Whether this is true depends on exactly what you mean by "wonky 'inner' loop".
The distinction between inner Loop Gain and inner Return Ratio is critical.
Was I unclear in my earlier post on this, or do you do not believe it, or is it a nomenclature problem?
Best wishes
David
Thank you very much Ihan !
I am reading it
Do you know if I can do niquist plot with ltspice ?I take this to mean the Tian probe on the output. I believe you've said this shows Bode's Return Ration around the OPS.
My apologies if my short term memory is as bad as my long term
Thank you very much Ihan !
I am reading it
When you do Bode plot(it is default) you left click on dB left y scale and then choose from Representation: Bode, Nyquist ot Cartesian.
A small disclaimer, for the record.
I have learned that at least one of the standard texts defines "open loop gain" as essentially the same as "loop gain".
Not a sensible definition in my opinion but it would mean the LTSpice comments make sense.
One more inconsistency to watch out for.
Best wishes
David
I choose low values to minimize noise, a concern of mine because I have ultra efficient compression drivers.
So I have looked at this and was pleased to find other benefits.
Nice to see someone else has come to the same conclusions.
A handful of extra metal film feedback resistors is quite inexpensive, a very worthwhile trade-off.
A trim pot probably makes it practical to dispense with the servo and/or the substantial feedback capacitor.
Haven't decided on that yet.
Best wishes
David
Try it with a servo. Will it make a difference to the sound ? I would be curious to know.
Thx-RNMarsh
Hi all,
Attached are the schematic and loopgain plots of a proposed alternative way of stabilising 'pure Cherry'. Notice that much less VAS degeneration can be used, and the additional compensation network at the input to the VAS.
Of course, one could argue that this is no longer 'pure Cherry' due to the additional network.
Best wishes,
Ian
Attached are the schematic and loopgain plots of a proposed alternative way of stabilising 'pure Cherry'. Notice that much less VAS degeneration can be used, and the additional compensation network at the input to the VAS.
Of course, one could argue that this is no longer 'pure Cherry' due to the additional network.
Best wishes,
Ian
Last edited:
Hello all,
My latest (and perhaps most illuminating) thoughts on the effect of the VAS degeneration resistor with Cherry compensation:
I was under the impression that the degeneration resistor's main role was to decrease the VAS transconductance.
However, we should also remember that the VAS degeneration resistor also influences the VAS input impedance. The VAS input impedance with standard miller compensation is not of great relevance.
However, with 'pure cherry' and no VAS degeneration, the input impedance is too low, and 'loads' the IPS current mirror, diverting current away from the Cherry compensation capacitor. Adding a VAS degeneration resistor increases the VAS input impedance allowing current to flow through the compensation capacitor.
When I get more time (over the next couple of days), I hope to produce plots of the current in the compensation cap and the VAS input impedance with respect to the VAS degeneration.
I'm out for a couple of beers this evening. Have a good weekend, Ian
My latest (and perhaps most illuminating) thoughts on the effect of the VAS degeneration resistor with Cherry compensation:
I was under the impression that the degeneration resistor's main role was to decrease the VAS transconductance.
However, we should also remember that the VAS degeneration resistor also influences the VAS input impedance. The VAS input impedance with standard miller compensation is not of great relevance.
However, with 'pure cherry' and no VAS degeneration, the input impedance is too low, and 'loads' the IPS current mirror, diverting current away from the Cherry compensation capacitor. Adding a VAS degeneration resistor increases the VAS input impedance allowing current to flow through the compensation capacitor.
When I get more time (over the next couple of days), I hope to produce plots of the current in the compensation cap and the VAS input impedance with respect to the VAS degeneration.
I'm out for a couple of beers this evening. Have a good weekend, Ian
Last edited:
I think you are perfectly entitled to call this a Holy 'pure Cherry' amp
Your little network on the VAS input is just a little local tweaking like Bob Cordell's on his IPS output for his supa dupa 1984 amp.
Having it there is much better than sullying the Holy HiZ nature of the VAS output.
Have you done some 50W 8R THD20k sims?
But be careful that your network doesn't divert current from the Cherry cap
Last edited:
Ian, you gonna tell us what's it with using 3503/1381 for all small BJTS?
Who's models are you using for the MOSFETs?
I think you are perfectly entitled to call this a Holy 'pure Cherry' amp
Your little network on the VAS input is just a little local tweaking like Bob Cordell's on his IPS output for his supa dupa 1984 amp.
Having it there is much better than sullying the Holy HiZ nature of the VAS output.
Have you done some 50W 8R THD20k sims?
But be careful that your network doesn't divert current from the Cherry cap
Hi Richard,
Yes, I've done some distortion simulations. Simulated distortion is less than 10ppm at 20kHz into a 2ohm load, 40Vpk output (which is my target for this design). TMC achieves around 30ppm (dependent upon stability margins). I usually aim for at least 60 degrees and 12dB.
10ppm is less than I can measure with my test equipment, but I usually find that the performance is determined by my board layout (particularly the routing of the supply conductors). If I can get Cherry compensation stable, it will be interesting to see if it reduces the THD20 by a measurable amount compared to TMC.
Incidentally, I believe that these very low distortion figures (<10ppm) are not required for good sound quality. I view this compensation investigation as an interesting learning exercise, not as a statement regarding sound quality. In the real world, I suspect that <100ppm is more than adequate.
I use 1381 and 3503 transistors since they are low capacitance, high voltage transistors, which can dissipate around 0.6W without a heatsink. My latest design will run on rails upto +/-100V, so power dissipation becomes an issue. In practice, at places in the circuit where I do not that level of power dissipation, I use KSC1845 and KSA992 (also low capacitance).
I use my own models for the mosfets, which I created some years ago using MicroSim parts. The models are far from perfect, but they do simulate with higher distortion than BJTs , meaning that they are least semi-representative. In practice, I like my choice of output devices. They are robust, relatively cheap, powerful, and importantly, suit my output stage protection system.
Best wishes,
Ian
Last edited:
Attached are plots (and schematic) indicating how VAS input Z, VAS input base current and the compensation cap current change with degeneration resistance. I have done plots for 1mohm (no degeneration resistor), 10R and 100R (green is 1mR, red 10R, blue 100R). Yellow 1mR, Magenta 10R, Cyan 100R.
Last edited:
Could you please post the Return Ratio plots for the three conditions?
As the RR alters this will alter the input impedance also, of course.
I am curious if your plots have any alteration that is not already explained by Blackman's results.
Best wishes
David
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.