Like what you see?
Subscribe to the Korf Blog!
- Be notified when new articles are posted
- Access members-only secret preview sales
- Get deals and discounts on Korf Audio products
Expect 1-4 emails a month. We will never ever spam you. You can always unsubscribe—no hard feelings!
The Korf Blog
The inside story: our research,
development and opinions
development and opinions
21 February 2023
Turntable Drives. Part VI, Direct Drive Feedback Loops
Last time, we spoke about direct drive torque ripple, and how we need to constantly adjust the driving current to compensate for the imperfections that exist in any motor. We also need to periodically check on our direct drive motor to make sure it's still rotating at the prescribed speed, right?
This appears to rule out what the specialists call "open loop motor control"—the way to run a motor by feeding it appropriate current, and simply ignoring whatever data might be given back on what the motor is actually doing. This is how most belt and idler drive motors operate.
The opposite of open loop control is "closed loop motor control"—a type of control that considers the feedback of motor signals like current and position. As far as I know, all historic and current turntable DD motors use closed loop control. The deviation from an ideal is measured and is used to correct the speed in a negative feedback loop.
This appears to rule out what the specialists call "open loop motor control"—the way to run a motor by feeding it appropriate current, and simply ignoring whatever data might be given back on what the motor is actually doing. This is how most belt and idler drive motors operate.
The opposite of open loop control is "closed loop motor control"—a type of control that considers the feedback of motor signals like current and position. As far as I know, all historic and current turntable DD motors use closed loop control. The deviation from an ideal is measured and is used to correct the speed in a negative feedback loop.
Negative Feedback Loop
If you read "Flat Earth" audiophile literature, you might think that negative feedback (NFB) is some kind of an audio universe's Ernst Stavro Blofeld, recurring villain of many a James Bond movie. Of course it isn't. Like any powerful tool, it is indispensable in skilled hands, and can wreak havoc when misapplied. Much of NFB's bad reputation stems from its uncanny ability to (almost) turn sow's ears into silk purses.
What follows is a deliberate oversimplification. Yes, I know that's not how electronic engineers think about NFB. If you can formulate it better while keeping the message accessible to lay persons, please do tell me.
NFB takes an output error of some device, inverts it (hence "negative") and feeds it right back into the input. In an ideal world, this subtracts the error from itself, and assures error-free output. Bingo!
In the real world, there's a pesky complication called "time". While an error is being measured, inverted and fed back, things have moved on. NFB loop is always chasing its tail. But if NFB works much quicker than the actual change we're tracking, this tail chasing is all but unnoticeable.
In the real world, there's a pesky complication called "time". While an error is being measured, inverted and fed back, things have moved on. NFB loop is always chasing its tail. But if NFB works much quicker than the actual change we're tracking, this tail chasing is all but unnoticeable.
Negative feedback is a great way to trade bandwidth for linearity
Or, put differently: negative feedback is a great way to trade bandwidth (speed of change) for linearity. For example, with a well-designed solid state audio amplifier, there's a lot of bandwidth to trade. We only need 30-40 kHz, while building an amp with 100+ kHz bandwidth is trivial. But do the direct drive turntables have that extra bandwidth?
Let's look at our DD diagram again. This time, I have added time constants τ to the elements that limit the bandwidth.
Let's look at our DD diagram again. This time, I have added time constants τ to the elements that limit the bandwidth.
There's a low pass filter with a time constant \(\tau _{0}\), and the motor plus the platter have a time constant \(\tau _{1}\) defined by the relationship between their combined inertia and the motor's power. Any NFB loop's bandwidth will be constrained by the sum of those two.
How big are those time constants?
Examining the Dual EDS 900's schematics will give us \(\tau _{0}\) of 0.004 seconds. \(\tau _{1}\) is a lot harder to estimate, and with our Dual it's almost an order of magnitude larger with 0.01 seconds. Adding the two will give us 14000 µs time constant and 11.4 Hz bandwidth of our negative feedback loop. To be effectively eliminated by the NFB loop, the errors have to come slower than that.
What does this mean? A typical direct drive turntable has absolutely no excess bandwidth to trade for the desired speed stability. This is why the momentary speed histogram of a typical DD turntable looks like a Golden Gate bridge:
How big are those time constants?
Examining the Dual EDS 900's schematics will give us \(\tau _{0}\) of 0.004 seconds. \(\tau _{1}\) is a lot harder to estimate, and with our Dual it's almost an order of magnitude larger with 0.01 seconds. Adding the two will give us 14000 µs time constant and 11.4 Hz bandwidth of our negative feedback loop. To be effectively eliminated by the NFB loop, the errors have to come slower than that.
What does this mean? A typical direct drive turntable has absolutely no excess bandwidth to trade for the desired speed stability. This is why the momentary speed histogram of a typical DD turntable looks like a Golden Gate bridge:
Here, t is reaction time to a typical disturbance, defined by the system's total τ and the magnitude of disturbances. Looks familiar? Yes, this is exactly the type of chart we have seen at the end of the introductory post on DD drives.
What Can be Done to Fix This?
For one thing, we can try to decrease τ as much as possible. Build a more powerful motor, and make the platter as light as possible. This would allow the use of a less aggressive low pass filter.
Good examples of this are JVC/Victor and Denon with their UFO series. And, of course, the utterly mad EMT 950. But the potential of this approach was very quickly exhausted. There are hard limitations on how light the platter and moving parts of the motor could be. And don't forget the mass of the LP itself. The EMT 950, built on a virtually unlimited budget, did not sound materially better than a much cheaper SP-10Mk.III
Good examples of this are JVC/Victor and Denon with their UFO series. And, of course, the utterly mad EMT 950. But the potential of this approach was very quickly exhausted. There are hard limitations on how light the platter and moving parts of the motor could be. And don't forget the mass of the LP itself. The EMT 950, built on a virtually unlimited budget, did not sound materially better than a much cheaper SP-10Mk.III
But what if we focus on fixing the future instead of the past? Eliminate τ completely rather than merely decrease it? What if could correct the speed deviations before they actually happen? The platter is smoothly rotating time after time after time. The irregularities repeat every turn (unless our system is underpowered and grossly disturbed by the variation in stylus drag).
For every angular position of the platter, we will pre-fill a table with the required correction. Every time the platter rotates, we will know what correction to apply before it is called for.
This is called feed forward correction.
For every angular position of the platter, we will pre-fill a table with the required correction. Every time the platter rotates, we will know what correction to apply before it is called for.
This is called feed forward correction.
It is still closed loop control; to apply these correction we do need to measure the current angle of the platter. But it's not a negative feedback loop, and the bandwidth of the whole system is irrelevant to its correct operation.
It's easy to see why feed forward correction wasn't applied in the 1970s or 1980s. The microcontrollers available back then were just not capable of anything remotely as sophisticated.
Is anyone using feed-forward control now? Technics SP-10R has an interesting undocumented calibration mode. A combination of keypresses gets it to rotate its platter really slowly a few times. I have a hunch that it's filling a feed-forward table this way.
It's easy to see why feed forward correction wasn't applied in the 1970s or 1980s. The microcontrollers available back then were just not capable of anything remotely as sophisticated.
Is anyone using feed-forward control now? Technics SP-10R has an interesting undocumented calibration mode. A combination of keypresses gets it to rotate its platter really slowly a few times. I have a hunch that it's filling a feed-forward table this way.
Direct Drive—Summary
That was quite a handful, wasn't it? DD's simplicity is deceptive, and the more you dig the more bewildering complexity comes to the surface. Of course we haven't excavated all the way; there's a lot more to direct drive than what we covered.
As usual, here are the key strengths and weaknesses of a directly driven turntable. Please keep in mind that those describe existing turntables, not to some ideal DD that is yet to come.
As usual, here are the key strengths and weaknesses of a directly driven turntable. Please keep in mind that those describe existing turntables, not to some ideal DD that is yet to come.
+
Potentially perfect (traditionally) measured performance
+
Mechanical simplicity, very few moving parts
+
Very easy to design a bearing for (no side loading)
+
Good reaction to load variation, no elasticity in the drivetrain
-
Most existing implementations subjectively underwhelming
-
(typically) poor momentary speed control
-
(typically) poor torque ripple rejection
-
Electronics complexity
Direct Drive is audio's child prodigy
Turntable Direct Drive is one of audio's child prodigies. Amazing promise of its early years did not translate to anything but mediocrity (or worse) in maturity. The technology of the time was simply not ready for the ideas it brought forward.
However, the insurmountable challenges of 1970s are trivial schoolboy stuff now. Even bottom-of-the-line microcontrollers can create a passable DDS sine wave, or store a few kilobytes of a feed forward table .
Is anyone really developing DD turntables today? Matsushita (Technics) certainly does. VPI, Grand Prix Audio and Brinkmann also contest the high end price niche. Down below, it gets lonely—Technics and... Technics 40 years ago (that's what most Chinese OEM DDs are). Just like with the idler drive, I think there's an opportunity here.
Is anyone really developing DD turntables today? Matsushita (Technics) certainly does. VPI, Grand Prix Audio and Brinkmann also contest the high end price niche. Down below, it gets lonely—Technics and... Technics 40 years ago (that's what most Chinese OEM DDs are). Just like with the idler drive, I think there's an opportunity here.
Please subscribe to receive blog updates in your inbox!