His original design's are over 30year old.
For the pre-drivers indeed TO-5 and the drivers TO-39, both with a small heatsink.
My old Leach (single ouput pair) runs at +/- 42 V and stays very cool.
For the offset-adjustment i made an extra circuit, connected to the tail of the 33k base input resistor. Works fine to me.
Loek
For the pre-drivers indeed TO-5 and the drivers TO-39, both with a small heatsink.
My old Leach (single ouput pair) runs at +/- 42 V and stays very cool.
For the offset-adjustment i made an extra circuit, connected to the tail of the 33k base input resistor. Works fine to me.
Loek
Re: His original design's are over 30year old.
I've been looking to do a 'mini-Leach' with 42V rails and one pair of outputs, but I feared it wouldn't be enough for the input stage. How do you get enough current through the zeners?
loek said:
I've been looking to do a 'mini-Leach' with 42V rails and one pair of outputs, but I feared it wouldn't be enough for the input stage. How do you get enough current through the zeners?
zeners current
I used the first schematic, the zeners are 24V so with adjusting R16 and R17 (sch. 1976) it is no problem at all.
Loek
I used the first schematic, the zeners are 24V so with adjusting R16 and R17 (sch. 1976) it is no problem at all.
Loek
You could reduce the voltage of the zener stack or eliminate them completely. IIRC on version 4.2 Leach suggested no lower than 40V rails.
Another solution is to provide a higher (regulated) voltage for the front end. Some will argue that this is not good because you want the output stage to clip first.
I tend to avoid clipping by ensuring more than adequate reserve power. I'm guessing that you are either looking for either a tweeter amp or driving a high efficiency system since you want low rails.
As long as your load doesn't drop significantly below 8 ohms or you don't expect to push it hard you could probably use slightly higher rails (say 48-50/ 35 VAC input) and avoid front end issues. Run the SOA calculations to be sure. You need to do that to figure your protection resistors anyway.
Another solution is to provide a higher (regulated) voltage for the front end. Some will argue that this is not good because you want the output stage to clip first.
I tend to avoid clipping by ensuring more than adequate reserve power. I'm guessing that you are either looking for either a tweeter amp or driving a high efficiency system since you want low rails.
As long as your load doesn't drop significantly below 8 ohms or you don't expect to push it hard you could probably use slightly higher rails (say 48-50/ 35 VAC input) and avoid front end issues. Run the SOA calculations to be sure. You need to do that to figure your protection resistors anyway.
Hi,
as Loek says, change to lower voltage Zeners and redo the calculations to ensure at least 10% dissipation in the Zener.
Re: zeners current
ah - theyre 40V in the newer version.
Can you post a schematic for the version you're using?
Good guess - it's for a Tannoy DC tweeter, 15 ohms and 10dB more efficient than the bass unit.
Yes, I am thinking the same way as you - I initially didn't want to give the input more volts than the output so that it clips before the output, but I reckon I'd go deaf before it clips in this application.
Reducing the zeners to 24V would be neater than having to use a high voltage on the input though.
loek said:
ah - theyre 40V in the newer version.
Can you post a schematic for the version you're using?
Good guess - it's for a Tannoy DC tweeter, 15 ohms and 10dB more efficient than the bass unit.
Yes, I am thinking the same way as you - I initially didn't want to give the input more volts than the output so that it clips before the output, but I reckon I'd go deaf before it clips in this application.
Reducing the zeners to 24V would be neater than having to use a high voltage on the input though.
Hi Bob,
bc546c/556c would be excellent. I used bc550c/560c.
VAS and Pre-Driver: 2sa1360/c3423
Driver: 2sb649/c669
Output: MJL4302/4281
The Driver To126 package selection completely solves the heatsink shorting problem mentioned some time ago. Mine are up in the air, isolated from all around, but "live", no mica needed helps with dissipation, but still the sink is cool.
My selections are completely at odds with the low fT and low hFE philosophy promoted by many including Leach, but he was using commercially available transistors affordable 30years ago.
ps. the 546c/556c only became available after the other GB. Thanks.
bc546c/556c would be excellent. I used bc550c/560c.
VAS and Pre-Driver: 2sa1360/c3423
Driver: 2sb649/c669
Output: MJL4302/4281
The Driver To126 package selection completely solves the heatsink shorting problem mentioned some time ago. Mine are up in the air, isolated from all around, but "live", no mica needed helps with dissipation, but still the sink is cool.
My selections are completely at odds with the low fT and low hFE philosophy promoted by many including Leach, but he was using commercially available transistors affordable 30years ago.
ps. the 546c/556c only became available after the other GB. Thanks.
AndrewT said:
Yep - Mouser seemed to be waiting for a shipment. Before they ran out they were available in small quantity. I had to place the group buy order for 10K, then Mouser would sell singles as soon as I got mine. Arrgh!
Thanks for the insight into your thinking. I agree that these devices go against Leach's low open loop gain philosophy. We're headed into Self land. Hmm. 😉
Hi Bob,
Leach's open loop gain is unchanged by the device gains. It is set by the topology and the component values.
I believe the choices I have made do not affect the 8MHz bandwidth that Leach engineered into LoTim.
My reasoning goes along the lines of; make sure the topology/values do the job they are meant to do and the chosen device characteristic should not interfere with your circuit aims. Individual variation in hFE SHOULD not affect circuit gain and similarly fT SHOULD not apply extra delay to the signal loop, otherwise second order effects are starting to complicate the feedback loops. I may have this wrong because I am right at my limits of understanding in this HF/NFB/feedforward/poles/zeros part of my education.
I wish I knew enough to say you have drawn the wrong conclusion, but I don't.
Leach's open loop gain is unchanged by the device gains. It is set by the topology and the component values.
I believe the choices I have made do not affect the 8MHz bandwidth that Leach engineered into LoTim.
My reasoning goes along the lines of; make sure the topology/values do the job they are meant to do and the chosen device characteristic should not interfere with your circuit aims. Individual variation in hFE SHOULD not affect circuit gain and similarly fT SHOULD not apply extra delay to the signal loop, otherwise second order effects are starting to complicate the feedback loops. I may have this wrong because I am right at my limits of understanding in this HF/NFB/feedforward/poles/zeros part of my education.
Andrew,
Changing transistors may mean that your amps need to have their compensation circuit reworked also. The parasitic properties of the semiconductors implemented will influence on the open loop performance of the amp, and thus may require different values in the compensation network.
I altered a bit in the feedback network, to make the compensation fit the devises I specified in the original BOM.
Where did you get 2sa1360/c3423? they look nice!
\Jens
Changing transistors may mean that your amps need to have their compensation circuit reworked also. The parasitic properties of the semiconductors implemented will influence on the open loop performance of the amp, and thus may require different values in the compensation network.
I altered a bit in the feedback network, to make the compensation fit the devises I specified in the original BOM.
Where did you get 2sa1360/c3423? they look nice!
\Jens
schematic
Hi Bremen Nacht, this is the link to the old schematics:
http://users.ece.gatech.edu/~mleach/papers/lowtim/feb76feb77articles.pdf
Andrew, it is a passive adjustment, a 100 ohm pot(10trn) with on both sides a resistor (15k) to DC and also 2 caps to ground. Very simple.
Loek
Hi Bremen Nacht, this is the link to the old schematics:
http://users.ece.gatech.edu/~mleach/papers/lowtim/feb76feb77articles.pdf
Andrew, it is a passive adjustment, a 100 ohm pot(10trn) with on both sides a resistor (15k) to DC and also 2 caps to ground. Very simple.
Loek
Hi Loek,
is the DC feed into the inverting or non-inverting input?
Did you think about copying the Krell (and others) style DC feed using a pair of diodes to generate a stable DC voltage to then feed into the inverting input?
is the DC feed into the inverting or non-inverting input?
Did you think about copying the Krell (and others) style DC feed using a pair of diodes to generate a stable DC voltage to then feed into the inverting input?
Hi Jens,
I selected 1360 from a range of suitable Qs from the Dalbani online catalogue. They have many that perform better, but at greater cost.
I thought 1360 gave me a reasonable performance/price ratio.
You may prefer 1405/6/7 with the 1406 standing out as superb but at 5times the cost.
I tried to source 1380 but the complement was unavailable at the time.
I have not experimented with Miller comp cap yet.
Not even tested the amp on the scope nor sustained power into dummy loads.
But I will BEFORE I start assembling the other 4 PCBs.
I selected 1360 from a range of suitable Qs from the Dalbani online catalogue. They have many that perform better, but at greater cost.
I thought 1360 gave me a reasonable performance/price ratio.
You may prefer 1405/6/7 with the 1406 standing out as superb but at 5times the cost.
I tried to source 1380 but the complement was unavailable at the time.
I have not experimented with Miller comp cap yet.
Not even tested the amp on the scope nor sustained power into dummy loads.
But I will BEFORE I start assembling the other 4 PCBs.
DC feed
Hi Andrew, to the bottom of R3 in the feb 76/77 schematic. With 2 diodes you will have about +/- 0,6 V and you don't need this range.
I buid this amp in begin '80s and in between thinking of modifiyng this old amp. Its is a stereo amp with one transformer and 2 15000uF caps.
It is to much work to renew this amp, so i keep it as it is.
I have swapped some caps last year to keep things ok.
But i have also the 3 pairs Leach version pcb's that Jens designed.
And indeed also the Krell and an other Gainclone on staple, so depending what is coming to my mind has priority.....
Greetings,Loek
Hi Andrew, to the bottom of R3 in the feb 76/77 schematic. With 2 diodes you will have about +/- 0,6 V and you don't need this range.
I buid this amp in begin '80s and in between thinking of modifiyng this old amp. Its is a stereo amp with one transformer and 2 15000uF caps.
It is to much work to renew this amp, so i keep it as it is.
I have swapped some caps last year to keep things ok.
But i have also the 3 pairs Leach version pcb's that Jens designed.
And indeed also the Krell and an other Gainclone on staple, so depending what is coming to my mind has priority.....
Greetings,Loek
Where did you get 2sa1360/c3423? they look nice!
The company NEDIS sells them for about 50 eurocent. I order there now and then.
Loek
The company NEDIS sells them for about 50 eurocent. I order there now and then.
Loek
There are a few parameters that set GBP and slew rate (SR). They are the emitter resistors (Re) in the input pairs, the quiesent current (Iq) in the input pair, and the lag compensation cap (Cc).
SR=Iq/Cc.
GBP= 1/[2*pi*((2*Vt/Iq)+Re)]
Therefore,
GBP goes up if: Cc or Re go down, or Iq goes up, where SR will also go up (Re does not affect SR)
GBP goes down if: Cc or Re go up, or Iq goes down.
Increasing Iq will increase the diff amp linear voltage region (up to a certain point, and you don't want to starve the available current to charge Cc).
Note that Cc is not simply the value of the added lag cap. It also includes the capacitance of the VAS transistor used, and probably some other capacitance in the circuit.
Leach reduced GBP from 10MHz to 8 to 8.5MHz over the years. He felt it was more stable, and made a better sounding amp. This may have been due to the transistors used. I've wondered if by using modern transistors with higher ft, the GBP may be effectively raised again. Anyhow, with higher fts, it would seem to me that the amp may be even more stable compensated to the same GBP as Leach uses, as the phase rolloffs would be pushed to higher frequencies, or just as stable at higher GBPs.
SR=Iq/Cc.
GBP= 1/[2*pi*((2*Vt/Iq)+Re)]
Therefore,
GBP goes up if: Cc or Re go down, or Iq goes up, where SR will also go up (Re does not affect SR)
GBP goes down if: Cc or Re go up, or Iq goes down.
Increasing Iq will increase the diff amp linear voltage region (up to a certain point, and you don't want to starve the available current to charge Cc).
Note that Cc is not simply the value of the added lag cap. It also includes the capacitance of the VAS transistor used, and probably some other capacitance in the circuit.
Leach reduced GBP from 10MHz to 8 to 8.5MHz over the years. He felt it was more stable, and made a better sounding amp. This may have been due to the transistors used. I've wondered if by using modern transistors with higher ft, the GBP may be effectively raised again. Anyhow, with higher fts, it would seem to me that the amp may be even more stable compensated to the same GBP as Leach uses, as the phase rolloffs would be pushed to higher frequencies, or just as stable at higher GBPs.
Hi Loek,
Sounds like you already know the Krell style DC bias cancelling circuit. It also uses resistors to cut the 700mV available from the diodes, but eliminates the rail noise more effectively than a 100r pot on it's own.
you really meant to the top of R3 in fig2 feb '76? this is the non-inverting input.
Sounds like you already know the Krell style DC bias cancelling circuit. It also uses resistors to cut the 700mV available from the diodes, but eliminates the rail noise more effectively than a 100r pot on it's own.
you really meant to the top
Andrew, as i wrote it is the bottom of R3.
I write this out of memory, i have to open the box to sea what i really did at that time....And the noise is no problem.
Loek
Andrew, as i wrote it is the bottom of R3.
I write this out of memory, i have to open the box to sea what i really did at that time....And the noise is no problem.
Loek
Hi Pooge,
you are confirming my thoughts exactly. Can I rely on you to be absolutely correct?
Taken further, could higher fTs allow a slightly smaller Miller comp cap to be used and still retain the same GBW product and the same phase margin?
My chosen VAS has slightly less Cob, so effectively my PCB has slightly less Ccomp already, so further reduction in Miller comp cap will have to be investigated carefully.
you are confirming my thoughts exactly. Can I rely on you to be absolutely correct?
none of this is affected by transistor characteristics.
suggesting that higher fT produces a more stable amplifier due to slightly reduced phase angle in the feedback signal.
Taken further, could higher fTs allow a slightly smaller Miller comp cap to be used and still retain the same GBW product and the same phase margin?
My chosen VAS has slightly less Cob, so effectively my PCB has slightly less Ccomp already, so further reduction in Miller comp cap will have to be investigated carefully.
- Status
- Not open for further replies.