Krell KSA 100mkII Clone

[Flodstroem]

Originally posted by TomWaits
I've been thinking about my build attempt at this amp and I really think the pre/driver board from Watts/Gulbrandsen should have a regulated supply for optimum performance. I'm not certain I can draw up a good performance circuit for this task. I'm not even certain of the power consumption of the driver board on its own?

So I'm wondering if there is someone out there that has a schematic up to the task. Also, is there trouble if the pre/driver board circuit has a separate power supply including a separate transformer? Any advice to would be great!

Cheers,

Shawn.

Shawn, this could be a starting point to go for. This regulated power supply did Krell use in the KMA160 and could easilly be modified to suit the KSA100. Guess there is a simple task to switch the zeners for to get lower voltage for the drivers (60-70V)

Regards

[AndrewT]

Hi Flod,
using the 1.2Vgs gives about 20V across drain-source of the FET.
I too am not sure if this is close or far from an incircuit measurement.
But, thanks for the response.

The reason I ask, is trying to get a feel for how voltages and currents around the FET produce voltage swings in the next two stages of the amp.

I plan to use 3.5mA ccs for D5 & D6. I may insert a 47k to 51k from rail to D5 to shunt 0.5mA past the FETs to sink and thus keep the FETs at their nominal 3mA current. But, this may not be necessary, it depends on tolerance (+-16%) of the ccs and gain through the FETs.

[TomWaits]

I read your posts guys and got to thinking a bit more about regulation.

If the driver board has its own power supply including transformer, that in it self could provide a tremendous amount of regulation as the output stage would only be pulling down on its own supply.

I guess the fault is the AC mains fluctuations as the output stage power supply surges in demand. I assume it would in turn affect the AC mains supply feeding the driver stage power supply?

PWatts, your regulation scheme sounds awesome but it may be a bit over the top for me. It would require some cutting of the pcb trails I assume and perhaps a few component changes or removal. I guess the first two stages as they are on the driver board now have their own zener diodes for regulation. As Mr. T asks would you use these zeners or remove them?

Mr. T, are you suggesting the entire driver board will consume less than an amp of current under any situation? I thought the PWatts/Gulbrandsen boards would draw much more than 550mA?

I have these 1kVA transformers that will provide 60-0-60 VRMS DC rails. Of course I want to use them on the output stage only. A little high, I know but all should be fine. Just wondering if the separate power supply feeding the driver board can have a slightly different rail voltage? Should it be different and if so what would be ideal?

Thanks Flodstroem for posting the KMA160 regulator. I recall now that you folks were discussing it here before. If the current demand is with in an amp or two, how bad would it be to try using one of those IC's from Nation Semi/Fairchild?

Besides having good KSA100 build quality I feel the regulation for the driver board could be the next most important factor that would influence performance and listening.

Cheers,

Shawn.

[PWatts]

Feeding the drivers from a separate supply will definitely help, but the LTP is far more important. The supply I'll be using is a 100VA per channel that will supply the drivers, frontend and LTP, and depending on the results which ones would be regulated, tapped directly from the DC or subregulated etc. All in all I think a simple zener-follower such as the KMA160 is sufficient for the drivers while something a little more exotic may suffice for the rest.

AndrewT, I plan to experiment with lots of different permutations but the idea would be to leave out the 1k resistor and 39V zeners and replace it with a proper regulated stage, give the rest of the front-end their own (i.e. leaving the jumpers open), and depending on the results the drivers either unregulated (still from a separate supply though), zener-regulated or extensively regulated.

Nothing wrong with the idea of the nice IC supplies from Natsemi except for the issue that high-voltage negative supplies are always hard to find. Positive there are lots of excellent solutions but negative is always the problem. When I get the time I still want to design a small split-rail halfbridge switching supply, but for the time being linear regulation would have to suffice.

[AndrewT]

Originally posted by AndrewT
LTP + DC offset stage requires 15mA.
VAS + prior to driver requires 21mA.
Drivers varies with load current. Quiescent about 100mA (50+50 for the pair) but can go to 500mApk for a 10A to 15Apk output to the load. 500mA is adequate for a 4r0 load, anything less than 4r0 and more driver current will be drawn.

If any of the component values or supply voltages are changed then all the preceeding current values must be re-calculated.

Originally posted by TomWaits
are you suggesting the entire driver board will consume less than an amp of current under any situation? I thought the PWatts/Gulbrandsen boards would draw much more than 550mA?

Hi Tom,
as shown above the front end excluding the drivers consumes just 32ma to 37mA from +-52Vdc supplies. With unregulated these current values move about. With regulated they are fixed fairly closely. The signal will cause a little bit of current modulation. I can't sim it, but I suspect it wll be well under 10%, so absolute max from a regulated supply will be about 38mA.

The drivers are a completely different story.
Quiescent draw is about 100mA, but this depends heavily on the Vre across the drivers' resistors. As the bias voltage is changed Vre changes with it. More output bias also gives more driver bias.
Then one has the driver output current feeding the bases of the output stage. Again more output bias leads to more driver output.
Once the amp exceeds the ClassA of the drivers then they move into ClassB and the current in the drivers nearly follows the output current divided by the output stage current gain (~=Iout/hFE). I showed values for 4r0. 2r0 would be roughly double. A severe 4ohm speaker could get close to the worst figure that 2r0 predicts (3r0 for less severe 4ohm speakers), but as you know the KSA100 is reputedly able to drive all the way down to 1ohm. That puts an enormous load on the drivers and presumably why Krell doubled them up.

Decision time. Design for 1ohm or 2ohm or 4ohm and then provide that level of current to the drivers, regulated or not.

[PWatts]

AndrewT is 100% correct. Besides the fact that high-current regulators are harder to design (transient response being one of the problems), they are related to the bias as well. All in all I don't think they will benefit much from regulated supplies, and can even degrade performance.

I've simulated different implementations of regulation by changing the ideal DC supplies with random fluctuating DC with heavy 50Hz and harmonics of it superimposed. Regulating the LTP makes a big difference as is to be expected, but it should be carefully designed. At high frequencies proper regulators work well, but at low frequencies the zener actually can work better if one isn't careful. Changing the driver supply to an ideal voltage source didn't make a large difference, except when pushing the output to almost clipping levels, as one would expect.

Throwing out the entire DC offset circuitry had a very big impact on the distortion. The offset actually wasn't too bad without it, but it would be a bit risky in real life. Adding a DC servo made it much better of course, and raised the distortion significantly less than the standard circuit.

[Jozua]

Mark


I received the boards and diodes today. Many thanks for the excellent packaging. I do appreciate it.

Hopefully we can kick-off this project in a couple of days.



Jozua

[Jozua]

Mark

I nearly forgot, do I owe you money for the extra postage ? I seem to remember a smaller amount was quoted for postage.

Let me know.


Jozua

[jacco vermeulen]

Originally posted by PWatts
Adding a DC servo made it much better of course, and raised the distortion significantly less than the standard circuit.

Glad at least one here agrees, a JC-1 DC servo does wonders for a KSA50.

Look Ma, no Caps =>

[PWatts]

Jacco,

Good to hear that the servo improves performance in reality. Did it make an audible difference too, and to what extent? (i.e. is it worth hacking the board besides achieving perfect offset). A pretty solution would be cutting a fibreglass/aluminium board the same size as the KSA board. Build the servo on a piece of veroboard and mount that on the board, and then mount this board on the standoffs. Then mount the KSA on top of that with another set of 10mm standoffs. This way the mod would be invisible, fairly close to the board and only raises the KSA board by 10mm from the heatsink or whatever substrate it's mounted on.