Guesstimating lag comp capacitor value

[pooge]

Is there any easy way to predict a starting point for guestimating a lag cap value to minimize the amount of cap values to purchase for optimization?

For example, Leach specifies a 15pf value for his amp, which has a 8.5mHz GBP. The VAS caps have an ft of 15mHz, and a Ccbo of 25pf.

2N5415

If I were to replace that VAS transistor with another one of higher ft and much lower Ccbo, is there a rote way of getting a good estimate of the needed comp cap?

I'm looking to replace Leach's 2N3439/2N5415 VAS pair with a Toshiba 2SA1360/2SC3423 pair, with ft's of 200mHz and Cob's of ~2pf

Any wisdom appreciated.

[wg_ski]

fT only matters a little, Cob matters a lot. If the original transistor had a Cob of 25pF and an external cap of 15pF, then you just need a total of 40 pF. If your new transistors are 2pF, then you should add 38. Your choices are 33 (maybe 36) and 39. The new VAS will perform better not strictly because of higher fT and lower Cob, but because of lower Cob variation with signal swing. The portion inside the transistor changes with voltage, the external part does not (if you don't use a class 2 ceramic).

[pooge]

I believe the internal Cob is also affected by gain. In other words, I didn't think you could just count on the specified Cob without considering the multiplying effect on Cob due to the placement of the transistor in circuit.

[AndrewT]

Hi,
start with 100pF to ensure stability.
Then reduce in E6 values.
It should start peaking on slowish squarewaves, and may show some ringing.
When you get to this stage go back up one E6 value and try listening to the result.
Fine tuning, I'm told, should move you on from there.
But, that bit I'm not too good with, so don't listen to me.

This means you need 22pF, 33pF, 47pF, 68pF, 100pF as a minimum.

[pooge]

Appreciate the suggestions, Andrew.

No doubt it will work. However, I was hoping to find a good close approximation to avoid too many replacements, as the Leach supplied boards are single sided and don't have plated through holes. There is only so much heat these boards can take before pad lifting.

[wg_ski]

Originally posted by pooge
I believe the internal Cob is also affected by gain. In other words, I didn't think you could just count on the specified Cob without considering the multiplying effect on Cob due to the placement of the transistor in circuit.

Cob is not affected by gain. The effective input capacitance is. If the gain increases (by loading, change in Early voltage, etc), so does the input capacitance. But open loop GBW and slew rate will remain relatively constant with a given Miller cap value. If you wanted to get the same volatge gain back (you don't need to) trim the emitter resistor value until you're satisfied with the result.

[sreten]

Originally posted by pooge

If I were to replace that VAS transistor with another one of higher ft and much lower
Ccbo, is there a rote way of getting a good estimate of the needed comp cap?

Hi,

Basic Vas compensation is beating it round the head with a baseball
bat so it is so sluggish that loop gain gain is low when the output
starts complaining about the input being too fast for it, or in other
words Cdom depends on the output stage bandwidth / ft.

This does not change with faster Vas transistors, Cdom is the same.
Cdom is the transistor capacitance || with the external capacitor.

Seems you are looking at 40pF to 50pF total, the latter more stable.

/sreten.

[darkfenriz]

Calculate cap value for feedback factor of 30dB at 20kHz

[wg_ski]

That's just another way of stating "Set GBW at 20 MHz". I thought Leach set his a little lower, say 4 to 10 MHz. Too much above the output trannies' fT and non-dominant poles start eating into the phase margin.

[AKSA]

I'd go with Andrew's suggestions, though start at 68pF, and work down.

However, I'd add listening tests to your assessment.

Select a busy track with a female vocal. Humans really know voices well, it's instinctive.

Too much will sound leaden, muffled, slow.

Too little will be bright, sibilant, and 'hot'.

It takes a little practice but using E12 values you can generally home in on the two limits.

Select midway values, trimming either way a few pF, over a period of days of assessment. Then check the CRO. Hosfelt Electronics in Steubenville OH have an excellent range of cheap silver micas for this task. A little ringing is acceptable, and you should not aim for perfect square waves, some rounding is best, but this method works well.

It is irritating and unscientific to have to do it this way, but LC is not easily calculated because you are dealing with tiny capacitances and layout has profound effect. On one single sided board I did, I finished up with 6pF of phase lead. When I moved to double sided, there was now a pad on both sides of the board. After changing phase lead to 3pF it all clicked into place. You could clearly hear the difference.

I have found PSpice useful to establish limits, but of course it cannot account for layout. The final tuning, and certainly the most sonically pleasing results, are achieved empirically.

Hugh