Not me. The average purchase price for the test equipment on my (Tubelab) workbench is about $50 each.
My bench at work is a different story. There is an easy $250K worth of test equipment on my bench. It is mostly all RF stuff though. I do RF and a little digital or DC distribution work, no audio at all. The 8903A on my bench doesn't get used.
I used to get under 1% at 1 watt rising gently to about 2% as you approach clipping. Then the distortion goes up rather quickly. You get about 40% when grossly overdriven. The distortion and the power output is lower in triode mode. It is also lower with CFB. I haven't measured a Simple SE in about 2 years. One of my amps gets experimented on and is never in stock condition. It is currently non - functional. The other is hooked up to my speakers and sound card and is used for listening only. I am going to swap it out for a Simple P-P soon.
I have an older HP 8903A which measures THD, and audio voltage level. Built in arithmetic functions make frequency response measurements easy. I also used an Audiophile 2496 sound card and a rather crude home made interface. It was in a dedicated computer that ran Win MLS software. The measurements from these two systems matched pretty good. Unfortunately I robbed that computer for parts about a year ago.
I bought one of Petes boards, grabbed an Audiophile 192 off of Ebay, and got a $99 bare bones computer from Newegg, so I will be back in business when I get a chance to build it all.
I recently purchased Pete's kit, and still have to order the various parts. I probably will skip the rms voltmeter as I have both a Keithley 2002 (2MHz true rms converter) and a Fluke which are good enough.. I purchased an antec aria case with Athlon 64 3000, and an Abit KV-80 mobo with a bad psu. I am familiar with both the case and the processor as I use both in other systems. (The seller indicated that the mobo tested ok with another psu, and examining the failed psu indicated it failed benignly - I'm debating whether to fix it or not.) I've ordered a replacement supply, dvd-rw drive, and have an almost unused 160GB HD.. The Aria/mobo/processor ran me $44.. I also scored an Audiophile 24192 on sale at amazon recently.
At one point my old setup functioned satisfactorily, but the new(er) laptop doesn't seem to play nicely with the m-audio transit I used previously. The attraction of 192kHz sampling rate and a PC able to run large FFTs without choking was too hard to resist once I saw Pete's board.
It's nice to be able to make consistent and repeatable measurements, something that my current FFT setup no longer seems capable of doing..
.
Some of my amplifiers have had enough bandwidth to need a 192 KHz card for frequency response measurement. The old 2496 card was within 0.1 db out to about 42KHz. If the Audiophile 192 will go to 80KHz I'll be happy. The 8903A goes from 20 Hz to 100 KHz, although the AC voltmeter seems accurate down to 5Hz.
Big FFT's need big processing power. No problem with the Athlon, but the dedicated PC that I used to use was P3 based with 1 G of memory and a bit on the wimpy side. The new box uses a core II duo with 4 G, so I should be able to process some 1 M point FFT's in less than a week!
We think alike..
I also have a 2496 which lives in the media server and provides analog i/o for speaker measurement, as well as sound for the home theater set up.. The digital spdif output goes to my dac and is actually supporting the primary purpose of the media server to serve high quality digital to my stereo system. (I spent a lot of time getting it "bit perfect." Strangely enough these days I use my Shigaclone a lot more. I like selecting and handling the disks I guess, the difference in sound quality is probably non-existent.)
Precede the 'scope with an analogue distortion meter and take the oscilloscope output to the 'scope for FFT. That way, you have the dynamic range of the analogue meter with the detail of the FFT. Or just use Pete's card and a good sound card.
.
To make this statement a bit more clear one needs to understand how an analog THD analyzer works:
A signal from a pure sine wave source is applied to the unit under test. Its output is connected to a suitable load and also connected to the THD meter.
First the meter takes a reading of the total signal voltage.
Then a notch filter is engaged to remove as much of the original sine wave signal as possible.
The meter then takes a reading of the remaining signal voltage after the original fundamental tone is removed. The remaining voltage is assumed to be THD but may also be hum and noise.
The meter will display the ratio of the original signal to the signal remaining after the original fundamental tone is removed by the notch filter. These results can be displayed as % distortion, or as db's.
These functions can be completely manual as with the old HP 331A's or automated as in the HP 8903's.
Most analog distortion analyzers have an output that is taken after the notch filter so that the signal remaining after the fundamental is removed can be looked at with a scope, or analyzed by FFT. It is also possible to chain the output from one THD analyzer to another for IMD measurements. I keep a pair of old HP331A's around for this purpose.
The 331A does not have a built in audio source. Any imperfections in the source signal will show up as THD in the measurements. I use an old HP204D for a source. I have two for IMD measurements. Some of these are better than others in the THD department. One of mine measures .05% distortion which is good enough for tube amps. The other one is .12%.
There are software audio generators for use with a sound card. There are also test CD's that can run in a CD player. Both are good alternatives for an audio test source.
The output of the THD analyzer can be connected to the scope or sound card. Set the THD analyser up for level measurements and measure its output on the scope or sound card. Absolute accuracy is not needed, only a relative level. Then engage the filter in the analyzer to remove the fundamental tone. The output of the analyzer will drop by 10's of db because the main signal has been removed. The individual harmonics remaining can then be measured and compared to the original fundamental tone. The additional dynamic range boost afforded by this measurement technique will depend on the THD present in the amplifier under test, but 30 to 50 db is certainly possible.
To some extent we here at DIYAUDIO do think alike but not to the extent that we have nothing to share. Perhaps in this thread we have similar tools.
So here is a particular question. How do we use our dedicated test PC and Millett interface to measure an amplifier’s output impedance?
And a more general question. What amplifier performance parameters can we measure with this set of tools? (dedicated test PC and Millett interface)
BTW; it is interesting that the tools cost more than the $0.63 6BQ6’s that I have on the breadboard to test.
DT
All just for Fun!
So here is a particular question. How do we use our dedicated test PC and Millett interface to measure an amplifier’s output impedance?
And a more general question. What amplifier performance parameters can we measure with this set of tools? (dedicated test PC and Millett interface)
BTW; it is interesting that the tools cost more than the $0.63 6BQ6’s that I have on the breadboard to test.
DT
All just for Fun!
Wow that was some good info! I do happen to have a 331A, but never attempted using it as I heard what a PITA it was. Is it really that bad for infrequent usage? Does a PC setup offer a better alternative, or any alternative at all for THD measurements?
I'm still eyeballing that Rigol for a toy fix! Perhaps just the cheaper 50Mhz version. Reading suggest that the FFT funtion would not be as good as a Soundcard. Somethong about 8 bit resolution only offering a -70 db bottom.
I'm still eyeballing that Rigol for a toy fix! Perhaps just the cheaper 50Mhz version. Reading suggest that the FFT funtion would not be as good as a Soundcard. Somethong about 8 bit resolution only offering a -70 db bottom.
The children are all tucked snugly into their beds and I am having visions of oscillations in my head, Merry Christmas.
To visualize oscillation imagine that shopping cart we all have had the occasion to push through the produce section, just as we reach critical velocity that pesky front wheel begins flopping all about making the cart hard to push and creating a terrible noise. Magic? No.
It is not just high frequency current feedback op-amps and shopping cart wheels that oscillate. Even vacuum tubes will oscillate well into the Radio Frequency range. Check out the data sheets for the 807 and similar tubes. They can operate and oscillate well past 125 M Hz. A feedback loop can shift from negative to positive feedback due to phase shift. The wheel can fall off the cart and we can destroy the radio reception for two blocks away. We may not even know it at the time.
This is the thinking that makes me want to maximize the bandwidth of my next O-scope, one that can resolve RF oscillation.
DT
All just for fun!
To visualize oscillation imagine that shopping cart we all have had the occasion to push through the produce section, just as we reach critical velocity that pesky front wheel begins flopping all about making the cart hard to push and creating a terrible noise. Magic? No.
It is not just high frequency current feedback op-amps and shopping cart wheels that oscillate. Even vacuum tubes will oscillate well into the Radio Frequency range. Check out the data sheets for the 807 and similar tubes. They can operate and oscillate well past 125 M Hz. A feedback loop can shift from negative to positive feedback due to phase shift. The wheel can fall off the cart and we can destroy the radio reception for two blocks away. We may not even know it at the time.
This is the thinking that makes me want to maximize the bandwidth of my next O-scope, one that can resolve RF oscillation.
DT
All just for fun!
Tubelab: Thanks for taking the time and trouble to explain precisely why combining an 8 bit FFT with an analogue distortion meter works.
TubeMack: As far as oscilloscopes are concerned, there's no such thing as too much bandwidth. As you point out, valves can easily oscillate at tens of MHz, so a 100MHz 'scope is entirely justified (possibly even a bit marginal), and 300MHz is nice.
Merry Christmas, one and all.
TubeMack: As far as oscilloscopes are concerned, there's no such thing as too much bandwidth. As you point out, valves can easily oscillate at tens of MHz, so a 100MHz 'scope is entirely justified (possibly even a bit marginal), and 300MHz is nice.
Merry Christmas, one and all.
Not to throw sand in any one's gears, but there are a couple of HP 3581's for sale on the Bay for around $20 (one is a parts unit -- still valuable for the crystal filter and mixer) -- I was looking at a preamp yesterday with my refurbished analyzer -- 1kHz FFT -- and took it up to clipping. I measured each of the harmonics. Then switched on my HP 3581 and used it to manually measure each of the harmonics up to #6 -- wouldn't you know that the HP3581 is just as accurate! What the old fashioned test equipment can tell you is how much of the THD+N% is N. With any of this measurement equipment, don't scrimp on test cables.
A cherry one went for $120 at ETF this year. It's a wonderful piece of gear, low noise, versatile, and accurate. I use it routinely as a reality check against the sound card/ FFT software method; the two methods agree remarkably well. The variable bandwidth filters allow great flexibility in noise measurement. Every home lab should have one.
I don't see those HP's on the Bay. Someone must of read that and snagged them!
I reread Tubelabs post, and now see that the Analog Analyzer boosts the resolution of the 8 bit FFT so to speak. Thats good news if I understand it correctly
I don't want to continue to take this thread to far off topic, but since on the subject, I'd love to see the experts here post some more info, (In a new thread!), that perhaps could be a sticky disscussing all this. I've searched the forum quite a bit, and there's just bits, and trinkets here and there. Googling the net in general produces just a very few basic examples. Perhaps there's a good new, or old book on the subject?
I'd love to see more info on:
Most usefull new, and used (Ebay) equipment for a home lab.
Exactly what is all does, and how to use it,
What role can a PC set up play with these.
How to interconnect it all.
And a big one for me right now would be the specific connections, and adapters to purchase for best results.
For example:
Connecting to a audio generator with banana jacks, splitting the signal, sending it to one trace of the scope, and to the input of the amp. I see there are a few ways to do this, but whats the best? Use mainly RCA cables, or BNC cables? Make the adaptions at what point? Preferred method to make the split? Do you connect to the dummy load with the probes, or a dedicated adapter cable? Stuff like that. Maybe none of that matters, and thats being over thought? Perhaps everything matters? Us new guy's just don't know.
I reread Tubelabs post, and now see that the Analog Analyzer boosts the resolution of the 8 bit FFT so to speak. Thats good news if I understand it correctly
I don't want to continue to take this thread to far off topic, but since on the subject,
I'd love to see the experts here post some more info, (In a new thread!), that perhaps could be a sticky disscussing all this. I've searched the forum quite a bit, and there's just bits, and trinkets here and there. Googling the net in general produces just a very few basic examples. Perhaps there's a good new, or old book on the subject?I'd love to see more info on:
Most usefull new, and used (Ebay) equipment for a home lab.
Exactly what is all does, and how to use it,
What role can a PC set up play with these.
How to interconnect it all.
And a big one for me right now would be the specific connections, and adapters to purchase for best results.
For example:
Connecting to a audio generator with banana jacks, splitting the signal, sending it to one trace of the scope, and to the input of the amp. I see there are a few ways to do this, but whats the best? Use mainly RCA cables, or BNC cables? Make the adaptions at what point? Preferred method to make the split? Do you connect to the dummy load with the probes, or a dedicated adapter cable? Stuff like that. Maybe none of that matters, and thats being over thought? Perhaps everything matters? Us new guy's just don't know.
Last edited:
Well, I just opened a Christmas present, " Morgan Jones: Building Valve Amplifiers". Wanna guess what the back half of the book is about?
Perhaps it will answer some of my "connection" questions!
Well, your nice comment is a Christmas present for me as well - I hope you enjoy it.
Nice work. Of course, no two of us have the same ideas. I use a mic preamp now, and it is terrible. Most often, just probes made with resistors in them. Anyway, always need two channels at least. What I really wish is someone would so some PC scope software like Zalescope but to use 196K inputs like are now available for USB. That would save me from dragging out the scope every now and again.
Anyway, my idea is a buffer gain/attenuator stage that plugs directly into the front of something like a Firewire 610 with BNC's and high Z input so I can use my scope probes. Had not thought about an output buffer. Good idea. My sound card can not drive my MENG into clipping.
Anyway, my idea is a buffer gain/attenuator stage that plugs directly into the front of something like a Firewire 610 with BNC's and high Z input so I can use my scope probes. Had not thought about an output buffer. Good idea. My sound card can not drive my MENG into clipping.
To measure output impedance we need a signal source to drive the amplifier that produces a relatively clean sine wave of a constant amplitude. This can be the sound card interface, an audio signal generator, or a CD player with a test disc.
We also need a device to measure the output voltage that the amplifier produces. This can be the sound card interface, an AC voltmeter, a distortion analyzer, or even a cheap multimeter IF it has been tested and shown to be reasonably accurate at the frequency chosen for the test (usually 1KHz). Many cheap meters were designed to work at 50 or 60Hz, They may or may not read correctly at 1KHz (most of mine are OK) but very few are even close at 10KHz.
We also need at least two different values of load resistors, or a pair of identical resistors to test one channel. I commonly use a pair of 8 ohm resistors. One is used as an 8 ohm load and both in parallel are used for 4 ohms.
We operate the amplifier at a reasonable power level into the 8 ohm load and write down the voltage. I like to use 8 volts for easy arithmetic. Without changing anything else change the load impedance to 4 ohms. Write down the output voltage with the lower impedance load. The math to determine the output impedance involves solving two simultaneous equations based on thevinin equivalent resistance. I have an Excel page that does it, but it is on a computer sitting 1200 miles away in Florida. I can post it when I get home.
There is a much simpler method that is NOT generally recommended since it involves operating the amp without a load, or possibly a very low load impedance. It also requires a variable resistance capable of handling the amplifiers output power. It does illustrate Thevenin's theorem and can be used on low power tube amps.
Step 1, operate the amp at a low power with a 1KHz tone and NO LOAD (open circuit) and measure the output voltage. Keep the voltage very low to avoid possible damage and non-linearity which will corrupt the measurement. If a scope is available verify that the output remains a sine wave. I use 1 or 2 volts. Connect the variable load and adjust it until the output voltage drops to exactly HALF of the unloaded value. Remove the load and measure its resistance. By Thevenin's theorem this value will be equal to the output impedance of the amp.
Warning: many tube amplifiers can be damaged if they are operated at medium or high power without a load. The output impedance of many solid state amps is very low. Some solid state amps can be damaged if operated at normal power levels with very low load resistances.
I thought it was supposed to be a big secret.
The usual suspects are frequency response, distortion, and output voltage (which easilly computes to power output). Depending on the software being used one of these can be graphed against another, like distortion VS power output. What is more important is the ability to see a visual representation of the individual frequency components of the output spectrum. The THD analyzer will lump everything that is not the fundamental tone together and quantify it ALL as THD. We have no way of knowing if the "distortion" is all second harmonic (relatively benign) or a mixture of 5th, 7th, and 60Hz hum (real ugly sounding).
When I was designing the Tubelab SE I was trying to figure out what was causing a vague "fogginess" in the sound. I was using a 45 tube with AC powered filaments, and the hum balance pot had been adjusted to bring the hum below the measurement capability of my scope. It was at this time that I discovered the FFT analyzer. I quickly discovered that I was experiencing IMD distortion, but not the textbook IMD distortion created when two tones intermingle as they meet each other inside the amp. This was a single 1KHz tone entering the amp but the tones that came out of the amp (1KHz, 2KHz and 3KHz) each had 60 Hz sidebands (940 and 1060 Hz tones). These are not musically pleasing but not plainly audible since they were at a relatively low level, and "masked" by the fundamental tone. Operating the tubes filament on pure DC eliminated this effect, so the Tubelab SE uses DC on all filaments.
Further investigation revealed that the IMD can be nulled out with the hum balance pot, but now the hum was audible and very obvious on the FFT analyzer. I found this effect to be dependent on the output tube being used, and some tubes did exhibit a null in hum and IMD at the same point.
In many cases the operating point of the individual tubes can be adjusted to affect the levels of the individual harmonics. Many "tweakers" have made careers out of blindly optimizing the harmonic spectrum without fully understanding what is going on. Spend some serious time tweaking, listening and then measuring with the FFT and you can gain an understanding of some of the secrets of vacuum tube "magic".
I agree that there is no such thing as too much bandwidth, but bandwidth isn't cheap. My well used Tek is a 100 MHz scope. It has been pointed out that tubes can oscillate at VHF frequencies, and mosfets used as source followers to drive tube grids can oscillate well into the VHF region as well. Often this oscillation occurs at specific drive levels, sometimes just as the amp enters or leaves clipping. Invariably the oscillation will go away as soon as a scope probe gets stuck into the circuit. My experience as an RF engineer has offered me another solution. I use an RF spectrum analyzer to poke around inside every one of my tube amp designs. The little 5842 that I use in the Tubelab SE can easilly sing away at 400 MHz. Just touching the grid pins with the scope probe kills the oscillation, but it is plainly visible with the spectrum analyzer's input placed in the vicinity of the tube, without physical contact. It would be beyond the expectations of a casual amp builder to do this, but many of these oscillations can be detected with a radio or analog TV placed in close proximity to the amp.