Hi!
Newbie here! I recently started dabbling in amplifier design. What an addictive hobby! I've learned more than I ever could have imagined.
Anyway, I put together the amplifier in the attached schematic. It's fairly basic and classic design. I like the way it sounds and I get respectable values in both simulations and real measurements. So far, so good!
However, I was recently told to lose the current mirrors on the LTP by someone (who's probably on this site, so feel free to chime in). The reason was that they supposedly make the sound "grainy". I tried to replace the current mirrors with 2k resistors in SPICE and it increased my THD from 0.004% to 0.15%. Suffices to say I wasn't impressed.
I guess my question is whether there's any truth to current mirrors having a detrimental effect on sound. I'm here to learn, so lay it on me!
Newbie here! I recently started dabbling in amplifier design. What an addictive hobby! I've learned more than I ever could have imagined.
Anyway, I put together the amplifier in the attached schematic. It's fairly basic and classic design. I like the way it sounds and I get respectable values in both simulations and real measurements. So far, so good!
However, I was recently told to lose the current mirrors on the LTP by someone (who's probably on this site, so feel free to chime in). The reason was that they supposedly make the sound "grainy". I tried to replace the current mirrors with 2k resistors in SPICE and it increased my THD from 0.004% to 0.15%. Suffices to say I wasn't impressed.
I guess my question is whether there's any truth to current mirrors having a detrimental effect on sound. I'm here to learn, so lay it on me!
Why on earth should they? Their only bad consequence is higher open loop gain, so make sure you get the compensation correct. Don't try for a few extra "0"s in the THD. The good effect is the gain and supply rejection
I was wondering the same: "Why on earth would they do that?". And I agree that anything below 0.1% THD is mostly vanity and bragging rights. The compensation is fine. I think I have very close to 90 degrees phase margin in SPICE if I remember correctly. It behaves nicely without any ringing on square waves or other shenanigans.
Yeah. my goal with this design was basically "a wire with gain". I think I achieved that to some extent at least.
Not just second order harmonics, BTW. All the harmonics!
The only way you’re getting .004% is if Q1 and Q2 are matched as well as Q3 to Q4. The offset pot is a band aid and unnecessary with matching (including the degen resistors). But it’s not THAT tall an order, so worth doing. And keep the current mirror.
The SE input stage allows one to get something between .15 and .004%, without anything being critical. The collector resistor in the LTP is critical if the mirror is dispensed with. The current through one side will be forced to whatever is needed to bias the VAS, and they may no longer be equal. To get decent differential performance the CURRENTS need to be the same. To get world class, the vbe’s need to be identical too. The only way to achieve both is hand matching devices.
The SE input stage allows one to get something between .15 and .004%, without anything being critical. The collector resistor in the LTP is critical if the mirror is dispensed with. The current through one side will be forced to whatever is needed to bias the VAS, and they may no longer be equal. To get decent differential performance the CURRENTS need to be the same. To get world class, the vbe’s need to be identical too. The only way to achieve both is hand matching devices.
@wg_ski I agree that 0.004% is a theoretical number assuming everything us perfectly matched. I don't have great equipment for measuring THD at that low, but my ballpark from looking at the oscilloscope FFT and doing some napkin math is that I'm at about 0.05%. It's below 0.1%, which is all I'm asking for. Now way that's audible. Pardon the noise. It's due to poor probing.
And yes, the pot on the LTP emitters is actually pointless and I'll remove it if I ever build it again. DC offset is about 20mV and it doesn't change if I turn the pot, thanks to the current mirrors. As I said, I just started dabbling in this and I'm very much still learning. But the pot was a mistake for sure.
And yes, the pot on the LTP emitters is actually pointless and I'll remove it if I ever build it again. DC offset is about 20mV and it doesn't change if I turn the pot, thanks to the current mirrors. As I said, I just started dabbling in this and I'm very much still learning. But the pot was a mistake for sure.
You can approach theory if you DO match parts. .05% is par for the course just sticking in random parts that come out of the same bag. Just buying modern types that tend to be consistent goes a long way, when lacking sophisticated equipment. And you won’t even get .05% without the mirror unless you trim the collector resistor for balance.
The reason was apparently "poor base charge removal" on the transistors. While I understand what the words mean, I'm not sure I get how is applies to my design.
I had to look up the term and found a thread on this site. I think it means harsh and screechy. Some people talked about intermodulation (which seems to be a favorite term to throw in to sound like you know what you're talking about).
Yeah, I only had about 20 transistors of each kind, so it's hard to get an exact match. I tried to get the the HFE within 15-20 units. Not ideal, but this is only my second design and build (not counting one purely experimental design), so I'm pretty happy with the result. It sounds good enough for very enjoyable everyday listening.
BTW - “grainy” is high order IMD. If someone builds a Blameless topology and complains of “grain” they are either hearing things that aren’t there or they did something fundamentally wrong. There are lots of ways for the latter to happen.
@wg_ski To clarify, I’m not complaining about grain. I like the way this amplifier sounds. The person who recommended removing the mirrors also told me to get rid of the Miller cap and put in lag compensation instead. I wasn’t crazy about what that looked like in SPICE either.
The reason I went for a blameless-inspired design was that they seem idiot proof and I still consider myself an idiot when it comes to amplifier design. 😛
The reason I went for a blameless-inspired design was that they seem idiot proof and I still consider myself an idiot when it comes to amplifier design. 😛
The transistor choices were based on the dozen or so transistors I’m familiar with. Bear with me as I’m getting the hang of this. 🤣
I'm a big fan of Doug Self and the Blameless designs but watch out for one detail. Check the clipping behavior on the top and bottom of the waveform. Look for recovery time issues and oscillation during clipping. I did a similar design and it was very difficult to solve those issues.
Mirrors don't sound grainy
Its just basic DC current for balance.
Be easier and better to just use normal BJT stage for outputs.
Verticals can be rugged devices and easy to drive.
For actual measurable performance at high frequency it involves much more current and methods
For a straight forward common design as shown.
High frequency with normal BJT will be improved.
You always use excessively high values for your VBE multiplier C6 just needs to be 100n
And these are not lateral mosfet so they do not need excessively high gate resistors to actually be stable
R8 , 18 not really more than 33 to 47 ohms needed. High values could actually cause stability issues
and will ruin the slew of the already not so great high frequency.
Its just basic DC current for balance.
Be easier and better to just use normal BJT stage for outputs.
Verticals can be rugged devices and easy to drive.
For actual measurable performance at high frequency it involves much more current and methods
For a straight forward common design as shown.
High frequency with normal BJT will be improved.
You always use excessively high values for your VBE multiplier C6 just needs to be 100n
And these are not lateral mosfet so they do not need excessively high gate resistors to actually be stable
R8 , 18 not really more than 33 to 47 ohms needed. High values could actually cause stability issues
and will ruin the slew of the already not so great high frequency.
It’s not that difficult to solve those issues. Provide a current clamp for the VAS and 99% of the problems go away. If you include SOA limiting for the output stage it becomes MANDATORY so it’s a moot point. When you can actually hear the 500 ns of residual rail stick I’ll eat that hat. It’s far easier to hear 120Hz IMD products from the supply ripple when clipping.