When I found out the good professor passed away , I was sorry that I never got his "creation" beyond a crude prototype. The original amp is truly a classic , I'm sure more have built it and more have survived than all the other DIYA amps combined.
Simulating the original (LT .asc file "leach 4.5.zip"-below) to see what's going on is a good first step. You might have to "fix" that LT file , get models .. etc. BUT , it simulates well ...but is nothing to "write home about" ,performance wise. Many hobbyist's love the sound and durability of the design and have written many nice things about it even as it is "normal" in its performance. Below in the (first picture attachment) I have beta enhanced it. I ran this simulation with the original VAS configuration , beta enhanced (BE) , and BE with TMC.
I got .01% across the audio spectrum in normal mode , .004% with BE and .001% or less with BE and TMC. All tests were done at 40v p-p/ 8r load -800K UG point. "Normal" had 60 db loop gain and the BE's had 80+DB.
This topology , on first impressions... seems to be very linear - independent of frequency and load (triple is used) ,it is easy to keep the slew very fast and symmetrical , and the design seems to tolerate a wide range of devices/ values. We have a lot of blameless type designs .. TOO many symasym types , why not a good "leach type" design to round out our collection of projects.
The second LT sim file, (LX.1.0.zip) is the one on the schema with BE and TMC , it DOES work with included models (.txt in zip)
Initially , I will build and run this as a voltage board in EF2 mode then design a "all in one" Triple / complete project with 4 pair plastic OP's (1 OP pair driver -njw0281/0302/ 3- NJW21193/4 as main OP's) .. all to run on 55-65v rails.
ANY ideas are VERY welcome and appreciated.
OS
Simulating the original (LT .asc file "leach 4.5.zip"-below) to see what's going on is a good first step. You might have to "fix" that LT file , get models .. etc. BUT , it simulates well ...but is nothing to "write home about" ,performance wise. Many hobbyist's love the sound and durability of the design and have written many nice things about it even as it is "normal" in its performance. Below in the (first picture attachment) I have beta enhanced it. I ran this simulation with the original VAS configuration , beta enhanced (BE) , and BE with TMC.
I got .01% across the audio spectrum in normal mode , .004% with BE and .001% or less with BE and TMC. All tests were done at 40v p-p/ 8r load -800K UG point. "Normal" had 60 db loop gain and the BE's had 80+DB.
This topology , on first impressions... seems to be very linear - independent of frequency and load (triple is used) ,it is easy to keep the slew very fast and symmetrical , and the design seems to tolerate a wide range of devices/ values. We have a lot of blameless type designs .. TOO many symasym types , why not a good "leach type" design to round out our collection of projects.
The second LT sim file, (LX.1.0.zip) is the one on the schema with BE and TMC , it DOES work with included models (.txt in zip)
Initially , I will build and run this as a voltage board in EF2 mode then design a "all in one" Triple / complete project with 4 pair plastic OP's (1 OP pair driver -njw0281/0302/ 3- NJW21193/4 as main OP's) .. all to run on 55-65v rails.
ANY ideas are VERY welcome and appreciated.
OS
Attachments
Here it is , I thought all knew of it , it is akin to D. self passing away.
W. Marshall Leach, Jr. 1940 - 2010 - Obituary - Tributes.com
I intend on making a "leach amp" that will be smooth-musical and cheap so many will build economically.
I will use his 3 diode Vbe and as many of his original design choices (split ac/dc NFB) as can be useful. I won't diverge too far (except using the excellent NJW TO-3P's for triple OP stage )
OS
Last edited:
Hi OS,
Nice work and a nice tribute to Marshall. We all miss him and cannot believe he is gone. Over 35 years, we have all learned a lot from Marshall Leach. In addition to being certifiably smart as a PhD, we had an incredibly good ability to communicate ideas and was very practical. These latter traits are not always found in PhD's.
Cheers,
Bob'
Thanks , Bob. Marshall's style was similar to yours (not strokin' any ego's here). Just easy to understand yet informative. I wish I could afford your book , I'm driven to spend on parts instead (mouser is my "dealer"
).I'm not done yet , I still have to decide how much of the good doctor's work to leave original and what changes are necessary to make the newer devices perform at their optimum.
The new NJW21193/4's are "in -between" they have almost the beta (most exhibit Hfe 60-80) of the 0281/0302 pair but have almost the SOA of the original TO-3 devices. I am running them (only 1.80/2.70 ea.) on my 60w amp and CAN NOT burn them out (shorted on purpose) I'm impressed. Ft/Cob is 4mhz/500pF , close to the 15003/4. The njw0281/0302 is a great choice of driver for 3 pair of the 21193/4 and has some "muscle" for ultra low imp. loads.
The input stage- The professor used mps8099 which is similar to the 2n5401/5501 but with less Vceo. For ease of construction/sourcing / higher rails, the 2N-xxxx pair would be a good substitute. Since we are cascoded , low noise ss9014/15 or any other suitable input pair can be used. The LED current sources are a simple and cheap improvement worthy of addition as well ...
VAS - This is the "great decision" , Leach's original voltage stage works ... it even works real good. This might be the source of all the favorable reviews , untouched - as is. I know this will disallow the addition of TMC and keep my loop gain low but I am after the classic's "flavor" (sound).
Feedback Compensation - "split" compensation (AC and DC - taken from both the driver and final OPS). I will reread the professors explanation of this to refresh my memory. I also want to experiment in LT to see how this method is better or different than straight global NFB.
Vbe - thermal comp. , from what was said in the professor's article and the fact that DIY'ers everywhere have reported minimal issues , the doctor's approach is a good - (practical)method , even as it uses a simple 3 diode string to bias the main Vbe.
I read the leach amp article before I even found this site (DIYA). At first , looking at the design , it "blew my brain" ... I wondered how college students could build this thing ?? It seems so simple now , not quite as simple as a "blameless" , but close. A lot of different "varients" can be derived from this design ... cascoded VAS (like the luxman) , beta enhanced cascoded VAS , even a folded cascode symmetrical VAS. As tempting as all this is , it would not be "classic".
NO thermaltrack designs either. I am a little uninformed as how does one correct any DC offset in a complimentary differential ? , better put .. what is the BEST way to correct/minimize it ??
Any comments - idea's from leach amp owner's or builder's would be greatly heeded/acknowledged. (original below)
OS
Attachments
Last edited:
The Leach Amp is one of the first articles I read online when I got the interest in building audio about 6 years ago. Many of the other designs I found were just schematics, with no other explanation other than "it just works like that, dont touch". The Leach Amp was thoroughly documented, with the reasons behind its design explained at every stage.
This amp can run happily even on 80+ V rails. With the KSA/C's ,or 2n5404/5551's , we are derated by at least 1:2 on the "close to final" (below 1). This time I went "classic" , just to see how good it can get ... I'm impressed.
What is strange , is that as I get closer to the original topology , it only gets better. The accuracy of the fairchild models assures me that what is below will work close to the "real world" , 100 DIY amps and a new scope are proof.
At 1khz the spectra is "golden" with thd dropping 6-10db per octave , not as good as the blameless , but still a good showing. H2 is dominant , surprising for this type amp as well. At 20K , still a good amp ... at .001% total , H3 shoots up a little , but not much.
Next post , I will explore Slewing (where this design should "rule") and the effects of the good Doctors "split compensation" scheme.
I also have to obtain the SS9014/9015 models (input devices) to make the simulation as "honest" as possible. OS
Attachments
Last edited:
Me too. For a long time, ordering a set of boards was on my "to do" list but maybe it's too late now.
Ray
Not too late , just the beginning. A COMPLETE leach amp using mouser/digikey ON semi's is on the way that will have the "classic Bart Locanthi" triple OPS. The only changes I foresee are the fairchild to-126 VAS/predrivers , more power (160/250w) and a single 100 X 200mm FR-4 PCB. If you can find a HP laser printer , your all set.
OS
No issue with that , why ?? a certain HS ??
I have 2 styles.. A and B (below) I will use the same styles but include the whole amp (one board)."A" will go on a 75mm HS (3") "B" is supported by the OP's and needs a 127mm ( 5") HS. On these amps , one pair output devices njw0281/0302 with be the main drivers , this will allow it to be scaled into a MONSTER leach - 8-10 pair NJW21193/4 - 80V rails = 400w+ / 700W+ "super leach"
OS
Attachments
No issue with that , why ?? a certain HS ??
I have 2 styles.. A and B (below) I will use the same styles but include the whole amp (one board)."A" will go on a 75mm HS (3") "B" is supported by the OP's and needs a 127mm ( 5") HS. On these amps , one pair output devices njw0281/0302 with be the main drivers , this will allow it to be scaled into a MONSTER leach - 8-10 pair NJW21193/4 - 80V rails = 400w+ / 700W+ "super leach"
OS
Do you mind me asking what PC board software you are using?
Ray
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.