Bipolar-MOSFET amp conversion: seeking opinions

[jeannotvh]

I am looking at converting an existing, functioning bipolar-output amplifier into a mosfet-output, and I am seeking opinions as to the best way to do this. I will be asked why I want to do this: 1- These amps are already in a large case with a ~1KVA power supply and massive heatsinks. 2-The input stages are high-quality, wide-bandwidth and already powered by their own supply. 3-The output of the input stages (at the bias-circuitry level) is obviously high current.

The conversion will involve replacing everything on the right of the bias circuitry by the MOSFETs (I'm looking at 3 pairs of 2SJ201 and 2SK1530) with appropriate resistors, and may be frequency-taming capacitors.

Also, I will need to modify the bias circuitry so it creates a larger voltage.

I need recommendations, or opinions on the approach. Included is the schematic of my canvas.

Thanks in advance!

[ostripper]

You seem the have the same Vbe as all my "mongrels" (2.2k/200+680/620r) therefore you must have 6-7ma VAS current. I ran my classic k135/j50's latfet's by just replacing the Vbe with a simple 1k resistor in parallel with a 25 turn 500R trimmer. At about 350R on the trimmer my fet's were biased at 100ma. I actually liked the simplicity of the arrangement , the sound was slightly improved ... but not much.

I would suggest you simulate the input stage with the latfet models. It should work well w/ proper layout and component values.

OS

[jeannotvh]

Originally Posted by ostripper

therefore you must have 6-7ma VAS current.

The service manual specifies 12-13mv across the .27 ohms emitter resistor, which gives about 44ma per output trans.

I ran my classic k135/j50's latfet's by just replacing the Vbe with a simple 1k resistor in parallel with a 25 turn 500R trimmer.

I assume the Vbe would be the resistor at the gate?

I would suggest you simulate the input stage with the latfet models. It should work well w/ proper layout and component values.

I was hoping not having to use a simulator, and just starting with adding a pair of outputs with the safest circuitry around it (and today's insufficient bias), and work my way from there.
But you're saying it can be done and has been done before. That's good news to me.

Thanks for the input!

[ostripper]

Originally Posted by jeannotvh

I assume the Vbe would be the resistor at the gate?

I was hoping not having to use a simulator, and just starting with adding a pair of outputs with the safest circuitry around it (and today's insufficient bias), and work my way from there.
But you're saying it can be done and has been done before. That's good news to me.

Thanks for the input!

I assume the Vbe would be the resistor at the gate?

No , the vbe would be everything between the collectors of q109 and q111.

OS

[ostripper]

BTW , I will have your voltage stage on a module soon , thanks for schema.

OS

[jeannotvh]

Originally Posted by ostripper

No , the vbe would be everything between the collectors of q109 and q111.

OS

I'm sorry, I just realized you were talking about the bias circuitry. I see now!
Thanks!

[jeannotvh]

Is the Bias circuitry really the only thing I need to worry about?

[ostripper]

Also , 220R to 560R gatestoppers to SK/SJ devices.And zener gate protection..

Look below for example. PS... you don't need the 56pF's , that is just how this amp is compensated.

OS

[ilimzn]

Originally Posted by jeannotvh

The service manual specifies 12-13mv across the .27 ohms emitter resistor, which gives about 44ma per output trans.

This shows some misunderstanding about the circuit.
VAS = voltage amplifier stage, which is basically the transistors above and below the bias set pot.
10mA or so is NOT enought for your selected MOSFETs, even with only one pair.
You can't just simply replace everythin right of the bias generator with no less than 3 pairs of MOSFETs as that's loading the VAS with a capacitive load two orders of magnitude larger than it's originally loaded with - the result, it's bandwidth will become abysmally low, which i would consider a big problem even all other things (frequency compensation?!) being equal.

The conversion itself is questionable on several levels, if we are talking a large amp:
1) Idle heat increases by about 3 times per output pair. Optimum bias for your chosen parts is about 150mA per pair. You can get away with less but with reduced performance.
2) Headroom decreases - even when the output stage is driven from a separate supply which has a higher voltage. This is because the input capacitances of the MOSFETs are non-linear, and become very nonlinear when Vgd or Vds is low (below 5-10V depending on actual parts). This translates to a rather drastic increase in HF distortion as maximum power is approached. The problem can be lessened by higher drive current.
3) Lower transconductance of the MOSFETs compared to BJTs increases output impedance and decreases damping factor. If the drivers have too low a current, the problem increases drastically towards HF. It may be a moot point as feedback will lower output impedance, but the excess gain available for feedback may quickly be exhausted at HF. In nay case, the actual distortion profile will be more dependant on the changing load impedance. Again, measures can be taken to lessen this problem.
4) Unless you match MOSFETs, it's more difficult to insure good current sharing in MOSFETs of the type you chose (lateral MOSFETs are an exception to an extent). They require higher value source resistors (to replace the emitter resistors for the BJTs), which further reduces already lower transconductance, so you are back to problem (3).

On the plus side, if you do things right, you can get more speed and better feedback performance, as well as a lot more robust output stage. Soundwise, opinions differ so i will not comment on that.

So, you just have to do things right.
First of all, do NOT remove the driver stages completely. I would even go so far as to leave them as they are now, perhaps optimize the current. just replace the output BJTs with MOSFETs. Remove the base chokes in parallel with the base resistors, increase resistors to 33-100 ohms, these will be gate resistors now. Select output transistors, and be sure to use non-inductive source resistors. Adjust the components around the bias servo for slightly increased bias voltage - the Toshiba MOSFETs you chose have a low treshold voltage, so the adjustment should not be too large. You need to increase the C-B resistor on the bias servo transistor, by about 15-20%. See how it fares as it is (replace trimmer with 10-turn unit as suggested earlyer), you may not need to change anything.

[jeannotvh]

I jumped in this idea with more misconceptions than I can shake a stick at.

Why the 10-turn trim pot? Is for ease of precise adjustment, or stability?

I'm glad I asked for opinions here, thanks guys!!!