Slew Rate in Preamps

Introduction

You may probably already asked yourself: what is this “slew rate” business? Has this guy lost his mind? Probably I have, but not due to this interesting phenomenon. The second question that should have probably popped in your brain is: why? Well, writing about this stuff came up by sheer coincidence.

We tested my friend Tony’s 01a preamp which has an older gyrator board I made for him about 3 or 4 years ago. The preamp (as well as his system) sounds extremely good in my opinion and the 01a has brought a new clarity which is what you’d expect as a result of the introduction of a DHT stage. We decided to run some frequency response tests since Tony has made some interesting mods to his Push-Pull amp. We encountered an unexpected challenge as I didn’t have the right XLR connectors for my testing gear so we run the FR tests on the power amp with the 01a.

The 01a preamp measured as expected with a flat response up until 40kHz before the sound card rolloff kicked in – I’m aware of this and this is the limited bandwidth of my current portable measurement gear. The distortion was also really low (H2 predominantly ) with less than 0.02% for 2 Vrms output signal. This test was done with Pete Millett’s interface which has an input impedance of 100kΩ:

Tony's 01a FR
Tony’s 01a FR
Tony's 01A THD
Tony’s 01A THD

 

 

It was all going as planned until we found surprisingly an HF roll-off of about 15kHz in the amp, which wasn’t expected at all. This was a result of driving the output amp to 5W so the amp input voltage level was not low (but I didn’t record which level was anyway).

Tony's PP amp FR at 4W output into 8R test load
Tony’s PP amp FR at 4W output into 8R test load

Someone was to blame here for the 15kHz roll-off! At first glance, this could be either:

  1. The output TX – to be discarded as these are Lundahl’s LL1620
  2. The input TX – LL1676 wired in 1:1 phase splitting into a 7N7 LTP driver stage. I could suspect of the leakage capacitance of the transformer
  3. The 01a was running below 3mA – I started to think of a potential slew rate issue here, but Tony didn’t know what was the quiescent current 🙂

After several tests, we opted to bypass the input transformer and found what I was suspecting: a nice and flat response of the PP amp:

Tony's PP amp FR at 2W output into 8R test load - input step up TX (LL1676) bypassed
Tony’s PP amp FR at 2W output into 8R test load – input step up TX (LL1676) bypassed

 

The 01a was struggling to drive the amplifier’s input transformer. Why?

Because of Slew Rate limitations!

 

Some Theory (don’t skip this section)

There are too factors which influence the dynamic and frequency response of our amps or preamps and some people tend to get confused about:

  1. Output impedance: the output impedance of the amplifier forms an HF pole with the next stage’s input impedance and introduces a high-frequency roll-off.
  2. Slew Rate: in simple terms the slew rate is the ability of the stage to change the voltage of the output fast enough to meet the load’s current requirements. There is plenty of technical theory stuff on this around which is mainly focused on Op Amps and solid state but you can get a good background if you’re interested in.

These two factors are independent of each other which means that despite we may have a low output impedance, the response of the stage at HF may be affected at specific output levels. Why? well, slew rate is dependent on frequency, capacitance of the load, current capability of the stage but is not dependent on the output impedance.

Slew Rate (Oh Dear, here we go again!)

Any stage has some limitation of the amount of current they can provide – source (or take – sink) to a specific load. If the load has some capacitance (e.g. input capacitance of the valve due to Miller effect, leakage capacitance of a transformer, etc.) then the current charging the capacitor will follow the well known formula:

I_{C}=C\cdot \frac{\mathrm{d} V}{\mathrm{d} t}

In other words, so the larger the capacitor or the fastest the voltage changes in time, the more current flows in (or out) of the capacitor.

This is the main reason why driving output stages is generally a challenge. Many poor implementations of 300B, 2A3 and even transmitting valves (845, 211, etc) are driven by wimpy valves which have a large mu but high output resistance but what is worse just a few milliamps of current (e.g. 6SL7). The input capacitance of these output stages is generally greater than 200pF and lots of volts are needed (150 or 200V peak to peak) so you can figure out why they sound poorly and some people complain about the 300B being dull or not good enough! Before you jump straight on, yes: what you’re thinking is right:

The introduction of a source follower or cathode follower to the output of a driver stage  is a wise approach to mitigate the Slew Rate problem in most of the valve drivers when there isn’t enough current driving capability available

I’m not going to bore you with calculus ,but the interesting formula which we want to use is the following one if we look specifically at sine waves:

I_{C}=C\cdot2\pi\cdot f\cdot V_{peak}

The above tells you the charging current as a function of the peak voltage of the stage, capacitance of the load and frequency. Again, what is clear from this equation is:

The current demand of a driver is proportional to the capacitance of the load, frequency and peak voltage of the output signal

If we re-arrange the formula we could determine the maximum capacitance that the stage can drive effectively at a specific frequency and output peak voltage:

C_{max}=\frac{I_{max}}{f\cdot2\pi\cdot V_{peak}}

So what is the “rule of thumb” here? You want to consider the following two practical points:

  1. Frequency: at least we should design or evaluate at the maximum frequency: 20kHz.  50kHz may be an overkill, but if you have the headroom, then yes.
  2. Maximum current of the stage: 5 times (14dB) headroom in the current is ok to ensure we can cope with any transients in the music. Again, x10 may be an overkill in my view.

You can use the formula in the way that best suits your needs. If you know the maximum current of your stage (Imax), the driver output voltage (Vpeak) and the load, you can figure out either the roll-off frequency (f) or the maximum capacitance that the stage can drive.

Analysis of the 01a Slew Rate

The issue we experienced here is that the 01a cannot sink more than 3mA of current hence limits the speed at which it can discharges a capacitive load. This is evident when swinging more volts. During our tests we measure the frequency response of the amp using the 01a preamp but driving it at 5W output power, which was closer to its limits. The slew rate of the 01a limited the driving capabilities of the amplifier and it depended solely on the peak voltage, capacitance of the load and frequency. Nothing to do with the output impedance of the stage, which is low thanks to the gyrator.

The maximum load capacitance for this 01a preamp circuit is 2.4nF for a 10V peak output at 20kHz. However, the recommendation is to have x5 (14dB) headroom to accommodate well any transient peaks in the music, so 470pF would be maximum.
When we listed to the 01a driving a 4P1L SE amp this stage sounded fantastic. No evidence of HF rolloff. So I looked at my notes when measuring the Miller capacitance of the 4P1L in triode, and the total input capacitance (Ci=Cgk+Cm) was around Ci=180pF.
So the 01a can drive the 4P1L well as a preamp, but I don’t think it could be used as a driver when you’d expect to swing 20V or more.
Spice simulation of the 01A preamp experiencing Slew Rate limitations
Spice simulation of the 01A preamp experiencing Slew Rate limitations

For example if you were considering to use the 01a as a driver the 4P1L will need 20Vpeak, so the distortion introduced due to the “Slew rate” limitation is considerable. The THD at 20kHz for 20Vpeak output with a 200pF load is around 0.13% which is about 5 times more. This may not be noticeable, but I’d not recommend pushing the 01a into a role which isn’t suitable for.

Key take aways  (doggy bag)

  • The slew rate of the 01a was responsible here for limiting the treble for a large capacitance load. A single stage will not a problem, an input transformer with higher leakage capacitance may be an issue (as we found here).
  • The slew rate has nothing to do with the output impedance of the preamp OR the bandwidth. It has to do with the peak output voltage, capacitance and frequency.
  • The 01a slew rate will only affect the treble at very high output power levels (when 01a output is high), at normal listening levels is unlikely to be a problem. Despite this, we found Tony’s system to sound really well.
  • Proof will be in the pudding. Tony is planning a minor mod as described below, and we will run again the measurements and the listening tests 🙂

 A simple workaround

A quick solution to address this (and many other low-current valve stages) is to add a follower to the 01a. A simple follower will be a MOSFET follower as shown below:

THD of 01a preamp with follower @ 20kHz / CL=2,000pF
THD of 01a preamp with follower @ 20kHz / CL=2,000pF
01a Preamp with output follower driving a heavy load!
01a Preamp with output follower driving a heavy load!

 

A more sophisticated follower could be implemented with a tail CCS, but this will do to prove the point. In the above simulation, the follower is driving 2,000pF at 20kHz. With the 10mA of sinking current, it can now drive the capacitance without any challenges.

Hope you found this useful, unfortunately I ran out of time and can’t type further on this interesting subject…

Merry Christmas!

Author: Ale Moglia

"A mistake is always forgivable, rarely excusable and always unacceptable. " (Robert Fripp)

3 thoughts on “Slew Rate in Preamps”

  1. Very nice article Ale. Very nice.
    One other possible way to tackle the problem is to add a CCS from the mu output of the gyrator to ground. So there is, say, 23mA running thru the gyrator, 3mA of that through the 01a and the other 20mA to ground. The gyrator FETS will be happier at higher currents too. Higher transC, lower impedance etc.
    No idea which might sound “better”.
    Merry Christmas
    tim

    1. Hi Tim, good feedback. I’ve seen the CCS to ground and looks like another simple way of addressing the same problem. Need to test it to see how it sounds 🙂
      Cheers
      Ale

  2. Excellent article Ale. This is the appropriate way to approach a good design with valves. This answers the question: ” can I drive this amp/cable with the preamp?”. It would be interesting to create a table with most popular power tubes and their best/worst drivers.
    I tried the 01A to drive different capacitive loads at different Vpeak levels and I can confirm that our findings are similar. Soundwise, there is a noticeable difference between 01A and 4p1l when driving for example the F4: at lower levels the 01A sound is gorgeous, but at higher levels it loses dynamics and the sound is thinner. 4p1l on the opposite sounds excellent at both low and high levels
    Best regards,
    Radu

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.