DIY Audio Projects Forum
 NEW  Matt presents bias and operation data for the 6V6 tube in SE operation - 6V6 Single-Ended (SE) Ultra Linear (UL) Bias Optimization.

DIY Audio Projects Forum

Welcome to the DIY Audio Projects Message Forum. Use these forums to discuss Hi-Fi audio and to share your DIY Audio Projects. Registration is free and required to post messages and view the file attachments. Registration will only take a minute and registered users do not see any advertisements. After you have completed the online registration process, check your email (including spam/junk folder) for the verification email to activate your account. New members are under moderation - so your posts will not be visible until approved by a moderator. See the Read Me 1st, Forum RULES and Forum FAQ to get started on the forum.

It is currently 25 Sep 2018, 01:55

All times are UTC - 6 hours [ DST ]




Post new topic Reply to topic  [ 103 posts ]  Go to page Previous  1, 2, 3, 4, 5, 6 ... 11  Next
Author Message
PostPosted: 09 Jul 2018, 01:33 
Offline

Joined: 26 Dec 2016, 03:46
Posts: 119
Location: Bayarea
I have a third module of the same circuit that I use for experiment and change parts to develop the board. Today, out of desperation, I just put that module in place of the bad module, IT WORKS!!! BUT when I put the bad module on the test stand, it tested out fine again!!!

When I built the pcb, I built in batch of 6. I measure every single resistor, every single capacitor in the signal path before putting it in. I was very careful. For the life of me, I don't know what went wrong with the bad module. ALL DC voltage measured out correct, tested out to be absolutely stable. There is nothing wrong other than measure 0.012%THD compare to all the others that are in 0.004% or lower. And that happens only in the amplifier chassis, not even on the test stand.

This starting to go into X-Files!!!! I am going to just put the third module in and take my time to look at the bad one. I am running out of patience.

Again sorry for the ranting, got nobody to vent to, can't tell my big boss as she'll just give me a blank look!!!


Top
 Profile  
 
PostPosted: 15 Jul 2018, 04:07 
Offline

Joined: 26 Dec 2016, 03:46
Posts: 119
Location: Bayarea
Been putting in quite a bit of hours. Made a lot of inroad into the problem. The grounding is very critical for high current device like power amp. even with double sided 2oz copper ground plane, moving the soldering point for the front end reference makes a difference. Less than 1" can cause problem even on a solid plane. Still have observation that doesn't make sense, but it's a big improvement already.

High current circuit is very tricky, just a little imperfection can cause the THD reading double or tripple when getting down to 0.005% at 15K or 20KHz.


Top
 Profile  
 
PostPosted: 27 Jul 2018, 18:29 
Offline

Joined: 26 Dec 2016, 03:46
Posts: 119
Location: Bayarea
I solved the problem, it is two fold:

1) Grounding of the OPS to the system is very critical, the return of the speaker is ultra critical to set up the ground reference for the front end stage. It has to be perfectly star ground or else the slight modulation of the ground causes the increase of THD.

2) The Quantasylum QA401 I used don't have isolated ground from output driving the amp to the input from the output of the amp to read the THD, the slight ground bounce causing the false reading.

The grounding is so much more critical in power amp. I was working as signal integrity engineer that specialized on grounding, groundloops, crosstalk and emissions. This is so so much more sensitive and critical in power amp..

After I fixed the grounding, there is an obvious improvement of the sound. This has a bigger effect on the sound than different front end topology can produce. This make me starting to wonder why put so much effort designing the front end. It's the OPS and grounding, filtering that make or break the power amp.


Top
 Profile  
 
PostPosted: 29 Aug 2018, 03:44 
Offline

Joined: 23 Feb 2017, 02:02
Posts: 193
Thanks for sharing your schematic, sorry for the late reply.

With regards to circuit troubble shooting.
I always recomend using isopurrpple alchool (prefered) or use actetone to clean your pcb. With a medium hard brush. Flux can sometimes cause leakages reducing sound quality or at extremes causing a ampflier to blow up.

With regards to your ranting its perfectly fine I don't think theres anything wrong with it.

Ground wirering is deffintly importnat which I aggree, from my exprience I tend to get plorbems when the feedback resistor and input ground has large resistance or large currents flowing in them.
By directly grounding to mains the end of the feedback resistor which connects to ground via capacitor or direclty usually gives my amps lowest noise.

With regards to schematic shown, you can improve perfomance by removing r81 and c67, given if the amp can operate without oscillation This will increase open loop voltage and thus more feedback lower distortion.

By setting the base voltage of q46 q47 to less than half VCC preferabley at about 6-12v. You reduce the voltage accross q48 q49. You can now direclty replace these boys with a hifi input fet or mostfet. My choicse is cheapo from china 2sk30a works well have rated voltages of 50v and is hifi. Howver again amp needs to be checked for oscillations.
By using mosfets or fets you get infnity input resistance and improve sound quality, gain maybe lower if r90 and r91 is a short, but with r90 r91 gain approx about collector impedance divided by r90 in both fet and transitor diffrential pair.
Hint 2sk30a tends to have some voltage at gate if you have more than 12 v accross it.
By replacing to mosfet you can also directly remove your input capacitor and have direct input.

By replacing your current compensation type to indirect compensation with RHP zero nulling resistor you also instantly increase high frequency gain and increase gain at human vocal range thus higher fi.
http://cmosedu.com/jbaker/students/thes ... ifiers.pdf

If I was building the schematic I would replace r75 or r76 with a mulit turn 100ohm in series with another 100ohms to allow for output voltage dc adjustments to zero. Its more safte to have pot in parallel with resistor but from expreince even cheapest chiense multi turn pots are highly reliable so no need.

I would also increacse c84 to 220uF or something like that, to reduce current noise at diffrential pair, by doing this less noise inside op amp and higher fi.
Q50 which is acting as dymanic load for voltage gain stage deffintly needs allot of noise reduction, the diode biasing will deffinlty result in allot of noise esp without any capacitors accross them

Fruthermore if you highly seek more sound quality I would recomend you to just copy and paste one of the thee stage op amps inside the pdf, and redisign it to drive amps using transitors.
I was going to do this for one of my projects if I run out of ideas, however I was lucky to succeed in a 4 voltage gain stage op amp which is pure uniuqe and geniune design.


This post has a file attachment. Please login or register to access it. Only Registered Members may view attached files.



Top
 Profile  
 
PostPosted: 03 Sep 2018, 16:23 
Offline

Joined: 26 Dec 2016, 03:46
Posts: 119
Location: Bayarea
Hi
I have not check here for a while.

Regarding DC offset adjustment, I don't do that, less circuit the better. Here is what I've done and consistently gave me less than 25mV offset at the output. I don't want to use dual match pair because beta are too low on those transistor. The one I am using has beta of over 400. So I first use a test fixture to pick out match pairs to below 3mV. I layout the pcb so they are butt up to each other. Then I put sticky copper tape to thermally tie them together as shown below.
Attachment:
IPS matching temperature pair.JPG

I cover the pair with copper tape and solder over it to form a metal cap over the two matching transistors. This gives very good temperature tracking and thereby reduce offset drift with temperature. I have very good luck with this. I am not planning to sell it, so I can customize it. If I ever try to sell, it's going to go to very high end market, so hand made is not an issue. It's not that I afraid of using SMD parts as you can see everything I do is SMD. It's the matter of finding the best transistor and not compromising anything.

I was an analog IC designer before, I am very familiar with all different opamp designs. Power amp is just a big discrete opamp. The biggest difference is large signal and that's the most challenge part of power amp. All the front end designs are the easy part and really doesn't make that big a difference. It's all on the OPS section and the power supply.

I have already in mind designing a version using MOSFET transistors. I even picked out some matched pairs.
https://www.digikey.com/product-detail/en/advanced-linear-devices-inc/ALD1101SAL/1014-1003-ND/2414279
https://www.digikey.com/product-detail/en/advanced-linear-devices-inc/ALD1102ASAL/1014-1005-ND/2414281
And then use MOSFET for the VAS stage.

I find 2 stages gain is plenty. Most high end amps on the market are just 2 stages with lower gain than my amp. I have not seen an amp that has higher gain than mine. That's the reason I achieve below 0.004% THD at 60Wrms at all audio frequency. You don't want ultra high feedback gain as the main way of lower THD and get high damping factor. Loop gain should be used as final touch. The OPS is the most important thing. Have a good OPS design, you should not need 4 stages to jack up the loop gain. It's too long to explain here. My OPS stage has output damping fact over 200 without the need of any closed loop feedback. There is no short cut on the output stage, more transistors, run at high current to get a big class A region.

If you look at my OPS, it's a 3EF output stage that has very high input impedance. This jacks up the loop gain tremendously. Also, I use darlington VAS to jack up the input impedance so it does not load down the input IPS stage. These are all very well explained in the book by Bob Cordell.

Here are a few THD graphs and waveform of the amp I posted.
Attachment:
Rev 5 1kHz 45Vpp 4ohm.PNG
Attachment:
Rev 5 20kHz 45Vpp 4ohm.PNG

The noise floor is the analyzer's limit, it dose not change even if I turn off the amp. So the amp has lower noise than the test equipment. You can see in the middle of the pictures it shows THD of 0.00146% at 1KHz and 0.00329% at 20KHz at 63Wrms output power into 4ohm load.

Attachment:
Overdrive.jpg
Attachment:
Squarewave.jpg


The BW of the amp is about 350KHz with slew rate of about 33V/uS.

The magic is on the output stage in page 1 of the schematic. You don't need 4 gain stages. Most of the distortion is from OPS. This is not the lowest THD amp, my other design is below 0.002% across audio frequency. Also with just two stages.

Remember, most of the gain stages are really transconductance amp ( voltage to current amp). The voltage gain mainly depends of the impedance that is loading the output. The key here is to present very high impedance load for the gain stage and you get the high loop gain with only two stage. Get the Bob Cordell book if you are interested in designing power amp. IC opamp book is only the introduction into power amp design. It's the OPS that is everything.


This post has a file attachment. Please login or register to access it. Only Registered Members may view attached files.



Top
 Profile  
 
PostPosted: 03 Sep 2018, 21:17 
Offline

Joined: 26 Dec 2016, 03:46
Posts: 119
Location: Bayarea
BTW, I did a lot of closed loop feedback designs dealing with poles and zeros and stability designs. High initial loop gain ONLY affect the final precision of the voltage. It's the poles and zeros that govern how high the frequency response you can get. You want to have single pole roll off at the -3dB crossover ( that is the top of the bandwidth). No matter how good you are, you are at the mercy of the available components. Audio transistors are very low frequency. fT or about 150MHz for small signal transistors, 30MHz for big transistors. This means the parasitic inside the transistors will cause phase shift at relative low frequency.

I think this kind of effect is NOT include in the simulation models. It is definitely NOT included in LTSpice as it is almost useless to do stability simulation with LTSpice. I have been searching up and down for the highest speed transistors and use them. The only transistor I use that's not the fastest is the big output transistors. The ones from Sanken can go as high as 50MHz instead of the ones I use that are 30MHz. Problem is the Cob and Cib is very high compare to the ones I use. Also, if you look at the datasheet on the fT plot, it's really not that big a difference. The higher frequency the transistor, the less it will interfere with the stability of the amp. It would be so much better to use transistor with fT>500MHz. I design RF circuits, I am not afraid of high speed devices.

OK, all these talk is going to lead back to the closed loop feedback. You know beyond a few hundred KHz, all the transistors will add addition phase shift. So you have to have compensation so the -3dB is before the effect of those transistors. So just say the best you can do is -3dB 500KHz. At that point, loop gain is 0dB or unity. That means you have only loop gain of 20dB at 50KHz and 40dB at 5KHz. You can play with very fancy two poles roll off first and then put in zeros at around 400KHz to bend the phase back at crossover. ( lag-lead network) ( I use those already) to get a little higher loop gain at 20KHz or below. But there is a limit to this. That's the extra compensation you see in my design.

What does all these means? This mean, if you put more stages to jack up the loop gain, the more aggressive you have to compensate. At the higher frequency range ( above say 5KHz), it is very likely you are not gaining anything with all the high gain you put in as you have to roll off earlier to fit to -3dB at 500KHz to ensure stability. All you get out of 4 stages gain is your amp is more precise at low frequency of 100Hz or so. You are not gaining much if any. A good design of 2 stage gain will get the best out of it as in my amp. 0.0015% at 1KHz and 0.0033% at 20KHz is very good by any standard.

The more stages, the more aggressive compensation, the impulse response might suffer. this will show up at square wave test that you might have over shoot and ringing. If you look at the picture of the square wave of my amp, no over shoot, no ringing.

This is what I have been emphasis all along, the large signal output stage is the KEY and is the hardest to deal with. At large signal, Cib and Cob change drastically along the waveform. It is all too common you see oscillation on part of the waveform. It is so easy to compensate the amp to get stability just sitting there or in small signal situation. It's the large signal situation that is very trying. This is where the normal opamp design goes out the window. It's only a good start. To test stability, it's another can of worm. That's the reason so many popular amps oscillate with certain speakers or speaker cables. They did not do their homework and the amp is only conditionally stable. This is where you earn your keeps.


Top
 Profile  
 
PostPosted: 04 Sep 2018, 16:13 
Offline

Joined: 23 Feb 2017, 02:02
Posts: 193
Thats a very smart way in reducing offset drift I like, IMO becasue the current mirror with transistors controbute most of the impedance, changing the emitter resistor a bit shouldn't change the open loop gain much or at all so I personally like offset adjustments.

That seems like a nice choice for a diffrential pair the mosfet you listed.

Not trying to show off, but unfortunately that's what I observed.
I had used excatly same output stages in many head amp builds, I observed large sound quality diffrences from going 2 stage op amp to 3 stage, then 4 stages. By each increacing stage having extra low frequency gain of arround 200 times meaning distortion reduction of 200x more per stage after feedback.
IMO the output stage is as important as gain stage.

With 8R load. Cause I use 8R phones. Also same transitors used.
The 4 stage ampfiler showed higher frequency response, higher phase margin very close to 90 degree and much higher gain compared to my 3 stage amplifier.
And the 3 stage amplifier also showed higher frequency response than 2 stage, the 3 stage one is fully pure capacitive load stable but the 4 stage is not.

The DC stablity is very high due to openloop gain in excess of 163dB and is highly immune to temperature diffrences due to high loop gain.
It was very hard to design becasue the more stages of gain the harder to get stable.
The 4 stage only required compensation on two nodes so the high frequency gain is large.
I've designed and built 7 models of these 4 stage op amps and was going to give up but luckly succeeded in 8th try.

I don't use simulators at all because I had plorbems with it not matching up in real life and its not helping me at all to design anything new.


Top
 Profile  
 
PostPosted: 04 Sep 2018, 19:32 
Offline

Joined: 23 Feb 2017, 02:02
Posts: 193
Higher frequency response after compensating by pF capacitor in parallel with feedback resistor to fine tune so that sine wave distortion is minmal while flat voltage response accross 20-20khz.
Freq response no load and no external compensation is simliar in all systems, however open loop gain is much larger at 100khz with more stages.
4 stages had slightly better maximum frequency response after feedback so gain is set to about 3.2x

With regards to scope probes, do you use ones with attenuation, or do you use ones thats direct wire 1x attenuation?

I've found in the past a 10x probe acted as a filter and the scope couldn't pick up the parasitic oscillation from a oscilalting amplifier, therfore I always use 1x probes direct in.
I wounder if you've tired something simlar.


Top
 Profile  
 
PostPosted: 05 Sep 2018, 01:08 
Offline

Joined: 26 Dec 2016, 03:46
Posts: 119
Location: Bayarea
KochiyaYamato wrote:
.....................
Not trying to show off, but unfortunately that's what I observed.
I had used excatly same output stages in many head amp builds, I observed large sound quality diffrences from going 2 stage op amp to 3 stage, then 4 stages. By each increacing stage having extra low frequency gain of arround 200 times meaning distortion reduction of 200x more per stage after feedback.
IMO the output stage is as important as gain stage.

With 8R load. Cause I use 8R phones. Also same transitors used.
The 4 stage ampfiler showed higher frequency response, higher phase margin very close to 90 degree and much higher gain compared to my 3 stage amplifier.
And the 3 stage amplifier also showed higher frequency response than 2 stage, the 3 stage one is fully pure capacitive load stable but the 4 stage is not.

The DC stablity is very high due to openloop gain in excess of 163dB and is highly immune to temperature diffrences due to high loop gain.
It was very hard to design becasue the more stages of gain the harder to get stable.
The 4 stage only required compensation on two nodes so the high frequency gain is large.
I've designed and built 7 models of these 4 stage op amps and was going to give up but luckly succeeded in 8th try.

I don't use simulators at all because I had plorbems with it not matching up in real life and its not helping me at all to design anything new.


What do you mean by "same output stages in many head amp builds"? Same as mine?

What is you -3dB frequency response?

What is your power supply voltages, that is what is the max swing of your amp?

How do you measure that you have 90deg phase margin?

You have means to measure THD?

It's impossible to talk without schematic.


Last edited by Yungman on 05 Sep 2018, 01:21, edited 2 times in total.

Top
 Profile  
 
PostPosted: 05 Sep 2018, 01:13 
Offline

Joined: 26 Dec 2016, 03:46
Posts: 119
Location: Bayarea
KochiyaYamato wrote:
Higher frequency response after compensating by pF capacitor in parallel with feedback resistor to fine tune so that sine wave distortion is minmal while flat voltage response accross 20-20khz.
Freq response no load and no external compensation is simliar in all systems, however open loop gain is much larger at 100khz with more stages.
4 stages had slightly better maximum frequency response after feedback so gain is set to about 3.2x

With regards to scope probes, do you use ones with attenuation, or do you use ones thats direct wire 1x attenuation?

I've found in the past a 10x probe acted as a filter and the scope couldn't pick up the parasitic oscillation from a oscilalting amplifier, therfore I always use 1x probes direct in.
I wounder if you've tired something simlar.


I use 10X probe. I never see problem with 10X probe not seeing high frequency oscillation. This is audio amps, even oscillation are below 10MHz or so. I use 10X probe to look at circuits into 200MHz with absolutely no problem. Been working on amplifier designs with BW of over 300MHz. Higher than that, got to have 50ohm coax termination.

What's the BW of your scope? Mine is a 400MHz scope, no oscillation can escape that.

You cannot probe high impedance nodes with probe, even a few pF will change everything.


Top
 Profile  
 
Display posts from previous:  Sort by  
Post new topic Reply to topic  [ 103 posts ]  Go to page Previous  1, 2, 3, 4, 5, 6 ... 11  Next

All times are UTC - 6 hours [ DST ]


Who is online

Users browsing this forum: Garygo and 2 guests


You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot post attachments in this forum

Jump to:  
cron
DIY Tube Projects :: DIY Tube Amp Kits :: DIY Speaker Projects :: DIY Solid State Projects :: DIY IC / Op-amp Projects :: DIY Phono Projects :: DIY Cable Projects :: Hi-Fi Audio Schematics
© diyAudioProjects.com - Terms of Service - Privacy Policy