Here's my latest crazy project, intended to answer the question 'How low can you go?' in power supply impedance. No listening tests yet, I've just finished assembling this stack of hexagon capacitor arrays out of cheap electrolytics. Total capacitance of the top 4 tiers is comfortably above 1F. The top tier is designed for lowest ESR at HF using Nichicon 1500uFs, the next tier is a trade between ESR and uF. The last two are the highest capacitance density I could find (a total of 182 4700uFs). I've yet to connect the bottom two hexagons to the upper tiers but the top two are connected by wires running through the gaps between caps.
The whole is intended to push the limits of low frequency noise in powering a couple of TDA1545s which will sit on a small PCB supported on thick wires from the centre of the top tier. I am aiming to get a capacitive reactance (at 50Hz) at the DAC supply pins below 3mohm and ESR below 1.5mohm.
I have a pair of tired Goodmans Magnum SLs. They are in reasonable condition, but one mid is blown, and they have no front covers. I tried grafting a new mid on to the broken one, expecting I would need to do some crossover work, but it really did not sound good. A very unpleasant sound that I found quite fatiguing and unpleasant to listen to coming from the upper bass lower mid area, quite a relief to turn it off actually. I put bigger and bigger choke on the woofer to try and filter it away, but that didn't work. I took of the dust cap in case that was causing it but no.
I am thinking of re-purposing the boxes now, but first did just check the woofer to make sure it has a problem. Below are some distortion plots, for a good one, the bad one and also for a Wharfedale Dovedale 3 woofer, with fixed rubber (Note the different scale). These plots were done at quite low level, like normal neighbour friendly listening levels.
So you have a small handful of parts and want to build a (simple) discrete voltage regulator instead of using an IC. What to do?
For line-level audio circuits, especially op amp based (IC or discrete) preamps with high PSRR, something like the Z-reg is generally sufficient. Robust, works well, has enough ripple rejection to cut power line noise from the preamp output.
If you add just a couple more parts, however, you can add feedback to the Z-reg circuit, a simple error amplifier in the form of an additional transistor Q2, with the output-sampling voltage divider R1,R2.
The ripple rejection is not vastly superior to the circuit without the feedback unless some additional bypass capacitors are added as shown in the first version of the circuit below. The output impedance, however, improves from a few ohms to a few tenths of an ohm as a result of the feedback. Which could, in principle, be of use.
I'm not going to spend too much time on this one. The idea is to increase the input impedance of the pass transistor by buffering it with a jFET so it will support a high-impedance passive CRCRC filter section that generates a low noise reference voltage. The reference is defined not by a Zener or diode stack, but by a simple voltage divider. There is a LM317 pre-regulator on the front, but it is traditionally configured and works independent of the following circuit so it is omited here together with the additional transistor that speeds up the charging of the reference voltage filter capacitors.
The basic problem is that lowering the noise of the reference cannot lower the output noise indefinitely. After a point the output noise is defined by the performance of the pass transistor instead.
Two versions are presented, one with all the protection diodes and a simplified version with extraneous components removed.
LTSpice simulation shows so-so performance into a light load, with about 70 dB of ripple rejection and a fairly high output impedance, but the drop out voltage is respectably low and we must factor in - coming directly from the Jung Super Regulator - that this is just a two transistor circuit, with no error amplifier to provide feedback.
As a frame of reference, it is quite similar in performance to the Z-reg we looked at back in part III.
The k-multipler is of a class of voltage regulators where the output is referred to the input voltage, rather than to ground. It provides "X volts less than the input", rather than the traditional regulator which provides "X volts above zero"....
The Kmultiplier is as far as I know the best power filter you could possibly squeeze out of just 2 active components. Not only this, but due to its simplicity it's RF PSRR is greater than the majority of regulators. It has output impedance lower than most lytics, doesn't oscillate into film bypass, and won't glitch on fast load signals.
Posted 13th February 2014 at 04:21 AM byrjm Updated 14th February 2014 at 10:52 AM byrjm(clean up)
In the next part of the series, I'll be presenting various published regulator circuits.
Today we have the "Jung Super Regulator" (2000 version) on deck, thanks to Tangentsoft's excellent write-up.
In translating the circuit to LTSpice, I've made some concessions. While I have kept the protection diodes so as to be consistent with the original - even if they do nothing in this simulation - the op amp, transistors, voltage regulator and reference have been substituted with working equivalents from the LTSpice libraries. I've been approximate in the resistance and capacitance values, and tuned the circuit to output 10 V at 10 mA to keep in line with the previous circuits I've uploaded.
It works though, and, under simulation at least, it works extremely well. Putting it together in LTSpice gave me a new appreciation for just how much work and refinement went into its design. Now, its an open question whether such over-the-top performance...
Posted 12th February 2014 at 04:10 AM byfas42 Updated 12th February 2014 at 09:55 PM byfas42
I mentioned in a thread a while ago about doing an exercise of engineering an amp capable of delivering 2k watts into a 1 ohm load, with distortion aiming at the magical ppm figure. This is still happening, and making progress ... key problem as I saw it was managing crossover distortion intelligently - I'm looking at a conventional class AB output stage at the moment - and it didn't make sense to try and control it using a classic global negative feedback approach.
By nature I'm an excellent scavenger, I look to see what ideas are already out there - so I'm trying out some concepts in using local negative feedback. This is evolving, step by step - and showing promise: in a simulation of the output stage only, getting effectively 2kW in 1R, at 200kHz with reasonable stability - and the waveform at this stressful frequency looks pretty good, there is still some crossover glitching, but I'm reducing the visible level of it steadily.
To show just how lazy I am, the main function is the file titled "dig_cross.c" - as that was the main function I edited as the base of this code. There is also a file "ad1940.c" which contains a bunch of the SPI stuff. This is yet another illustration of my bone idle-ness - as this module is probably a decade old. It is used, but has nothing to do with an AD1940 IC....
There is a bunch of comments in this, but some general overview comments are:
- About 95% of the source code is about:
- Running the user interface
- Generating the display (rather utilitarian implementation)
- Reading from the EEPROM, and doing limit checks on data
- Writing to EEPROM