cancellation that I would expect you to easily pick up on.
    Referencing to the circuit colored in red.

    Do you see, what I see ?

    The cancellation circuit pictured is an easy add-on (with appropriate mods where needed) to pretty much any push-pull topology, I'd say. The circuit is due its AC coupling with just 50 some pico effective at higher frequencies only and is fully transparent at lower frequencies.
    The transistors colored in red shall be nicely paired, e.g. been on the same die would be preferable.

    The approach of error cancellation is one of delicate finesse in my books.
    Adding 8 more pairs of OPS or another 100000uF charge caps or jacking up the idle current to a ludicrous level is filed under BRUTE FORCE with diminished value of return on money spend ! Would you agree ?

    Anyway, there are other means of taking care of distortion using similar techniques, but I believe pictured above shall suffice to guide you out of your bind.

    Now the ball is in your corner, fire up your sim and add 4 more transistors and see what you get.

    Talking about getting out of jail free card or so....
     
    Alan0354 likes this.
  16. Alan0354

    Alan0354 Super Member

    Messages:
    2,087
    Location:
    Bay Area
    Hi Technics-goy
    That's an interesting circuit. This circuit actually support my suspicion in the way of higher amplitude induces odd harmonics.

    Analyze error cancellation.jpg

    I did a quick analysis. I labeled all the transistors and I drew a step of 1V on both the base of Q1 and Q6. If you follow the path from Q1, Q2, Q3 and the 0.1ohm resistor to the output. I am going to ignore the 51p and just short it out. As voltage increase, more current through Q3 and source resistor and more voltage drop across them. I labeled only 0.9V step at the output from 1V step at the base of Q1. Then look at Q13( lower current mirror of the error cancellation circuit). The voltage drop across Q13+the emitter resistor ( I call it 1K) drop by 0.1V. So the current through them drops 10%. This current mirrored through Q12 which decrease the current drop at the point of base of Q1 and Bias1. Same idea on the bottom where current increase 10% at junction of base of Q6 and Bias1.

    Since the conventional VAS is common emitter circuit using current drive into the input of Q1 and Q6. The current change literally pull the voltage on base of Q1 and Q6 up some to compensate the lost of voltage on Q3 and the source resistor. Am I getting this right?

    This is actually quite easy to mod into my circuit. If you look at my OPS schematic, my circuit is exactly the same except I use BJT output transistors, not MOSFET. I need to simulate this first as I have a concern. This is positive feedback, Usually there is price to pay for positive feedback. But if this works, I think I can add the 4 transistors to try it. If this works, it should work for normal 3EF output stage also. I did try to design some sort of enhanced class A with something like the idea of the New class A. I ran into positive feedback in simulation and causing ringing and even oscillation. Any positive feedback should be looked with concern.

    Still the most pressing problem is the interconnection. The THD gone from 0.008% to 0.019% just going through the two red wires and the banana connector. This will destroy any benefit out of the error correction circuit.

    I don't have a lot of RF experience, I venture out for a little over 2 years into RF world but not a lot of achievement. I did a lot of study in RF, distribution element design like using strip and microstrip as inductor and capacitor. I used Microwave Office at the time to do simulation. I did all the layout myself and I did made my design very close to simulation on the Smith Chart when I tested on the network analyzer. But I only ventured out for like 2 years and the original company( the one I worked before and the same one called me back for the contracting job in 2015) called me and I went back to semiconductor metrology design again. I got spoiled by the environment of the company. The people are mostly PhDs and ultra bright. People are a lot more creative and accept new ideas. Yes, RF is more challenging and interesting on paper, BUT.....BUT, when I ventured out, I found people lack of creativity. They literally took the circuit from the application notes and put it on the board. I don't thing they stop and thing whether the circuit is optimized. The application engineer is human, they make mistake also. I would never accept any circuit at face value. But it seems like that's how the industry become......at least two companies I ventured to. I rather go back to the very creative environment.

    The company I worked for is a small company, used to call Charles Evans and Assoc. Later sold to Physical Electronics and now becomes Revera Inc. But the top people are the same few very bright people. We sit around, exchanging ideas, literally created new concept. I published two papers in the America Institution of Physics, Review of Scientific Instruments while working with them and my idea landed a phase II grant from the DOE worth $500K. I got a patent under my name while working with them. I like this kind of environment of working. That's why when they called me in 2015, I dropped the amp and went back. I am sure the technology is not as challenge as yours, but it's pretty exciting to be in that kind of environment. I retired for 10 years before they called me back, in my retirement, I spent a few years studying math and electromagnetics, then I started designing guitar amps, I designed two ( not just copy Fender and change a few resistors) channel switching and with power scaling that I can crank it down to bed room level and still sounds like at gig level. I even came up with a guitar noise cancellation concept and I wrote and got a US patent on my own in 2014. Electronics is my one passion of my life. I do push myself, I don't take it easy. I am still almost as busy in my retirement as when I was working. Good thing is my company let me work at home, I work till 3am a lot of times.

    But as I said, I want to finish my amp first. I am very goal and result oriented. I don't want to keep changing and never get it done. My company might have phase II design again, but I even told them not in the next month and half. I don't want anything to stop me from completing my first amp.....for the better or worst.

    Thanks for the idea, I'll definitely look into this and also the new class A. I stuffed 4 pcb of each, after I completed the first amp, I have more pcb ready for adding the error correction circuit. Who designed this?

    EDIT:

    One problem with your schematic. The diamond is not stable at all. I started out with this in my first version pcb, it oscillated, I traced, the driver (Q2) feed back the signal to the collector of Q6, so is Q5 back to Q1. I wasted time chasing the circuit, finally I put Q1 collector to -ve rail and Q6 to +ve rail. Everything works like a champ. You can refer to my schematic in the first post. As I said, I layout as if it's RF circuits with ground plane and power planes, I am sure it's not my layout.I actually test this without IPS and VAS, I just drive the voltage spreader directly without any NFB. It oscillated by itself. Not good. Cordell showed this in his book, I question he ever actually tried that. Simulation looked good, but not in real circuit. I am sure someone might monkey with it enough to make it work. But it's not a good circuit.

    Alan
     
    Last edited: Mar 31, 2017
    kjello and Hyperion like this.
  17. Alan0354

    Alan0354 Super Member

    Messages:
    2,087
    Location:
    Bay Area
    Tonight I am busy sorting transistors, I am matching power transistor to within 2 to 3mV in the set. I will hook up the setup and look at the other side of the load after the speaker cable and see what is the THD looks like. Ha ha, I have a lot of Monster cable built, I'll take a look at the effect of single vs multiple pairs, long vs short. I just forgot to do that last night.
     
  18. Technics-goy

    Technics-goy Active Member

    Messages:
    258
    Location:
    KW, Ontario
    Yes, it is an interesting and very useful circuit when applied properly yields about 10 to 20dB improvement in THD at 20kHz. That should do the trick.

    The concept of cancellation is neat. BTW. are you familiar with a term called 'zeroing in' ? That too is part of a neat concept, but for another day....

    I believe you do not want to short these small caps, you are DC coupling now and you are exposing the possibility of DC drift, furthermore this circuit may now become effective over the entire band, you don't want that, you want this circuit to chime in only on higher frequencies where the GNF is lower and not as effective. GNF performs fine at lower frequencies and THD there is likely not an issue for you.

    Yes, as I said, it lends itself to be an easy Add-On circuit to pretty much any Push-Pull topology. I know I am repeating myself.

    Well very obviously this circuit as is, is not an exact copy, cut & paste directly into your OPS, you will have to massage this a little bit to suit your needs. This was meant as a close example how cancellation can work.
    Your sim will guide you.

    If you were to pause for a minute (which I know you will not) ;)and reflect and think a bit about this approach you might realize you could make this easily work by adding only two transistors not 4.

    I believe by design as is, this circuit operates as localized negative feedback not positive as you concluded, so I disagree here. Positive feedback here makes no sense to me. I simmed this circuit some time ago and there is no hint of oscillation, this circuit is stable provided you maintain AC coupling via the small caps.
    This circuit is a little bit of a mind bender if you try to follow it in your mind. Notice the double signal crossing and location of the PNPs vs. NPNs.

    Altho there is criss crossing going on it keeps ones mind straight :D

    Yeah, I hear ya, well 2 years in devils kitchen is good experience, you have now an good appreciation to the importance of trace width, board material, loop area and parasites and so on.... I watched my chief engineer in amazement when he did his 'magic' waving his hand over the RF amps while observing the power meter...
    This gave him indication how well the circuit performed and if there any annoying RF leaks.
    Physical location of RF transistors on the amps were beyond crucial, location variance of as little as 0.1mm would make a pass or fail and cause a nasty RF leak. When working with kW RF power 0.1dB more or less is a lot of power that needs to be accounted for....

    Oh boy if RF people there were just able to do copy, cut & paste 1:1 from the app sheet is sad, very sad. Probably highly underpaid and not motivated or total duds with some sort of a degree. Way to go.o_O

    Well the first thing that Rick Hartley drilled into me was not to trust any application paper whatsoFever. Rightfully so, it seems these days the students write the datasheets and application papers, riddled with errors. They manage to get something barely to work with a stackup that is totally wrong and then they proceed to document they're ineptitude in the app sheet. Oh boy.... Freescale is really bad.:no:

    Well, I am glad you realized the brilliance and potential this little circuit offers. I don't know who designed this, I have no idea but if I had to bet money, I'd bet it was not a PhD that came up with this but a real engineer, most likely one out of Japan.;)

    To my knowledge this circuit was widely used in high-end op-amps, I suspect there could be patents on this.

    Ehm, no, I can not claim credit for this circuit, it's not mine.

    Referencing one more time to the screen shot you took when the amp came out of clipping and exhibiting a little wavyness on the edge there, I believe you may still want to investigate what is causing that. This is likely still a lingering problem.

    If you have observed oscillation, that is not good at all, you need to locate the cause for this and take measures to prevent oscillation.

    I found that transistor lead length was in part problematic in some circuits, shortening the transistor legs helped, also adding small caps of few picos across the offending transistor helps to stabilize.

    As a final remark, I found horsing through a design process generally leads to inferior solutions. But as long as management is happy with the outcome of an endless trail of ECO's then things are just peachy....

    If management would ever push me into such a situation to horse through, skipping every corner in the book to meet some over hopeless committed deadline, I would tell them something....
     
    Hyperion likes this.
  19. Hyperion

    Hyperion Roobarb & Custard Subscriber

    Messages:
    39,121
    Location:
    Hertfordshire, UK
    This is a very interesting thread, you guys are amazing, I wondered when you two would bump into each other. :)
     
    Last edited: Mar 31, 2017
  20. Alan0354

    Alan0354 Super Member

    Messages:
    2,087
    Location:
    Bay Area

    Please expand the above. I commented inside what you wrote in blue color.
     
    Last edited: Apr 1, 2017

Share This Page