|
|
PartTimeProject Clone of the Krell KSA-50
In the end, after about a year, I ended up with a powerful class A amplifier, about 50 watts per channel, that has a great sound and can drive any load.
Key parameters are 37 VDC per channel, four 64,000 uF caps, so 128,000 uF per channel, 4 pairs of MJ21193 and MJ21194 per channel, fan cooled.
Of course, this amp is not a "clone" of the famous Krell KSA 50 in the sense that it uses different circuit boards, it has my goofy parts placement and hand-soldered wires vs. the original PCB connections, and more modern parts (such as the output transistors). It does, however, have the same basic amplification circuit, and more capacitance in a bigger power supply. It runs at the same voltage, and because I have twice the number of output transistors as the original, my version will hopefully outperform the original in the power department.
In the end, who really knows how close it is to the original, all I can say is that it was fun building and really performs well.
I'm going to go through this in sequence rather than posting a chart like with the other web pages. Just trying something different, the images will open in their own window.Update 1/07
Low-Quality Movie Added.
Click Here For a QuickTime movie of the amp running with the fan controller cycling through the temperature zones.
Update 2/06
Resistors To Power Supply |
 |
 |
 |
I've updated the amplifier as follows: (a) new on/off switch & AC inlet, (b) increased thermister temporary load for soft start (hence the reason for (a)!), (c) 10K resistors across the power supply capacitors to drain the capacitors after power is turned off, (d) Solen 1 uF decoupling capacitors across the main power capacitors to reduce the effective ESR of the capacitor bank. I also (e) changed the power supply capacitor grounding a small amount. Most important change- (f) the fan mounts have been spaypainted yellow to match the rest of the internal metal! |
Prior Version
First, some pictures of the completed amplifier. You see that I borrowed some inspiration from the original nameplate of the KSA-50, as well as the triangle of lights from newer Krell series amplifiers. You can tell this is a Krell derived product even though the amp itself does not even closely resemble a Krell. I've got a computer fan controller in here to sense and display temperatures throughout the unit and control fan speed. Side and bottom panels are black walnut, as with the rest of my DIY projects.
| Zoom Front Panel (click to enlarge) | Angle View (click to enlarge) | Perspective For Scale (click to enlarge) | Side View (click to enlarge) |
 |
 |
 |
 |
You see the Aerogate 3 Fan Controller (review). Kind of feel like one
of those computer modders with this thing. Had to hack into it and make a 12V and 5V regulated power supply to
power it up. Using this, I can adjust each of the 2 fan speeds separately (can have up to 4 fans) and have status
information of 4 zones of temperature. The back panel is pretty plain, using WBT RCA connector, some binding posts, and
a switched-fused-AC inlet. Not sure what the case used to be, but you see that it mentions some serious voltages..
Some more pics.
 |
 |
 |
 |
 |
After building the amplifier, I did some testing to get some data.
 |
 |
 |
 |
Biased 60WPC Class A Graph Shows Heatsink Temps | 20 KHz Square Wave 5 ohm load 10V/div. |
2 KHz Sine Wave 3 Ohm Load 10 V/div. V+ on Top Trace |
40 Hz Sine Wave 5 Ohm Load 10 V/div. V+ on Top Trace |
The data looks good. The amp has no problem putting out 25 V pk-pk into 3 ohms, and there is not any visible oscillation or "ringing" on the square wave response. As I did not put in any current limiting circuitry, I'm not going to test maximum output into less than 3 ohms for saftey reasons. As you can see the temperature of the output heatsinks goes up to the 60C range and slowly increases thereafter. I did not measure the room temperature, but it was also going up slowly. So I think that after 3 hours we can say that the output heat sinks reached thermal equilibrium, only getting hotter due to the room getting hotter. Lowering the Bias would obvious lower the heat, I may measure temperature over a few hours at a lower bias setting someday.
OK, starting out the project, I had to find a case and some big heatsinks for fan cooling. Found them at a surplus store after many hours of looking and on-line shopping. Turns out that this case is awesome, all panels come off independently for easy access. The front panel is perfect for attaching a custom faceplate, and the case is very very strong when all panels are inserted. I wanted to ensure adaquate spacing below the amp for fan air intake so used large feet fastened to a wood baseplate. I installed steel furniture threaded feet so its height can be adjusted and it can be made level.
 |
 |
 |
I had to make the back opening fit a IEC inlet/fuseholder, so used a file as shown. I also had to make a bottom plate for the case, and get some other metal pieces for shielding, so these were ordered from Metal Supermarkets .
Next was drilling and filling the case. Yeah, my drill holes are not that nice, but you try drilling over 100 large holes in 4mm thick aluminum in a day.
(click to enlarge) |
click to enlarge) |
(click to enlarge) |
 |
 |
 |
You now see some parts showing up. Here is what they are:
| Transformer | two 700 VA dual 28 V secondary Transformers from Victoria Magnetics. |
| Capacitors | four 64,000 uF 85 C computer grade from JEA Capacitors |
| Bridge Rectifiers | two 400V 35A, part MB345-ND from Digikey |
| Output Transistors | 8 each of MJ21193 and MJ21194 (4 pairs per channel), from Digikey |
After figuring out where everything was going to be in the case; I made my own soft start board. The soft start board is shown in more detail on the the Soft Start Web Page. The softstart board was set for 12VDC to power the relay. However, I'm using a computer fan controller that you can see in some of the pics, and it needs 5 VDC as well as 12VDC. So I put a 5 VDC regulator on the soft start board-- connected to the 12V regulator. Just hacked into my own board as shown in the pictures below. The components on the underneath of the board are for the added 5 V regulator. Also shown are the steel shielding plates after they were painted yellow. Why? To make them look better, the plain cold rolled steel was kind of rusty.
(click to enlarge) |
(click to enlarge) |
(click to enlarge) |
 |
 |
 |
The soft start board was then tested and worked. I then finished up the L/R input driver boards; shown below.
(click to enlarge) |
(click to enlarge) |
(click to enlarge) |
 |
 |
 |
I spent a few days hooking all this up and testing one driver board. It appeared to work based on the DC values matching those in the Wiki. This is a good thing, because as you can see from the first picture, available parts did not quite match the "specified" parts. I got the closest thing I could get at 1% tolerance (for resistors).
Before turning to the output stage, I designed the front panel-- so while it was being made I could work on the output stage. As I normally do I downloaded the software from Front Panel Express and designed the panel as I wanted it to be. This time I wanted a Krell style badge and a triangle of Krell style LEDs. I also wanted room for the fan controller. After carefully measuring everything down to .1 mm, I sent away for the panel and it came back....perfect! You really feel good after spending all that design time getting something made by someone else and it turns out that your measurements and assumptions were accurate.
(click to enlarge) |
(click to enlarge) |
 |
 |
The output stage was a lot of work, I worried about having live heatsinks but ultimately decided that it would be ok. Every metal piece that comes in contact of any part of the heatsinks (except the transistors of course) is insulated using various methods. Hopefully no one will short out the heatsinks, but we will have to wait and see. Detaching the output stage from the input driver boards means that there are a LOT of wires that need to connect between the two. You need, at a minimum, 8 connections: Vin+, Vin-, VCC+, VCC-, sense B, sense E, sense C (from sense transistor), Vout. I also added (not shown in these pics, but visible in steps 5 and 6 below): VemitterResistor+, VemitterResistor- (for external bias probing), and Ground (to ground some shielded cables).
(click to enlarge) |
Picture 2 (click to enlarge) |
 |
 |
Here are some pics of testing of the various parts. With all the external wires you need to connect everything its very difficult and time-consuming to hook this up for testing.
(click to enlarge) |
(click to enlarge) |
(click to enlarge) |
 |
 |
 |
You see some things are painted yellow. I used steel plates to separate the case and hold the boards, but they were a little rusty and nasty looking, so I gave them some color. Hope you like it. Also, here is a cutaway of the amp internals after many steps of assembly and testing.
(click to enlarge) |
(click to enlarge) |
(click to enlarge) |
 |
 |
 |
Finally, assembly of the entire amp. I added wood side panels, threaded the case for mounting bolts and added a wood bottom panel. Test points from two of the emitter resistors are brought out to the back panel to aid in setting bias. I'm not sure what kind of connector I have back there, it might be for a tube or something; got it cheap at the surplus store. The back panel can be probed and bias set without needing to get into the output stage itself. It still looks messy inside but I'm willing to live with it like that. The case is flexible but the 2nd floor on the front prevents access to the faceplate bolts. I need to
(click to enlarge) |
(click to enlarge) |
(click to enlarge) |
 |
 |
 |
(click to enlarge) |
(click to enlarge) |
(click to enlarge) |
 |
 |
 |
(click to enlarge) |
(click to enlarge) |
(click to enlarge) |
 |
 |
 |
(click to enlarge) |
(click to enlarge) |
(click to enlarge) |
 |
 |
 |
Free Site Counter