Jack:
The Boonton's seem to be too easy to acquire. I think I have 4 of them all told. The KH's residual distortion is around .0002% it seems, similar to the Tek. The Tek stuff was designed by the guys that started AP. The Boonton's servo loop for the frequency may be part of the limitation. On the spectrum analyzer you can see the skirts of the frequency shifting poking up on each side of the notch. The KH doesn't show this. There are some tricks in the KH that may be applicable to reducing the residual of the Boonton.
I have a Wavetek 7530 fft connected to the output of the Boonton. its a good combination. The analyzer removes the fundamental and the fft allows a close look at whatever I want to see. I have 3 7L5's which have some strengths as spectrum analyzers but are too slow for this application usually.
I would love to get the firmware for the Boonton revised to fix that problem but its more than I want to reverse engineer. On one of my Boontons the relay welded shut from the abuse. I had to contrive a replacement.
What software do you have to run the Boonton? I have the NI PCMCIA GPIB stuff for it (and my new laptop ditched the PCMCIA slot for an expresscard slot, drat) that I got to interface with my scope.
The Boonton's seem to be too easy to acquire. I think I have 4 of them all told. The KH's residual distortion is around .0002% it seems, similar to the Tek. The Tek stuff was designed by the guys that started AP. The Boonton's servo loop for the frequency may be part of the limitation. On the spectrum analyzer you can see the skirts of the frequency shifting poking up on each side of the notch. The KH doesn't show this. There are some tricks in the KH that may be applicable to reducing the residual of the Boonton.
I have a Wavetek 7530 fft connected to the output of the Boonton. its a good combination. The analyzer removes the fundamental and the fft allows a close look at whatever I want to see. I have 3 7L5's which have some strengths as spectrum analyzers but are too slow for this application usually.
I would love to get the firmware for the Boonton revised to fix that problem but its more than I want to reverse engineer. On one of my Boontons the relay welded shut from the abuse. I had to contrive a replacement.
What software do you have to run the Boonton? I have the NI PCMCIA GPIB stuff for it (and my new laptop ditched the PCMCIA slot for an expresscard slot, drat) that I got to interface with my scope.
The HP 400 a-l tube/SS meters all have the output off the meter circuit. I have been through most of them. Its a cleaver circuit with a lot of gain wide bandwidth (4 MHz) and stable with not too many parts. Thats why the meter is in the feedback circuit.
1audio said:
By the servo loop I take it to mean the error detection circuit? I don't think that they had the voltage controlled amplifiers like the SSM2018 "in the day". Some of the chips on that "source" board are truely unobtanium.
I just acquired the ProLogix USB-GBIP Controller so haven't set it to work yet. Conrad uses PowerBasic. I have VB6, VB.NET and QB. Softwire works in Visual Basic dot Net but has been obsoleted by National Instruments. One of the Softwire guys wrote a GPIB routine and I have the code around here somewhere.
I have a National Instruments GPIB Controller, but haven't used it in years.
I was playing around with my HP339A over the weekend --- nice but slow to use. The THD is around 0.0017% There is a wee bit of 60Hz feed through on the HP339A and I can't locate the source -- but you can see it as a spur on the fundamental.
The Boonton has a servo loop for the frequency that passes through the microprocessor. It has really good frequency accuracy long term but some jitter from the PLL circuits.
Jack:
Someone had created a small app for controlling the Boonton and posted it. I was going to have a programmer student working for me revise it but he got too swampped with school to make progress. I will try to find the link and post it or the program and source code if I can't find the link but it will be a few days.
Someone had created a small app for controlling the Boonton and posted it. I was going to have a programmer student working for me revise it but he got too swampped with school to make progress. I will try to find the link and post it or the program and source code if I can't find the link but it will be a few days.
Hi Demian,
-Chris
You do know how old and out of date that series is, don't you? The 33X series and 8903x types have evolved a fair amount from the earlier types.
-Chris
Hi Jack,
If I can get some FFT happening here, I'll check mine for the same 60 Hz spurs. Did you play with the ground lift switch? Might be some stuff in there creating leakage too.
-Chris
I use that model as well. The residual THD and noise is spec'd at 0.0018%. I see mostly noise.
At those levels the source may be difficult to find. Anything from a shield screw to some power transformer leakage could be your source.
If I can get some FFT happening here, I'll check mine for the same 60 Hz spurs. Did you play with the ground lift switch? Might be some stuff in there creating leakage too.
-Chris
Yes I do, and how expensive the new stuff is. But the point was a meter to use as an input conditioner for a DIY distortion analyzer. Not to get an old analyzer. (Earlier discussion than you were responding to)
The HP 331-334 are OK but the manually tuned ones will be a challenge to get optimally notched. And I'm not sure the S/N will be good enough to not be the limit in any measurements today. Plus calibration of the chain is a pain since the attenuation and sensitivity of each stage needs to be worked out. This is the problem I have with the Boonton- its distortion output autoranges and doesn't say what that range is, so my FFT is only relative. If its important to someone I have an HP333 somewhere around here that I can test, if I can find it.
I didn't like my HP 8903 much. Its power meter on the front was nice but the rest was clunkier to use than the Boonton.
Hi Demian,
The 333A and 334A are auto nulling once you get it close. The 339A (love it!) is not fast, but it is accurate (and auto nulling). The ranges are manually switched so you shouldn't have much trouble with figuring out what range you are on. The voltmeter on mine is deadly accurate. I am shocked to see how close it is actually. Don't sell these things short.
Since the 333 / 334 are 70's units, an overhaul of the power supply would be in order. I can not understand how these instruments stay in as good condition as they seem to. To much to ask for if you want to work with them without going over them. I would not touch the cal if I were you. They hold calibration very well.
I'd have to say that those were worth every single penny. On Ebay, you can sometimes get one at less than $100 USD. 339A's tend to run $400 + and will need switch cleaning at the very least. You do not spray anything into these things, beware or you will create issues. Considering that this includes a very good oscillator and high bandwidth AC voltmeter - not too shabby. The earlier models (33x) had even wider bandwidth I think. I have a 654A (oscillator) and 334A that I am going to fix up for exactly that reason.
-Chris
The 333A and 334A are auto nulling once you get it close. The 339A (love it!) is not fast, but it is accurate (and auto nulling). The ranges are manually switched so you shouldn't have much trouble with figuring out what range you are on. The voltmeter on mine is deadly accurate. I am shocked to see how close it is actually. Don't sell these things short.
Since the 333 / 334 are 70's units, an overhaul of the power supply would be in order. I can not understand how these instruments stay in as good condition as they seem to. To much to ask for if you want to work with them without going over them. I would not touch the cal if I were you. They hold calibration very well.
I'd have to say that those were worth every single penny. On Ebay, you can sometimes get one at less than $100 USD. 339A's tend to run $400 + and will need switch cleaning at the very least. You do not spray anything into these things, beware or you will create issues. Considering that this includes a very good oscillator and high bandwidth AC voltmeter - not too shabby. The earlier models (33x) had even wider bandwidth I think. I have a 654A (oscillator) and 334A that I am going to fix up for exactly that reason.
-Chris
The 331-334 will null to 500 KHz I think- matches the 200 CD oscillator. FWIW the 200CD special version for low distortion has its output turned down to 7V, otherwise its the same as the standard model. The 331-334 have a meter response to something like 2-4 MHz.
The 654 is a well made instrument, but its distortion floor is pretty high. It has a very flat response. Is that the one with the fancy attenuator?
The 654 is a well made instrument, but its distortion floor is pretty high. It has a very flat response. Is that the one with the fancy attenuator?
Hi Demian,
Those figures sound roughly correct. They are amazing instruments, especially when you stop and think about when they were designed and put into production.
-Chris
Those figures sound roughly correct. They are amazing instruments, especially when you stop and think about when they were designed and put into production.
Yes, yes, yes and yes. I plan to investigate the distortion aspect a little. There must be a cure for that. Modern parts may make all the difference. It does look pretty cool. I like these better than the 200 series. Having an output meter is very handy.
-Chris
The 650 series also have a lot of space in the boxes. The distortion relates to the AGC and the available gain in the loop. Years ago I started to look into replacing the circuit with a new one. Its harder than it may seem. And transistors are not substantially different. I got distracted by the Krohn Hite oscillators which are higher performance.
The error correction term is discussed in Bruce Hofer's paper for the AES in 1979. Bob Cordell referenced this in in his THD% analyzer from Audio Magazine decades ago. Bob's article is on his website. Hofer designed the SG505.
There is a Boonton 1110 "oscillator only" unit for sale on EBay at the moment -- you'll have to search under boonton as the URL isn't inserting correctly -- the starting bid is $14 which is ridiculously low for an instrument of this quality. If I bid for it my wife would shoot me. I think that the seller actually has 2 of these.
Demian -- my spectrum analyzer is really an HP3577 network analyzer -- I didn't notice the "uncertainty" -- you mentioned.
There is a Boonton 1110 "oscillator only" unit for sale on EBay at the moment -- you'll have to search under boonton as the URL isn't inserting correctly -- the starting bid is $14 which is ridiculously low for an instrument of this quality. If I bid for it my wife would shoot me. I think that the seller actually has 2 of these.
Demian -- my spectrum analyzer is really an HP3577 network analyzer -- I didn't notice the "uncertainty" -- you mentioned.
While it's not a professional solution, I am quite pleased with what a soundcard + FFT-analyser can do. I'm using a semiprofessional 8ch-I/O 24bit/96kHz card, not the newest design (AKM AK4524 codecs, MC33078 buffers, electrolytic coupling caps and all). Still I can have distortion, noise and alias components more than 110dB down in a loopback measurement, at 1kHz and 48k sampling rate. This worsens of course when test freq. increases and also varies with sapmling rate (40k...60k work best for this card). With a modern card, featuring 192kHz rates to get a decent 80++kHz audio bandwidth I guess the results would be really comparable to a typical hardware solutions (not AP, R&S etc precision levels), especially when using a notch filter. A collegue of mine has compared the results of a soundcard based system with good hardware, R&S UPL66, Sprectral Dynamics, HP (don't know the models, though) and found them very similar. Care has to be taken not to kill the inputs with overvoltages, though.
The qood thing with a software solution ist one can easily sync the generator to the analyser and if one uses an exact bin center frequency one can avoid using FFT windows and can get really sharp and skirtless needles, only a single bin in width. With averaging this gives quite a good resolution, especially when a prime multpiple of the bin bandwidth is used and the test frequency aliases and/or IM products with fs give close sidebands. For example, I use 1001.220703125Hz as a 1kHz test frequency at 48kHz fs for a 64k FFT, this is a prime multiple (1367) and the IM sidebands are only +-58Hz. The problem is to find a generator software which handles this input value correctly -- i.e. Samplitude does.
- Klaus
The qood thing with a software solution ist one can easily sync the generator to the analyser and if one uses an exact bin center frequency one can avoid using FFT windows and can get really sharp and skirtless needles, only a single bin in width. With averaging this gives quite a good resolution, especially when a prime multpiple of the bin bandwidth is used and the test frequency aliases and/or IM products with fs give close sidebands. For example, I use 1001.220703125Hz as a 1kHz test frequency at 48kHz fs for a 64k FFT, this is a prime multiple (1367) and the IM sidebands are only +-58Hz. The problem is to find a generator software which handles this input value correctly -- i.e. Samplitude does.
- Klaus
Attachments
This seems to be essentially the same technique given by Walt Jung and Scott Wurcer in ADI AppNote 245. It differs mainly in the way the signals are subtracted -- they used an instr. amp, Renardson uses current summing. Clever approach, without doubt. I used the Jung/Wurcer method, with additional INAs ahead of the main subtractor, to be able to measure differential amp I/O.gaetan8888 said:
- Klaus
Using a sound card w/ good FFT software is very appealing, however there are a few challenges. First is input conditioning. You need to protect the input from getting fried, you need to provide optimum range adjustment to get the full performance of the card and some type of passive analog output adjustment so the output signal is at its best performance. above the idea of using either an HP ACVTVM or and HP distortion analyzer for this were proposed and they both have merit. The ACVTVM however has too much internal distortion. It could be modified but it would be a big chore. The distortion analyzer has more potential but it needs careful review.
Alternatively an external passive attenuator, balanced, could be fabbed pretty easily and compensated for flat response. Adding a passive notch filter at a few key frequencies will extend the systems performance way down into the mud, limited only by the noise of the soundcard.
I have the Juli@ sound card which does 192/24 and seems to be close to the state of the art and only around $120.
Second challenge is software. I use Praxis and RMAA and get a lot of utility from them. However Praxis is pretty expensive and RMAA is not easy to use for anything other than soundcards. But its possible and RMAA is free.
Alternatively an external passive attenuator, balanced, could be fabbed pretty easily and compensated for flat response. Adding a passive notch filter at a few key frequencies will extend the systems performance way down into the mud, limited only by the noise of the soundcard.
I have the Juli@ sound card which does 192/24 and seems to be close to the state of the art and only around $120.
Second challenge is software. I use Praxis and RMAA and get a lot of utility from them. However Praxis is pretty expensive and RMAA is not easy to use for anything other than soundcards. But its possible and RMAA is free.
Hi Demian,
I intend to use RMAA and a sound card in the short term running off the monitor output from an HP 339A. Lowering the fundamental has a lot of potential to increase your range, and you still get an idea where your fundamental is. I would much rather tie a spectrum or signal analyzer to the project. Sound cards are too limited in their bandwidth in my estimation. Active circuitry can misbehave up in the MHz as well. You should be able to see this.
I did play with some DSOs with FFT as a feature. This was useful, but the resolution was not good enough to see 60 Hz skirts. Otherwise they did provide useful information.
-Chris
I intend to use RMAA and a sound card in the short term running off the monitor output from an HP 339A. Lowering the fundamental has a lot of potential to increase your range, and you still get an idea where your fundamental is. I would much rather tie a spectrum or signal analyzer to the project. Sound cards are too limited in their bandwidth in my estimation. Active circuitry can misbehave up in the MHz as well. You should be able to see this.
I did play with some DSOs with FFT as a feature. This was useful, but the resolution was not good enough to see 60 Hz skirts. Otherwise they did provide useful information.
-Chris
RMAA (the free version, I don't know anything about the pay version) doesn't like to work in a one channel mode or much out of its basic loopback mode. I have made it work, with difficulty.
I have a Picoscope that does a fair job on the output of an analyzer and is good to 50 MHz. My Wavetek FFT goes to 100KHz and that close to the limits of the analyzers. When I hook the 7L5 up I get the same results. On analog audio stuff thats working OK there isn't much to see way up there. Add digital and who knows? Digital by definition should have nothing. However bad execution can leave a lot of nothing up there.
Most of my analog stuff has a power bandwidth of 500KHz+ so I'm familiar with the problems. I was looking at a DC to 100MHz 40 dB gain amp recently. Input Z was 1 MegOhm. It was extreamly unstable and would oscillate if the terminated coaxes got too close to each other. And the oscillation was near the top of the 300 MHz range of my scope. Invisible when the BW was switched to 20 MHz.
I have a Picoscope that does a fair job on the output of an analyzer and is good to 50 MHz. My Wavetek FFT goes to 100KHz and that close to the limits of the analyzers. When I hook the 7L5 up I get the same results. On analog audio stuff thats working OK there isn't much to see way up there. Add digital and who knows? Digital by definition should have nothing. However bad execution can leave a lot of nothing up there.
Most of my analog stuff has a power bandwidth of 500KHz+ so I'm familiar with the problems. I was looking at a DC to 100MHz 40 dB gain amp recently. Input Z was 1 MegOhm. It was extreamly unstable and would oscillate if the terminated coaxes got too close to each other. And the oscillation was near the top of the 300 MHz range of my scope. Invisible when the BW was switched to 20 MHz.
Hi Demian,
One of my dreams would be to have an HP 3562, or like product along with a 3585. I do some RF work.
Until then, a sound card will have to do.
-Chris
One of my dreams would be to have an HP 3562, or like product along with a 3585. I do some RF work.
Until then, a sound card will have to do.
The answer is another 339A then. Either that or different software. I'll figure something out some time.
-Chris
Hello
Once a wild, I have seen, for sale , some HP-851B spectrum analyser, did it can do audio spectrum analysis ?
The little I found about it on the web are not clear, it seem to do RF frequency but I don't know if it's go down to audio spectrum and how low distortion it can analyse.
Thank
Gaetan
Once a wild, I have seen, for sale , some HP-851B spectrum analyser, did it can do audio spectrum analysis ?
The little I found about it on the web are not clear, it seem to do RF frequency but I don't know if it's go down to audio spectrum and how low distortion it can analyse.
Thank
Gaetan
- Status
- Not open for further replies.