Actually most of the time it's for private use, but once in a while for PA. So I do want to make it rugged and ready for possible abuse.
Where can I get decent models for those 15034/35??? They would suffice for the needs and likely be much better than the 30/31 and 32/33.
Simulating? Bah! You know they will be better than the model you have, just build with them.
Really?? You think the real thing would perform better than the sims? That would be great.
What I don't understand is why the leach amp sim doesn't perform as well as others that I've done, and that includes the 3055 amp, that does perform rather well with some mods.
Which project/pcb/schematic are you referring to?
If this is on my current design, I need to re-check. But I just looked at my schematic (eagle) and it's correct. So you must be referring to something else then.
Hey , spooky .
If you could give me the .asc - a good bet someone in my "camp"
would build the IPS - I would keep it VERY close to your design ...
(just change to 4.5-5.5ma VAS I) We already have the OPS (EF3).
"spookyamp leach ips ?? "
OS
Ha! Yes. Good catch. It was wrong back then, but I corrected that since.
You mean you would drive the amp's input right off the opamp and move the input filter before it?
I wouldn't know how to do this right, since this is a diff input and both plugs need to be in parallel so a signal can go right through unchanged if need be.
All this is highly likely to change anyway. For one thing, I will do away with the output relays and replace them by mosfet based SSRs, and I will also use SSRs instead of fuses on the rails.
I want to have an input mute as well, and I am looking into having a clipping prevention circuit there as well, which would be switchable, so I think that INA may be changed to something else or it may be moved much farther ahead of the amp's input, if the clipping prevention sits there in between and can be used as a mute at the same time could be nice.
This amps must be able to withstand abuse from PA use, so I want to build in all the features that are needed (basic ones) and usually found in commercial amps.
I am quite likely to re-think the leach amp itself, at least the choice of parts, but this initial design will stay with mj15003/4. I have planned for using other devices later.
I've seen commercial amps with little transformers on those xlr inputs, although often switchable (bypass). I wonder what they are for.
Perhaps the hf filtering could be put on the feedback on the INA???
If you're designing your own board, I'm not sure why you want to use the INA at all. Just lift the feedback cap from ground and use that as your negative input. Tie the input to ground with 100K or so just so the input isn't completely floating. For single ended operation just ground the negative input.
If you could give me the .asc - a good bet someone in my "camp"
would build the IPS - I would keep it VERY close to your design ...
(just change to 4.5-5.5ma VAS I) We already have the OPS (EF3).
"spookyamp leach ips ?? "
Cool!
I'll attach the sim and model files.
I am not totally set on its exact current design, except that for this first edition I do want to keep using the MJ15003/4. I know they aren't the best, but they'll do fine and leach himself used them as well.
They do contribute to some good amount of thd, as I found out when swapping them for MJL4281/4302, but they'll have to do for now, I'll make others later with other devices.
But as far as the other parts are concerned, those can easily change, although I do have the MJE15030/31 needed on hand, it would be worth getting better ones if that makes a serious difference.
The MJE15034/5 seem like much better parts, although lower powered, they should do find driving 3 pairs of MJ15003/4.
I have been trying to decide which small signal types to use. My latest schematic in eagle is using the mpsa42/92, but having simulated with bc546b and then trying bc550c instead later, I think I should try making use of those high gain types, and they are also low noise at the same time.
The bc546 has been used in leach amps (Jens uses them) with much higher rails than the original leach amp v4.5 (at 58V), so I suppose the cascoding on the diff amps helps limiting the exposure. But how far can that go?
Using the bc550 instead of 546 lowers the vce0 at 45 instead of 65, so is this doable? I wish I was more advanced in spice to help figure this out properly.
My current schematic also calls for MJE340/50 for the vas and pre-drivers, but unless my models are not so good, I found them rather inferior. The KSA1381/KSC3503 seem better by far (because of their speed???)
I am flexible on most parts and will consider improvements in the schematic.
I definitely don't want to keep the diodes as sensors for the heatsink/bias. I want to use a TO126 at least, or perhaps a TO220, but that doesn't seem warranted. I want a part bolted on the heatsink and not some wires going to some parts in through holes in the heatsink or a pcb with wires. I know the heat sensing may have a bit of a lag, but since the sensing transistor is right on the heatsink near the TO3s, this may be limited.
What really surprised me when doing the leach amp sim, is how much worse it performed thd wise compared to other inferior amps that I've tried. I've tried quite a few, from cordell's examples from his book, some other low tim by a french engineer who published it in a magazine, the elektor 3055 amp that we'll do as a grounded bridge.... Even took a peek at the honey badger and the symasym...
(wink wink)I was able to reach below 100ppm thd on that elektor amp, using 3055s, and the leach comes up around .76% with MJ15003/4 and the other parts as in that sim attached. How come???
That will change anyway, since I want input muting and a clipping prevention inserted before that amp. So this will change things quite a bit.
I do want both xlr plugs there, one female and one male, as a pass through (I use this sometimes) and I do want the real balanced input. Plus I do design my boards and everything else, including the metal case in cad. I have certain goals, but am open to many things.
Having simulated the leach amp now, I think there may be some easy improvements to make.
I'm designing the boards so nearly everything is on it, and I will replace the mechanical relays by SSRs and also replace the rail fuses with SSRs.
I'll need to design a good protection command logic...
btw: that leach amp sim I attached earlier doesn't really match my current amp design, as it's just about the original leach v4.5 with 2 pairs and the sensing diodes are there.
Besides the semiconductor choices that have changed in my new design, doing away with the heat sensing diodes and going for 3 pairs of MJs on the outout, the main thing that changes is the rails are going to about 61V, because I have toroids on hand for it already.
Other things that I will use because I have them are the non inductive dual power resistors for the outputs, which are 0.22 and not 0.33ohms.
In any case, the soa calculations for the protection locus must be done again to match.
Besides the semiconductor choices that have changed in my new design, doing away with the heat sensing diodes and going for 3 pairs of MJs on the outout, the main thing that changes is the rails are going to about 61V, because I have toroids on hand for it already.
Other things that I will use because I have them are the non inductive dual power resistors for the outputs, which are 0.22 and not 0.33ohms.
In any case, the soa calculations for the protection locus must be done again to match.
What I proposed is a real balanced input - rather than convert to single ended to feed one leg of a differential pair. It takes the balanced signal to positive and inverting inputs, just like the Aleph J. (see attached)
Of course, to use it this way you'll want a balanced attenuator...
Cascoding will work up to as close to 40+ the VCE of the front end devices as you dare go.
Of course, to use it this way you'll want a balanced attenuator...
Cascoding will work up to as close to 40+ the VCE of the front end devices as you dare go.
Last edited:
Yes, it's a good ans sensible idea. Plus it saves on parts.
But this will no longer be possible if there are more things inserted there, unfortunately.
What would the attenuator be for?
So the 40V is what the cascode operating point is set to by the zeners then.
So, for example in this case with the 61V rails, which could go up a few volts when unloaded or if the power lines are a little "rich", that would only expose the diff amps transistors to some 21V of vce then? Which remains quite within the bc550/60 capabilities at 45V vce0.
If this is correct, then we can use those high gain low noise tranies instead of the 546/56 for rails up to some 80V or a tiny bit more...
This would not be the case on the vas, but we need beefier ones there though.
Shouldn't there be a cap like C1 as well on the +IN (same value) and logically, this would also apply to C5????
Just trying to make sense of it.
You'd mentioned an attenuator or other circuitry to limit overdriving the amp. Implement it balanced and save balanced to single ended conversion.
No, the 40V string keeps the input differential at 39.4V, then you've got the balance of the rail voltage to take care of. Take a look at the Honey Badger - it is cascoded on both legs of the differential and uses ~15V cascode reference to keep the input pair at 14V and then something else above. As Pass says in the F5T article, cascode transistors don't need to do much other than not break. In practice they are level shifters. In the Leach one side of the differential isn't cascoded, but the load resistor manages the voltage protection.
No, the 40V string keeps the input differential at 39.4V, then you've got the balance of the rail voltage to take care of. Take a look at the Honey Badger - it is cascoded on both legs of the differential and uses ~15V cascode reference to keep the input pair at 14V and then something else above. As Pass says in the F5T article, cascode transistors don't need to do much other than not break. In practice they are level shifters. In the Leach one side of the differential isn't cascoded, but the load resistor manages the voltage protection.
Hi Guys
As far as having balanced inputs, it is best to keep that as a circuit separate from the PA feedback loop, especially if you want low-THD/TIM.
The whole TIM issue has been pretty much torn apart and relevant facts determined. Basically, anything that reduces THD20 reduces any risk of slew-induced effects. Input stage degeneration is all you need to be able to avoid TIM otherwise - noting that "degeneration" really means either actual degeneration, or more generally that there be low-gm. Jfets have low gm but are difficult to find in complements. BJTs exhibit low gm when operated at low currents. Bandwidth limiting the input is always a good idea and helps reduce worries about TIM.
As long as you stay away from Miller compensation, then slew rate will be independent from the compensation itself.
Device matching is the best (and hardest) path to achieving the lowest THD in any fully symmetric circuit. Most builders do not want to take the time to match, despite tonnes of evidence of its benefit.
I like the massive heat sinks and such like of the chassis, but believe you are building a service nightmare that will have difficulties to construct as well. Consider the narrowness of each PCB and the "canyon" you have to work within to bolt the output transistor carriers onto the heat sinks. A friend of mine used allen-head bolts and a ball-end tool, claiming good off-axis access to the bolts. I found allen-heads easy to strip.
Cascoding the VAS is problematic with respect to stability. If you want to gain about 10dB open-loop gain and reduce THD much more dramatically, use the EF-VAS - just two BJTs more than the basic design and entirely stable. The collectors of these BJTs should be decoupled.
Remember that you have to build this AND service it. The leads from the toroids under the boards should be long enough to provide service access.
I see you corrected the XLR wiring in post-79.
Note that Eagle can't assure correctness of your connections. It will only assure the board and schematic match - to a point. It will not warn you if two traces cross that should not.
12mil trace spacing only provides 100V spacing. If you are using higher than +/-40V rails spacing should be doubled to 25mil (0.025"). This assumes a solder mask is applied on both sides. Also, use at least 2-ounce copper.
Have fun
Kevin O'Connor
As far as having balanced inputs, it is best to keep that as a circuit separate from the PA feedback loop, especially if you want low-THD/TIM.
The whole TIM issue has been pretty much torn apart and relevant facts determined. Basically, anything that reduces THD20 reduces any risk of slew-induced effects. Input stage degeneration is all you need to be able to avoid TIM otherwise - noting that "degeneration" really means either actual degeneration, or more generally that there be low-gm. Jfets have low gm but are difficult to find in complements. BJTs exhibit low gm when operated at low currents. Bandwidth limiting the input is always a good idea and helps reduce worries about TIM.
As long as you stay away from Miller compensation, then slew rate will be independent from the compensation itself.
Device matching is the best (and hardest) path to achieving the lowest THD in any fully symmetric circuit. Most builders do not want to take the time to match, despite tonnes of evidence of its benefit.
I like the massive heat sinks and such like of the chassis, but believe you are building a service nightmare that will have difficulties to construct as well. Consider the narrowness of each PCB and the "canyon" you have to work within to bolt the output transistor carriers onto the heat sinks. A friend of mine used allen-head bolts and a ball-end tool, claiming good off-axis access to the bolts. I found allen-heads easy to strip.
Cascoding the VAS is problematic with respect to stability. If you want to gain about 10dB open-loop gain and reduce THD much more dramatically, use the EF-VAS - just two BJTs more than the basic design and entirely stable. The collectors of these BJTs should be decoupled.
Remember that you have to build this AND service it. The leads from the toroids under the boards should be long enough to provide service access.
I see you corrected the XLR wiring in post-79.
Note that Eagle can't assure correctness of your connections. It will only assure the board and schematic match - to a point. It will not warn you if two traces cross that should not.
12mil trace spacing only provides 100V spacing. If you are using higher than +/-40V rails spacing should be doubled to 25mil (0.025"). This assumes a solder mask is applied on both sides. Also, use at least 2-ounce copper.
Have fun
Kevin O'Connor
I got it ...
I really wanted to add this to the IPS lineup as it is a VFA that
HAS the CFA slewrate (from my first encounter with it.)
I'll keep it as close to your design as possible except where I
have to "conform" to my EF3's "rules". (4-6ma VAS I).
I have the EF3 "solved" flatpack opposed PCB layout with the dual Vbe
no leads for the Vbe and just cm's of trace between Vbe pre/driver.
Mje 15032/3 ?? hah ! NJW to-3P as driver !! My baby EF3 uses to-220,
but NJWxxxx is just as cheap as the 15032/33 with 3X SOA.
Mje15032/33 are stressed by (over) 3 pair MJL21193/4. A pair of NJW can
do 10pair with ease.
About to splice your IPS onto the 3-pair "slewmaster" EF3 ...
I'll show the results ...
Thanks - OS
I really wanted to add this to the IPS lineup as it is a VFA that
HAS the CFA slewrate (from my first encounter with it.)
I'll keep it as close to your design as possible except where I
have to "conform" to my EF3's "rules". (4-6ma VAS I).
I have the EF3 "solved" flatpack opposed PCB layout with the dual Vbe
no leads for the Vbe and just cm's of trace between Vbe pre/driver.
Mje 15032/3 ?? hah ! NJW to-3P as driver !! My baby EF3 uses to-220,
but NJWxxxx is just as cheap as the 15032/33 with 3X SOA.
Mje15032/33 are stressed by (over) 3 pair MJL21193/4. A pair of NJW can
do 10pair with ease.
About to splice your IPS onto the 3-pair "slewmaster" EF3 ...
I'll show the results ...
Thanks - OS
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.