I've used a Tek DPO4034 for a long time, eventually switched to a Tek DPO5034 ~7 years ago for which I paid ~$5000 back then. I'd consider it as a mid-end scope. I've also used a number of low-end scopes (the usual sub-$500 Siglent, Rigol scopes, and years ago even a TDS2xx etc.). Unfortunately I don't have a lot of experience on Agilent scopes.
Of course what you need strongly depends on what you're doing with it. I love long sample depth for automated captures (I've been doing DPA - differential power analysis - which requires sometimes million of captures, and segmented capture with 120M sample depth, a fast CPU that can write data at almost Gigabit speed to a network drive is super helpful in improving the acquisition time by an order of magnitude over what other scopes give you). I love "complex" measurements with statistics ("what's the latency standard deviation from this input signal to that output signal" etc. type). I use protocol triggers, protocol decoders for interactive debugging. I've used a scope to trigger on, and capture, a wide bandwidth RF signal (using a mixer) to debug a wideband artifact of a Wifi radio that happened every x thousand packets, and was hard to catch with a standard spectrum analyzer.
I don't do much analog stuff, but occasionally I need to check if a power supply rail is noisy or not.
In theory, the "cheap scopes" can do all of this as well. They are useful for interactive debugging (probe this testpoint, display some I2C transfers...), but you relatively quickly run into limitations of their protocol decoders or trigger logic (even though they do digital triggers, which in theory should be way more flexible). You can use them for automatic captures, but very quickly curse at them because they either crash every 10k transfers, or are just veeeery slow, or a combination of both. They can do measurements, but then miss exactly that type of measurements you just need. But of course they are a fraction of the cost.
I know people here hate Tek instruments - and yes, there are a few stupidities, especially around the UI of the DPO5xxx-series - but:
- They are rock solid stable, at least in my experience. They don't randomly crash, or worse: display wrong results.
- They have pretty decent analog frontends.
- They have protocols decoders that are solid (and work better than, for example, Saleae Logic decoders)
- They can easily be scripted (via SCPI) to do what I want them to do, _and_ they are stable when doing that.
- Their UI performance is okay, except for really long waveforms. This is actually my biggest issue - using the UI with 120M waveforms is borderline painful.
- Active probes are available on eBay
- I can use the Scope remotely over the network without limitations.
But the reason why I keep going back to this instrument is that it's versatile enough that it can do _all_ of this, and quickly switch between usage cases. For example, after finding test points for an I2C bus I can check the analog voltages, then quickly setup a long capture, and decode into a CSV file. I don't need to re-wire my setup to use a logic analyzer. I don't need to buy a different scope if I want an "analog-like experience" (instead I just press the DPX button).
On the negative side, I've never been a big fan of the integrated MSO port, some of the trigger limitations are annoying, and they seriously miss a few more physical keys (like the DPO7xxx have for trigger setup). But aside from that, I'm very happy with this instrument and haven't felt a need to upgrade in the past 7 years (which, combined with my TEA, means a lot).
(For the record I was also pretty happy with the DPO4034, but going back to it now I miss some of the flexibility that I've learned to love.)