Leach amp question

This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Well, I've been perusing this site for quite some time, now, and have found more good info than I can digest! Great site!

Well, as a Geek Group project (www.thegeekgroup.org, site is being redone, unfortunately), we're designing and building a custom DIY audio amp. The idea was to choose a really good DIY project, make any improvements possible, and build it. For "Version 2," we'd make a few more improvements (heftier power supply, more power output, more bells and whistles, etc). We've chosen Leach's amp, and are in the process of making it better. I had no idea this amp was so popular amongst the DIYers.

I've come across some interesting information on this site that I'd like clarified/elaborated. Several people have stated that the ver 4.5 boards sound much better than his previous boards. Do you mean his PCB literally, or the v4.5 schematic? The reason for asking is that we intend to make our own set of PCBs to more efficiently pack the components, and to utilize the MJ21193/4 output transistors (plus an extra pair :)

Can anyone recommend any other good design improvements over his v4.5 design?

FWIW, I'm a mechanical engineer. Although I can understand most mechanical concepts, I cannot grasp the advanced concepts of linear amplifier design.... I'll be most pleased when we get all the design details squared away so I can begin PCB and chassis work and assembly.

Mark Broker
FWIW: I recently upgraded my 20 year old chassis to the ver. 4.5
board, with some minor parts modifications:

--as per Self and Sloan, added a 1 uF film capacitor across R36, to
speed up turn-off time of the output transistors; not sure if the
10 ohm base resistors impact this modification; this reduces high
frequency distortion

--used polystyrene caps when I could find them in those values for all pF-sized capacitors; otherwise used silver mica; there are
probably better (audiophile preferred) brands for these parts--I simply used what I had or could get locally; small value polystyrene caps are getting very hard to find

--C4, 5, 6; used Black Gate electrolytics; C6 is a bipolar N series; I couldn't afford to use Black Gate for all of the electrolytics, so I put them in where I thought they were most critical

--better quality IRC wirewounds used for emitter resistors in place of the usual sandcast devices; I recommend Mills or similar
quality resistors with lowest possible Tc

--MJ21193 and -94 output transistors; might be interesting to try 2SA1302/2SC3281 or their improved versions; driver transistors might be good candidates for substitution with lower beta droop; alternately, try additional pairs of output transistors up to four pairs which should make possible higher output power and improved distortion for sub 4 ohm impedances (some parts values may have to be changed if supply rails are pushed over 65 volts) Self reports that additional pairing does reduce output
staget distortions due to 'beta droop' which is actually caused by
the driver transistors having to work harder into the inevitable curve

--matched all mirrored resistors as closely as possible with my DVM, not necessarily for absolute value; matched zeners for absolute value from a batch of 50; you could try a 39 volt zener in series with a regular silicon rectifier to improve thermal tracking
(thermal coefficients cancel each other out)

--matched all differential/input transistors with curve tracer and
strapped together the differential pairs with a small dab of thermal compound to improve thermal tracking

--tweaked bias with THD meter, viewing the residuals on an oscilloscope to monitor the crossover glitch; I used minimum bias
to eliminate the glitch; Self appears to recommend setting bias
at 5 kHz or so; there's probably room for improvement in this procedure; my IM meters either don't work or don't have an oscilloscope output monitor; an Audio Precision test set is the
ne plus ultra of audio test equipment and can plumb the depths
of noise and distortion of almost any amplifier in great detail

Things I've Been Thinking About:

--radical power supply modifications: use a separate, regulated
power supply for the driver and earlier stages, possibly a higher
rail voltage, too. Not essential, as Leach's design has very good
power supply ripple rejection due to two stages of RC decoupling
and zener regulation

--PCB layout modifications; to eliminate as much external wiring as possible, such as using plastic case output transistors with the
leads directly soldered to the board and input connector on the
board with the board oriented to a hole in the chassis so the
cable plugs directly onto the board; there may be slight improvements possible in grounding schemes, but I'm not yet
clear on the details
Another Leach amp question

I just finished assembling my Leach amp. With 100 ohm resistors in place of fuses, the voltage drop was <2.5V for both channels (0.76 and ~1.7) when full rail voltage applied (used Variac to ramp up slowly).

However, when I attempt to set the bias I get no response. That is, the ammeter does not respond (for either channel). Any suggestions for a puzzled chemist?

My own personal opinion is that the Leach Amp is one heck of a good design. I can't help but believe that with audiophile grade components, and possibly a few tweaks, that it can't be a top line amplifier.

I am hoping to eventually get around to replacing my Version 3 with the version 4.5 and, like the rest of us, have been looking at possible enhancements. Damon does a great job in summarizing many of the tweaks one can make.

One area that Leach did change between the Version 3 and 4 amps is the addition of base resistors to the driver transistors (R43 and R44). These are 10 Ohm, and Leach said he added them to prevent possible oscillations which occured in the Super Amp (but never in the Leach Amp). I am a little concerned about these resistors and have considered leaving them off my rebuild for the reason that they induce distortion. Allow me to explain.

The output transistors have a typical gain of around 50. At typical output levels, there will be at least one to two amperes of peak collector current from each output transistor. If even one ampere, with a current gain of 50, the base current into each conducting driver transistor will be 20 mA peak. The voltage drop across each base resistor, therefore, will be 200 mV peak. In other words, the base resistors will induce a gain error. Of course, feedback will reduce the gain error in proportion to the amount of global feedback, but in a low feedback design, there will still be a number of millivolts of gain error. Is this audible? Heck if I know, but why install a component that has the potential to alter the sound if it isn't really needed?

My plan, therefore, would be to omit the base resistors and install them only if there is a problem. I doubt there will be a problem, especially if lead lengths and interconnect distances are kept short.

Other tweaks? I have head that increasing the size of C12, the capacitor across the Vbe multiplier transistor, to 1 uF, maybe even much larger (100 uF), helps.

Also, I have heard that reducing the output emitter resistors (R45-R48) to .1 Ohms helps (though you would have to recalculate the overcurrent protection circuit resistances). If you consider this change, it might be best to build the amp per plans, then disable the protection circuit and install the lower value resistors. If you notice an improvement in sound, then you can recalculate the protection circuit resistances (or ask us here to help you).

Finally, I think that L1 is unnecessary, and probably R50/C25 as well. Again, build it per plans then see what happens if you remove them. If you do use L1, wind it with some good wire. Too many of us go all out on our speaker wire, but we still have a foot of normal, 22 gauge hookup wire used in L1. I guess we have to use such small wire in order to get the desired windings around R49. Still, if L1, R50 and C25 really aren't needed, why use them?

Good luck.
Also, I have heard that reducing the output emitter resistors (R45-R48) to .1 Ohms helps (though you would have to recalculate the overcurrent protection circuit resistances).

Be aware that the choice of emitter resistors in the output stage has the surprising effect of changing the optimum bias current that minimizes crossover distortion. I've done some SPICE simulations of crossover distortion using the models for MJL3281a and MJL1302a output transistors. I calculate the DC Vout vs Vin characteristic and save it in a file. Then I do a linear regression on it to find the best-fit straight line through the data. Subtracting the actual data from the best-fit line allows viewing the distortion at high resolution. Using the 0.33 Ohm resistors, I tweaked the bias current in the simulator until the distortion is minimized. I found that for the 0.33 Ohm case, crossover distortion is minimized at a bias current of 40-45 mA per transistor. This is the same result that Leach arrived at using a distortion analyzer on the finished amp. Repeating the experiment with 0.1 Ohm emitter resistors, I found that the distortion plots showed low gain near the zero crossing point (indicating an underbiased condition) until the bias was increased to around 85 mA per transistor. So for this case the target bias current for the overall amp would be 85 mA * 2 devices, plus the quiescent current of the remainder of the devices on the PCB.
andy_c said:

Be aware that the choice of emitter resistors in the output stage has the surprising effect of changing the optimum bias current that minimizes crossover distortion.

Nice work. I forgot to mention that the bias current would have to be adjusted to a different value than Leach recommended. Thanks for pointing that out, and for making a logical recommendation for where to sit it (though listening tests will always be the final guide).
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.