[SnoopyRocks]

There seems to be a fair amount of confusion about how to calculate multiloop gain and bandwidth. I did a quick analysis that simplifies the problem into a form that most people here should be able to work with and thus be able to intelligently deviate from the standard resistor values in a multiloop amplifier such as a pimeta or ppa.

MultiLoop.pdf

In general, the simplified expressions on the first page are more than good enough if you are using the topology the way it was intented (i.e. with a high inner loop gain). There are more exact expressions for the feedback factors on the second page which I got by rearranging tohpu's (on headwize) expression for the loop gain. The bottom line is that if R3 is more than 10x larger than the other resistors, it's a waste of effort (IMO) to resort to the more complicated expressions. After all, engineering is not about exact answers. If it were, very little would work in the real world.

<rant>As a side note, I'd just like to say that I'm am not convinced of the multiloop topology's claimed bennefits. Jung's description in the electronic design article amounts to hand waving. That is to say that he does not back up his claims there. If someone has reference to a propper treatment of the topology, I'd like to see it. I do understand that the place where it was published was not appropriate for a rigorous treatment of the topic. Yet, it seems like the bennefits of the jung multiloop architecture are taken as gospel around here without good reason. If anyone has measurements to support jung's claims, I'd like to see those as well. </rant>

Knock yourselves out.

 

[Ob3ron]

Thanks very much for that Snoopy. I've been interested in learning what makes the multiloop topology tick but I've been too busy preparing for exams. I'm sure this will be quite a help to me in a few weeks.

And hopefully this thread will shine some light on the matters you bring up.

About the jung article... I read it a couple of times quickly and i also found it to be lacking in substance, though I assumed just because some of it was over my head.

Is the jung multiloop used much elsewhere? In general audio or otherwise? I've seen no mention of any multiloop opamp topologies in my big giant huge Sedra & Smith Electronic Circuits textbook.

 

[nikongod]

what is your source for the various bandwidth calculations?

my admitedly non-exhaustive searches led me to nothing of great value with regards to selecting the innerloop gain.

 

[SnoopyRocks]

Quote:

Originally Posted by nikongod what is your source for the various bandwidth calculations? my admitedly non-exhaustive searches led me to nothing of great value with regards to selecting the innerloop gain.

The gain bandwidth product for opamps is constant. So if the closed loop gain increased 5x, then the closed loop bandwidth is reduced 5x.

The inner loop bandwidth should be f_IL=fo*(1+beta_i*Ao)/Ao~fo*beta_i.

Mental note: check result before posting next time. I put a band-aid on it until I can rescan it in the morning. It sure is hard to write with a mouse.

 

[drewd]

You might want to take a look at this thread. I'm not sure if it will answer your question, but it does provide some references to a couple of additional technical resources that may help you make a decision.

The traditional Head-Fi answer to the multiloop question is that the high inner loop gain allows one to use an uncompensated amplifier with unity gain and reduces phase errors and TIM. However, I don't know if anybody has produced any definitive measurements to back up these claims. I know that the first claim is not true. I can't speak to any others with any specific data, but I'm coming to believe, based on experience, that an audio amplifier using an open loop buffer (like the BUF634) gains no audible benefit from two loops.

The easiest way to resolve this problem might be to just run some SPICE simulations using available models from some vendors. If I have time tomorrow, I'll set that up at work and run a few simulations with various multiloop and single loop configurations.

-Drew

 

[ppl]

Quote:

Originally Posted by drewd You might want to take a look at this thread. I'm not sure if it will answer your question, but it does provide some references to a couple of additional technical resources that may help you make a decision. The traditional Head-Fi answer to the multiloop question is that the high inner loop gain allows one to use an uncompensated amplifier with unity gain and reduces phase errors and TIM. However, I don't know if anybody has produced any definitive measurements to back up these claims. I know that the first claim is not true. I can't speak to any others with any specific data, but I'm coming to believe, based on experience, that an audio amplifier using an open loop buffer (like the BUF634) gains no audible benefit from two loops. The easiest way to resolve this problem might be to just run some SPICE simulations using available models from some vendors. If I have time tomorrow, I'll set that up at work and run a few simulations with various multiloop and single loop configurations. -Drew

Oh yes folks I have critics! Hard to believe isn’t it. People that I believe are honest in there belief however incorrect it may be. When the reality of the entire system is treated and not lumped into one easy to simulate model, the subtleties of the system are revealed. Regarding weather Multi-loop works or sounds better is something I myself have consistently proved and later confirmed via measurements that performance is at least as good as Amps employing high overall Global feedback and operating at low open-loop bandwidth. THD vs frequency plots for a given op amp buffer pair operated in a correctly implemented multi-loop aspect results in a THD only slighter higher or the same as the same chipset operated normally however is devoid of the Rising THD above typically 1KHz as happens when the chipset is operated normally as in a OPA627 plus BUF634 chipset as depicted in the BUF634 data sheet. The THD at 20 KHz can be 10 times what it is at 1KHz and this is what Olta and the like were getting at now there was mention of Walt Jung not understanding Olta in the Linked thread? Well the advantages of Multi-loop feedback to aid in stability is actually the recommended method of stability enhancement by the esteemed Malcolm Halksford http://www.essex.ac.uk/ese/research/...distortion.pdf
Pg.8 Fig 7

Also consider
“The dominant advantages of this approach is that
only very modest local negative feedback need be applied
via R1and that the high-level distortion is partially
compensated by the error amplifier. Such a technique
allows good signal power coupling to Ti, yet permits
an acceptable high-level distortion characteristic. It
therefore follows that T1is exercised over a wide range
of its operating characteristic while retaining good
overall linearity.”

No Way Malcolm! Really.

Again there are people that do not fully understand these concepts and thus will not see any benefit to em. Ok I can understand that it can be difficult to see the forest through the Trees, HOWEVER to totally discredit solid design and engineering just because it don’t follow ones understanding of the universe is a regressive backward way of thinking and has resulted in plenty of discoveries not being realized because of the old boy network within some societies. See http://www.gre.ac.uk/~bj61/talessi/tlr22.html

If you want to find out more on Multi-Loop Feedback do a search about 100 solid concepts come up and about 1000 slightly relevant topics

 

[tangent]

Quote:

Originally Posted by drewd The traditional Head-Fi answer to the multiloop question is that the high inner loop gain allows one to use an uncompensated amplifier with unity gain

Anyone who says that doesn't know what they're talking about. The only thing in this regard that I've noticed is that buffering the op-amp does seem to let you get away with technically unstable configurations, sometimes. But that doesn't have anything to do with multiloop.

 

[rickcr42]

guys,you come close to being personal here.Standing up for one topology over another is fine but leave the "edge" off.

thanks,

rickster

 

[drewd]

Quote:

Originally Posted by ppl Oh yes folks I have critics!

No criticism intended - and, for my part, no discrediting of any design at all. Perhaps I should have said in my previous post that I was leaning toward single loops because of my listening experience. I simply cannot hear a difference.

But I most definitely did not aim any criticism at you nor at anyone else. For the most part, I simply regurgitated comments that have appeared in a number of posts on Head-Fi and pointed out, as others have, that at least one of the conclusions is wrong.

My apologies if you thought that I was attacking you - it certainly wasn't intended - I wasn't even thinking about you

-Drew

 

[rickcr42]

Quote:

My apologies if you thought that I was attacking you - it certainly wasn't intended - I wasn't even thinking about you

I don't see where any apology is required for posting an opinion that was in no way directly aimed at any one person or group.You don't require multiloop topology /

so what is the friking problem in that ?

To apologize smacks of intimidation and fear of being different.not cool at all.

the only thing keeping the DIY forum from becoming so boring it would die a slow death due to lack of interest is the amount of difference,not sameness and one size fits all.

'nuff sid by me on this topic but I am monitoring so keep it on the high road


another public service annopuncement from your freindly neighborhood monsterator

 

[SnoopyRocks]

Thanks for the responses and links everyone.

PPL, could you clarify how a mutliloop should be set up. As I understand it the inner loop should have its gain set so that the 3dB bandwidth is outside the audio range. Roughly 100kHz or so. The outer loop sets the overall gain. It's as simple as that right? Or is there more to a correctly implemented multiloop?

Quote:

Originally Posted by PPL THD vs frequency plots for a given op amp buffer pair operated in a correctly implemented multi-loop aspect results in a THD only slighter higher or the same as the same chipset operated normally however is devoid of the Rising THD above typically 1KHz as happens when the chipset is operated normally as in a OPA627 plus BUF634 chipset as depicted in the BUF634 data sheet. The THD at 20 KHz can be 10 times what it is at 1KHz and this is what Olta and the like were getting at now there was mention of Walt Jung not understanding Olta in the Linked thread?

So you are saying that an advantage of a multiloop topology is that the harmonic distortion level is now constant over the audio range? Yet, if I interpret what you 're saying, it doesn't just raise the overall harmonic distortion level to what it was at 20KHz in a single loop arrangement, but is actually lower? I find this a bit counter intuitive. If this is really so, it certainly makes a multiloop approach attractive. Please correct me if I am misinterpretted your post. Can you point me to a reference for this?

Does anyone have a reference to the aforementioned Olta paper?

 

[rickcr42]

do a google search.all the Otala papers AND Otala amp designs resulting from them are online for viewing/downloading though can't at the moment direct you to them.Memory is the first thing to go you know

 

[ppl]

Quote:

Originally Posted by SnoopyRocks Thanks for the responses and links everyone. PPL, could you clarify how a mutliloop should be set up. As I understand it the inner loop should have its gain set so that the 3dB bandwidth is outside the audio range. Roughly 100kHz or so. The outer loop sets the overall gain. It's as simple as that right? Or is there more to a correctly implemented multiloop? So you are saying that an advantage of a multiloop topology is that the harmonic distortion level is now constant over the audio range? Yet, if I interpret what you 're saying, it doesn't just raise the overall harmonic distortion level to what it was at 20KHz in a single loop arrangement, but is actually lower? I find this a bit counter intuitive. If this is really so, it certainly makes a multiloop approach attractive. Please correct me if I am misinterpretted your post. Can you point me to a reference for this? Does anyone have a reference to the aforementioned Olta paper?

Yes it is really that simple and the 20 KHz thd is not reduced to the 1KHz level but rather the 1KHz THD is raised to the 20 kHz level resulting in a higher overall THD spec however a THD spectrum that is constant vs frequency as the amount of feedback is now constant throughout the Audio range as opposed to changing wildly from about 100 Hz upward. Some Olta info is hear http://home.online.no/~tsandstr/OtalaStory.htm
Also see http://hyperphysics.phy-astr.gsu.edu...io/amp.html#c2 For a description of Distortion including TIM. This site is also a great place for info on all things Physics and its from Georgia state so you know its accurate you might want to bookmark this http://hyperphysics.phy-astr.gsu.edu...nd/soucon.html
http://hyperphysics.phy-astr.gsu.edu/hbase/hph.html#hph

Directory of terms http://acoustics.a.dictonarypage.co.uk/
Non Linear Audio Distortions and http://www.rane.com/par-i.html

http://www.tvhandbook.com/support/pd...hapter13_3.pdf
Distortion in Hearing Aids http://www.ncbi.nlm.nih.gov/entrez/q...t_uids=7987021

Electronics Digest UK 1985 a 60 watt Amp http://www.tcaas.btinternet.co.uk/curtis60w.pdf

 

[SnoopyRocks]

Quote:

Originally Posted by ppl Yes it is really that simple and the 20 KHz thd is not reduced to the 1KHz level but rather the 1KHz THD is raised to the 20 kHz level resulting in a higher overall THD spec however a THD spectrum that is constant vs frequency as the amount of feedback is now constant throughout the Audio range as opposed to changing wildly from about 100 Hz upward.

I thought so...no free lunch here. This is much easier to accept.

You have quite a collection of bookmarks! I wonder who has more - you or rickcr42? Thanks for the links.

 

[ppl]

Rick by alot