
Home  Forums  Rules  Articles  diyAudio Store  Blogs  Gallery  Wiki  Register  Donations  FAQ  Calendar  Search  Today's Posts  Mark Forums Read  Search 
Headphone Systems Everything to do with Headphones 

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 

Thread Tools  Search this Thread 
16th November 2018, 10:35 PM  #71 
diyAudio Member
Join Date: Feb 2004
Location: Silicon Valley

My idea of fun is not groping around in the dark trying to juggle 3 sets of poles plus a couple of zeroes. I will be making the investment for a GP analyzer in the near future, most likely an HP4194A, which also doubles as an impedance analyzer. Both impedance and G_P functions go out to 40 MHz.
Right now, I'm using the Venable GP analyzer we have at work, which has 5 MHz capability and costs me nothing. The first scan I made with original compensation had practically no phase margin and crossed over way too high, directly at odds with the relatively benignlooking time domain response. This may be possibly be due to the LP introduced by the input mosfet pair gate and Miller capacitance and a limited impedance input source. Anyway, here's what I have now after a fair amount of twiddling. The compensation values I'm using now are in the same ballpark as the values I originally used in my simulation to stabilize the design. I'll do the time domain response when I get a chance. I'd like to twiddle the compensation a bit more to get a more benign crossover slope, but 70 degrees of phase margin is not too shabby. 
16th November 2018, 11:58 PM  #72 
diyAudio Member

Boy the HP4194A is a wonderful piece of gear! I hope you snag one that needs very little TLC and refurbishment before you can begin using it.
Looking at the blue "phase" line in the plot of #71, it seems to cross 45 degrees at 100 kHz. And that's likely to be a good approximation of the 3dB rolloff point. So we expect the 10%90% risetime to be somewhere in the vicinity of 3.5 microseconds. Gain of 10.5X ; bandwidth of 0.1MHz ; risetime of 3.5usec (predicted) ; phase margin of 70 degrees ; %overshoot of only 1.3% (predicted). 
21st November 2018, 04:01 AM  #73 
diyAudio Member
Join Date: Feb 2004
Location: Silicon Valley

I did some work over the weekend, subbing 3X smaller fets in the output stage to move out the pole introduced by the gate + Miller capacitance. I went from IRF630/9630 t0 610/9610, reducing the capacitance by about a factor of 3.
I also reduced the rails to 16V from 19V, and upped the bias current to 200mA. These changes allowed me to chip away at the compensation capacitors, increasing bandwidth while retaining (actually increasing) phase margin. The GP scan also looks more "classic" as a result. I want to try one more dodge before buttoning this thing up, tinkering with the stoppers on the input differential stage to perhaps push out the bandwidth a bit more. The attached GP plot shows a family of curves resulting from successive tinkering with the compensation values. The one of interest is plot #4 (grey/black). That compensation gets me 2.76 MHz gain crossover with 75 degrees phase margin. 
21st November 2018, 09:45 PM  #74 
diyAudio Member
Join Date: Feb 2004
Location: Silicon Valley

Here's the last pass of compensation tweaks. The amp is going home for some cleanup and installation in an enclosure.. The cases I have in mind for the finished unit will arrive early next week. I was thinking of routing out something nice in oak and sheet metal, but time is of the essence if I want to get this thing ready for Christmas.
I used ferrite beads on the gates of the input mosfet diff pair and got a little more bandwidth. Of the two gain phase curve sets shown, the one of interest is the orange/green pair, with a crossover frequency of 2.98 MHz and 62 degrees of phase margin. A second marker shows the frequency for 45 degrees of phase shift (313 kHz). The rise and fall times are a little under 1 usec. One more thing I'll be trying over the Thanksgiving holiday is mating the amp with its negative supply DCDC converter. 
21st November 2018, 10:06 PM  #75 
diyAudio Member

The plotted LF gain of +20dB suggests to me that this instrument has plotted the gain and phase of the control system with the loop closed. If were an open loop plot then the LF gain would be at least 30dB higher.
So, if this indeed IS a closed loop plot, you can't really read off the open loop gain margin or the open loop phase margin directly. You'll need to mathematically undo the loop closure. And I doubt the instrument does this for you automatically, because if it did then the plotted LF gain would be more like +50dB. 
22nd November 2018, 05:02 AM  #76 
diyAudio Member
Join Date: Feb 2004
Location: Silicon Valley

Of course I'm looking at the gain and phase with the loop closed. Here's what things look like if I break the loop between the output and gain setting network.

22nd November 2018, 03:13 PM  #77 
diyAudio Member

That open loop plot certainly shows enormous amounts of open loop phase margin, more than 90 degrees, at the crossover frequency of 200 kHz.
If I had an amplifier with that open loop plot, I would experiment with reducing the dominant pole compensation capacitor by a factor of about 5x. This would move the open loop crossover frequency up to about 1 MHz, where the open loop phase margin is still a quite respectable 75 degrees. More importantly it increases the amount of excess gain (a/k/a "the amount of feedback") by a factor of 5x, which gives a commensurate reduction in distortion. It also speeds up the square wave risetime response by the same factor of 5x. There may be other notquitedominant poles and zeroes which begin to impact the open loop phase margin as you decrease the dominant pole compensation cap, so you may not be able to get the full 5x increase. But more bigger is more better, even if it's not 5x better. Your amp has a gain of 10.5X and a 3dB bandwidth of 0.2 MHz, for a GBW product of 2.1 MHz. For purposes of comparison, the power amps in Bob Cordell's book have a GBW product of 16 MHz and the ones in Douglas Self's book have a GBW product of 20 MHz. So there is precedent for getting >2.1 MHz with thru hole parts and 2 layer PCBs and low priced nonmilitary components. Even if the synthesizedSIT topology used here limits ultimate GBW, we'd hope to get at least half of what the booksforbeginners get, namely 8 MHz or so. In an optimistic and hopeful mood we could multiply 2.1 MHz by "almost 5x" and get 9 or 10 MHz which would be in that ballpark. 
27th November 2018, 03:30 PM  #78 
diyAudio Member

Whoops, I double checked the GBW numbers in post #77 and discovered that my memory had failed me. In fact,
Very sorry for the errors!* The power amps in Bob Cordell's book have a GBW of 10 MHz 
Thread Tools  Search this Thread 


Similar Threads  
Thread  Thread Starter  Forum  Replies  Last Post 
All Discrete ClassA Headphone Amp  mnturner  Headphone Systems  33  6th June 2011 02:39 PM 
Discrete simple headphone amp  kubeek  Headphone Systems  15  7th September 2007 05:13 PM 
Discrete Headphone Amp Q  weinstro  Headphone Systems  1  24th September 2003 10:48 PM 
New To Site?  Need Help? 