FR measurements of tape machines

I just purchased a QA403 last week with the intention to eventually replace my aging AP S1. I generate a lot of FR graphs for my customers and while I’ve only been playing with the software a bit I can’t find a way to do sweeps with the generators. The Expo chip won’t work as there is a delay between in and out due to the physical spacing between the record and playbacks heads in the tape machines. I’m looking to sweep something like 30-20Khz with a fixed gain and graph the output like the attached. I tried the AMP FR automated test but it’s not quite what I need for tape. Can you point me in the right direction? Thanks,

Hi @jsantoro, you should be able to accomplish what you need. I think your FFT size is too small. But let’s walk through things in more detail because your task is a common one and key for folks to understand.

For the explanation below, I’m going to use a studio delay, set to 100% wet, no feedback, and 100 mS of delay.

First, let’s quantify the delay. For this, I run the output of the QA403 into the delay, and the output from the delay into the L+ input. The L- input has a shorting block in place. Then, I use the Automated Test->MISC Path Delay plugin.

The settings for the plugin are as follows. Note the -20 dBV level because I don’t want to clip the input. The noise floor generation can be enabled if you are dealing with a digital device that has a noise gate. The noise floor can be set low enough to keep the gate open. In this case, there is no noise gate.


The output from the plugin has some info on the graph. But really what we want to know is the estimated path delay. We can see that at the bottom of the page, and it’s 104 mS:

Now, let’s setup for a frequency response run.

  1. File->New Settings
  2. Increase FFT size to 64K
  3. In the GENERATORS section, select FREQ RESPONSE. Right click on that button to set the level appropriate for your setup. For the effects processor, I use -20 dBV.

Now run a single expo chirp (shift + spacebar). Switch to Time domain view (TIME button in DISPLAY control group). Also switch to OUTPUT. And then zoom in the front of the activity by click+dragging a rectangle with the mouse

Below we can see the TIME domain OUTPUT data (ie, the waveform leaving the analyzer), showing the activity starts at about 40 mS.

Next, press the INPUT button in the DISPLAY control group. Here, we can see the activity starts at about 150 mS. This is the audio waveform captured at the input of the analyzer. And since we had a 100 mS 100% wet delay in the path, we can see in the time domain that the signal was shifted by 100mS or so.

Now, press the time button again. This will unzoom everything. And then, on the unzoomed input time-domain waveform, click+drag to zoom in the tail of activity. Below we can see the chirp ends about 1.15 seconds or so. But more importantly, we can see the chirp doesn’t appear to be chopped off.

Next, push the FREQ button in the DISPLAY control group. This will compute the frequency response of the input. And we can see below, it’s relatively flat from 20 to 20 kHz.

If I activate the notch filter and a rolloff filter and run it again, the plot reflects the effects settings.

So, the key here is to use the time domain to make sure your captured waveform was indeed fully captured. If you have your FFT size set too small, then the chirp and delay won’t fit. Here’s a 4K FFT size (48ksps). This is the output waveform. You can see there’s about 50 mS of “room” left after the chirp for delay.

And here’s a 64K fft. You can see there’s about 400mS of “room” left after the chirp for delay.

And here’s a 256k fft. There’s about 1.4 seconds of “room” left after the chirp for delay. But, I’d use only as large of an FFT as needed to cover the delay. Large FFTs don’t always buy you much and they consume resources very quickly.

The summary is as follows: If your frequency response isn’t making sense, examine the time domain waveforms and ensure you are seeing what you expect. The chirp must be fully captured.

Please share a plot and DUT path delay if able. Thanks!

1 Like

Thanks Matt for the assistance. The delay when while recording and monitoring off the playback head is 87ms which is reasonable. I didn’t measure the gap and do the math against the tape speed but seems fine.

The weird part is that I tried the same path delay with the machine is source mode, not using tape at all, just in and out and I get 1.3 sec delay which is not possible.

After that I used 32K FFT as you suggested and the FR chirp looks fine. Here is the chip in Time and Freq domain with the machine in source mode.Looks flat and normal.

And here is the FR off the tape.

Need to clean it up a bit. I’ll try getting the traces into the Graphtool.Need to read up some.

Thanks for the help.

As a new user I can only embed 1 images. I’ll post the rest of them when restriction is lifted.

Hi @jsantoro can you explain the source mode a little more? If you still have it, please post the plot from the path delay. If DUT response is relatively flat, the path delay can give very good estimates. The more unflat the frequency response of the DUT, the harder it is for the path delay discern the true delay because the frequencies begin arriving at different times and you’ll see the blue energy detect trace go from a narrow spike to a smear.

Hi Matt,

Source mode on a tape machine bypasses the record and playback heads. What goes in comes right back out. It’s used to compare what the sources sounds like against what is being recorded and coming off the playback head. with no tape playing or being recorded is basically just amplifies the input signal a bit. There is no delay. On my post above that is the one while recording. The 87 ms delay is the distance between the record and playback head. This one below is in source mode, the machine is just sitting, not running any tape , just passing a signal through it. I don’t understand why is is any delay at all let alone 1.3 sec.

The FR response traces seem to be working fine:

The chirp seem to be fine also

Having some difficulty exporting a FR trace and importing into the Graphtool. A Time domain trace export and imports just fine, but it doesn’t like the FR CSV, dies at Line 6.The format appears diffeerent than the TD CSV.


Just an FYI, there is a great tape deck measurement software that is based on the Nakamichi T-100, that is in the Microsoft Store under NAK T-100. It is about $25. You would need to use an A/D converter with it like one of the Behringer units which are <$20…

Thanks, I was following that thread on Tapeheads. My problem with software based solution is calibration, fiddling with gain etc. I tried Audio tester with the Foscurite A/D. I do this most of the day so rely on discrete instruments. Trying to save bench space is main reason for going to the QA solution, plus the price is right. My Audio Precision takes up a bunch of room and is getting old. Right now the plots match my AP unit very well. What I can’t replace is my Sound Technology Tape Test System (and that is even older than the AP). Nothing I’ve found will do everything it does, I rely on it so much that I bought a backup.

All that you need is a “calibrated cassette” with 400 Hz and I think 3khz. It does have a little learning curve and is much easier to use with 3head cassette decks. I am going to use it before too long to see how one of my Pioneer cassette decks is doing after all of these years and put it on my youtube channel.

Yes, I think you are right this should change. This will be changed in the next release

I don’t understand why is is any delay at all let alone 1.3 sec.

This one is a bit more curious. The issue here seems to be that every so often, the QA40x hardware will start up off-by-one sample (~20uS) in the I2S stream. If it starts normally, then the delay is reported correctly. If it starts on the prior sample, then you see the peak is reported as the max value (65535) and the delay is assumed to be the max delay.

This will be addressed in the next release. In cases where the delay is 65535 samples, it will be reported as 0 samples.

What would you need the QA40x to do to replace it?

Hi Matt,
The ST1510 was geared towards tape which was the predominant recording medium when it was created in the 80’s. Most of my work is on reel to reel machines, some cassette and some tube gear. The ST does head azimuth, FR, MOL, W&F, Speed, Noise, channel separation, etc. I do have discrete W&F and AC millivolt meters and other test equipment that I use also and I like having redundancy. What it can’t do is create nice output, there is a video out that I connect to a video printer but that it. I bought an AP to replace it but I use it mostly for FR output and it takes up too much space just to use for that. Which led me to the QQ403. Take no space and it’s quite amazing for it’s size. Already I love the speed of the plotting as I can adjust pots and watch the output almost in real time. Is there a way that I can create any automated tests to spit out a plot?
Anyway the AP will probably be sold and I’ll start to base things on the QA.

Hi @JSantoro, Have you tried the wow and flutter visualizer yet? At some point it needs its own spectrum display and the ability to apply weighting. But it’d be nice to get your thoughts on that. (See Visualizer->Wow and Flutter).

What would it take for you to adjust azimuth with the QA40x. As I understand it, you are looking at phase differences between left and right channels, is that right? So, perhaps a Visualizer with a display that lets you tune in real time. And when the phase between L and R is identical you are done? Is this done at a single frequency, or any/many freqs?

Let me take a stab at some of the other items and as you work up the learning curve you can tell me if they fall short or not.

FR sounds like is OK.

MOL will probably be a bit involved, mostly learning on my end.

Speed could be done by using the Wow and Flutter plug-in. This will give very accurate frequency derived from interpolated zero-crossings over many cycles.

Noise should work out of the box.

Channel Separation should work out of the box.

Is there a way that I can create any automated tests to spit out a plot?

You can save configurations. For example, for measuring channel separation, presumably you will play a tape with a 1 kHz tone on the left channel only, capture the left and right channel, and then difference between those levels is your separation. So, in that case, prepare a measurement (FFT, weighting, full scale input levels, a left and right RMS measurement, and whatever else you need). And then save that as “Channel Separation”. And when you recall it, everything will be there as you like. Now, what would be nice is a MATH measurement, where you can ask for the delta between L and R. But that would give you a common starting point for a lot of measurements. And you can add a title to each graph (up to 4 lines–just enter in the Add Measurement box). And you can make the display dark on light (better for printing).

And if you are preparing a doc for a client, you can export (right click on display, select Copy as EMF) the main screen as a EMF (vector) for pasting in a Word or PowerPoint instead of PNG (raster). The EMF will carry the full resolution of the plot, even as you convert to PDF. So, no matter how much the client zooms in on your report, the plots will be crystal clear.

And if you know programming and have a work flow you need to repeat over an over, you can automate that too.

High-end tape machines are such impressive feats of electromechanical engineering…I am happy to learn whatever you can teach about these machines. :wink:

Thanks Matt,

Yes FR is great. I just tried the W&F visualizer, It seems to be accurate. Speed is fine also.Playing a 3Khz reference tape compares to my analog W&F meter which I just calibrated last month.An extra benefit of your Visualizer is that it does both channel at once, most of the old analog meter are single channel. I’d need a weighted measurement as that’s what we use.

Azimuth is typically done at a fixed frequency12K or whatever. But you would start at 1K to be sure the heads are not tilted enough to be 180deg out. And yes looking to get them in phase L+R. The scope visualizer may work if you could see both channels and trigger for a stable waveform. They waveforms move a bit back and forth at high frequencies given the nature of tape. I could use the display window in a pinch. I use a scope for it also and I’m not geting rid of it so not that big a deal. :slight_smile:

Is there a shortcut for expanding the horizontal time base in the display window in the Tme domain other than dragging? This is where I need to learn the interface.

All noted, thanks very much for posting.

Is there a shortcut for expanding the horizontal time base in the display window in the Tme domain other than dragging?

Currently, the process while viewing the time domain is “drag to zoom in until you see what you want” and then press TIME button again to zoom back out to the full waveform. Along with way, you can go back to your previous zoom level with the mouse back button OR the arrow that appears in the lower right. You can change the y scale with the YMIN and YMAX buttons in the AXIS control group.

In the freq domain, to get back to your unzoomed view, you press either the XLIN or XLOG button. And then zoom by dragging with mouse. As with the time-domain, you can go back to your previous zoom level with the mouse back button OR the arrow in the lower right.

1 Like

This thread is very much important for me, because I would like to calibrate Azimuth, Speed and Wow and Flutter on an old Sansui SC-3110 Tape Desk.

I already bought a cassette tape with test tones.

Is there a procedure to take measurements with QA403, please? Some instructions or reference materials I can follow to learn and apply to my case, please?

Thank you very much @matt and @jsantoro for your suggestions.

Was there any progress in the QA403 software to handle tape deck measurements?

I have the QA400 but it can’t handle the typical tape deck delays from input to output.
Depending on the tape speed and physical distance from record to playback head, the in/output delay can be as high as 400 ms.

Currently I own a hardware signal generator, analogue mV/dB meter, distortion meter and W&F meter.
I’m considering purchasing the QA403 if it can handle all the above flawlessly and generate nice plots.

Best regards from Belgium,

1 Like

The QA403 has no issues with the delay, you just need to make sure the FFT buffer is large enough for the slower speeds. Here’s a FR plot I did yesterday.


Yes, record/playback FR seems to work quite well.
Same for THD and noise.

I assume the QA403 can not handle a playback FR via test tones from a calibration tape?

What about W&F? Which standard is used: JIS, NAB, CCIR, DIN?
Matt mentioned a spectrum display for W&F which could be very helpful to detect which rotating part has an impact on W&F for open reel tape decks.


1 Like

Hi @DeAl66, currently there isn’t a way to collect an acquisition of test tones and convert that to a frequency response plot. For W&F, there’s not any weighting currently applied. More info at the link below. Note that the W&F follows the AES6-2008, just minus the weighting.