C3g Aikido Phono

[simon]

Here's R3 with a stepped load from 30R to 70R in 10R steps.
30R green
40R blue
50R red
60R cyan
70R magenta
.

BoozehoundLabs Pre-Preamp Sim 2 stepped Source resistor.png (194.95 KiB) Viewed 1226 times

.
Interesting that much less than 50R dramatically reduces gain, increasing it doesn't change (decreases) it much.

[simon]

Presumably to parallel the fet each will need a gate and source resistor? I can't quite work out in my mind before simming this whether it will increase the gain, or improve noise. Possibly both?

[Nick]

The gate resistor will be needed for each one. in case you don't know, its job is to protect the fet if the input is sent too +ve. The gate/source junction will become a forward biased diode (as shown in the symbol) and its to prevent excess current in that case. This is one of these bits of information that's hard to find, once you know it its obvious, but the net is full of people assuming (and stating confidently) that its to prevent instability as it would be in the case of a valve.

The source resistor, I would use 10R or so source resistor under each fet, then join them together and use a single resistor under those.

The main advantage is lower noise.

[Mike H]

Screenshot 2024-02-14 192036.png

What's this analysis you did Mike? (The sine wave jobbie.)

Where you've got your '.ac oct ...' etc.

Right mouse click on it, change it to '.tran 1' (transient, 1 second - or 5 seconds, or even a fraction of a second)

From the help file -

Perform a transient analysis. This is the most direct simulation of a circuit. It basically computes what happens when the circuit is powered up. Test signals are often applied as independent sources.

Syntax: .TRAN <Tstep> <Tstop> [Tstart [dTmax]] [modifiers]

.TRAN <Tstop> [modifiers]


LTspice's window for it: Menu > Simulate > Edit Simulation Command
-

Screenshot 2024-02-15 182548.png (41.61 KiB) Viewed 1167 times

-


If there is no simulation command on the schematic, more usually because the schematic is new and not been run yet, you should automatically get this window when you try to run it.

You can delete the run command or turn it into a comment, and the window will appear again.

Hope that makes sense.

[simon]

Thanks Mike. There's a lot to Spice, I've only scratched the surface so far. Just spent some time looking at noise...

[simon]

The gate resistor will be needed for each one. in case you don't know, its job is to protect the fet if the input is sent too +ve. The gate/source junction will become a forward biased diode (as shown in the symbol) and its to prevent excess current in that case. This is one of these bits of information that's hard to find, once you know it its obvious, but the net is full of people assuming (and stating confidently) that its to prevent instability as it would be in the case of a valve.

The source resistor, I would use 10R or so source resistor under each fet, then join them together and use a single resistor under those.

The main advantage is lower noise.

I didn't know, no, thanks.

I've had a go at modelling noise, though TBF just because I've produced some graphs it doesn't mean they're necessarily correct!

First, I added a second fet to the Boozehound
.

Boozehound Parallel.png (103.99 KiB) Viewed 1141 times

.
I gave it its own drain resistor too, rightly or wrongly.

So, noise for the TubeCAD
.

TubeCAD Noise.png (142.53 KiB) Viewed 1141 times

.
Next the Boozehound
.

Boozehound Noise.png (155.43 KiB) Viewed 1141 times

.
Can't seem to upload another image, so on to the next post.

[simon]

And finally the paralleled Boozehound noise
.

Boozehound Parallel Noise.png (170.2 KiB) Viewed 1140 times

.
It's a bit tricky comparing them so I scaled off the plots and graphed in Excel
.

Noise.png (141.16 KiB) Viewed 1140 times

.
It's been really interesting investigating noise. Paralleling the fet makes a big difference, presumably the law of diminishing returns sets in pretty quickly though?

The TubeCAD sims better than the paralleled Boozehound below 60Hz though. Whether this is important though I don't know.

I wonder how the paralleled Boozehound compares with step up trafos? Only one way to find out I suppose.

[Nick]

Yep the noise will reduce for every doubling of the FETs. I think there is a error though. The 40r source resistor is shorted out

[simon]

So it is. Nuts, it took ages to read off the noise levels.

[simon]

Updated for the shorted 40R resistor, a little light relief at lunch. Interesting results that needs more investigation. Noise was better with the shorted common resistor, but maybe gain was affected?
.

Noise2.png (162.59 KiB) Viewed 1060 times

[Nick]

Yep, you maybe need to compensate for any gain change. Its the S/N ratio that matters in this case. Also, you tend to get that sort of noise/f curve depending on the type of noise source (thermal, shot, partition etc ). Generally paralleling devices will reduce noise proportionally to square root of the number of devices.

I can send you some LSK170's if you want to try that circuit.

[simon]

Well, the gain had dropped around 3dB. Changing the lower resistor R9 to 20R was about optimal for gain, but the gain drops really quite quickly below this so I used 22R for a little safety margin in sim inaccuracy. This sims as a gain of 28dB, so just over 9x.

Then I tried increasing the coupling cap C1 from 10nF to 20nF - this decreased noise below 200Hz, significantly so below around 60Hz and puts it closer to the TubeCAD. TBH I only used 10nF as the smallest capacitance I could get away with whilst keeping roll off acceptable. It might be interesting to try 100nF or even 470nF.
.

Noise3.png (154.16 KiB) Viewed 1031 times

[simon]

I can send you some LSK170's if you want to try that circuit.

Well I think it would be rude not to, thanks . I'm kinda fascinated to find out how this sounds now.

[simon]

Doubling the voltage to 24V adds another .7dB, hardly worth it.

I suppose I should add another fet to see what the effect is, it's a bit of an effort with LTSpice though as I don't think you can copy groups.

[Nick]

Send me your address