The Leach moving coil pre-preamp

Here's what I have now. All signal grounds are left of the vertical blue line. Ground trace forms a large "E" shape.

gnd-a.jpg

Trying to achieve a true "star" topology isn't very feasible, given parts placement w/in board size limits, but this is. Is this any better than above?

gnd-b.jpg

Or a copper pour tying all the ground points together akin to this? Either on top or bottom board surface.

gnd-c.jpg

I don't see much difference between the first 2, but the copper pour might be good?
 
Here's what I have now. All signal grounds are left of the vertical blue line. Ground trace forms a large "E" shape.

View attachment 1308799

Trying to achieve a true "star" topology isn't very feasible, given parts placement w/in board size limits, but this is. Is this any better than above?

View attachment 1308800

Or a copper pour tying all the ground points together akin to this? Either on top or bottom board surface.

View attachment 1308802

I don't see much difference between the first 2, but the copper pour might be good?
Yes, I'd go for the third one.
 
Thanks Wyn, that's what I thought too.

I see no reason to cover the entire board as there aren't any more ground points.
 
Well, I thought I was done with silly questions, but while compiling a parts list for ordering, something cropped up.

I wanted to have a range of resistors on-hand for different gain settings, so ran LTSpice sims to determine best values, and discovered a capacitor phenomenon I wasn't expecting.

Turns out the value ratio of base caps (C4, C5) to output caps (C6, C7) has a significant effect on infrasonic frequency response (peaks of up to 0.7dB+ in the 1-4 Hz range), which get worse as gain decreases.

ie, for a given value of C4,C5, higher values of C6,C7 are needed to prevent the peaks. And vice-versa, for a given value of C6,C7, lower values of C4,C5 are needed to prevent the peaks.

Value pairs with flat response at high gain, exhibit increasing peaks as gain is reduced.

I didn't expect C4,C5 to affect frequency response but they obviously do, so my questions are:

1) What purpose do C4,C5 serve? (from Leaches descriptions, I thought they were just constant energy sources for transistor turn-on).

2) Is the absolute value of C4,C5 (100uF) critical? or would 47uF or 33uF work equally well?

And regarding the RFI suppression caps (C2,C3,C8), is the 100pF value important? or would 220/390pF work as well?

I've got lots of 220/390pf polystyrenes in my parts collection, but no 100s :(
 
Well, I thought I was done with silly questions, but while compiling a parts list for ordering, something cropped up.

I wanted to have a range of resistors on-hand for different gain settings, so ran LTSpice sims to determine best values, and discovered a capacitor phenomenon I wasn't expecting.

Turns out the value ratio of base caps (C4, C5) to output caps (C6, C7) has a significant effect on infrasonic frequency response (peaks of up to 0.7dB+ in the 1-4 Hz range), which get worse as gain decreases.

ie, for a given value of C4,C5, higher values of C6,C7 are needed to prevent the peaks. And vice-versa, for a given value of C6,C7, lower values of C4,C5 are needed to prevent the peaks.

Value pairs with flat response at high gain, exhibit increasing peaks as gain is reduced.

I didn't expect C4,C5 to affect frequency response but they obviously do, so my questions are:

1) What purpose do C4,C5 serve? (from Leaches descriptions, I thought they were just constant energy sources for transistor turn-on).

2) Is the absolute value of C4,C5 (100uF) critical? or would 47uF or 33uF work equally well?

And regarding the RFI suppression caps (C2,C3,C8), is the 100pF value important? or would 220/390pF work as well?

I've got lots of 220/390pf polystyrenes in my parts collection, but no 100s :(

See post 19 etc. concerning the peaking at LF. It's hard to explain in words, but here goes.
Essentially the problem is that the gain consists of the gm of the input devices into a load that consists of the parallel combination of the common load R (with C6/C7 in parallel) and the internal combination of the current setting Rs in series with the base caps in parallel. The current setting Rs are fixed, so as the set gain falls (the post output cap resistor gets smaller) the gain difference between the "internal" gain and the "external" gain increases, and at some point the extra internal gain is "exposed" to the output causing a peak.
As the feedback R/base R actually behaves like an integrator with feedback to the bases of the input devices you want to have the pole for this high enough that the peaking is reduced due to the loss in gain through the transistor (it essentially is unity without the base caps due to the bias Rs providing 100% negative feedback). In effect, you can use the base cap/output coupling cap ratio to set the bandwidth LF cut off of the amp.
So, let's see what happens when the common load R is reduced to say, 6k- with the same caps the response is +0.7dB at about 2.5Hz.
Halve the base caps and the peak disappears, and the -3dB point is c. 3.5Hz. Half the caps again and the cut off is about 7Hz.
So, the absolute value is not actually critical- but the function IS critical, but only in the sense that it can be used to tune the LF cutoff, and that depends on the gain setting circuit. Run simulations. See what you get. Yes, you can develop equations for this, but why bother?
As far as the 100pF caps are concerned- 220 pF is fine, 390 pF is fine- although the RF starts to come up a bit in the 2MHz or so region. 0pF is also fine in simulation. 390pF will provide an extra 15dB of attenuation at 10MHz compared to 0pF, and about the same as 100pF.
Whether that matters or not- who knows.
 
See post 19 etc. concerning the peaking at LF. It's hard to explain in words, but here goes.
Essentially the problem is that the gain consists of the gm of the input devices into a load that consists of the parallel combination of the common load R (with C6/C7 in parallel) and the internal combination of the current setting Rs in series with the base caps in parallel. The current setting Rs are fixed, so as the set gain falls (the post output cap resistor gets smaller) the gain difference between the "internal" gain and the "external" gain increases, and at some point the extra internal gain is "exposed" to the output causing a peak.
As the feedback R/base R actually behaves like an integrator with feedback to the bases of the input devices you want to have the pole for this high enough that the peaking is reduced due to the loss in gain through the transistor (it essentially is unity without the base caps due to the bias Rs providing 100% negative feedback). In effect, you can use the base cap/output coupling cap ratio to set the bandwidth LF cut off of the amp.
So, let's see what happens when the common load R is reduced to say, 6k- with the same caps the response is +0.7dB at about 2.5Hz.
Halve the base caps and the peak disappears, and the -3dB point is c. 3.5Hz. Half the caps again and the cut off is about 7Hz.
So, the absolute value is not actually critical- but the function IS critical, but only in the sense that it can be used to tune the LF cutoff, and that depends on the gain setting circuit. Run simulations. See what you get. Yes, you can develop equations for this, but why bother?
Thanks for that explanation Wyn, makes sense to me now. This circuit just doesn't lend itself to widely adjustable gain, it should really be optimized per cartridge.

A couple of things occurred recently that got me off on this track. 1) had some unexpected cash come in, so decided to put some of it into this project and do it right w/better parts, enclosure, hardware, etc. 2) while rummaging through my parts to see what I already had on-hand, came across an old rotary switch, which got me thinking about selectable gains.

So I set about in LTSpice to see what gain resistor values would be needed for 10 x 1dB steps, and that's what exposed the FR issue. I know you had brought it up earlier, but I had forgot about it while concentrating on the 34db gain.

I've run all the sims now, and I can still do it but only w/in a 6-7dB range before the cap values interfere. The plots below are all run with 22uF output caps (the largest physical size that readily fit in the box), and 4 different values of bias cap:
  • 100uF - green
  • 68uF - blue
  • 56uF - red
  • 47uF - cyan
First is at 34dB gain. The 100uF remains essentially flat to below 20Hz, but the others show some droop, with the 47uF worst at @ 0.35db down.

34db gain, 100,68,56,47u.jpg

The droop improves as gain is reduced and eventually all go flat to 20Hz, down to 28dB gain, which is the last good looking plot.

At 27dB gain, the 100uF cap has pronounced peak, @ 0.4dB, but the others still look good enough.

27db gain, 100,68,56,47u.jpg

At 26dB gain, even the 68uF is peaking, with the 56 close on its heels.

26db gain, 100,68,56,47u.jpg

I won't bother to show 25dB.

If I had my druthers I'd go with 68 or 56uF, but neither is available in the better audio caps (Silmic or Muse), and I don't know what characteristics are most desirable for this use: low ESR, low impedance, DA, DF, etc. so I'll probably stick with 100uF. Do you have any thoughts on polymer electrolytics? They seem to spec well in most regards.

With 22-24uF limit on the output caps, I can still do 7 gain steps (34-28dB), and will try it.

As far as the 100pF caps are concerned- 220 pF is fine, 390 pF is fine- although the RF starts to come up a bit in the 2MHz or so region. 0pF is also fine in simulation. 390pF will provide an extra 15dB of attenuation at 10MHz compared to 0pF, and about the same as 100pF.
Whether that matters or not- who knows.
Good to know I'll finally have a use for all those polystyrenes, thank you!
 
Thanks for that explanation Wyn, makes sense to me now. This circuit just doesn't lend itself to widely adjustable gain, it should really be optimized per cartridge.

A couple of things occurred recently that got me off on this track. 1) had some unexpected cash come in, so decided to put some of it into this project and do it right w/better parts, enclosure, hardware, etc. 2) while rummaging through my parts to see what I already had on-hand, came across an old rotary switch, which got me thinking about selectable gains.

So I set about in LTSpice to see what gain resistor values would be needed for 10 x 1dB steps, and that's what exposed the FR issue. I know you had brought it up earlier, but I had forgot about it while concentrating on the 34db gain.

I've run all the sims now, and I can still do it but only w/in a 6-7dB range before the cap values interfere. The plots below are all run with 22uF output caps (the largest physical size that readily fit in the box), and 4 different values of bias cap:
  • 100uF - green
  • 68uF - blue
  • 56uF - red
  • 47uF - cyan
First is at 34dB gain. The 100uF remains essentially flat to below 20Hz, but the others show some droop, with the 47uF worst at @ 0.35db down.

View attachment 1313935

The droop improves as gain is reduced and eventually all go flat to 20Hz, down to 28dB gain, which is the last good looking plot.

At 27dB gain, the 100uF cap has pronounced peak, @ 0.4dB, but the others still look good enough.

View attachment 1313936

At 26dB gain, even the 68uF is peaking, with the 56 close on its heels.

View attachment 1313938

I won't bother to show 25dB.

If I had my druthers I'd go with 68 or 56uF, but neither is available in the better audio caps (Silmic or Muse), and I don't know what characteristics are most desirable for this use: low ESR, low impedance, DA, DF, etc. so I'll probably stick with 100uF. Do you have any thoughts on polymer electrolytics? They seem to spec well in most regards.

With 22-24uF limit on the output caps, I can still do 7 gain steps (34-28dB), and will try it.


Good to know I'll finally have a use for all those polystyrenes, thank you!

100uF seems good.
Opinions on caps? That's a dangerous question.
The good thing is that the base caps are clearly DC biased and they only have a signal present at LF so any distortion components generally need to be of concern around the LF pole location. So, what then matters? Not DA- there's no AC signal to reappear magically in the future, and besides it's a linear effect and doesn't cause distortion, at least not in the non time-domain. DF is just the combination of ESR and the reactive impedance measured at a given frequency. Low ESR is generally good, long life is generally good. Non-linearity, in general, is extremely small as the signal is extremely small and distortion increases exponentially with signal size. Cap self resonant frequency could be problematic for a conventional electrolytic cap, but only at c.50kHz or above. Low impedance is captured by the ESR, SRF and the cap value. ESR is generally temperature and frequency dependent, but not voltage dependent.
As for polymer caps- they are not good for non-polar applications, they have lower ESR than conventional wet dielectric electrolytic caps and a better life for room temp applications. They also have "no DC bias characteristic of capacitance" so if true it implies low distortion, although I can find no distortion specs as of yet.
https://www.murata.com/en-us/products/emiconfun/capacitor/2015/02/24/20150224-p1

They would appear to be a good choice for this application.
 
Thanks Wyn!


The polymer caps looked good on paper, so I'm glad to get your opinion. Particularly since they're available in values that the so-called "Audio" caps are not: 56, 68 & 82uF.

I've run all through sims, and think I'll go with 82uF. Tames the lo-gain peaks somewhat reasonably, while maintaining flat passband at hi-gain.

In case you're wondering why I'm considering selectable gains: in playing around in LTSpice I've come to realize that, in addition to the use/non-use of emitter resistors affecting gain, so do the bias current and battery voltage!

So depending on what final bias current comes in at, being able to dial it up/down a dB or 2, to compensate over battery life, seemed a good idea.

And, if I should ever get another cartridge with higher output the the MC-20, I won't have to change parts on the board!

Some parts are on order, but I'm waiting til I have all in-hand before finalizing the board.
 
Thanks Wyn!


The polymer caps looked good on paper, so I'm glad to get your opinion. Particularly since they're available in values that the so-called "Audio" caps are not: 56, 68 & 82uF.

I've run all through sims, and think I'll go with 82uF. Tames the lo-gain peaks somewhat reasonably, while maintaining flat passband at hi-gain.

In case you're wondering why I'm considering selectable gains: in playing around in LTSpice I've come to realize that, in addition to the use/non-use of emitter resistors affecting gain, so do the bias current and battery voltage!

So depending on what final bias current comes in at, being able to dial it up/down a dB or 2, to compensate over battery life, seemed a good idea.

And, if I should ever get another cartridge with higher output the the MC-20, I won't have to change parts on the board!

Some parts are on order, but I'm waiting til I have all in-hand before finalizing the board.
if you are using the replica bias circuit then, to a first order, the gain will be proportional to the Battery voltage - 1diode drop, so the decline over life of a 9v battery to about 7v will change the gain by 2.5-3dB, say 3dB including the tolerances on the resistor, for the circuit using the THAT devices and 1% resistors with no degeneration resistors. That only corresponds to about 1.5dB in actual "loudness" over however long it takes for the battery to discharge. I doubt if you'll notice it- I certainly adjust my listening level with almost every piece of music.
Still, you seem to be enjoying the exercise and being able to use LTspice, but it's something that I wouldn't care about.
Is there some other project you'd like some design guidance on?
 
if you are using the replica bias circuit then, to a first order, the gain will be proportional to the Battery voltage - 1diode drop, so the decline over life of a 9v battery to about 7v will change the gain by 2.5-3dB, say 3dB including the tolerances on the resistor, for the circuit using the THAT devices and 1% resistors with no degeneration resistors. That only corresponds to about 1.5dB in actual "loudness" over however long it takes for the battery to discharge. I doubt if you'll notice it- I certainly adjust my listening level with almost every piece of music. Still, you seem to be enjoying the exercise and being able to use LTspice, but it's something that I wouldn't care about.
Yah, I tend to over think things and go overboard sometimes, but I'd rather have the option available than regret not having it later.

And you're right, I am enjoying this, having fun and learning a lot :)

Is there some other project you'd like some design guidance on?
One project at a time is all I can handle Wyn! But can you confirm I understand this correctly:

... This is with the EMITTER current of the CB transistors in the "current mirror" (replica bias) design set to about 133ua and the source resistance set to 100 ohms. The extra emitter degeneration resistors have been removed. If present they add 50% of the value of a single one linearly to the input resistance.

So in Leaches as-published CM circuit, w/Re = 100 ohms, amps effective input resistance is 150 ohms (@100 amp + 50% Re)? Then paralleled w/330 ohms R-in yields 103 ohm effective input impedance?

And that same formula applies to any Re? (eg. Re= 10 ohms, amp-in = 105; Re= 20 ohms, amp-in = 110, etc).

I do appreciate all the help you've been, thank you Wyn!
 
Yah, I tend to over think things and go overboard sometimes, but I'd rather have the option available than regret not having it later.

And you're right, I am enjoying this, having fun and learning a lot :)


One project at a time is all I can handle Wyn! But can you confirm I understand this correctly:



So in Leaches as-published CM circuit, w/Re = 100 ohms, amps effective input resistance is 150 ohms (@100 amp + 50% Re)? Then paralleled w/330 ohms R-in yields 103 ohm effective input impedance?

And that same formula applies to any Re? (eg. Re= 10 ohms, amp-in = 105; Re= 20 ohms, amp-in = 110, etc).

I do appreciate all the help you've been, thank you Wyn!
Yes, that's true. Each emitter, in this case, has about 200ohms re - the RE is added to that and then two are effectively in parallel. so it's 100 +RE/2.
 
Last edited:
Thanks Wyn!

Still fiddling w/PCB layout, trying to optimize. Will post again as progress continues.

Sean
 
It's still alive. All parts are finally in-hand. A couple of parts didn't match published specs, so board layout had to be adjusted a bit. Glad I waited before having the boards made!

Leach-parts-1.jpg
 
Well, did they turn out as expected?
Hi Wyn:

I've been sidetracked with other priorities so haven't got beyond electrical measurements, but those look good!

4.58 & 4.54 volts pos/neg to ground on both channels

133 & 135 uA bias currents each channel.

Almost ready to drill front & back panels for RCAs & switches, so final assembly will be in a week or 2.
 
Plodding along. My machinist who was going to do all the metalwork decided to take an extended holiday, so some parts awaiting his return. Couldn't wait tho so decided to do the back panel myself. Not perfect, the drill bit wandered a bit, but close enough.

Will temp in batteries & power switch this weekend for a first listen.

Back panel RCA.jpg Inside RCA.jpg Board in.jpg
 
Back
Top Bottom