M³ Project Announcement - Page 15

At last...a logic answer to a q....:)

Thnx Morsel and amb!
Nice to see that u know what ur talking about....
And, I ´ve to discovered ur findings about multiloop when simulating these circuits, but it's always nice to have it confirmed from the source..


Keep up the good work!
/qpedejo

AMB, thanks for the reinforcement. Please forgive my impetuous nature, but I'm still living with the original 1.0 SDS, so I am anxious to build the M³.

Can you give a short list of the major unresolved issues, since it looks like there is a 2 week break in updates?

Morsel is leaving town for two weeks but I'm continuing to work on the amp.

Currently the focus is on the power supply. The original schematic posted in this thread shows the use of a separate capacitance multiplier and TLE2426 rail splitter for each channel. We are testing to determine whether that is overkill. Perhaps a single TLE2426 is sufficient for all three opamps. We are also testing to determine whether capacitance multiplier or JFET isolation (a la PPA) is better, or maybe even none at all.

Even though this is a fairly "deluxe" amp, our philosophy for it is that every part has to earn its keep. This is to meet our goal of excellent sound, good price/performance and KISS. We do not want to throw extraneous circuitry into the amp when they drive up the cost, complicate the PCB layout while offering dubious benefits. Doing these tests require rebuilding portions of the prototype and doing various measurements, which takes time. I do this during my spare hours, so please bear with us.

The other open issues are PCB layout/component placement (not only for optimum trace routing, but also to allow for standard parts as well as audiophile boutique-grade stuff -- e.g., fancy capacitors, where feasible), board size and orientation, case selection, and other sundry related details. This will be worked out as we make further progress.

Thanks both for the update... I've adjusted my Thorazine intake accordingly.

was that over the complete audio range or just 1K as typicaly THD will rise with frequency and multiloop trades a over all higher THD for a constant THD over a wide frequency range.

Its is true that with several volts to eliminate is asking alot fom a low overall feedback topology. I had that exact same problem when operating Loudspeaker amps output open loop (Kind of a trend at the time and a nice one) a typical trippele Darlington controlled by one Vbe Multiplier i have seen preduce 200 or so mv of DC and this is bad evean for a speaker amp it would be real bad for a headphone.

In closing as AMB pointed out there is no one perfect topology as every eliment must be considered as a system.

For THD, yep, it's just at 1KHz because that's what RMAA does, and doing THD measurements at high fundamental frequencies aren't too meaningful with computer sound cards, because of the sampling rate limitation means that we aren't measuring many of its harmonics. What I really need is a good 'ol wide-band distortion meter and a very low distortion function generator, but my budget currently doesn't allow for that.

For what its worth, RMAA 5.4 has a IMD vs. frequency plot, and the M³ prototype measured very well, producing a flat line across the whole audio spectrum (no rise in IMD at high frequencies). I don't really know how RMAA internally does this test, though, so that's the reason for the FWIW.

The fact that this is a class A amp means that it should have good linearity at the output stage, which is good for keeping distortions low.

I did not imply that the M3 measures bad at all just clerifiying what the advantages of Multiloop or large open loop bandwdth. and AD-825 dose this all on its own as as sutch is a perfect OPA for conventionial Globial feedback applications.

Could you elaborate on this? Thanks!

I am sorry, I just caught this post. As I said earlier, I did prefer the m3, but I really would like to hold off on details until I can hear a more finalized version. I tend to talk in superlatives and I don't want to create a FOTM craze for something that is still sitting on a breadboard.

Also, I think amb answered this a little, so hopefully you got what you wanted to know.

Thanks Raif... I'm trying to extrapolate where my SDS v1.1 is from the modded v1.2, and then up to the M³. I read the verbage, but was after the superlatives to describe the sound....

I love my SDS, and realize the improvements that can be made, and greatly appreciate that Ti, Morsel, and others have found this is a worthy project to spend their time on to push forward.

Looking at the datasheet, the AD825 has relatively low open-loop gain (76dB) compared to most other opamps, which means it must use a good amount of internal local feedback. It also has fairly good output drive capability (50mA).

M³ Project status update

While Morsel is out of town I did more work with the prototype and more testing.

The prototype is now running in full three-channel mode. It also has a capacitance multiplier transistor on each supply rail, fed by a JFET current source to power the opamps (see schematic in the beginning of this thread). This in effect multiplies the 100uF capacitance by the DC current gain hfe of the pass transistors. Note that the capacitor is bridged across the bases of the pass transistors without being referenced to ground. This provides a good degree of rail isolation without the use of another voltage regulator. Use of IC voltage regulators cause more voltage drop each rail, and they must be referenced to ground which I'd like to avoid.

Note that this is a bit different than the parallel JFET rail isolation scheme found on the PPA. The BJT pass transistor fed by a JFET CCS here provides greater current capability, which allows me to power all three opamps. I am leaning with this approach rather than having one per channel, which leads to reduced parts count. The current draw in each opamp is quite steady and there is no advantage to having one capacitance multiplier per channel.

Furthermore, the signal ground is now a virtual ground, derived via a TLE2426 rail splitter, which is fed by the clean and isolated rails after the capacitance multiplier as shown in the schematic. Having a single capacitance multiplier and rail splitter for all opamps removes any possibility that multiple TLE2426s might "fight each other" while driving a common virtual ground.

I tested the amplifier running at maximum output into a 33 ohm dummy load, while watching for ripples on the op amp supply rails with an oscilloscope (across both rails, and from each rail to the virtual ground), before and after these modifications. The benefit of each of these changes was quite evident.

I also ran RMAA tests using an M-Audio Transit, and generated a report that compares the latest results against two previous measurements, as well as of the M-Audio Transit in loopback mode as a baseline. To view the report, look at the link below. I added some remarks to clarify the tests and results.

http://www.amb.org/mmm/comparison_20041019.htm

I am very happy with the outcome so far and am excited that the continuing work is producing excellent results. While these measurements alone do not tell you how the amp sounds, I am pleased to say that it's everything I've hoped for and more. I dislike using ethereal adjectives to describe the sound, but it is very, very good.

I am going to wait for Morsel to return from her trip to discuss the results of these developments, and hopefully we'd be making some important headway soon.

That looks very good AMB Keep up the good work!



/U.

Sounds good i know the cap multiplier you speak of works quit well as i have used this configured as you did for quite awhile and this was going to be incorporated into the PPA as was a High speed fully ground plane PCB however the other PPA team members overruled me on these aspects.

Another thing you might consider is that in my early tests with the PPA i found three isolation op amp rails gave more imaging i suspect this is due to some inter-capacitive effects and PCB Foil inductance.

You don’t have to reference a voltage regulator to ground referencing to the opposite supply rail is fine as long as the maximum input voltage to the regulator is not exceeded also two complementary regulators can be referenced at there midpoint as in a Dual tracking arrangement, Just a thought.

If the jfets are somewhat matched with regard to Idss the TLe’s will not fight each other as they are isolated by the Rds on of the fets . Since the M3 is not a portable amp any mismatch in Idss of the fets will only serve to raise the current drain of the amp and just might be of some benefit by serving as a preload or Bias for the capacitance multipliers.

The above are just my thoughts on some things you might want to investigate prior to the final PCB layout.