ProSoundWeb Community

Please login or register.

Login with username, password and session length
Advanced search  

Pages: [1]   Go Down

Author Topic: Switching amp hash filter  (Read 2453 times)

Frank Koenig

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 1318
  • Palo Alto, CA USA
Switching amp hash filter
« on: November 11, 2021, 11:42:18 AM »

The power amps in my life these days, both large (Powersoft K) and small (Parts Express hobby boards), are switch-mode. They all exhibit a few hundred mV to several V of switching hash on their outputs. This is of no consequence in normal operation but makes it pretty much impossible to see any meaningful low-level output on a scope.

With nothing better to do yesterday I cobbled together a simple second-order Butterworth low-pass and put it in a little box with banana/binding-post input and BNC output. It uses an ancient, 88 mH, air-core, basket-weave inductor, a 440 pF capacitor and a 10 kOhm load resistor which yields a cutoff frequency of ~27kHz.

At 20 kHz it's .6 dB down and introduces -72 deg of phase shift (lag?  ;) ) so cannot be used for critical frequency response measurements but it cleans up the hash and I hope will be useful for looking at low-frequency noise, DC offset, low-level distortion and other problems. We'll see.

--Frank
Logged
"Nature abhors a vacuum tube." -- John Pierce, Bell Labs

Guillermo Sanchez

  • Newbie
  • *
  • Offline Offline
  • Posts: 18
Re: Switching amp hash filter
« Reply #1 on: April 26, 2022, 10:08:12 AM »

The power amps in my life these days, both large (Powersoft K) and small (Parts Express hobby boards), are switch-mode. They all exhibit a few hundred mV to several V of switching hash on their outputs. This is of no consequence in normal operation but makes it pretty much impossible to see any meaningful low-level output on a scope.

With nothing better to do yesterday I cobbled together a simple second-order Butterworth low-pass and put it in a little box with banana/binding-post input and BNC output. It uses an ancient, 88 mH, air-core, basket-weave inductor, a 440 pF capacitor and a 10 kOhm load resistor which yields a cutoff frequency of ~27kHz.

At 20 kHz it's .6 dB down and introduces -72 deg of phase shift (lag?  ;) ) so cannot be used for critical frequency response measurements but it cleans up the hash and I hope will be useful for looking at low-frequency noise, DC offset, low-level distortion and other problems. We'll see.

--Frank

Hello Frank, any news on this? Did it worked as you intended?
Logged

Frank Koenig

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 1318
  • Palo Alto, CA USA
Re: Switching amp hash filter
« Reply #2 on: April 26, 2022, 12:15:09 PM »

Hello Frank, any news on this? Did it worked as you intended?

I haven't used it in anger yet but in initial testing on one of the little Parts Express amps it worked fine. If and when I get one of the big pro amps on the bench for some reason I'll take some scope screen shots and put them up here. --Frank
Logged
"Nature abhors a vacuum tube." -- John Pierce, Bell Labs

John Roberts {JR}

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 17172
  • Hickory, Mississippi, USA
    • Resotune
Re: Switching amp hash filter
« Reply #3 on: April 26, 2022, 12:35:04 PM »

I don't know what you are using for distortion measurements but if it has a distortion product output, feed that into a spectrum analyzer.

Back in the 70s I feed the product output from my cheap heathkit distortion analyzer into an old spectrum analyzer (that I paid hundreds of dollars for used). The spectrum analyzer only had a 50 dB dynamic range but the combination allowed me to measure the distortion introduced by the Heathkit audio path. By running the 0dB level through the Heathkit -10dB added 20dB net resolution to my test bench measurement floor. 

JR   
Logged
Cancel the "cancel culture". Do not participate in mob hatred.

Frank Koenig

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 1318
  • Palo Alto, CA USA
Re: Switching amp hash filter
« Reply #4 on: April 26, 2022, 07:26:32 PM »

I don't know what you are using for distortion measurements but if it has a distortion product output, feed that into a spectrum analyzer.

Back in the 70s I feed the product output from my cheap heathkit distortion analyzer into an old spectrum analyzer (that I paid hundreds of dollars for used). The spectrum analyzer only had a 50 dB dynamic range but the combination allowed me to measure the distortion introduced by the Heathkit audio path. By running the 0dB level through the Heathkit -10dB added 20dB net resolution to my test bench measurement floor. 

JR   


I'm using an HP 8903B distortion analyzer that a friend gave me when he got an Audio Precision (lucky dog). The HP indeed has a distortion product output on the back panel which I've fed into a Tektronix DPO 3034 digital scope and viewed in both time and frequency. (The scope includes an FFT-based spectrum analyzer function.) The thing that didn't make sense to me was that the fundamental wasn't as attenuated as I thought it should be. When I do a loop-back test on the analyzer (connecting its internal sine generator directly to the analyzer input) it reads within spec -- .03% or so, as I recall.

I got distracted and moved on. I need to go back and figure out why the spectrum analyzer display and the THD reading failed to agree. Maybe the distortion product output is after a pre-filter but before the synchronous detection, or whatever the analyzer usus, to make the super-notch?

One interesting thing about the analyzer is that it works with any roughly sine-wave input -- you don't have to use the internal generator. So it's useful for measuring distortion of whatever sine-source you've got or in cases where you don't have analog access to the system input, such as a Bluetooth receiver. It's a pretty cool box and it behooves me to understand better how it actually works internally.

--Frank
Logged
"Nature abhors a vacuum tube." -- John Pierce, Bell Labs

John Roberts {JR}

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 17172
  • Hickory, Mississippi, USA
    • Resotune
Re: Switching amp hash filter
« Reply #5 on: April 27, 2022, 10:37:30 AM »

Distortion analyzers are generally measuring THD + N(oise). Noise can significantly alter THD+N readings.

JR
Logged
Cancel the "cancel culture". Do not participate in mob hatred.

Frank Koenig

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 1318
  • Palo Alto, CA USA
Re: Switching amp hash filter
« Reply #6 on: April 28, 2022, 03:37:08 PM »

I looked at the service manual for the HP 8903B audio analyzer, which dates from the early 1990s, and reveals a good bit about the design of the notch filter. It is an analog tour de force in the form of an electronically-tunable, 2nd-order, state-variable analog filter. The tuning elements are analog multipliers that get a coarse tuning input from a frequency counter that measures the frequency of either the internal generator or the input signal. A fine tuning input comes from phase and amplitude detectors that compare the input and bandPASS output of the filter so as to maintain the bandreject filter's null at the spectral peak of the signal. Using the bandpass output provides a robust signal for the phase measurement -- very clever. The coarse tuning ensures that the phase detector hunts for the correct peak.

I assume they used very high quality components and advanced design ticks, which they do not detail in the manual, to get the high Q and stability needed to make this work. I think this might be one of those examples of an old technology that reached its climax well after it was being supplanted by its successor. Steam locomotives in the 1940s, piston-engine airliners in the 1950s, CRT displays in the 1990s and rotating magnetic storage today come to mind as other examples.

The block diagram shows the residual output downstream of the notch filter, as expected. I'll revisit the measurements one of these days when I get that test bench urge. The weather is too nice right now.

--Frank
Logged
"Nature abhors a vacuum tube." -- John Pierce, Bell Labs

John Roberts {JR}

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 17172
  • Hickory, Mississippi, USA
    • Resotune
Re: Switching amp hash filter
« Reply #7 on: April 28, 2022, 08:08:10 PM »

back in the 90s I had an 8903 on my test bench at Peavey... it is a solid citizen.... The more expensive computer controlled test gear often ended up out in factory doing automated testing (ATE). While the 8903 was too expensive for me to own back before Peavey hired me.  8)

JR

PS: For TMI I discovered a design flaw (?) in a competitors SKU during bench testing when I switched the 8903 output impedance to 600 ohms and saw the frequency response of the DUT change.   ::)
Logged
Cancel the "cancel culture". Do not participate in mob hatred.

ProSoundWeb Community

Re: Switching amp hash filter
« Reply #7 on: April 28, 2022, 08:08:10 PM »


Pages: [1]   Go Up
 



Site Hosted By Ashdown Technologies, Inc.

Page created in 0.044 seconds with 22 queries.