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

Analog Line Level Preamplifiers , Passive Pre-amps, Crossovers, etc.

New Doug Self pre-amp design...
New Doug Self pre-amp design...
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 16th May 2012, 06:51 AM   #251
owdeo is offline owdeo  Australia
diyAudio Member
 
Join Date: Jun 2003
Location: Sydney
Hi Chrisitan, not at all, interesting to hear your views. You seem to be making a few unflattering assumptions about the design though... this was no rough veroboard botch up - it seems a very nicely implemented design to me, and the PCB layout looks to have been very carefully done. Since you've hinted at a few interesting things I'm going to have to you ask you more about them now, so here goes:
Quote:
Originally Posted by ChristianThomas View Post
The first is to know your reference. Your signal is between one thing and another. And when it comes out it is by comparison to that same reference. Mostly this means a jolly good earth. Without it you are lost. So make sure your output is referenced to the same place. You really do have to look at where the currents go and make absolutely certain nothing has gone wrong. Worry if you see shared paths of any sort. (This actually doesn't mean that you need an elaborate star ground, just something decent and intellectually defensible. But stars are generally good, if in doubt.)
The gnd of the power input terminals on the board is used as the reference for all signal grounds, which are seperate planes around the opamps for each channel that connect back to this point, and the opamp decoupling cap gnds are seperately routed back to this point also.

Quote:
Originally Posted by ChristianThomas View Post
Secondly you need a decent power supply. This will come from a 317 and perhaps its mate. If these are going to be properly referenced supplies the have to go back to the same place. This is your ultimate reference. You will also need to smooth this supply. An RC will usually do, but do not think that the accepted RC on the way in is correct. It will likely not be. Make these choices for the op amps you are going to use.
The seperate PSU is indeed implemented using the 317/337 pair with all the usual bypass caps. There's no RC filtering though, just two 220uF electros at the power supply input of the preamp board. Why does it need RC filtering, why would the values be different for different opamps and/or how do you suggest appropriate values be chosen?

Quote:
Originally Posted by ChristianThomas View Post
And have a proper look at the bypassing needed by the op amps 100nf, though standard, is likely not the right value, even if spread around. These capacitors pass signals that are inside the op amps and you had better learn how the rejection ratio happens across their Miller cap. That is why it tails off. (You can perfectly well use any op amp without understanding this, but you will need to when you come to choose them). Knowing how the bypass caps work for different designs is probably key to getting the best out of them.
There is a 100nF polyester cap from each opamp power pin to gnd and they are right next to the opamps. Other than that, I'm afraid I have no idea what you're getting at. As far as I can remember, every opamp datasheet I've ever seen shows this value in the applications section, and while other values might work just as well, isn't the point that the value isn't critical? Aren't they just reducing the impedance of the supplies at AC and providing a path for unwanted noise to return to gnd before if gets into the opamp? As long as the thing's stable what else matters?

Quote:
Originally Posted by ChristianThomas View Post
Then you should look at the loading on the output. It's nice if it can do 600 ohms, but just bloody ignore that. There are about 3 op amps that can manage it. But it had better be able to do 2k, even if it is aiming for 10k.

And within the circuit, have a look out for loads that you don't even notice, like within feedback loops. Those are still loads and didn't get magiced away supposed op amp behaviour,
CT
Well the first opamp has 4k7 and 2k feedback resistors. It's driving a 10k pot, so the total load on its output is about 4k. There's also a 220p cap across the 4k7, which should be insignificant loading-wise. The second opamp is a unity gain buffer. It was driving the ~22k input impedance of my power amp via ~2m of sheilded audio cable. None of this seems like it should present a difficult load to either opamp type. Do you disagree and if so, how did you come to that conclusion, ie how do you determine what the minimum load should be for a given opamp?

Quote:
Originally Posted by ChristianThomas View Post
You probably know all this already, but the fact is that your really ought to be able to put three op amps down on a piece of veroboard, an input, a volume control and an output, and have it sound pretty much near perfect, irrespective of the amplifiers you chose.
CT
I couldn't agree more, that was my point. Perhaps I am particularly incompetant at constructing these things, but I don't really believe that. The trouble is I'm not sure that any of this helps to explain what I was hearing and why the measurements didn't correlate. Anyway, as I'm sure we all know it's a massive and complex subject and we've probably digressed too far in this thread already
  Reply With Quote
Old 16th May 2012, 07:10 AM   #252
owdeo is offline owdeo  Australia
diyAudio Member
 
Join Date: Jun 2003
Location: Sydney
Quote:
Originally Posted by Mooly View Post
You could always try FET opamps to eliminate DC bias currents (pot wiper currents) which was the main reason for all the electros in the 1996 precision preamp.

My own view on absolute noise figures is that they are one of the least important parameters in a domestic set up. Perhaps the LM4562 with its significanlty better DC specs might be an option. I suspect in practice that there would not be a problemone.

I think it's a superb design and one I would perhaps include in a new amplifier build.
Yeah I'd thought about that too and saw your post about having tried TL072s. Trouble is their distortion is going to be much higher at the impedances used in the design. I've also tried them in other things before and was not keen on the sound.

I think if we follow the teachings of the designer the only real contender is the 4562. I might try this in a few locations and see what sort of offsets result. This might be another good reason to use bipolar coupling caps (at least initially) as the input offsets will be different polarities between different opamps...

I totally agree re SNR in domestic hifi - I've found that even preamps I built many years ago with at least 20dB worse SNR don't produce any noticeable noise in practice, though my speakers are 85dB/W B&W801s which probably helps...

Very glad to hear you like the preamp. I'll let you know how mine goes when it's finished.
  Reply With Quote
Old 16th May 2012, 02:06 PM   #253
AndrewT is offline AndrewT  Scotland
R.I.P.
 
Join Date: Jul 2004
Location: Scottish Borders
Quote:
Originally Posted by owdeo View Post
.................... The published THD+N curve was below 0.001% from 20 to 20k at 1V.
.......................... But, massed loud violins sounded a bit screachy. The Blues Brothers cranked up high sounded harsh too, but much more enjoyable. Clearly I was hearing distortion of some sort that wasn't noticeable with the OPA2134.

At the time I had access to a Neutrik A1 analyser and I measured THD+N vs freq at 0.5 and 2V output with each opamp. It was below the resolution limit in all cases (-100dB) from 20Hz to 20kHz.

The distortion causing screachy violins must have been much less than 100dB below the signal, so it seems that the THD+N vs freq curve was not picking it up. .............
THD of continuous non changing sinewave test signals seem by repute to not give much indication of good/bad sound reproduction.

Badly implemented opamps and power amps that have poorly selected phase and gain margins do seem to sound quite different from correctly compensated opamps/power amps when reproducing audio/music signals.

I think that in there is the clue to why we hear different amps sounding differently.
I think I remember correctly in paraphrasing Cordell with "all amplifiers sound the same when they are asked to process signals within their specified limitations. It's when the amplifiers misbehave using signals outside those limiting values that amps sound different"
BC, please forgive me in getting that quite wrong and being too lazy to go and find the correct quote.

From all the above, I strongly suspect that undershoot and overshoot of changing signals, is what you hear as differences in sound performance.
__________________
regards Andrew T.
  Reply With Quote
Old 16th May 2012, 02:19 PM   #254
ChristianThomas is offline ChristianThomas  United Kingdom
diyAudio Member
 
Join Date: Feb 2006
Location: On a hill, in a wooden shack, next to the woods, in Somerset.
Quote:
Originally Posted by owdeo View Post
Hi Chrisitan, not at all, interesting to hear your views. You seem to be making a few unflattering assumptions about the design though... this was no rough veroboard botch up - it seems a very nicely implemented design to me, and the PCB layout looks to have been very carefully done. Since you've hinted at a few interesting things I'm going to have to you ask you more about them now, so here goes:

The gnd of the power input terminals on the board is used as the reference for all signal grounds, which are seperate planes around the opamps for each channel that connect back to this point, and the opamp decoupling cap gnds are seperately routed back to this point also.

The seperate PSU is indeed implemented using the 317/337 pair with all the usual bypass caps. There's no RC filtering though, just two 220uF electros at the power supply input of the preamp board. Why does it need RC filtering, why would the values be different for different opamps and/or how do you suggest appropriate values be chosen?

There is a 100nF polyester cap from each opamp power pin to gnd and they are right next to the opamps. Other than that, I'm afraid I have no idea what you're getting at. As far as I can remember, every opamp datasheet I've ever seen shows this value in the applications section, and while other values might work just as well, isn't the point that the value isn't critical? Aren't they just reducing the impedance of the supplies at AC and providing a path for unwanted noise to return to gnd before if gets into the opamp? As long as the thing's stable what else matters?

Well the first opamp has 4k7 and 2k feedback resistors. It's driving a 10k pot, so the total load on its output is about 4k. There's also a 220p cap across the 4k7, which should be insignificant loading-wise. The second opamp is a unity gain buffer. It was driving the ~22k input impedance of my power amp via ~2m of sheilded audio cable. None of this seems like it should present a difficult load to either opamp type. Do you disagree and if so, how did you come to that conclusion, ie how do you determine what the minimum load should be for a given opamp?

I couldn't agree more, that was my point. Perhaps I am particularly incompetant at constructing these things, but I don't really believe that. The trouble is I'm not sure that any of this helps to explain what I was hearing and why the measurements didn't correlate. Anyway, as I'm sure we all know it's a massive and complex subject and we've probably digressed too far in this thread already
Phew! Now you're asking the tricky questions. But I'll try to take them one by one.

Ground planes: These are modern luxuries, and they are fantastic, but I think they shouldn't actually be used as a universal path to ground. A danger is to regard them as the equivalent of the ground symbol on a schematic and to use them similarly. The current will take its own path to ground, possibly following the +ve path on the way in. And if it doesn't, then you have little idea of the route its taking. I would say that you should manage the route to ground always. I get the impression you are doing this anyway.

Bypassing: I suspect there are only about five people on the planet who understand this properly and I am not really one of them. How you do it actually depends on the op amp and where it takes its reference - to one of the rails or something in between. My suspicion is that it is this that accounts for the larger chunk of differences we hear between op amps. The fact is that if both voltages move then so will your signal, inevitably. For me, using a single supply, I can bridge the op amps with impunity, mostly because I have nowhere else to take them to! In dual supplies, as a rule of thumb, I believe it is important to resolve the currents close to the chip and then take the remainder to ground. There is little point in doing otherwise at HF because you simply introduce inductance so they can't do the job you want them to.

Not using 100nF: It's a perfectly good value, but probably not in a film cap. I dislike seeing them scattered around anyhow. Mostly I don't think they get the job done. With 317s, which I will come to, you have an inductive output - as would anything, somewhere - and I don't think you should be introducing a high Q drop in impedance. Much better to have one or more uF with some equivalent resistance (or add it) to make sure it doesn't resonate with the output of the power supply. I feel this makes for a more comprehensive job, rather than just tackling the possibility of oscillation alone.

The ubiquitous LM317: This deserves an article by itself. Jan? It has a reputation that goes far beyond what it is actually capable of. Most of this stems from the rather flukey way that Naim Audio use it. What happens in their implementation (with 10uF caps) is that you get some cancellation around 120Hz. What happens is that you have the Early voltage getting through the pass transistor being momentarily out of phase with the signal at its base, thus giving more rejection than it is otherwise capable of at mains frequencies. You can see this quite easily if you model it as a transistor with quite a low Early voltage and feed it with a zener coming from the same mucky supply. It seems to be capable of only about 35 or 40dB of actual rejection (which is fine for a cheap universal part) but it shouldn't be mistaken for the 60 or 70dB of rejection that you see on the datasheet. Most of that is simply filtering and it is why they say that values over 10uF (for the Adj capacitor) do not improve rejection. They will bring the rejection lower down in frequency, but won't get it below that 70dB threshold. And that is for the simple reason of the VAF of the pass transistor. Feed it with an absolutely perfect DC signal and it will still always let through whatever is at the collector, about 70dB down.

So the 10uF caps are doing their jobs, but are perhaps not the right way to use a 317. Another problem with a 317 is that, according to the models, it really needs 5 or 6V drop to really have it do its best regulation. So while I would say that you could use two of them at a wider bandwidth, this is a penalty you will have to pay.

Which CR, if any?: The R component of this is the important one. There is no smoothing, filtering, without a turnover RC frequency, and that requires an output R. To flatten the output impedance of a 317 requires an astonishing 1600uF, leaving you with a resonance (if you have managed it properly) smack in the middle of the audio band. And much over 10uF is getting you inside the audio band anyway. This is because it does do the one thing well, of keeping the resistive component of it output Z down. But, assuming we have a decent supply, what should the RC be? I would say that it should be directly correlated to the PSSR curve of the op amps being used. Some of these are flat to 20kHz, like the 2604, but some start right at the beginning. It all depends on what they did with their Miller compensation. I'm fairly sure that this is another component of op amp sound and, if you were Naim Audio, where I suspect their power supplies are a significant component of their house sound, the 604 would be an ideal amplifier to keep that intact. But it doesn't make it a good amp (personally I think it is dull, dull, dull.) So, if your chosen op amp starts to lose PSR at 100Hz, then that is where I would put the CR, within the limitations of current you need. In your situation you could have a go at a first approximation test of this. You have 220uF hanging off the end of an unknown output impedance (though we know it is a 317 and that value will bring any resonance into the audio band) so put 2 or 5 ohms in series with it and see what happens. You could be surprised. And even if you are not you will at least have a component value that is now chosen for a reason!

Op Amp Loading: I can't see any problem here, whatsoever. 22k into the power amp? Why do we go for these high resistances? Is it just to keep coupling capacitor values down or is it because we are too timorous? There isn't an op amp out there worth the name that isn't able to drive 2k but we are scared to lose these high input impedances. I saw that B&O felt the need to put a warning on their IcePower documentation that the amps have an input impedance of 8k. So? It should be a walk in the park.

I think that covers everything. I hope it helps.

CT
  Reply With Quote
Old 16th May 2012, 05:23 PM   #255
Bob Cordell is offline Bob Cordell  United States
diyAudio Member
 
Join Date: Sep 2006
Quote:
Originally Posted by ChristianThomas View Post
My principle objection to the shunt capacitor is almost aesthetic. Having known Kendall for what must be knocking on 30 years, and him having been my favourite tutor throughout, I'm inevitably imbued with a love for filters. And, as you struggle to put your own mark on designs, or alternatively put what you know to work, I instinctively dislike something that fails to meet what should be happening. Whether that has an effect on SQ I actually have no idea, because I have never tried the alternative.

But, if you take the gain down from 20 odd, to something smaller, which I think is the right way to go if you have decent op amps upstream with GBWPs in the 10s of MHz, then the filter shape becomes constrained. Yes, the -3dB point will be almost identical, but it is not much further down that the graph starts getting very different from the theoretical. I'm thinking x10 or x12 as gains - and that sets your platform. To me this is unattractive. It really means that just as you hit your -3dB point the amp is immediately heading off towards it own 0dB.

I don't know if I'm good enough to give an answer to whether the degradation is "about the same" for just the non-perfect aspects of the capacitors. Maybe Jan has an idea on this, since he looks into these sorts of things. The turnover frequencies have to point to the same thing - and if anything with a deficit my way. But I really don't know. Perhaps with a few pointers and suggestions from other people I might get something I could get a grip on, but for the moment its faith.

But if you might think the fact is that my amplifier is single supply and so must have a great big C which inevitably changes its response with load, so making all of this immaterial, it actually doesn't. It does have a great big C, but its response is defined by the bandwidth of the amp alone and it is entirely load invariant. I'm suspicious that there might be more going on than I know about but I have been fairly convinced so far that this load invariance at low frequencies, and my care of the LF transfer function, are significant reasons in it sounding so good. It certainly doesn't sound good because I have loaded it up with expensive components.

ATB

Christian
Hi Christian,

If the feedback shunt capacitor causes, let's say, a 3-dB LF roll-off at 3Hz, then in an amplifier with a closed-loop gain of 20 the zero in the transfer function caused by the unity-gain shelf will be at about 0.15Hz. That is why I don't think that unity-gain shelf is the cause of any SQ degradation.

As you know, I like DC servos anyway, as long as they are properly implemented as discussed in my book.

In your single-ended amplifier with a large output coupling capacitor, I'm guessing that you have enclosed that capacitor in the global feedback loop to keep the output impedance from rising at low frequencies. Is that correct?

Cheers,
Bob
  Reply With Quote
Old 16th May 2012, 06:00 PM   #256
Bob Cordell is offline Bob Cordell  United States
diyAudio Member
 
Join Date: Sep 2006
Quote:
Originally Posted by owdeo View Post
Thanks Bob. My ears would tend to agree with you on that. I am an engineer but audio is only a hobby so it puzzles me that many times I have been dissapointed by the sound of gear I've build even when it measures impeccably in terms of the usual measurements. You'd think that with subjective assessment, with psychology being such a big factor, everything I build, especially when it measures well, should sound brilliant to me! Very frustrating that...

A couple of years ago I built a line preamp design published in the local mag Silicon Chip. It uses an OPA2134 in a dual non-inverting amp configuration with the first section set to gain of 10dB, a 10k log pot in the middle, and the other opamp section unity-gain buffering the pot output. Bipolar coupling caps are used throughout incidently... The published THD+N curve was below 0.001% from 20 to 20k at 1V.

Trouble was, I didn't like the sound, smooth but very flat and boring with no sense of acoustic, in short I didn't want to listen to music with it. So I plugged in an NE5532, as it looked as though it should work just as well if not better in the circuit. Very different result, now there was more sense of acoustic and the brass section actually sounded like brass instruments. But, massed loud violins sounded a bit screachy. The Blues Brothers cranked up high sounded harsh too, but much more enjoyable. Clearly I was hearing distortion of some sort that wasn't noticeable with the OPA2134.

At the time I had access to a Neutrik A1 analyser and I measured THD+N vs freq at 0.5 and 2V output with each opamp. It was below the resolution limit in all cases (-100dB) from 20Hz to 20kHz.

The distortion causing screachy violins must have been much less than 100dB below the signal, so it seems that the THD+N vs freq curve was not picking it up. I realise this "distortion" was still a subjective assessment, but I'm quite sure I would still pick it in a double-blind test, and besides, at the time my wife complained that her favourite CD sounded harsh without my having mentioned anything about all this to her (as if she'd be interested ).

In the case of the Precision Preamp '96 the design is such that large coupling electros are essential anyway. Guess I'll build it as designed with polarised caps and can always experiment with bipolars should there be any incentive in terms of the way it sounds... Can't wait to have some good tone controls...
Hi Owedo,

You have just done a good job summarizing the mystery, challenge and fun of audio. I don't exactly know why the 5532 sounds different than the OPA213, but at least I cannot explain it by measurement, either distortion or frequency response.

The 5532 measures superbly, especially for a 30+ year-old part that is not made in a modern complementary IC process. Indeed, I use it, or the 5534, in a lot of instrumentation, even today. I used it in my original THD anayzer 30 years ago and it worked great. Last year I did some updating of the analyzer and tried other op amps in numerous locations, but could not do significantly better than the original 5534's. Yet it does not have a sterling reputation these days for sound quality, whether deserved or not.

I use mostly LM4562's these days if I'm using bipolar. They do, however, have significantly larger input current noise than the 5532/4.

Cheers,
Bob
  Reply With Quote
Old 16th May 2012, 06:42 PM   #257
ChristianThomas is offline ChristianThomas  United Kingdom
diyAudio Member
 
Join Date: Feb 2006
Location: On a hill, in a wooden shack, next to the woods, in Somerset.
Quote:
Originally Posted by Bob Cordell View Post
Hi Christian,

If the feedback shunt capacitor causes, let's say, a 3-dB LF roll-off at 3Hz, then in an amplifier with a closed-loop gain of 20 the zero in the transfer function caused by the unity-gain shelf will be at about 0.15Hz. That is why I don't think that unity-gain shelf is the cause of any SQ degradation.

As you know, I like DC servos anyway, as long as they are properly implemented as discussed in my book.

In your single-ended amplifier with a large output coupling capacitor, I'm guessing that you have enclosed that capacitor in the global feedback loop to keep the output impedance from rising at low frequencies. Is that correct?

Cheers,
Bob
That doesn't look so bad with figures attached, does it? I'll probably still hang onto my view of it, though now without any real justification.

On the amp, that is the end result - and effectively the root method. I'll email you a sketch. 1000 words and all that.
  Reply With Quote
Old 16th May 2012, 07:38 PM   #258
DouglasSelf is offline DouglasSelf  United Kingdom
diyAudio Member
 
DouglasSelf's Avatar
 
Join Date: Mar 2007
Location: London
Quote:
Originally Posted by Carl_Huff View Post
Doug,

If there is a way for you to publish your 'new and improved' tone control I'd like to see it. I for one (and I am sure many others here) would like to build it!
Well. here's one way. Perhaps Jan would like to put it in the next issue of Linear Audio? : )
__________________
Douglas Self
  Reply With Quote
Old 16th May 2012, 07:45 PM   #259
DouglasSelf is offline DouglasSelf  United Kingdom
diyAudio Member
 
DouglasSelf's Avatar
 
Join Date: Mar 2007
Location: London
Quote:
Originally Posted by Bob Cordell View Post
Doug,

If I'm not mistaken, the 5534 used in your design for the RIAA equalizer is in a non-inverting unity-gain configuration at high frequencies, yet you appear to be using a 4.7pF compensation capacitor when the spec sheets for the 5534 recommend a 22pF compensation capacitor for unity-gain non-inverting situations. Did I look at the schematic wrong or misunderstand something here? Can you comment on that?

Cheers,
Bob
Well Bob. this is a classic divergence between theory and reality. The 22pF is presumably a belt-and-braces value that will give stability under all possible feedback conditions, supply voltages, tempertaures, etc etc.

However, a 5534 in my circuit with 4p7 is definitely stable- I've been using that value for 16 years and haven't seen or heard of any problems yet. In fact I have found that a 5534 is usually unity-gain stable with no external compensation at all, though I wouldn't want to depend on that in high-volume production.
__________________
Douglas Self
  Reply With Quote
Old 16th May 2012, 07:56 PM   #260
Carl_Huff is offline Carl_Huff  United States
diyAudio Member
 
Join Date: Dec 2004
Location: Bakersfield, CA
Quote:
Originally Posted by DouglasSelf View Post
Well. here's one way. Perhaps Jan would like to put it in the next issue of Linear Audio? : )
Great!

If Jan agrees I will be sure to signup for a subscription.
__________________
Best Regards,
Carl Huff

Last edited by Carl_Huff; 16th May 2012 at 07:58 PM.
  Reply With Quote

Reply


New Doug Self pre-amp design...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

Forum Jump

Similar Threads
Thread Thread Starter Forum Replies Last Post
Doug Selfs NE5532 Power Amp. Thoughts anyone ! Mooly Chip Amps 218 17th July 2016 10:00 PM
blameless amp and doug self streetera Solid State 17 16th July 2007 10:06 PM
4th Ed. of Audio Power Amplifier Design Handbook (Doug Self) nickds1 Solid State 4 9th November 2006 12:02 PM
Doug Self Advanced Pre Amplifier PCBs macka Swap Meet 0 9th October 2006 07:41 AM
Help me design pre-amp PS hardcore Tubes / Valves 7 10th November 2004 06:32 AM


New To Site? Need Help?

All times are GMT. The time now is 06:23 AM.


Search Engine Optimisation provided by DragonByte SEO (Pro) - vBulletin Mods & Addons Copyright © 2019 DragonByte Technologies Ltd.
Resources saved on this page: MySQL 15.00%
vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2019 DragonByte Technologies Ltd.
Copyright ©1999-2019 diyAudio
Wiki