The Leach moving coil pre-preamp

[wyn palmer]

This is a Leach style CB amp with replica biasing and 34dB of gain into the same loads as before.. I hope that this will make some of the answers to prior questions less opaque.
Q2, Q6 are matched to Q8, Q7 respectively. The resistors R40 and R35 are used to set the bias current- the current is essentially
(battery voltage -Vbe-Vbe)/total resistance in the path: or approximately (9-0.7-0.7)/60k=7.6/60000=127ua.
R37, R41, R39 and R38 can be removed if the transistors are "adequately" matched.
I can't investigate what this means as Ltspice won't allow me to fractionally scale the models and I don't feel inclined to create my own scalable models for the transistors.
The load R in this case consists of the parallel combination of R40, R35 and R44.


Note that the LF response shows only a small amount of peaking. This is because the load R components before the coupling caps are a sufficiently large part of the overall load that the amp internal gain doesn't increase dramatically at the LF cut off and below.
The circuit above will alter the noise performance of the amp.

Compare this to the simple CB design with the lower bias Rs.

This is about a dB lower than the replica bias circuit, but most of this is due to the small difference in gain between the two designs so they are essentially equivalent.

 

[HalfApt]

Wyn, I'm blown away by your efforts on my behalf, thank you very much!

What they do need to do is to have the sum of the two Vbes match the sum of the "ideal" transistor Vbes that were used for originally defining the bias resistors- and who the heck knows what that was. Better to trim the bias.

Due to differences between manufacturers & batches, in his follow-up to the original article, Leach did recommend using a pot for R2 to adjust final bias, them measure it & replace w/fixed resistor (if desired). Still a good idea?

Your latest schematics ARE helping me understand it all better.

I've been under the impression that the output resistor needed to be significantly lower value than preamp input impedance, but your re-worked CB circuit uses 220k, so apparently was wrong in that regard.

I also see you've added a 2.2uF output coupling cap, is that required and/or desirable in your re-worked circuits?

If you want I'll try to drop in the ADI matched pairs into a simulation and see how they do. This assumes that I can get models...

I found PSpice models for both devices a couple of days ago, but didn't get them installed until last night (models have different pin assignments than the actual devices, but got it figured out).

They produce a tiny bit more gain (@ 3%), but otherwise seem to work the same as 4401/4403 (w/in my limited LTSpcice abilities).

Well matched transistors in the "current mirror" design eliminate the need for Re.

Are the ADI SSM devices adequately matched in this regard?

I'm gonna spend the weekend trying to educate myself better in the electronics, and get a PCB layout going. A single board can be made for either CB or CM configuration, and accommodate either the SSM devices or discrete transistors.

Sean

 

[Pio1980]

One of the few pieces I still have from the olde tymes is a Marcof PPA-1, interesting thread on this simple approach to LOMC adaptation.

 

[wyn palmer]

Wyn, I'm blown away by your efforts on my behalf, thank you very much!


Due to differences between manufacturers & batches, in his follow-up to the original article, Leach did recommend using a pot for R2 to adjust final bias, them measure it & replace w/fixed resistor (if desired). Still a good idea?

Your latest schematics ARE helping me understand it all better.

I've been under the impression that the output resistor needed to be significantly lower value than preamp input impedance, but your re-worked CB circuit uses 220k, so apparently was wrong in that regard.

I also see you've added a 2.2uF output coupling cap, is that required and/or desirable in your re-worked circuits?



I found PSpice models for both devices a couple of days ago, but didn't get them installed until last night (models have different pin assignments than the actual devices, but got it figured out).

They produce a tiny bit more gain (@ 3%), but otherwise seem to work the same as 4401/4403 (w/in my limited LTSpcice abilities).



Are the ADI SSM devices are adequately matched in this regard?

I'm gonna spend the weekend trying to educate myself better in the electronics, and get a PCB layout going. A single board can be made for either CB or CM configuration, and accommodate either the SSM devices or discrete transistors.

Sean

For the modified simple Leach CB amp with the low value bias Rs the main contributors to RL are the two collector base resistors. The output has to be AC coupled, so you need a resistor to ground at the node where the caps are connected to each other in order to reduce the potential impact of large transient voltages being applied to a follow up amp. The 220k was just a convenient large-ish value. The 2.2u/47k load is outside the "pin" of the amp and it just represents a typical RIAA stage input load.
Unfortunately the fixed R value needed to set the current is likely to be nowhere near a standard value, and a 3% error will just about half or double the current. There's also the issue of temperature drift, but I won't go into that...
The replica bias circuit greatly improves this situation.
The ADI matched transistors will have different Vbes than the 4401/3s for a given current and that causes the actual current and hence gain to be different.
The matching is good enough to eliminate the need for the low value "emitter degeneration" resistors that were added.
As far as the effort on your behalf is concerned- it's pretty minimal. Typing the postings takes much more time than anything, and I have sufficient spare time available. I'm glad to be of assistance.

 

[HalfApt]

... Unfortunately the fixed R value needed to set the current is likely to be nowhere near a standard value, and a 3% error will just about half or double the current.

So in the circuit I built, using his published 5.1k value, the bias current could still be way off optimal? I still don't have my head around transistor bias current. Does it flow into the Base, or through transistor C-E? If needed, I would put test points on the new PCB to accurately measure it.

The 220k was just a convenient large-ish value.

OK. I was surprised to see, as this (& the input R) on your modified schematic are orders of magnitude larger than Leach used, otherwise the circuit is essentially identical to the one I built (the Oct. '81 "Update").

Leach has Rin=39ohms for impedance = 30 ohms, calculated by Imp = Rin parallel to 133.5 ohms. I changed mine to 178 ohms yielding @75 impedance, to match Ortofons head-amp. I'm presuming though, that his "133.5" value was derived accounting for the 150 ohm emitter resistors in that "update".

The ADI matched transistors ... The matching is good enough to eliminate the need for the low value "emitter degeneration" resistors that were added.

Excellent! I think I'll go over-"board" on the PCB and include quality pin-sockets for some resistors so they can be changed/jumpered, just for experimenting/optimizing.

As far as the effort on your behalf is concerned- it's pretty minimal. Typing the postings takes much more time than anything, and I have sufficient spare time available. I'm glad to be of assistance.

Well, I do appreciate it very much, thank you! I would never have figured it out by myself.

I think I'm 95+% done with novice questions, so will do a PCB layout over the weekend, and post a first try for comments.

Thanks again Wyn!

Sean

 

[wyn palmer]

So in the circuit I built, using his published 5.1k value, the bias current could still be way off optimal? I still don't have my head around transistor bias current. Does it flow into the Base, or through transistor C-E? If needed, I would put test points on the new PCB to accurately measure it.


OK. I was surprised to see, as this (& the input R) on your modified schematic are orders of magnitude larger than Leach used, otherwise the circuit is essentially identical to the one I built (the Oct. '81 "Update").

Leach has Rin=39ohms for impedance = 30 ohms, calculated by Imp = Rin parallel to 133.5 ohms. I changed mine to 178 ohms yielding @75 impedance, to match Ortofons head-amp. I'm presuming though, that his "133.5" value was derived accounting for the 150 ohm emitter resistors in that "update".


Excellent! I think I'll go over-"board" on the PCB and include quality pin-sockets for some resistors so they can be changed/jumpered, just for experimenting/optimizing.


Well, I do appreciate it very much, thank you! I would never have figured it out by myself.

I think I'm 95+% done with novice questions, so will do a PCB layout over the weekend, and post a first try for comments.

Thanks again Wyn!

Sean

The external cartridge load R is in parallel with the input resistance of the amp.
To make it obvious what that is I'll run a sim and increase the cartridge source R until the input signal is attenuated by 6dB, then post it.
You can then add whatever R you need in parallel to get the required input R.
The Load R that sets the gain in conjunction with the gm of the input stage consists of several resistors in parallel, including the external 47k ohm load R.
There's no reason why the external load R would be the same as the Leach design as the gain is higher in this case and I've included other load elements.
Sims do not lie! (well, usually anyway).
By the way, I'm composing this while sitting in my "music room" listening to Miles Davis' "kind of blue"- I just obtained a 15IPS 1/4" tape 2nd generation copy of the stereo master from a friend in England and I'm comparing it against the original 6-eye vinyl copy and a relatively recent reissue. I'm multitasking as I listen and I can run the sims remotely from my laptop on my downstairs workstation so it really is no bother...

 

[wyn palmer]

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.
The attenuation is almost exactly 6dB, which tells you that the input resistance of the amp is pretty close to 100 ohms nominally, so an additional 300 ohm resistor to ground would yield 75ohms.
The emitter current equals the collector current plus the base current. The collector current is often referred to as the bias current for a transistor.
The base current is usually estimated at about 1/100 of the collector current- that ratio is actually a consequence of the physics/fabrication of the transistor and can differ substantially ( an order of magnitude or so in either direction) from 1/100, but 1/100 gives you the general idea. The base current occurs as a consequence of emitter/collector current flow and not vice versa- but the details are far beyond what you need to know.
I've tried explaining bipolar transistor semiconductor physics to the, how shall I say it, unititiated, in the past. It did not go well...

 

[HalfApt]

View attachment 1286927

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.
The attenuation is almost exactly 6dB, which tells you that the input resistance of the amp is pretty close to 100 ohms nominally, so an additional 300 ohm resistor to ground would yield 75ohms.
The emitter current equals the collector current plus the base current. The collector current is often referred to as the bias current for a transistor.
The base current is usually estimated at about 1/100 of the collector current- that ratio is actually a consequence of the physics/fabrication of the transistor and can differ substantially ( an order of magnitude or so in either direction) from 1/100, but 1/100 gives you the general idea. The base current occurs as a consequence of emitter/collector current flow and not vice versa- but the details are far beyond what you need to know.
I've tried explaining bipolar transistor semiconductor physics to the, how shall I say it, unititiated, in the past. It did not go well...

Thanks again Wyn!

I finally got LTSpice figured out, and was able to run frequency & noise analyses, settled on a schematic, and went on to design a board layout, when I hit a couple of roadblocks (It's always something isn't it - Murphys Law?).

While checking parts sizes & pin spacings for board layout, I discovered:

1) Eagle Lite (free) has too small a board size limit (I could squeeze it all in, but with little to no space between parts). I've been trying KiCad (no board size limit), but it has a very obtuse user interface, so it will take a while to figure out.

2) the ADI SSM-2220 is out-of-production/discontinued and unavailable from the usual vendors. I could go eBay/surplus, but I'm leary of that.

So went looking for alternatives and found a few, but am unsure of the specs and suitability, and could use some advice about these. Some have P-Spice models, but haven't got around to trying them yet. Will be doing so today.

Most intrigued by this one: THAT-340, all 4 transistors in one package, low noise, low Rbb, about 60% the cost of ADI SSM devices.

Then there are same-type pairs:

• Diodes Inc DMMT3904 and DMMT3906
• OnSemi NST45011 and NST65011
• NXP-BCM847 and BCM857

None of the pairs are touted as low-noise or quote an Rbb spec, all quote noise in dB, and don't know how to compare that to nV/Hz figure.

All above need a different board layout than the SSM devices, so I need to pick something before going any further on a PCB.

Any quick thoughts, suggestions, advice?

Sean

 

[wyn palmer]

Thanks again Wyn!

I finally got LTSpice figured out, and was able to run frequency & noise analyses, settled on a schematic, and went on to design a board layout, when I hit a couple of roadblocks (It's always something isn't it - Murphys Law?).

While checking parts sizes & pin spacings for board layout, I discovered:

1) Eagle Lite (free) has too small a board size limit (I could squeeze it all in, but with little to no space between parts). I've been trying KiCad (no board size limit), but it has a very obtuse user interface, so it will take a while to figure out.

2) the ADI SSM-2220 is out-of-production/discontinued and unavailable from the usual vendors. I could go eBay/surplus, but I'm leary of that.

So went looking for alternatives and found a few, but am unsure of the specs and suitability, and could use some advice about these. Some have P-Spice models, but haven't got around to trying them yet. Will be doing so today.

Most intrigued by this one: THAT-340, all 4 transistors in one package, low noise, low Rbb, about 60% the cost of ADI SSM devices.

Then there are same-type pairs:

• Diodes Inc DMMT3904 and DMMT3906
• OnSemi NST45011 and NST65011
• NXP-BCM847 and BCM857

None of the pairs are touted as low-noise or quote an Rbb spec, all quote noise in dB, and don't know how to compare that to nV/Hz figure.

All above need a different board layout than the SSM devices, so I need to pick something before going any further on a PCB.

Any quick thoughts, suggestions, advice?

Sean

Off the top of my head-
The THAT-340 looks like a winner to me with the caveat that it has no data sheet curves, no spice models that I can find and the specs are mostly at 1mA collector current which would not be a good operating point for the CB design- the input impedance would be c. 12 ohms- so there's no way of really knowing in advance of trying it out.
The DMT devices are specified at 100ua, but the noise figure is 5dB relative to a 1k source, so the noise voltage is about 7nv/rtHz I..e.4nv/rtHz (for the 1k source) increased by 5dB! so that won't work.
The BCMs are slightly better at 1.6dB for a 2kohm source, but still nowhere near good enough.
The NST devices are a complete non starter- 10dB NF for a 2k source- or 17nv/rtHz or there abouts.

 

[JoeESP9]

One of the few pieces I still have from the olde tymes is a Marcof PPA-1, interesting thread on this simple approach to LOMC adaptation.

I also have a Marcof PPA-1. Like you I have been following this thread with great interest.

 

[HalfApt]

Off the top of my head-
The THAT-340 looks like a winner to me with the caveat that it has no data sheet curves, no spice models that I can find and the specs are mostly at 1mA collector current which would not be a good operating point for the CB design- the input impedance would be c. 12 ohms- so there's no way of really knowing in advance of trying it out.
The DMT devices are specified at 100ua, but the noise figure is 5dB relative to a 1k source, so the noise voltage is about 7nv/rtHz I..e.4nv/rtHz (for the 1k source increased by 5dB)! so that won't work.
The BCMs are slightly better at 1.6dB for a 2kohm source, but still nowhere near good enough.
The NST devices are a complete non starter- 10dB NF for a 2k source- or 17nv/rtHz or there about.

Wyn:

There's model link at the bottom of the THAT page (linked here too, 300 Series Device Model). The download file contains all of their devices, but there's a "300 Series_Macro_01.lib" file w/in. Doesn't look like the other LTSpice files, and I haven't tried it yet, so don't know if it's compatible or even complete.

I didn't find the THAT-340 by generic pair internet searches, but found it mentioned by a few people here and in other audio forums as being good.

I didn't think the other matched pairs were going to be up-to-snuff, so thanks for that advice. I'll scratch them off the list.

Sean

 

[wyn palmer]

Wyn:

There's model link at the bottom of the THAT page (linked here too, 300 Series Device Model). The download file contains all of their devices, but there's a "300 Series_Macro_01.lib" file w/in. Doesn't look like the other LTSpice files, and I haven't tried it yet, so don't know if it's compatible or even complete.

I didn't find the THAT-340 by generic pair internet searches, but found it mentioned by a few people here and in other audio forums as being good.

I didn't think the other matched pairs were going to be un-to-snuff, so thanks for that advice. I'll scratch them off the list.

Sean

Thanks. I should be able to use the lib file as they look like real bjt models.
However, as you have only two choices in any case- to purchase a bunch of the original devices and match them up- which is easier than it was as you now know the requirements- or use the THAT devices- in essence the choice is still yours.
I can't guarantee that I will get to the THAT libs for a few days as I've other things on my plate at the moment, but I'll try if you wish.

 

[HalfApt]

I can't guarantee that I will get to the THAT libs for a few days as I've other things on my plate at the moment, but I'll try if you wish.

No, don't go out of your way Wyn, there's no rush. I'll give it a try myself later today or tomorrow.

I've got plenty of other thing to do as well, and probably won't get back to a board till weekend at least.

Thanks!

 

[wyn palmer]

No, don't go out of your way Wyn, there's no rush. I'll give it a try myself later today or tomorrow.

I've got plenty of other thing to do as well, and probably won't get back to a board till weekend at least.

Thanks!

I dropped the THAT models for the 300 units into the standard.bjt file in the LTspice XVII.lib cmp (component) folder. The simulator "sees" the devices but can't progress as it apparently doesn't recognize some of the extracted parameters in the model which are different from those in the other devices in the library.
I'll look at it some more, but the THAT models may just be fundamentally incompatible with LTspice.

 

[HalfApt]

I got it Wyn!

I've been on the web a lot to learn as much about LTSpice as I can/need, and found one site that made it simple!

1) Put 4 generic NPN/PNP transistors in the circuit

2) copy each ".model"s statements from the 300-series .lib file & paste onto the schematic

3) set the value of each transistor to the exact name as in the .model line

Boom! It works!

Still have much to learn, but I'm much more comfortable working in LTSpice now. Have run DC, AC & noise on all 3 configurations, with 3 different Re/R2 pairings, adjusted for same gain (as close as possible anyway).

I've found the THAT-340 is about 3 nV/Hz worse than the ADI SSM devices, which are about 2-3 worse than discrete 4401/4403.

Sean

 

[wyn palmer]

I got it Wyn!

I've been on the web a lot to learn as much about LTSpice as I can/need, and found one site that made it simple!

1) Put 4 generic NPN/PNP transistors in the circuit

2) copy each ".model"s statements from the 300-series .lib file & paste onto the schematic

3) set the value of each transistor to the exact name as in the .model line

Boom! It works!

Still have much to learn, but I'm much more comfortable working in LTSpice now. Have run DC, AC & noise on all 3 configurations, with 3 different Re/R2 pairings, adjusted for same gain (as close as possible anyway).

I've found the THAT-340 is about 3 nV/Hz worse than the ADI SSM devices, which are about 2-3 worse than discrete 4401/4403.

Sean

Good news, so the right thing to do here is to buy a few of each discrete transistor and match them in pairs.
I'll try and do the model thing the "right" way by including the .model files as part of the bjt library. That way you can, theoretically, simply select the model name from the pull down list. I've done it before so I've no idea why it didn't work this time- so it seems I'll have a bit of entertainment awaiting me.

 

[HalfApt]

I forgot to mention, the 300 series .lib has 4 different models in it. The first 2 are for their low-noise devices, the last 2 for hi-frequency devices. I only copied the first 2:

.MODEL QPNP_THAT_NS PNP
.MODEL QNPN_THAT_NS NPN

and changed the transistor values to:

QPNP_THAT_NS
QNPN_THAT_NS

 

[wyn palmer]

I forgot to mention, the 300 series .lib has 4 different models in it. The first 2 are for their low-noise devices, the last 2 for hi-frequency devices. I only copied the first 2:

.MODEL QPNP_THAT_NS PNP
.MODEL QNPN_THAT_NS NPN

and changed the transistor values to:

QPNP_THAT_NS
QNPN_THAT_NS

Yes, I understand that. I just included the noise optimized models.
A comment- this must be a product of the limitations of LTspice or more likely of the model extraction process performed by THAT. In my many years of designing low noise and wideband ICs using bipolars I never once had to split the design task up this way. In fact, for some far in the past requirements- such as making an optimal HF noise match for a GSM receiver designed in BICMOS - such a trade off would be impossible.
Still, I guess you do learn something new every day...

 

[wyn palmer]

I got it Wyn!

I've been on the web a lot to learn as much about LTSpice as I can/need, and found one site that made it simple!

1) Put 4 generic NPN/PNP transistors in the circuit

2) copy each ".model"s statements from the 300-series .lib file & paste onto the schematic

3) set the value of each transistor to the exact name as in the .model line

Boom! It works!

Still have much to learn, but I'm much more comfortable working in LTSpice now. Have run DC, AC & noise on all 3 configurations, with 3 different Re/R2 pairings, adjusted for same gain (as close as possible anyway).

I've found the THAT-340 is about 3 nV/Hz worse than the ADI SSM devices, which are about 2-3 worse than discrete 4401/4403.

Sean

I discovered my error- when I dropped the models into the bjt library for LTspice I compacted them and in doing so I made a syntax error.
I fixed that and now I can pull up the symbols/models directly from the library as per usual and insert them into the schematic for the design using the matched devices.
The result is that the THAT devices and the 2N4401 etc. circuits have almost the exact same performance in noise and gain with the 20ohm degeneration resistors included.




Red is the THAT design, green the 2N design. The THAT design has about 0.95% higher gain

Above is the THAT design output noise.
Below is the 2N design output noise.


Taking the gain difference into account the noise is negligibly different between the two designs.

 

[wyn palmer]

I forgot to mention, the 300 series .lib has 4 different models in it. The first 2 are for their low-noise devices, the last 2 for hi-frequency devices. I only copied the first 2:

.MODEL QPNP_THAT_NS PNP
.MODEL QNPN_THAT_NS NPN

and changed the transistor values to:

QPNP_THAT_NS
QNPN_THAT_NS

I have attached the modified library file which includes the THAT bjt models.
In order to use it you have to go to the */LTspiceXVII/lib/cmp folder (the * depends on where the LTspice installation was performed and the full path needs to be used) and replace the standard.bjt file with the one below where the .txt extension has been replaced by bjt.
The standard.bjt file that was there should have a standard.bjt.bak file (the backup) present in the same library. If not copy and rename the original .bjt into this location.
Once the new .bjt file is in place restart LTspice and it should be accessible.
No new or unusual commands or macros need be added to the schematic to use the new models.
You can just right click on the 2N devices and choose "pick new transistor". The THAT devices are at the end of the list.