Building a Leach Amp

Having built two Leach amps in the past, I have not been a fan of biasing the way Leach recommends. Pulling those fuses in and out, etc., to measure bias amps, to me, was always a cause of concern for, well, danger. Layout became an issue to make that marginally convenient. Also, the current seemed to always be drifting around a great deal, making it hard to pin down. Further, it has become apparent that biasing by measuring current may not be the best way to do it.

Self states that an optimum bias is about twice what Leach recommends. I found I like the higher bias. Andy_C, in another thread, determined that Self's optimum bias equated to the lowest variation in output impedance, which was also expressed by Oliver in the Feb '71 issue of the Hewlett Packard Journal. Self states that the optimum bias is really a function of the voltage across the output emitter resistors, also mentioned by Oliver.

Self's optimum Vq bias for .33 ohm emittor resistors is 47.6 mv across the emitter series pair, for about 75 ma current. For .47 ohm resistors, it is 54.8mv for 59ma current. Again, this is across the series combination. Therefore, to make biasing easy without opening the enclosure, connect the ends of a series pair of emitter resistors to connectors on the outside of the case, so you can insert probe tips of a volt meter into to measure the Vq voltage for biasing. Also, get a multi-turn trim pot that can be mounted on the enclosure so that a screwdriver can access it from the outside of the box, or layout the boards such that a hole drilled in the enclosure can provide access to the trim screw.

I found that the voltage Vq was less variable than measuring bias current, once warmed up.
 
Since Leach does not appear to care anymore about bringing his published parts selection up to date with easily available devices, I'd like to start a discussion about replacement devices.

For the output devices, some of these have been discussed and used:

MJ21193/4 for TO3 package
MJL3281A/MJ1302A and their updates 2SC5200/2SA1943
MJL4281A/4302A


Other interesting devices:

MJL0281A/MJL0302A, not yet available

These are sustained beta devices. The MJL3281 combo is sustained out to 8A vs. 5A for the MJL4281 combo.
 
Regarding drivers, it's hard to find any improvement over the MJE15030/31 combo originally specified by Leach, as the newer 32/33s, etc. are not as high current rated. It would seem that rating for a higher current could maintain their temperature and Vbe more steady.

The VAS and predriver are the big problems, since the ones picked by Leach are the hardest to find.

The MJE340/350 combo has been used with success, but with a dirth of specification.

I've seen the Toshiba 2SA1930/2SC5171 recommended here, but the voltage rating is much lower. Not sure of a rating of 180V in anything to worry about, though. With ft around 200Mhz, and hfe of 100 down to 10ma, this looks pretty good. Any others out there?

Also, the 2SA965/2SC2235 looks like it has potential as it's hfe is specified down to 3ma, but the voltage is even lower on this device, but still comparable to the 30/31 pair specified by Leach.
 
As for input transistors, the ones specified by Leach are still available, but there are no noise specs, and are not listed under the low noise groups by OnSemi.

I've seen the following recommended as low noise:

BC550C/BC560C (only 45V)
BC546B/BC556B (only 65V)

These are very high gain/high bandwidth, also, so they could improve DC offset. Although their voltage rating is low, they would seem safe under the cascode.

Another interesting pair is Toshiba's

2SA970/???? for higher voltage, although the pinouts are different.
 
pooge said:
Regarding drivers, it's hard to find any improvement over the MJE15030/31 combo originally specified by Leach, as the newer 32/33s, etc. are not as high current rated. It would seem that rating for a higher current could maintain their temperature and Vbe more steady.

The VAS and predriver are the big problems, since the ones picked by Leach are the hardest to find.

The MJE340/350 combo has been used with success, but with a dirth of specification.

I've seen the Toshiba 2SA1930/2SC5171 recommended here, but the voltage rating is much lower. Not sure of a rating of 180V in anything to worry about, though. With ft around 200Mhz, and hfe of 100 down to 10ma, this looks pretty good. Any others out there?

Also, the 2SA965/2SC2235 looks like it has potential as it's hfe is specified down to 3ma, but the voltage is even lower on this device, but still comparable to the 30/31 pair specified by Leach.



I've seen the Toshiba 2SA1837/2SC4793 pair recommended as a MJE1503X/x replacement, but max Ic is only 1A. Doesn't appear safe as a driver if there are problems with the output devices. However, I'm thinking it might be of good use in the predriver, as the hfe is specked down to 3ma, and the voltage rating of 230V looks good.
 
input transistors ...

I've happily used 2SA872 and 2SC1775 for these positions in some past designs.

Don't know if they' re still around, but they have good specs for the intended purpose and are cheap. I bought a bunch a long time ago and have been using them for years.

specs here:
http://www.datasheetcatalog.com/datasheets_pdf/2/S/A/8/2SA872.shtml

But they do have the pinout issue (compared to the 2N or BC parts) also.

mlloyd1
 
I simply set bias by running the amplifier at a low power level and monitoring the crossover notch on the oscilloscope output of a
harmonic distortion analyzer. Generally there is a null setting
of minimum distortion (especially high order products) as bias
is varied. Beyond a point, increasing bias increases distortion
and dissipation.

If the amplifier gets too hot in normal use, reset the bias a bit lower usually fixes that. Otherwise, there may be a problem with
the amplifier, or cooling may be inadequate.

I also monitor bias voltage across one of the output emitter
resistors, to catch any signs of thermal runaway or excessive
bias current while twiddling.

Once you've established reliable operation, you can vary bias
up or down to suit your listening tastes. But keep an eye on
heatsink temperature or a DVM across one of the emitter resistors to monitor current.

MJ21193/4 outputs work perfectly. That parts list could stand
updating.
 
Since Leach does not appear to care anymore about bringing his published parts selection up to date with easily available devices, I'd like to start a discussion about replacement devices.

I hope I didn't come off as disrespectful or ungrateful for Leach not updating his design lately. I didn't mean to, as Leach has been one of my heroes. It's a testiment to him that his design is still being built and respected 30 years after his initial publication of the amp. But the fact is that he has not been updating lately, for whatever reason. Sometimes he answers emails, sometimes he doesn't. That's his business. He once admitted an interest in looking into beta-sustained outputs, but apparently nothing became of it. We all know how that goes.

However, there has been a lot of these amps built out there, so there should be a lot of valuable experience to offer. While Leach has designed a stable, bullet-proof amp for any load, he doesn't impress as losing much sleep over parts quality (although he has mentioned appreciation of the ills of ceramic caps). As his focus is on teaching students with frugal budgets, this is understandable.

However, every time I build a new one, I want to make it better than the old one. It's just as easy to solder a superior part as an "inferior" one. That being said, I have different design priorities now. Unlike Leach, I am not as concerned about making the amp absolutely stable into any weird load at extremely high power in a safe operating range with a super high slew rate capability, no matter who builds it. Not that these things are bad. It's just that getting all of these things in one package may involve compromises, as they ALWAYS do, that I don't need to make.

For example, I am now using high efficiency horn speakers with 16 ohm input impedance. (Please don't bother asking why I then use a SS amp!) Therefore, I don't need a super high slew rate, because with a low voltage output requiremnet, the slew rate can be relaxed, since slew overload in the differential stage is a function of output level. By relaxing slew rate ability, which is already well more than needed, I can reduce the input degeneration resistors to reduce noise, since horn speakers will expose it. Thus, looking for low noise input transistors is a priority for me.

One of the biggest improvements I've made to my amp, or rather the amp interface, is the addition of Jensen input transformers to the inputs, which have totally and absolutely eliminately every last bit of hum, and allowed me to have a balanced interface from my DAC, all the way to the amp. The Jensens like a 10K ohm load to prevent ringing, so that means I can reduce the input resistor to improve noise. With this, the feedback impedance can be lowered in order to reduce DC offset brough about by changing the input resistor, while reducing noise even further. (After changing the input resistor to 10k, the DC balance only went up to ~35mv, which isn't bad. But it can be reduced to its lower level by lowering the feedback impedances to match the input again.) DC offset can also be improved with high gain transistors in the input, according to Self, while lowering distortion. So searching for a replacement here is in order.

I will probably eliminate the split feedback network also. Not because I think it is bad, per se. I just don't think it will be that valuable of insurance with a horn load not used for sound reinforcement for heavy metal. I can use newer high gain, high ft output transistors, which will keep the output impedance lower (as will the removal of the split feedback path), since higher ft devices will have lower output inductance, and a higher pole frequency for better phase margin and stability, and lower distortion where high gain ones are used. However, the main reason I want to eliminate the split feedback is to reduce the numbers of components in the feedback network, so use of very low-noise bulk foil resistors can be more affordable, as well as bring tighter available tolerances to keep down DC offset.

Although the newer high gain, high ft output devices do not appear to be as bullet-proof as the ones Leach picked, they are close enough in that regard for me and others to accept the trade-off of that advantage, especially where an extra pair are used, which again helps reduce distortion and impedance.

The newer high gain transistors, when used in the output triple, would provide lower output impedance, lower source impedance for the output transistors which they like, and a higher load impedance for the VAS stage, which it likes. They will also reduce distortion because of more available local feedback to them, since they operate 100% feedback with no voltage gain.

Higher gain, higher ft devices in the VAS stage will also improve distortion. Use of higher ft devices throughout will also improve phase margin, and/or allow for increasing the amp gain-bandwidth product or open-loop frequency response, or higher feedback, as desired. There has been much written about how high open-loop frequency response is not important to TIM or slew distortion, etc., as was once though--which is true. But it is also true that the output impedance rises above the pole frequency, and John Curl has stated that FM distortion is reduced if open-loop frequency response is higher. While I thought this was roundly pooh-poohed at one time, my memory is not 100%, and Curl is probably more in the know about it than me.

Leach has also stated that a capacitor across the driver emitter resistor will not help. Others said that Leach's rational is based on DC analysis, and not high transient analysis. Some say they can hear a difference, and Self showed a very slight improvement in distortion by using one here. Since there is NO argument that this capacitor can hurt, why not? It's not so much a topology change as an easy part addition.

In sum, I am not trying to diss Leach or his amp. It's just that by simply changing parts or their values, the amp may be optimised for different criteria and priorities, and I would like to experiment with sound engineering changes. My priorities happen to be lower noise and lower distortion, and maybe a better sounding amp, with the SAME pc boards and topology. Whether any of these changes will make it sound different, I don't know. But maybe making ALL of them WILL make it sound better.

Note that I didn't raise any issues about "audiophile approved" parts. What I did raise is the issue of how different parts can theoretically improve the present topology from sound objective criteria.

What I'd like to do in this thread is discuss aspects of part selection and values that would achieve a better amplifier using the SAME topology, and the reasons for doing so. I'm not opposed to discussion of observations of parts quality, as such, and/or where to get them cheap, or group buys for same--but I don't care to see this thread break down into off-topics about feedback vs. no feedback, symmetrical vs. current mirrors, tubes vs. transistors, etc. Take that somewhere else, please! I want to discuss the topology at hand, using our present circuit boards. However, since the great majority of amps use at least some aspects of this topology, observations and experience about how to improve various subsets of the topology would also be welcome.
 
amplifierguru said:
Hi Damon,

And what is the bias level across those emitter resistors that makes the crossover artefacts go away on the THD analyser CRO output?

Across what value emitter resistors?

Cheers,
Greg


I posted Self's optimum voltages in the first post of this thread, which corresponds with the minimum output impedance variation. That works out to about 150ma per channel, according to the Leach method.
 
pooge:

Taking note of your notes. Some interesting ideas to think about
and perhaps try out. Other than using some better quality parts
(such as Black Gates for the feedback and diffamp power supply circuits), I've built my units essentially stock.

Amplifierguru:

I generally don't worry about a specific voltage when setting
bias while monitoring with a distortion analyzer and oscilloscope.
In fact, I've forgotten what voltages I've been measuring! I was
using .33 ohm emitters, but inadvertantly bought into a large
batch of .27 IRC wirewounds that appeared to be better quality
than the usual sandcast jobs.

My primary amplifier is 'down' at the moment waiting for me to
get around to installing a new set of 'stealth' rectifiers; I'll try
to remember to make note of my measurements.

I don't know that setting to a specific voltage is really ideal. I'd
like to try this technique using my computer's soundcast as a
spectrum analyzer to see how the distortion products are
spread out at different bias settings. But I think I get a fairly
good idea just looking at the 'scope display off the analyzer. High
order products will show up as a sharper 'spike' on the waveform,
and even my geriactric Heathkit gear will go down far enough into the noise to give an idea of what's happening. This is why I like
this method of setting bias.
 
Hi Damon,

Yes I too have always set BJT output stage current for minimum apparent xover artefacts visually - unless the currents way high then a redesign is called for.

I was hoping you had your settings to x ref with self's as stated by Pooge. I used CFP's and CFT's in my BJT amps which had a lower Iq opt in the range 25-50mA with typ 0.22R , if my memory serves me well...hmm... good lyrics for a song.

Cheers,
Greg
 
Does anyone know how much the bias current would vary when building similar amps? I'm thinking that most of us don't have a distortion analyzer nor access to one. So assuming we follow similar BOM, would the ideal point be essentially the same or does it vary with each batch of active devices??
 
According to Self and Oliver, it is the voltage that matters, and not the particular current, and varies with topology and emitter resistor value. Oliver calculated a range the the optimum fell in. Leach's bais current seems to be at one end of the voltage range stated by Oliver, based upon his finding of 40-45mA, and Self's at the other end of Oliver's voltage range. Why Leach and Self found different bias settings looking at a distortion analyzer, I don't know. The details of how they analyzed the distortion is not evident. However, as I stated before, Andy_C did a circuit analysis of Leach's topology that evidenced a minimum variation of output impedance at Selfs optimum bias, and this is what Oliver stated the characteristic of optimum bias should be. Leach, however, welcomes increasing the bias current to 150ma (which is close to Self's bias), or more. I preferred the higher bias to Leach's, done by listening, before discovering Self's optimum.
 
I've tried running my amplifier at the higher current ratings, but it ran uncomfortably hot for my preferences. I'll have to change the physical design of my amplifiers to feel comfortable with the higher currents. Wasn't sure I liked the sound when I did that,
but that was before I rebuilt my amplifier with the latest version (went from 2.something to 4.5). No fans. Don't like 'em.

I feel like I've taken the Leach design as far as I can on my limited knowledge but haven't found any other semiconductor designs compelling enough to explore. But I would still like to build a 'final' Leach amplifier that actually >looks< good and has decent air circulation around some oversized heatsinks so I can
safely explore the higher bias.

Lacking a sheet metal shop and money, that hasn't happened so far.
 
For anyone interested, I put a feeler out for a group purchase for Jensen input transformers here:

Group purchase

By using this at the input, series input resistor R2 can be removed, as it's purpose for RF filtering will be better performed by the transformer with less noise.

Further, since the transformer likes a 10K load, R1 and R19 can be reduced to 10K, R17 and R18 to 5K, and R19 to 500, thereby reducing noise further and helping to reduce DC offset.

Off course, a few caps need changing, too. C6 needs to increase to 470uf, C8 to 100pf, and C9 to 400pf or closest value. Aternatively, C8, C9 and R20 can be removed, and R17 and R18 can be combined into a 10K value with one of these positions jumpered out.

The transformer also eliminates DC input from a source component. Further, balanced or unbalanced inputs can be used, as desired.
 
Ignore the previous post. R2 shouldn't be removed. It needs adjustment, though. So does C2. I know the values, but only for a Jensen transformer. The present value of C2 will improperly load the transformer, especially with the present value of R2. I have appropriate values of these confirmed by Jensen and Leach. I will ship that info with the Jensens purchased in the group buy. I don't want to be blamed for someone using the wrong values with other amps.
 

Luke

Member
2001-09-24 11:12 am
Im nearly ready to start my Leach. I have an old amp chassis with a 55 0 55 transformer. I expect to get rails of around 70 - 75V.
Is this too high for this amp and if so whats a good wat to lower the rails(dont want to build a regulator). Did I mention the transformer is potted:)
 
Try building the Super Leach instead--those are close to the
right rail voltages. :)

Otherwise, put in higher voltage devices for drivers and output
stages, and adjust several resistor values according to instructions.
If you don't try to drive low-impedance loads, you might get
away with such high rails.

You could try half-wave rectification, but the rails will still be high
and ripple will be considerably higher. The amplifier has very
good ripple rejection, so that will help mitigate the ripple. A
bucking transformer could help, but I've always avoided that approach.

Nah, take the direct approach and build the Super Leach!

--Damon, who needs to finish upgrading his Leach amp