I exported the data from LTSpice and plotted it in Python, scaling the magnitude logarithmically outside the unit-circle to allow one plot to show the whole thing. I guess I independently developed this approach to Nyquist plots because its really useful!
See posts #31 and #33 in https://www.diyaudio.com/community/threads/nic-opamp-vas-topology.348646/page-2
See posts #31 and #33 in https://www.diyaudio.com/community/threads/nic-opamp-vas-topology.348646/page-2
Interesting solution, i like it, although Bode plot gives instant result, especially when you play with .STEP to find best match. Or you automate export and Pyton processing with one simple action?
My posts #14 and #16 refer to the current mirror formed by Q4 and Q17 in post #13.You mean Q13 not Q17.
Ed
Q4 and Q17 from post #13 are NOT in current mirror configuration. In fact both collectors conducts only DC current, not AC. So they act as constant current sources, not mirrors. This configuration (I hope this time it is novelty design) allows for high gain without high voltage drops (contrary to resistors). To improve its efficiency (in case of not fully symetrical AC voltages on both collectors one may put several microfarads capacitor between base and emitter.
The circuit relies on the currents in Q4 and Q17 being the same. My point was that these transistors need emitter resistors.
Ed
Ed
I see that during copy-paste of transistors there is a mismatch in part numbering on schematics, so it may be confusing.
Yes, emitter resistors will cancel asymmetry of transistor parameters. As I wrote earlier it is not a final version and some things still need to be refined, but anyway, thanks for pointing this.
Interesting thing is, that slightly modifying this circuit we can get further gain increase (now transistors acts like current amplifiers), controlled by resistance R26 (no gain when R26=0). Here with emitter resistors.
With R26=510 Ohm we've got extra 14dB gain and amplifier reaches 138dB open loop gain (170dB without feedback network), which also results in THD as low as 0.000002%
Yes, emitter resistors will cancel asymmetry of transistor parameters. As I wrote earlier it is not a final version and some things still need to be refined, but anyway, thanks for pointing this.
Interesting thing is, that slightly modifying this circuit we can get further gain increase (now transistors acts like current amplifiers), controlled by resistance R26 (no gain when R26=0). Here with emitter resistors.
With R26=510 Ohm we've got extra 14dB gain and amplifier reaches 138dB open loop gain (170dB without feedback network), which also results in THD as low as 0.000002%
Last edited:
Noob question: What are the advantages to a Nyquist diagram over Bode in this context? The phase margin is the angle at which the curve exits the unit circle, right? Isn't it just as easy to glean that from the Bode plot?
None, yes, and yes. It's only a matter of habits.Noob question: What are the advantages to a Nyquist diagram over Bode in this context? The phase margin is the angle at which the curve exits the unit circle, right? Isn't it just as easy to glean that from the Bode plot?
Hi Zbig, im following with interest. The amps ive build like to have about 65 degrees and 15db of margins for good stability in reality.
For your active ips load: why not connect the bases of q18 and q19 and connect them to the common emitters of your vas. Now you have regulated vas current and active ips load.
For your active ips load: why not connect the bases of q18 and q19 and connect them to the common emitters of your vas. Now you have regulated vas current and active ips load.
Quite interesting, but this will not allow for extra gain (+14dB), because q18 and 19 will act only as current sources, not amplifiers.
I always find the LTspice Nyquist plots painful to use for the reasons you state. Interesting design ideas BTW. Thanks for sharing.Interesting idea about modified Nyquist plot.
Especially our amplifiers have huge gain, and zooming Nyquist plot to reach unity circle level takes some time, and also requires manual drawing of this circle.
It is how Nyquist diagram looks for my amplifier. LTSpice should have option to do it automatically.
View attachment 1429707
I'd like to point also novelty feedback network:
In typical FB network (on the right) capacitor C3 is connected in series with R11 to prevent R20/R11 attenuation at DC, reducing output DC offset. But C3/R11 forms also filter for AC, and requires relatively large C3 to keep low bandwith of the amplifier.
In my design (on the left) DC negative fedback is provided through R19, and C3 acts as high pass filter for AC together with R19, so its capacity may be reduced R19/R11 times with this same cut frequency, or easily go down to sub-hertz 🙂.
This configuration allows also for smaller values of R20/R11 reducing impact of serial resistance to base Q9 at high frequencies.
I've added also C10, to add zero to R20/C4 pole (forward correction). This improves gain characteristics below unity gain.
In typical FB network (on the right) capacitor C3 is connected in series with R11 to prevent R20/R11 attenuation at DC, reducing output DC offset. But C3/R11 forms also filter for AC, and requires relatively large C3 to keep low bandwith of the amplifier.
In my design (on the left) DC negative fedback is provided through R19, and C3 acts as high pass filter for AC together with R19, so its capacity may be reduced R19/R11 times with this same cut frequency, or easily go down to sub-hertz 🙂.
This configuration allows also for smaller values of R20/R11 reducing impact of serial resistance to base Q9 at high frequencies.
I've added also C10, to add zero to R20/C4 pole (forward correction). This improves gain characteristics below unity gain.
Last edited:
But this capacitor can be exactly this same as input one (also floating).
Btw. could you recall any reference for such FB network? I'm curious abut it.
Btw. could you recall any reference for such FB network? I'm curious abut it.
Similar approach was used in Hafler DH200/220. The intention is to use low value feedback resistors for AC feedback (low noise), while the DC resistance stays high to match the input impedance of the amp in order to minimize the output DC offset.
Take a look at this PDFBtw. could you recall any reference for such FB network? I'm curious abut it.
The supergainclone from Cordell's book also has a nice trick to avoid large value feedback caps but it's for an inverting config.
Actually you triggered my memory about a most relevant article I almost forgot about, in my own Linear Audio series! I must be getting old.
Here it is, complementary 😎
His figure 5b is what Jonathan showed in post # 26. But this time with a narrative ;-)
Jan
There is no such network in Hafler DH200/220 (or I've googled wrong scheamatics?)Similar approach was used in Hafler DH200/220. The intention is to use low value feedback resistors for AC feedback (low noise), while the DC resistance stays high to match the input impedance of the amp in order to minimize the output DC offset.
Interesting, this was employed for reducing capacitor sizes in miniature portable audio circuitry. For bigger amplifiers nobody cares about capacitor sizes 🙂.Take a look at this PDF
Version 3 of Leach's amp had the floating capacitor. https://leachlegacy.ece.gatech.edu/papers/lowtim/feb76feb77articles.pdf
Ed
Ed
Thank you. Interesting: "LowTIM3" amp has such feedback network, but "Low TIM4" not, it seems to be "classic" in 4, but with forward compensation with pole, zero and another pole.
Try here:There is no such network in Hafler DH200/220
https://audiocircuit.dk/downloads/hafler/Hafler-DH200-pwr-sm.pdf
Another example:
https://www.docdroid.net/seYd1tG/borbely-a-60w-mosfet-power-amplifier-pdf#page=7
Interesting, this was employed for reducing capacitor sizes in miniature portable audio circuitry. For bigger amplifiers nobody cares about capacitor sizes
Indeed, two different "problems" but the same solution.
- Home
- Amplifiers
- Solid State
- Power amp design by Zbig with new compensation method