I was bored this winter and did a small DIY project

Technics-goy

Super Member
Hi everyone,

well I am new here on the forum and thought I post a little about my recent winter project. As I have become very much a fan of Technics amplifiers and since these beauties with these wonderful VU meters seem to become really rare these days I thought why not revive the 80's technology a bit and make it a cool DIY project.
And so I decided to replicate the trusty SE-A5 New Class A circuit and give it a new spin and add a few minor features to it. :) My goal was to have 5 identical channels housed in one case so I had to scale down the output power a bit to fit it all in. I settled for 100W sine CW into 8 Ohm. The reduced power (and operating voltage) offers a really relaxing operating environment for the semiconductors which I expect to outlast me at this point. I added a staggered gentleman's soft-start to minimize inrush currents and I replaced the output relays with two 200A power mosfets. I added an option for balanced input as well. The amp modules are monitored and controlled by a Cypress FPGA chip.
Eventually to most part I was able to source original semiconductors altho I had to substitute a few for some newer vintage I also selected where needed more powerful transistors as I did not like to see them run so hot as in the original A5. Where it made sense I matched all semiconductors as far as possible which I doubt was actually done in the original amps. I had posted on the DIYAudio forum but interest, with exception of one :D, seemed to be NIL. I guess the cool and wonderful Class D is where it is at these days. :( Very very little do people actually know what they are missing when opting for a Class D.
I have a restored and modernized SE-A5 and was able to do direct comparison of my amp modules to the original A5. Using a set of DSS-7 and streaming FLAC I can not tell the difference to the good old trusty A5.
I measured flat frequency response up to 65kHz and step response with a 1us rise/fall time pulse. Group delay is about 0.5us as seen below. Currently I can not measure spectrum and had to resort to the cheesy channel math with the scope. But the measurement is encouraging anyway as the difference between the input signal to the output signal is effectively nil. Whatever little difference may be registered seems due to group delay more so than distortion, I'd say.
All boards are gold plated to improve somewhat the prettiness factor :D and the main circuit boards have 2oz. copper clad.

I wanted find out for myself if I can pull it off and preserve the 80's sounds as I was quite dissatisfied with newer vintage off the shelf amps like Denon, Bryston and AudioNet which surprisingly disappoint considering they are advertised with very reasonable spec's and sell anywhere between 6000 and 9500 suggesting they are serious competitors, which they are not (at least in my books), but at fortunately they look pretty on a stack.;)



Here are some snaps

20161025_193426.jpg


20161025_193720.jpg


20161120_201336.jpg


20161127_122633.jpg


20161127_121936.jpg


20170129_195300.jpg


20170113_192253.jpg


20161227_191531.jpg


 
Thanks mate.
Building an amp is not really that hard, however you do need some tools and some basic electronics knowledge to do it. In this case all I had to do was to replicate what the very smart and innovative Japanese engineers already had figured out and in my opinion to this day they are still leading the pack - untouched.

My deepest respect to the chief engineer at Technics (1985). :bowdown:
 
That's a really cool build. Do you have a link to your thread over on DIY? I don't have the knowledge to build an amp but the process is something I'd definitely like to read about.
 
Thanks for posting this. This is cool as hell! Like others, I wish I had these skills. I'm doubly impressed when people do custom boards like that. They look sharp!
 
Hats off to your clean work well organized and slick. As I own all Technics pro equipment myself I like their skill also. Great job.
 
So how does it sound? Are you at that stage yet?

I tested it for myself with the speakers as posted above, I am very pleased with it. Now it is in the progress of been integrated into a custom case. This will take a while to get parts machined. If there is interest I will post more pictures once it is all integrated and taken into operation.
Here is another picture of the initial stereo test setup. :)

_28141488547744541.jpg
 
If there is interest I will post more pictures once it is all integrated and taken into operation.

I think there would definitely be interest on this site. There was recently a discussion on a Technics thread (cant remember which one) about doing just this - a circuit re-boot of New Class A.

Have you also made some improvements or changes to how the circuit operates?

I have been hinting to the moderators that a lot of people would like to see a dedicatedTechnics forum set up - if it happened I would become a subscriber......but so far, a lot of resistance.
 
That's what I call DIY! Very impressive sir.

Thanks for your kind words very much appreciated.

Have you also made some improvements or changes to how the circuit operates?

Well, as I had indicated I beefed up some transistors that I felt were a little under-engineered as they operate hot and often failed in the original A5. I also opted for doubling the power capability of all resistors to make this thing bullet proof besides lowering the power rails and so lowering the output power to a relaxed 100W @ 8 Ohm. Otherwise I have not modified the original Technics A5 circuit. I did add a 200A power mosfet circuit on the output to function as 'relais' to connect the speaker to the output. When I did the layout I realized that the Technics engineers that done the original layout have not considered much the thermal coupling between certain components. And thermal coupling there is with the idle current transistors is not done very well in the original A5 using some sort of metal bracket. Also the original A5 has to most part unnecessary over the top wire bridges on the main board and lots of wiring to connect to the pre-amp.
I fixed these deficiencies for this design.:) Wires are at a bare minimum, input, output, AC in and status signal.
 
Nice to see someone DIY the complete amp. I am doing mine also. I design my own. I use Blameless concept to start and add on my own thing.

Since you layout your own boards, do you mind posting the schematic. Do you take their original schematic, change to newer components and beefed up the resistors and transistors?

I got my amp running and did a second revision. Just finished doing stability test and moved onto distortion optimizing. Hopefully I can optimize the THD and start building into the chassis. I am building a class AB amp but with the first few watts in class A to lower the heat. I will create a thread here after I get my result I am shooting for.

One question, you said you are going to put 5 100W class A amps into one chassis. How do you manage to do that given the power dissipation is so high even at idle. I have to go high bias class AB for two channels and I am using a Krell 50W class A clone chassis for that. I am doing about 100W into 8ohm and 200W into 4ohm, my bias current is 1A to get 16W of class A for 8ohm and only 8W into 4ohm load. Dissipation is over 60W/ch already.
 
Last edited:
Again thanks for your kind words.

One feature of the Technics New Class A is that they operate remarkable cool and efficient. The A5 is not a true Class A amplifier it is much more a very clever combination of a low power Class A and high power Class AB or B, hence much better total efficiency than a pure Class A.
Technics decided to call they're approach New Class A which is nothing more than a sales gig and code for saying we have got a very fancy Class A/AB over the more common Class AB.
The A5 operates with about 22mA idle current per transistor pair and totals to about 15W @ 30deg. C idle dissipation per channel. That is very reasonable considering this amplifier offers virtually Class A audio performance with unmeasurably low TIM.
The custom case will be quite large, there is going to be sufficient air for dissipation in normal operation.
 
Again thanks for your kind words.

One feature of the Technics New Class A is that they operate remarkable cool and efficient. The A5 is not a true Class A amplifier it is much more a very clever combination of a low power Class A and high power Class AB or B, hence much better total efficiency than a pure Class A.
Technics decided to call they're approach New Class A which is nothing more than a sales gig and code for saying we have got a very fancy Class A/AB over the more common Class AB.
The A5 operates with about 22mA idle current per transistor pair and totals to about 15W @ 30deg. C idle dissipation per channel. That is very reasonable considering this amplifier offers virtually Class A audio performance with unmeasurably low TIM.
The custom case will be quite large, there is going to be sufficient air for dissipation in normal operation.
Now you got me very interested. Can you post the schematic? If you are not willing to share yours, just post the original schematic. I want to see how they do it. There are a few patents on extending the class A region, Nelson Pass even had a patent on it in the 70s. I simulated a lot of them and none works as they all create their own distortion and ending up not improving. I heard Krell has their own version also. I tried on and off to design one my own, no luck so far.

Sunfire did something by making the rails follow the signal to keep the voltage across the amp low so they can make the amp class A. People are using class D to drive the rails and keep the signal amp in pure class A. I have not done any simulation, I bet there are pit falls on this as you modulating the rails and that will have adverse effect also.

I am doing FFT on my amp, Most of the distortion is generated in the OPS. BUT I tried adjust the bias up a lot, I can tell you the effect is not as much as I expected. My OPS has 9 pairs of transistors running about 120mA each, idle is about 1A or so normal, I increased to over 1.5A, THD lowered are not too significant at over 10KHz. Lower freq does not matter as it's covered by high loop gain already, I am already getting down to 0.002% at large signal.

That reminds me to test the frequency response of my amp, I never stop and tested it. I designed to be over 200KHz, I got over 30V/uS slew rate and rise time of less than 1uS. I go for high loop gain so I still have at least 20dB loop gain at 20KHz to lower distortion.
 
Last edited:
Hi Alan,
You can find the Technics SE-A5 service manual on the net free downloads.
See I am not so much into trying to reinvent the wheel. I'm lazy and figured I'd go with something proven and just add my spin on it but leaving the main circuit alone.
PS tracking using switch mode or Class D is popular I guess and there are many more approaches and variants one can think of.
Class S, Class AC come to mind to name a few interesting.
To my knowledge Technics developed 3 different variants they called:
New Class A
Class AA
Class A+
They are all different approaches to obtain Class A performance at much better efficiency than Class A. Each one by itself is verx intriguing and extraordinary clever in my opinion. I find particular the New Class A approach intresting. Long ago I learned simple is better, less stuff that can mess things up is better. Technics managed to achieve outstanding performance with only two output pairs.
Adding more output pairs lowers the dampening factor alright at the expense of frequency response or other parameters. Technics A5 is specd with DC to 200kHz @-3dB if I am not mistaken.
My simulation indicated THD 0.000% @ 1kHz. I guess that's the limit of what Multisim can do. Spice Simulation is cool and helps to prove the concept and to get things close but it does not account for the circuit board effects. As pictured above I measured about 0.5us group delay. I hear you, cranking up the bias is not the cure for improving THD. There are much more intelligent solutions to that such as error correction and cancellation.
A while ago I found some very inspiring ideas while surfing the net all aimed at lowering THD @ 20kHz. Perhaps if I have finished this project I might try out someting new. But first I want to finish this and measure THD to see where I am with this compared to the original A5.
You can find the service manual here:
http://www.eserviceinfo.com/downloadsm/54926/Technics_SE-A5.html
 
Looks like it's adjustable bias type. I have to study and do some simulation with LTSpice.

I found simulation with LTSpice is not very accurate. I used the model file by Cordell already. Both THD and stability are not that accurate. I know layout is very important, I layout as if it's RF circuits with ground and power planes, I design as if it's for hundreds of MHz. Closed loop feedback compensation is quite different from simulation.

Also, effect of what increase or decrease harmonics is different between simulation and real circuit. Simulation showed lower distortion than real circuit. Cordell put a lot of emphasis on Oliver's condition on bias, I design to oliver's condition, but when I adjust up and down, it really does not make as much difference as it cracked up to be.
 
Hi Alan,
I hear ya, that's what I meant the simulation gets you close. The last little bit is layout, actual parts and parasitics. Return path and loop area are most important so as the board material. I guess we share a common background.
When I run the simulation I knew the spice models that come with the library in Multisim are sub basic they get you going but that's about it. Trying to analyze some phenomenon at or below the noise floor is not going to be much successful with standard models. More refined models are needed to study the effects and temperature dependency as well. And sadly board parasitic are not part of the sim.
In the New Class A they opted for two independent idle current adjustments. One to set such that the transistors barely start to conduct and the other to set the actual idle current. Quite clever once you realize...
The Japanese did much much more with finesse than brute force a rediculous idle current on the transistors. Altho Japan seems pockmarked with reactors I think the Technics engineers were sensable and opted for finesse.
Designing the layout as if you would for RF is good but mind you some ground connections you do not want to connect straight down into your ground plane. You would still want to use a fat trace above your ground plane to meet the reference point.
My boards are two layer, top all signals and bottom solid ground.
Keep traces short where possible to keep parasitics low. Also over the years I learned that a certain amount of symmetry helps a lot. Of course I have seen some people taking this to an extreme and the wrong way by just locating parts neatly in some symmetrical fashion, but that's is not it.
I believe there is an advantage to keep the GNF reasonable instead of increasing it. It is better to reduce the distortion locally at the output and driver devices than trying to relay on the GNF. GNF will be too slow. By the time the feedback tries to correct the distortion it's already to late the 'damage' was done. Now the GNF is correcting to little and way to late, now actually making things worse and adding distortion.
Like I indicated simple is better (and much faster). 0.5us Group delay :D
 
Back
Top Bottom